ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-19 04:19:36 +00:00
Code Issues Packages Projects Releases Wiki Activity

Refactor device op implementations into impl subdirectory. (#420)

Browse Source 
* Move kernel implementation files under impl directory.

* Update examples paths.

* Update device kernel impl include paths.

* Update tensor operation instances include paths.

* Update profiler and tests include paths.

* Clang-format

* Update include paths for batched gemm reduce

* Refactor UnitTest ConvNDBwdWeight.

* Refactor fwd and bwd data convND UT.

* Fix used test macro.

* Fix include path.

* Fix include paths.

* Fix include paths in profiler and tests.

* Fix include paths.

Co-authored-by: Adam Osewski <aosewski@amd.com>

[ROCm/composable_kernel commit: 3048028897]
This commit is contained in:
Adam Osewski
2022-10-13 16:05:08 +02:00
committed by GitHub
parent 1dcaa3991f
commit c747be612f
305 changed files with 1153 additions and 884 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1037
include/ck/tensor_operation/gpu/device/impl/device_batched_contraction_multiple_d_xdl_cshuffle.hpp Normal file
View File

File diff suppressed because it is too large Load Diff

683
include/ck/tensor_operation/gpu/device/impl/device_batched_gemm_e_permute_xdl.hpp Normal file
View File

@@ -0,0 +1,683 @@
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_batched_gemm_e_permute.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
/*
* \brief Wrapper function of GridwiseGemm::Run to realize BatchedGEMM.
*
* \tparam ComputePtrOffsetOfBatch Class that computes the base pointer offsets of A, B, C matrix
* given the batch. For example, ComputePtrOffsetOfStridedBatch() computes the offsets of evenly
* strided batched, but we can easily extend to other layouts. The returned offset can be either \p
* index_t or \p long_index_t. If it returns \p long_index_t, we are not subject to the 2GB
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
* limitations.
*
* \tparam Block2ETileMap Block2ETileMap::CalculateBottomIndex() takes in id of a workgroup and
* returns the 2D index of the tile that it computes. \see
* GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r3::Run().
* \note Using \p ComputePtrOffsetOfBatch gives us the flexibility that 2 workgroups can compute 2
* tiles from different matrices. Keep in mind that these 2 matrices can share the same grid
* descriptor (like in BatchedGEMM), or use their own grid descriptors (in GroupedGemm). \link
* impl/device_conv3d_fwd_xdl_ndhwc_kzyxc_ndhwk.hpp kernel_gemm_xdlops_v2r3_for_conv3d \endlink for
\link
* DeviceConv3d \endlink uses the same concept, but currently does NOT encapsulate the computing of
* pointer offset into \p ComputePtrOffsetOfStridedBatch.
*
* \note \p Block2ETileMap allows customized mapping between a workgroup and the C-tile it computes.
* Together with \p ComputePtrOffsetOfBatch, we can reuse GridwiseGemm (and GridwiseGemm fusion ) to
* realize BatchedGemmCPermute and GroupedGemm (and the corresponding GEMM fusion).
*
*/
template <typename GridwiseGemm,
typename ABDataType,
typename EDataType,
typename AGridDesc_AK0_M_AK1,
typename BGridDesc_BK0_N_BK1,
typename EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
typename ComputePtrOffsetOfBatch,
typename Block2ETileMap,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_batched_gemm_e_permute_xdl(const ABDataType* __restrict__ p_a_grid,
const ABDataType* __restrict__ p_b_grid,
EDataType* __restrict__ p_e_grid,
const index_t batch_count,
const AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1,
const BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1,
const EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock
e_grid_desc_mblock_mperblock_nblock_nperblock,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const CDEElementwiseOperation cde_element_op,
const ComputePtrOffsetOfBatch compute_ptr_offset_of_batch,
const Block2ETileMap block_2_etile_map)
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__))
const index_t num_blocks_per_batch =
__builtin_amdgcn_readfirstlane(get_grid_size() / batch_count);
const index_t g_idx = __builtin_amdgcn_readfirstlane(get_block_1d_id() / num_blocks_per_batch);
const long_index_t a_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_ptr_offset_of_batch.GetAPtrOffset(g_idx)));
const long_index_t b_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_ptr_offset_of_batch.GetBPtrOffset(g_idx)));
const long_index_t e_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_ptr_offset_of_batch.GetCPtrOffset(g_idx)));
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
GridwiseGemm::template Run<HasMainKBlockLoop>(p_a_grid + a_batch_offset,
p_b_grid + b_batch_offset,
ck::Tuple<>{},
p_e_grid + e_batch_offset,
p_shared,
a_element_op,
b_element_op,
cde_element_op,
a_grid_desc_ak0_m_ak1,
b_grid_desc_bk0_n_bk1,
ck::Tuple<>{},
e_grid_desc_mblock_mperblock_nblock_nperblock,
block_2_etile_map);
#else
ignore = p_a_grid;
ignore = p_b_grid;
ignore = p_e_grid;
ignore = batch_count;
ignore = a_grid_desc_ak0_m_ak1;
ignore = b_grid_desc_bk0_n_bk1;
ignore = e_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = a_element_op;
ignore = b_element_op;
ignore = cde_element_op;
ignore = compute_ptr_offset_of_batch;
ignore = block_2_etile_map;
#endif
}
template <typename ALayout,
typename BLayout,
typename ELayout,
typename ADataType,
typename BDataType,
typename AccDataType,
typename CShuffleDataType,
typename EDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
GemmSpecialization GemmSpec,
index_t NumPrefetch,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock,
index_t KPerBlock,
index_t AK1,
index_t BK1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_K0_M_K1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_K1,
index_t ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_K0_N_K1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_K1,
index_t BBlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CDEBlockTransferScalarPerVector_NPerBlock,
LoopScheduler LoopSched = make_default_loop_scheduler()>
struct DeviceBatchedGemmEPermuteXdl : public DeviceBatchedGemmEPermute<ALayout,
BLayout,
ELayout,
ADataType,
BDataType,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation>
{
using DeviceOp = DeviceBatchedGemmEPermuteXdl;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto matrix_padder =
MatrixPadder<GemmSpec, index_t, index_t, index_t>{MPerBlock, NPerBlock, KPerBlock};
static auto MakeAGridDescriptor_M_K(index_t MRaw, index_t KRaw, index_t StrideA)
{
const auto a_grid_desc_mraw_kraw = [&]() {
if constexpr(is_same_v<tensor_layout::gemm::RowMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(StrideA, I1));
}
else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(I1, StrideA));
}
}();
return matrix_padder.PadADescriptor_M_K(a_grid_desc_mraw_kraw);
}
static auto MakeBGridDescriptor_N_K(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
return matrix_padder.PadBDescriptor_N_K(b_grid_desc_nraw_kraw);
}
static auto
MakeEGridDescriptor_M_N(index_t MRaw, index_t NRaw, index_t stride_M, index_t stride_N)
{
const auto e_grid_desc_mraw_nraw =
make_naive_tensor_descriptor(make_tuple(MRaw, NRaw), make_tuple(stride_M, stride_N));
return matrix_padder.PadCDescriptor_M_N(e_grid_desc_mraw_nraw);
}
static auto MakeEGridDescriptor_G0_G1_M_N(index_t G0,
index_t G1,
index_t MRaw,
index_t NRaw,
index_t stride_G0,
index_t stride_G1,
index_t stride_M,
index_t stride_N)
{
const auto e_grid_desc_g0_g1_mraw_nraw = [&]() {
return make_naive_tensor_descriptor(
make_tuple(G0, G1, MRaw, NRaw),
make_tuple(stride_G0, stride_G1, stride_M, stride_N));
}();
const auto M = math::integer_divide_ceil(MRaw, MPerBlock) * MPerBlock;
const auto N = math::integer_divide_ceil(NRaw, NPerBlock) * NPerBlock;
const auto MPad = M - MRaw;
const auto NPad = N - NRaw;
if constexpr(GemmSpec == GemmSpecialization::MNPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad M and N
return transform_tensor_descriptor(
e_grid_desc_g0_g1_mraw_nraw,
make_tuple(make_pass_through_transform(G0),
make_pass_through_transform(G1),
make_right_pad_transform(MRaw, MPad),
make_right_pad_transform(NRaw, NPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
}
else if constexpr(GemmSpec == GemmSpecialization::MPadding ||
GemmSpec == GemmSpecialization::MKPadding)
{
// pad M, but not N
return transform_tensor_descriptor(
e_grid_desc_g0_g1_mraw_nraw,
make_tuple(make_pass_through_transform(G0),
make_pass_through_transform(G1),
make_right_pad_transform(MRaw, MPad),
make_pass_through_transform(NRaw)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
}
else if constexpr(GemmSpec == GemmSpecialization::NPadding ||
GemmSpec == GemmSpecialization::NKPadding)
{
// pad N, but not M
return transform_tensor_descriptor(
e_grid_desc_g0_g1_mraw_nraw,
make_tuple(make_pass_through_transform(G0),
make_pass_through_transform(G1),
make_pass_through_transform(MRaw),
make_right_pad_transform(NRaw, NPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
}
else
{
// not pad M or N
return e_grid_desc_g0_g1_mraw_nraw;
}
}
using AGridDesc_M_K = decltype(MakeAGridDescriptor_M_K(1, 1, 1));
using BGridDesc_N_K = decltype(MakeBGridDescriptor_N_K(1, 1, 1));
using EGridDesc_M_N = decltype(MakeEGridDescriptor_M_N(1, 1, 1, 1));
using EGridDesc_G0_G1_M_N = decltype(MakeEGridDescriptor_G0_G1_M_N(1, 1, 1, 1, 1, 1, 1, 1));
struct ComputePtrOffsetOfStridedBatch
{
ComputePtrOffsetOfStridedBatch(index_t Batchstride_A,
index_t Batchstride_B,
EGridDesc_G0_G1_M_N e_grid_desc_g0_g1_m_n)
: Batchstride_A_(Batchstride_A),
Batchstride_B_(Batchstride_B),
e_grid_desc_g0_g1_m_n_(e_grid_desc_g0_g1_m_n)
{
}
__host__ __device__ constexpr long_index_t GetAPtrOffset(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(Batchstride_A_);
}
__host__ __device__ constexpr long_index_t GetBPtrOffset(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(Batchstride_B_);
}
__host__ __device__ constexpr long_index_t GetCPtrOffset(index_t g_idx) const
{
const index_t G1 = e_grid_desc_g0_g1_m_n_.GetLength(I1);
index_t b0 = g_idx / G1;
index_t b1 = g_idx - b0 * G1; // g_idx % G1
return e_grid_desc_g0_g1_m_n_.CalculateOffset(make_multi_index(b0, b1, 0, 0));
}
private:
index_t Batchstride_A_;
index_t Batchstride_B_;
EGridDesc_G0_G1_M_N e_grid_desc_g0_g1_m_n_;
};
using GridwiseGemm = GridwiseGemmMultipleD_xdl_cshuffle<
ADataType, // TODO: distinguish A/B datatype
AccDataType,
CShuffleDataType,
ck::Tuple<>, // DsDataType,
EDataType, // EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
InMemoryDataOperationEnum::Set,
AGridDesc_M_K,
BGridDesc_N_K,
Tuple<>,
EGridDesc_M_N,
NumPrefetch,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
AK1,
BK1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_K0_M_K1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_K1,
false, // AThreadTransferSrcResetCoordinateAfterRun,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_K0_N_K1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_K1,
false, // BThreadTransferSrcResetCoordinateAfterRun,
BBlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CDEBlockTransferScalarPerVector_NPerBlock,
LoopSched>;
using AGridDesc_AK0_M_AK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(AGridDesc_M_K{}))>;
using BGridDesc_BK0_N_BK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(BGridDesc_N_K{}))>;
using EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock = decltype(
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(EGridDesc_M_N{}));
using Block2ETileMap = typename GridwiseGemm::DefaultBlock2ETileMap;
// Argument
struct Argument : public BaseArgument
{
Argument(const ADataType* p_a_grid,
const BDataType* p_b_grid,
EDataType* p_e_grid,
index_t M,
index_t N,
index_t K,
index_t stride_A,
index_t stride_B,
index_t batch_stride_A,
index_t batch_stride_B,
BatchedGemmEPermuteDesc batched_gemm_e_permute_desc,
index_t BatchCount,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op)
: p_a_grid_{p_a_grid},
p_b_grid_{p_b_grid},
p_e_grid_{p_e_grid},
BatchCount_(BatchCount),
a_grid_desc_m_k_{DeviceOp::MakeAGridDescriptor_M_K(M, K, stride_A)},
b_grid_desc_n_k_{DeviceOp::MakeBGridDescriptor_N_K(K, N, stride_B)},
e_grid_desc_m_n_{
DeviceOp::MakeEGridDescriptor_M_N(batched_gemm_e_permute_desc.M_,
batched_gemm_e_permute_desc.N_,
batched_gemm_e_permute_desc.stride_M_,
batched_gemm_e_permute_desc.stride_N_)},
a_grid_desc_ak0_m_ak1_{
GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(a_grid_desc_m_k_)},
b_grid_desc_bk0_n_bk1_{
GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(b_grid_desc_n_k_)},
e_grid_desc_mblock_mperblock_nblock_nperblock{},
e_grid_desc_g0_g1_m_n_{
DeviceOp::MakeEGridDescriptor_G0_G1_M_N(batched_gemm_e_permute_desc.G0_,
batched_gemm_e_permute_desc.G1_,
batched_gemm_e_permute_desc.M_,
batched_gemm_e_permute_desc.N_,
batched_gemm_e_permute_desc.stride_G0_,
batched_gemm_e_permute_desc.stride_G1_,
batched_gemm_e_permute_desc.stride_M_,
batched_gemm_e_permute_desc.stride_N_)},
compute_ptr_offset_of_batch_{batch_stride_A, batch_stride_B, e_grid_desc_g0_g1_m_n_},
block_2_etile_map_{GridwiseGemm::MakeDefaultBlock2ETileMap(e_grid_desc_m_n_)},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
cde_element_op_{cde_element_op}
{
if(GridwiseGemm::CheckValidity(a_grid_desc_m_k_,
b_grid_desc_n_k_,
ck::Tuple<>{},
e_grid_desc_m_n_,
block_2_etile_map_))
{
e_grid_desc_mblock_mperblock_nblock_nperblock =
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
e_grid_desc_m_n_);
}
}
void Print() const
{
std::cout << "A[M, K]: " << a_grid_desc_m_k_ << std::endl;
std::cout << "B[N, K]: " << b_grid_desc_n_k_ << std::endl;
std::cout << "C[M, N]: " << e_grid_desc_m_n_ << std::endl;
}
// private:
// pointers
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
EDataType* p_e_grid_;
// batch count
index_t BatchCount_;
// tensor descriptors for problem definiton
AGridDesc_M_K a_grid_desc_m_k_;
BGridDesc_N_K b_grid_desc_n_k_;
EGridDesc_M_N e_grid_desc_m_n_;
// tensor descriptors for block/thread-wise copy
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock e_grid_desc_mblock_mperblock_nblock_nperblock;
EGridDesc_G0_G1_M_N e_grid_desc_g0_g1_m_n_;
// for calculating Batch offset
ComputePtrOffsetOfStridedBatch compute_ptr_offset_of_batch_;
// block-to-e-tile map
Block2ETileMap block_2_etile_map_;
// element-wise op
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CDEElementwiseOperation cde_element_op_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_m_k_,
arg.b_grid_desc_n_k_,
ck::Tuple<>{},
arg.e_grid_desc_m_n_,
arg.block_2_etile_map_))
{
throw std::runtime_error(
"wrong! GridwiseBatchedGemmCPermute_km_kn_m0m1n0n1_xdlops_v2r3 has invalid "
"setting");
}
const index_t grid_size =
arg.block_2_etile_map_.CalculateGridSize(arg.e_grid_desc_m_n_) * arg.BatchCount_;
const auto K =
arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) * arg.a_grid_desc_ak0_m_ak1_.GetLength(I2);
auto launch_kernel = [&](auto has_main_k_block_loop_) {
const auto kernel = kernel_batched_gemm_e_permute_xdl<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
EDataType,
remove_reference_t<DeviceOp::AGridDesc_AK0_M_AK1>,
remove_reference_t<DeviceOp::BGridDesc_BK0_N_BK1>,
typename GridwiseGemm::EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
ComputePtrOffsetOfStridedBatch,
remove_reference_t<Block2ETileMap>,
has_main_k_block_loop_>;
return launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_e_grid_,
arg.BatchCount_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.e_grid_desc_mblock_mperblock_nblock_nperblock,
arg.a_element_op_,
arg.b_element_op_,
arg.cde_element_op_,
arg.compute_ptr_offset_of_batch_,
arg.block_2_etile_map_);
};
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
return launch_kernel(integral_constant<bool, true>{});
}
else
{
return launch_kernel(integral_constant<bool, false>{});
}
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
return GridwiseGemm::CheckValidity(arg.a_grid_desc_m_k_,
arg.b_grid_desc_n_k_,
ck::Tuple<>{},
arg.e_grid_desc_m_n_,
arg.block_2_etile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const ADataType* p_a,
const BDataType* p_b,
EDataType* p_e,
index_t M,
index_t N,
index_t K,
index_t stride_A,
index_t stride_B,
index_t batch_stride_A,
index_t batch_stride_B,
BatchedGemmEPermuteDesc batched_gemm_e_permute_desc,
index_t BatchCount,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op)
{
return Argument{p_a,
p_b,
p_e,
M,
N,
K,
stride_A,
stride_B,
batch_stride_A,
batch_stride_B,
batched_gemm_e_permute_desc,
BatchCount,
a_element_op,
b_element_op,
cde_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_a,
const void* p_b,
void* p_e,
index_t M,
index_t N,
index_t K,
index_t stride_A,
index_t stride_B,
index_t batch_stride_A,
index_t batch_stride_B,
BatchedGemmEPermuteDesc batched_gemm_e_permute_desc,
index_t BatchCount,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op) override
{
return std::make_unique<Argument>(static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
static_cast<EDataType*>(p_e),
M,
N,
K,
stride_A,
stride_B,
batch_stride_A,
batch_stride_B,
batched_gemm_e_permute_desc,
BatchCount,
a_element_op,
b_element_op,
cde_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceBatchedGemmEPermuteXdl"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

747
include/ck/tensor_operation/gpu/device/impl/device_batched_gemm_gemm_xdl_cshuffle.hpp Normal file
View File

@@ -0,0 +1,747 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_batched_gemm_gemm.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_batched_gemm_gemm_xdl_cshuffle_v1.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
#include "ck/host_utility/io.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <typename GridwiseGemm,
typename FloatAB,
typename FloatC,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename AccElementwiseOperation,
typename B1ElementwiseOperation,
typename CElementwiseOperation,
typename AGridDesc_AK0_M_AK1,
typename BGridDesc_BK0_N_BK1,
typename B1GridDesc_BK0_N_BK1,
typename CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename Block2CTileMap,
typename ComputeBasePtrOfStridedBatch,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_gemm_gemm_xdl_cshuffle_v1(
const FloatAB* __restrict__ p_a_grid,
const FloatAB* __restrict__ p_b_grid,
const FloatAB* __restrict__ p_b1_grid,
FloatC* __restrict__ p_c_grid,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const AccElementwiseOperation acc_element_op,
const B1ElementwiseOperation b1_element_op,
const CElementwiseOperation c_element_op,
const AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1,
const BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1,
const B1GridDesc_BK0_N_BK1 b1_grid_desc_bk0_n_bk1,
const CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
c_grid_desc_mblock_mperblock_nblock_nperblock,
const Block2CTileMap block_2_ctile_map,
const index_t batch_count,
const ComputeBasePtrOfStridedBatch compute_base_ptr_of_batch)
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__))
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
const index_t num_blocks_per_batch =
__builtin_amdgcn_readfirstlane(get_grid_size() / batch_count);
const index_t g_idx = __builtin_amdgcn_readfirstlane(get_block_1d_id() / num_blocks_per_batch);
const long_index_t a_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetABasePtr(g_idx)));
const long_index_t b_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetBBasePtr(g_idx)));
const long_index_t b1_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetB1BasePtr(g_idx)));
const long_index_t c_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetCBasePtr(g_idx)));
GridwiseGemm::template Run<HasMainKBlockLoop>(p_a_grid + a_batch_offset,
p_b_grid + b_batch_offset,
p_b1_grid + b1_batch_offset,
p_c_grid + c_batch_offset,
p_shared,
a_element_op,
b_element_op,
acc_element_op,
b1_element_op,
c_element_op,
a_grid_desc_ak0_m_ak1,
b_grid_desc_bk0_n_bk1,
b1_grid_desc_bk0_n_bk1,
c_grid_desc_mblock_mperblock_nblock_nperblock,
block_2_ctile_map);
#else
ignore = p_a_grid;
ignore = p_b_grid;
ignore = p_b1_grid;
ignore = p_c_grid;
ignore = a_element_op;
ignore = b_element_op;
ignore = acc_element_op;
ignore = b1_element_op;
ignore = c_element_op;
ignore = a_grid_desc_ak0_m_ak1;
ignore = b_grid_desc_bk0_n_bk1;
ignore = b1_grid_desc_bk0_n_bk1;
ignore = c_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = block_2_ctile_map;
ignore = batch_count;
ignore = compute_base_ptr_of_batch;
#endif // end of if (defined(__gfx908__) || defined(__gfx90a__))
}
// Computes C = A * B0 * B1
// ^^^^^^ (Acc0)
// ^^^^^^^^^^^ (Acc1)
template <typename ALayout,
typename BLayout, // B0Layout
typename B1Layout,
typename CLayout,
typename ADataType,
typename BDataType,
typename B1DataType,
typename CDataType,
typename GemmAccDataType,
typename CShuffleDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename AccElementwiseOperation,
typename B1ElementwiseOperation,
typename CElementwiseOperation,
GemmSpecialization GemmSpec,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock, // Gemm0NPerBlock
index_t KPerBlock, // Gemm0KPerBlock
index_t Gemm1NPerBlock,
index_t Gemm1KPerBlock,
index_t AK1,
index_t BK1,
index_t B1K1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
index_t Gemm1NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
bool ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
bool BBlockLdsExtraN,
typename B1BlockTransferThreadClusterLengths_BK0_N_BK1,
typename B1BlockTransferThreadClusterArrangeOrder,
typename B1BlockTransferSrcAccessOrder,
index_t B1BlockTransferSrcVectorDim,
index_t B1BlockTransferSrcScalarPerVector,
index_t B1BlockTransferDstScalarPerVector_BK1,
bool B1BlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CShuffleBlockTransferScalarPerVector_NPerBlock,
LoopScheduler LoopSched = LoopScheduler::Default>
struct DeviceBatchedGemmGemm_Xdl_CShuffle : public DeviceBatchedGemmGemm<ALayout,
BLayout,
B1Layout,
CLayout,
ADataType,
BDataType,
B1DataType,
CDataType,
AElementwiseOperation,
BElementwiseOperation,
AccElementwiseOperation,
B1ElementwiseOperation,
CElementwiseOperation>
{
using DeviceOp = DeviceBatchedGemmGemm_Xdl_CShuffle;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto matrix_padder =
GemmGemmPadder<GemmSpec, index_t, index_t, index_t, index_t>{
MPerBlock, NPerBlock, KPerBlock, Gemm1NPerBlock};
static auto MakeAGridDescriptor_AK0_M_AK1(index_t MRaw, index_t KRaw, index_t StrideA)
{
const auto a_grid_desc_mraw_kraw = [&]() {
if constexpr(is_same_v<tensor_layout::gemm::RowMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(StrideA, I1));
}
else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(I1, StrideA));
}
}();
const auto a_grid_desc_m_k = matrix_padder.PadADescriptor_M_K(a_grid_desc_mraw_kraw);
const auto M = a_grid_desc_m_k.GetLength(I0);
const auto K = a_grid_desc_m_k.GetLength(I1);
const auto AK0 = K / AK1;
return transform_tensor_descriptor(a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_pass_through_transform(M)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
static auto MakeBGridDescriptor_BK0_N_BK1(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
const auto b_grid_desc_n_k = matrix_padder.PadBDescriptor_N_K(b_grid_desc_nraw_kraw);
const auto N = b_grid_desc_n_k.GetLength(I0);
const auto K = b_grid_desc_n_k.GetLength(I1);
const auto BK0 = K / BK1;
return transform_tensor_descriptor(b_grid_desc_n_k,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_pass_through_transform(N)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
// Args: Gemm1KRaw, Gemm1NRaw, StrideB1
static auto MakeB1GridDescriptor_BK0_N_BK1(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b1_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, B1Layout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, B1Layout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
const auto b1_grid_desc_n_k = matrix_padder.PadB1Descriptor_N_K(b1_grid_desc_nraw_kraw);
const auto N = b1_grid_desc_n_k.GetLength(I0);
const auto K = b1_grid_desc_n_k.GetLength(I1);
const auto B1K0 = K / B1K1;
return transform_tensor_descriptor(
b1_grid_desc_n_k,
make_tuple(make_unmerge_transform(make_tuple(B1K0, B1K1)),
make_pass_through_transform(N)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
static auto MakeCGridDescriptor_M_N(index_t MRaw, index_t NRaw, index_t StrideC)
{
const auto c_grid_desc_mraw_nraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(StrideC, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(I1, StrideC));
}
}();
return matrix_padder.PadCDescriptor_M_N(c_grid_desc_mraw_nraw);
}
struct ComputeBasePtrOfStridedBatch
{
ComputeBasePtrOfStridedBatch(index_t BatchStrideA,
index_t BatchStrideB,
index_t BatchStrideB1,
index_t BatchStrideC)
: BatchStrideA_(BatchStrideA),
BatchStrideB_(BatchStrideB),
BatchStrideB1_(BatchStrideB1),
BatchStrideC_(BatchStrideC)
{
}
__host__ __device__ constexpr long_index_t GetABasePtr(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideA_);
}
__host__ __device__ constexpr long_index_t GetBBasePtr(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideB_);
}
__host__ __device__ constexpr long_index_t GetB1BasePtr(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideB1_);
}
__host__ __device__ constexpr long_index_t GetCBasePtr(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideC_);
}
private:
index_t BatchStrideA_;
index_t BatchStrideB_;
index_t BatchStrideB1_;
index_t BatchStrideC_;
};
using AGridDesc_AK0_M_AK1 = decltype(MakeAGridDescriptor_AK0_M_AK1(1, 1, 1));
using BGridDesc_BK0_N_BK1 = decltype(MakeBGridDescriptor_BK0_N_BK1(1, 1, 1));
using B1GridDesc_BK0_N_BK1 = decltype(MakeB1GridDescriptor_BK0_N_BK1(1, 1, 1));
using CGridDesc_M_N = decltype(MakeCGridDescriptor_M_N(1, 1, 1));
// GridwiseGemm
using GridwiseGemm = GridwiseBatchedGemmGemm_Xdl_CShuffle<
ADataType, // TODO: distinguish A/B datatype
GemmAccDataType,
CShuffleDataType,
CDataType,
AElementwiseOperation,
BElementwiseOperation,
AccElementwiseOperation,
B1ElementwiseOperation,
CElementwiseOperation,
InMemoryDataOperationEnum::Set,
AGridDesc_AK0_M_AK1,
BGridDesc_BK0_N_BK1,
B1GridDesc_BK0_N_BK1,
CGridDesc_M_N,
NumGemmKPrefetchStage,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
Gemm1NPerBlock,
Gemm1KPerBlock,
AK1,
BK1,
B1K1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
Gemm1NXdlPerWave,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
true,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
true,
BBlockLdsExtraN,
B1BlockTransferThreadClusterLengths_BK0_N_BK1,
B1BlockTransferThreadClusterArrangeOrder,
B1BlockTransferSrcAccessOrder,
B1BlockTransferSrcVectorDim,
B1BlockTransferSrcScalarPerVector,
B1BlockTransferDstScalarPerVector_BK1,
false,
B1BlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CShuffleBlockTransferScalarPerVector_NPerBlock,
LoopSched>;
// Argument
struct Argument : public BaseArgument
{
Argument(const ADataType* p_a_grid,
const BDataType* p_b_grid,
const B1DataType* p_b1_grid,
CDataType* p_c_grid,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t Gemm1NRaw, // = ORaw
index_t Batch,
index_t StrideA,
index_t StrideB,
index_t StrideB1,
index_t StrideC,
index_t BatchStrideA,
index_t BatchStrideB,
index_t BatchStrideB1,
index_t BatchStrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
AccElementwiseOperation acc_element_op,
B1ElementwiseOperation b1_element_op,
CElementwiseOperation c_element_op)
: p_a_grid_{p_a_grid},
p_b_grid_{p_b_grid},
p_b1_grid_{p_b1_grid},
p_c_grid_{p_c_grid},
a_grid_desc_ak0_m_ak1_{DeviceOp::MakeAGridDescriptor_AK0_M_AK1(MRaw, KRaw, StrideA)},
b_grid_desc_bk0_n_bk1_{DeviceOp::MakeBGridDescriptor_BK0_N_BK1(KRaw, NRaw, StrideB)},
b1_grid_desc_bk0_n_bk1_{
DeviceOp::MakeB1GridDescriptor_BK0_N_BK1(NRaw, Gemm1NRaw, StrideB1)},
c_grid_desc_m_n_{DeviceOp::MakeCGridDescriptor_M_N(MRaw, Gemm1NRaw, StrideC)},
c_grid_desc_mblock_mperblock_nblock_nperblock_{},
block_2_ctile_map_{GridwiseGemm::MakeDefaultBlock2CTileMap(c_grid_desc_m_n_)},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
acc_element_op_{acc_element_op},
b1_element_op_{b1_element_op},
c_element_op_{c_element_op},
batch_count_(Batch),
compute_base_ptr_of_batch_{BatchStrideA, BatchStrideB, BatchStrideB1, BatchStrideC},
raw_lengths_m_n_k_o_{MRaw, NRaw, KRaw, Gemm1NRaw}
{
if(GridwiseGemm::CheckValidity(a_grid_desc_ak0_m_ak1_,
b_grid_desc_bk0_n_bk1_,
b1_grid_desc_bk0_n_bk1_,
c_grid_desc_m_n_,
block_2_ctile_map_))
{
c_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeCGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
c_grid_desc_m_n_);
}
}
void Print() const
{
std::cout << "A[AK0, M, AK1]: " << a_grid_desc_ak0_m_ak1_ << std::endl;
std::cout << "B0[BK0, N, BK1]: " << b_grid_desc_bk0_n_bk1_ << std::endl;
std::cout << "B1[BK0, N, BK1]: " << b1_grid_desc_bk0_n_bk1_ << std::endl;
std::cout << "C[M, N]: " << c_grid_desc_m_n_ << std::endl;
}
// private:
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
const B1DataType* p_b1_grid_;
CDataType* p_c_grid_;
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
B1GridDesc_BK0_N_BK1 b1_grid_desc_bk0_n_bk1_;
CGridDesc_M_N c_grid_desc_m_n_;
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
c_grid_desc_mblock_mperblock_nblock_nperblock_;
typename GridwiseGemm::DefaultBlock2CTileMap block_2_ctile_map_;
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
AccElementwiseOperation acc_element_op_;
B1ElementwiseOperation b1_element_op_;
CElementwiseOperation c_element_op_;
index_t batch_count_;
ComputeBasePtrOfStridedBatch compute_base_ptr_of_batch_;
// For robust IsSupportedArgument() check
std::vector<index_t> raw_lengths_m_n_k_o_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(!DeviceOp::IsSupportedArgument(arg))
{
throw std::runtime_error("wrong! unsupported argument");
}
const index_t grid_size =
arg.block_2_ctile_map_.CalculateGridSize(arg.c_grid_desc_m_n_) * arg.batch_count_;
// Gemm0_K
const auto K =
arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) * arg.a_grid_desc_ak0_m_ak1_.GetLength(I2);
float ave_time = 0;
auto launch_kernel = [&](auto has_main_k_block_loop_) {
const auto kernel = kernel_gemm_gemm_xdl_cshuffle_v1<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
AElementwiseOperation,
BElementwiseOperation,
AccElementwiseOperation,
B1ElementwiseOperation,
CElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
DeviceOp::B1GridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::DefaultBlock2CTileMap,
ComputeBasePtrOfStridedBatch,
has_main_k_block_loop_>;
return launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_b1_grid_,
arg.p_c_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.acc_element_op_,
arg.b1_element_op_,
arg.c_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.b1_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_ctile_map_,
arg.batch_count_,
arg.compute_base_ptr_of_batch_);
};
// Gemm1_K is split into Gemm1_K0/K1 where K1 is known at compile time, so we only need
// to concern Gemm0's loop
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
ave_time = launch_kernel(integral_constant<bool, true>{});
}
else
{
ave_time = launch_kernel(integral_constant<bool, false>{});
}
return ave_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
if(!(ck::get_device_name() == "gfx908" || ck::get_device_name() == "gfx90a"))
{
return false;
}
// Note: we need raw lengths since threadwise copy can not handle vector load when part of
// vector is out of bounds
const auto MRaw = arg.raw_lengths_m_n_k_o_[0];
const auto NRaw = arg.raw_lengths_m_n_k_o_[1];
const auto KRaw = arg.raw_lengths_m_n_k_o_[2];
const auto Gemm1NRaw = arg.raw_lengths_m_n_k_o_[3];
// Check scalar per vector requirement
const auto a_extent_lowest =
is_same_v<tensor_layout::gemm::RowMajor, ALayout> ? KRaw : MRaw;
const auto b_extent_lowest =
is_same_v<tensor_layout::gemm::RowMajor, BLayout> ? NRaw : KRaw;
const auto b1_extent_lowest =
is_same_v<tensor_layout::gemm::RowMajor, B1Layout> ? Gemm1NRaw : NRaw;
const auto c_extent_lowest =
is_same_v<tensor_layout::gemm::RowMajor, CLayout> ? Gemm1NRaw : MRaw;
if(!(a_extent_lowest % ABlockTransferSrcScalarPerVector == 0 &&
b_extent_lowest % BBlockTransferSrcScalarPerVector == 0 &&
b1_extent_lowest % B1BlockTransferSrcScalarPerVector == 0 &&
c_extent_lowest % CShuffleBlockTransferScalarPerVector_NPerBlock == 0))
{
return false;
}
return GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.b1_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const ADataType* p_a,
const BDataType* p_b,
const B1DataType* p_b1,
CDataType* p_c,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t Gemm1NRaw,
index_t Batch,
index_t StrideA,
index_t StrideB,
index_t StrideB1,
index_t StrideC,
index_t BatchStrideA,
index_t BatchStrideB,
index_t BatchStrideB1,
index_t BatchStrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
AccElementwiseOperation acc_element_op,
B1ElementwiseOperation b1_element_op,
CElementwiseOperation c_element_op)
{
return Argument{p_a, p_b, p_b1, p_c, MRaw,
NRaw, KRaw, Gemm1NRaw, Batch, StrideA,
StrideB, StrideB1, StrideC, BatchStrideA, BatchStrideB,
BatchStrideB1, BatchStrideC, a_element_op, b_element_op, acc_element_op,
b1_element_op, c_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,
const void* p_b,
const void* p_b1,
void* p_c,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t Gemm1NRaw,
index_t Batch,
index_t StrideA,
index_t StrideB,
index_t StrideB1,
index_t StrideC,
index_t BatchStrideA,
index_t BatchStrideB,
index_t BatchStrideB1,
index_t BatchStrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
AccElementwiseOperation acc_element_op,
B1ElementwiseOperation b1_element_op,
CElementwiseOperation c_element_op) override
{
return std::make_unique<Argument>(static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
static_cast<const B1DataType*>(p_b1),
static_cast<CDataType*>(p_c),
MRaw,
NRaw,
KRaw,
Gemm1NRaw,
Batch,
StrideA,
StrideB,
StrideB1,
StrideC,
BatchStrideA,
BatchStrideB,
BatchStrideB1,
BatchStrideC,
a_element_op,
b_element_op,
acc_element_op,
b1_element_op,
c_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceBatchedGemmGemm_Xdl_CShuffle"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< AK1 << ", "
<< BK1 << ", "
<< MPerBlock << ", "
<< Gemm1NPerBlock << ", "
<< Gemm1KPerBlock << ", "
<< B1K1 << ", "
<< getGemmSpecializationString(GemmSpec) << ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

716
include/ck/tensor_operation/gpu/device/impl/device_batched_gemm_multi_d_xdl.hpp Normal file
View File

@@ -0,0 +1,716 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_batched_gemm_multi_d.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
#include "ck/host_utility/io.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
/*
* \brief Wrapper function of GridwiseGemm::Run to realize BatchedGEMM.
*
* \tparam ComputePtrOffsetOfBatch Class that computes the base pointer offsets of A, B, C matrix
* given the batch. For example, ComputePtrOffsetOfStridedBatch() computes the offsets of evenly
* strided batched, but we can easily extend to other layouts. The returned offset can be either \p
* index_t or \p long_index_t. If it returns \p long_index_t, we are not subject to the 2GB
* limitations.
*
* \tparam Block2ETileMap Block2ETileMap::CalculateBottomIndex() takes in id of a workgroup and
* returns the 2D index of the tile that it computes. \see
* GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r3::Run().
*
* \note Using \p ComputePtrOffsetOfBatch gives us the flexibility that 2 workgroups can compute 2
* tiles from different matrices. Keep in mind that these 2 matrices can share the same grid
* descriptor (like in BatchedGEMM), or use their own grid descriptors (in GroupedGemm). \link
* impl/device_conv3d_fwd_xdl_ndhwc_kzyxc_ndhwk.hpp kernel_gemm_xdlops_v2r3_for_conv3d \endlink for
* \link DeviceConv3d \endlink uses the same concept, but currently does NOT encapsulate the
* computing of pointer offset into \p ComputePtrOffsetOfStridedBatch.
*
* \note \p Block2ETileMap allows customized mapping between a workgroup and the C-tile it computes.
* Together with \p ComputePtrOffsetOfBatch, we can reuse GridwiseGemm (and GridwiseGemm fusion ) to
* realize BatchedGemm and GroupedGemm (and the corresponding GEMM fusion).
*
*/
template <typename GridwiseGemm,
typename ABDataType,
typename DsPointer,
typename EDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
typename AGridDesc_AK0_M_AK1,
typename BGridDesc_BK0_N_BK1,
typename DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
typename ComputePtrOffsetOfBatch,
typename Block2ETileMap,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_batched_gemm_xdl(const ABDataType* __restrict__ p_a_grid,
const ABDataType* __restrict__ p_b_grid,
DsPointer p_ds_grid,
EDataType* __restrict__ p_e_grid,
const index_t batch_count,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const CDEElementwiseOperation cde_element_op,
const AGridDesc_AK0_M_AK1 a_grid_desc_k0_m_k1,
const BGridDesc_BK0_N_BK1 b_grid_desc_k0_n_k1,
const DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
ds_grid_desc_mblock_mperblock_nblock_nperblock,
const EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock
e_grid_desc_mblock_mperblock_nblock_nperblock_,
const ComputePtrOffsetOfBatch compute_ptr_offset_of_batch,
const Block2ETileMap block_2_etile_map)
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__))
const index_t num_blocks_per_batch =
__builtin_amdgcn_readfirstlane(get_grid_size() / batch_count);
const index_t g_idx = __builtin_amdgcn_readfirstlane(get_block_1d_id() / num_blocks_per_batch);
const long_index_t a_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_ptr_offset_of_batch.GetAPtrOffset(g_idx)));
const long_index_t b_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_ptr_offset_of_batch.GetBPtrOffset(g_idx)));
const long_index_t e_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_ptr_offset_of_batch.GetEPtrOffset(g_idx)));
const auto ds_batch_offset = compute_ptr_offset_of_batch.GetDsPtrOffset(g_idx);
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
DsPointer p_ds_grid_grp;
static constexpr index_t NumDTensor =
DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock::Size();
static_for<0, NumDTensor, 1>{}(
[&](auto i) { p_ds_grid_grp(i) = p_ds_grid[i] + ds_batch_offset[i]; });
GridwiseGemm::template Run<HasMainKBlockLoop>(p_a_grid + a_batch_offset,
p_b_grid + b_batch_offset,
p_ds_grid_grp,
p_e_grid + e_batch_offset,
p_shared,
a_element_op,
b_element_op,
cde_element_op,
a_grid_desc_k0_m_k1,
b_grid_desc_k0_n_k1,
ds_grid_desc_mblock_mperblock_nblock_nperblock,
e_grid_desc_mblock_mperblock_nblock_nperblock_,
block_2_etile_map);
#else
ignore = p_a_grid;
ignore = p_b_grid;
ignore = p_ds_grid;
ignore = p_e_grid;
ignore = batch_count;
ignore = a_grid_desc_k0_m_k1;
ignore = b_grid_desc_k0_n_k1;
ignore = ds_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = e_grid_desc_mblock_mperblock_nblock_nperblock_;
ignore = a_element_op;
ignore = b_element_op;
ignore = cde_element_op;
ignore = compute_ptr_offset_of_batch;
ignore = block_2_etile_map;
#endif
}
template <typename ALayout,
typename BLayout,
typename DsLayout,
typename ELayout,
typename ADataType,
typename BDataType,
typename AccDataType,
typename CShuffleDataType,
typename DsDataType,
typename EDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
GemmSpecialization GemmSpec,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock,
index_t KPerBlock,
index_t AK1,
index_t BK1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
bool ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
bool BBlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CDEBlockTransferScalarPerVector_NPerBlock,
LoopScheduler LoopSched = make_default_loop_scheduler()>
struct DeviceBatchedGemmMultiD_Xdl : public DeviceBatchedGemmMultiD<ALayout,
BLayout,
DsLayout,
ELayout,
ADataType,
BDataType,
DsDataType,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation>
{
using DeviceOp = DeviceBatchedGemmMultiD_Xdl;
static constexpr index_t NumDTensor = DsDataType::Size();
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr auto matrix_padder =
MatrixPadder<GemmSpec, index_t, index_t, index_t>{MPerBlock, NPerBlock, KPerBlock};
static auto MakeAGridDescriptor_M_K(index_t MRaw, index_t KRaw, index_t StrideA)
{
const auto a_grid_desc_mraw_kraw = [&]() {
if constexpr(is_same_v<tensor_layout::gemm::RowMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(StrideA, I1));
}
else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(I1, StrideA));
}
}();
return matrix_padder.PadADescriptor_M_K(a_grid_desc_mraw_kraw);
}
static auto MakeBGridDescriptor_N_K(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
return matrix_padder.PadBDescriptor_N_K(b_grid_desc_nraw_kraw);
}
template <typename ELay>
static auto MakeEGridDescriptor_M_N(index_t MRaw, index_t NRaw, index_t StrideE)
{
const auto e_grid_desc_mraw_nraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, ELay>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(StrideE, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, ELay>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(I1, StrideE));
}
}();
return matrix_padder.PadCDescriptor_M_N(e_grid_desc_mraw_nraw);
}
static auto MakeDsGridDescriptor_M_N(const std::array<index_t, NumDTensor>& MRaws,
const std::array<index_t, NumDTensor>& NRaws,
const std::array<index_t, NumDTensor>& DsStride)
{
return generate_tuple(
[&](auto i) {
using DLayout = remove_cvref_t<tuple_element_t<i.value, DsLayout>>;
return DeviceOp::MakeEGridDescriptor_M_N<DLayout>(MRaws[i], NRaws[i], DsStride[i]);
},
Number<NumDTensor>{});
}
using AGridDesc_M_K = decltype(MakeAGridDescriptor_M_K(1, 1, 1));
using BGridDesc_N_K = decltype(MakeBGridDescriptor_N_K(1, 1, 1));
using DsGridDesc_M_N = remove_cvref_t<decltype(MakeDsGridDescriptor_M_N({}, {}, {}))>;
using EGridDesc_M_N = decltype(MakeEGridDescriptor_M_N<ELayout>(1, 1, 1));
struct ComputePtrOffsetOfStridedBatch
{
ComputePtrOffsetOfStridedBatch(index_t BatchStrideA,
index_t BatchStrideB,
std::array<ck::index_t, NumDTensor> BatchStrideDs,
index_t BatchStrideE)
: BatchStrideA_(BatchStrideA),
BatchStrideB_(BatchStrideB),
BatchStrideDs_(BatchStrideDs),
BatchStrideE_(BatchStrideE)
{
}
__host__ __device__ constexpr long_index_t GetAPtrOffset(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideA_);
}
__host__ __device__ constexpr long_index_t GetBPtrOffset(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideB_);
}
__host__ __device__ constexpr auto GetDsPtrOffset(index_t g_idx) const
{
std::array<long_index_t, NumDTensor> ds_offset;
static_for<0, NumDTensor, 1>{}([&](auto i) {
ds_offset[i] = g_idx * static_cast<long_index_t>(BatchStrideDs_[i]);
});
return ds_offset;
}
__host__ __device__ constexpr long_index_t GetEPtrOffset(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideE_);
}
private:
index_t BatchStrideA_;
index_t BatchStrideB_;
std::array<ck::index_t, NumDTensor> BatchStrideDs_;
index_t BatchStrideE_;
};
using GridwiseGemm = GridwiseGemmMultipleD_xdl_cshuffle<
ADataType, // TODO: distinguish A/B datatype
AccDataType,
CShuffleDataType,
DsDataType,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
InMemoryDataOperationEnum::Set,
NumGemmKPrefetchStage,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
AK1,
BK1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
false,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
false,
BBlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CDEBlockTransferScalarPerVector_NPerBlock,
LoopSched>;
// desc for blockwise copy
using AGridDesc_AK0_M_AK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(AGridDesc_M_K{}))>;
using BGridDesc_BK0_N_BK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(BGridDesc_N_K{}))>;
using DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock = remove_cvref_t<decltype(
GridwiseGemm::MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(DsGridDesc_M_N{}))>;
using EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock = remove_cvref_t<decltype(
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(EGridDesc_M_N{}))>;
// block-to-e-tile map
using Block2ETileMap =
remove_cvref_t<decltype(GridwiseGemm::MakeDefaultBlock2ETileMap(EGridDesc_M_N{}))>;
// Argument
struct Argument : public BaseArgument
{
Argument(const void* p_a_grid,
const void* p_b_grid,
std::array<const void*, NumDTensor> p_ds_grid,
void* p_e_grid,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t Batch,
index_t StrideA,
index_t StrideB,
const std::array<ck::index_t, NumDTensor>& StrideDs,
index_t StrideE,
index_t BatchStrideA,
index_t BatchStrideB,
const std::array<ck::index_t, NumDTensor>& BatchStrideDs,
index_t BatchStrideE,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op)
: p_a_grid_{static_cast<const ADataType*>(p_a_grid)},
p_b_grid_{static_cast<const BDataType*>(p_b_grid)},
p_ds_grid_{},
p_e_grid_{static_cast<EDataType*>(p_e_grid)},
Batch_(Batch),
a_grid_desc_m_k_{DeviceOp::MakeAGridDescriptor_M_K(MRaw, KRaw, StrideA)},
b_grid_desc_n_k_{DeviceOp::MakeBGridDescriptor_N_K(KRaw, NRaw, StrideB)},
ds_grid_desc_m_n_{},
e_grid_desc_m_n_{DeviceOp::MakeEGridDescriptor_M_N<ELayout>(MRaw, NRaw, StrideE)},
a_grid_desc_ak0_m_ak1_{
GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(a_grid_desc_m_k_)},
b_grid_desc_bk0_n_bk1_{
GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(b_grid_desc_n_k_)},
ds_grid_desc_mblock_mperblock_nblock_nperblock_{},
e_grid_desc_mblock_mperblock_nblock_nperblock_{},
compute_ptr_offset_of_batch_{BatchStrideA, BatchStrideB, BatchStrideDs, BatchStrideE},
block_2_etile_map_{GridwiseGemm::MakeDefaultBlock2ETileMap(e_grid_desc_m_n_)},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
cde_element_op_{cde_element_op}
{
// populate pointer, desc for Ds
static_for<0, NumDTensor, 1>{}([&](auto i) {
using DLayout = remove_cvref_t<tuple_element_t<i.value, DsLayout>>;
using DDataType = remove_cvref_t<tuple_element_t<i.value, DsDataType>>;
// D pointer
p_ds_grid_(i) = static_cast<const DDataType*>(p_ds_grid[i]);
// D desc
ds_grid_desc_m_n_(i) =
DeviceOp::MakeEGridDescriptor_M_N<DLayout>(MRaw, NRaw, StrideDs[i]);
});
// populate desc for Ds/E
if(GridwiseGemm::CheckValidity(a_grid_desc_m_k_,
b_grid_desc_n_k_,
ds_grid_desc_m_n_,
e_grid_desc_m_n_,
block_2_etile_map_))
{
ds_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
ds_grid_desc_m_n_);
e_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
e_grid_desc_m_n_);
}
}
void Print() const
{
std::cout << "A[M, K]: " << a_grid_desc_m_k_ << std::endl;
std::cout << "B[N, K]: " << b_grid_desc_n_k_ << std::endl;
static_for<0, NumDTensor, 1>{}(
[&](auto i) { std::cout << "Ds[M, N]: " << ds_grid_desc_m_n_[i] << std::endl; });
std::cout << "E[M, N]: " << e_grid_desc_m_n_ << std::endl;
}
// private:
// pointers
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
typename GridwiseGemm::DsGridPointer p_ds_grid_;
EDataType* p_e_grid_;
// Batch
index_t Batch_;
// tensor descriptors for problem definiton
AGridDesc_M_K a_grid_desc_m_k_;
BGridDesc_N_K b_grid_desc_n_k_;
DsGridDesc_M_N ds_grid_desc_m_n_;
EGridDesc_M_N e_grid_desc_m_n_;
// tensor descriptors for block/thread-wise copy
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock
ds_grid_desc_mblock_mperblock_nblock_nperblock_;
EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock e_grid_desc_mblock_mperblock_nblock_nperblock_;
// for calculating batch offset
ComputePtrOffsetOfStridedBatch compute_ptr_offset_of_batch_;
// block-to-e-tile map
Block2ETileMap block_2_etile_map_;
// element-wise op
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CDEElementwiseOperation cde_element_op_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceBatchedGemmMultiD_Xdl::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_m_k_,
arg.b_grid_desc_n_k_,
arg.ds_grid_desc_m_n_,
arg.e_grid_desc_m_n_,
arg.block_2_etile_map_))
{
throw std::runtime_error("wrong! GridwiseGemm has invalid setting");
}
const index_t grid_size =
arg.block_2_etile_map_.CalculateGridSize(arg.e_grid_desc_m_n_) * arg.Batch_;
const auto K =
arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) * arg.a_grid_desc_ak0_m_ak1_.GetLength(I2);
auto launch_kernel = [&](auto has_main_k_block_loop) {
constexpr bool has_main_loop = has_main_k_block_loop.value;
const auto kernel =
kernel_batched_gemm_xdl<GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
typename GridwiseGemm::DsGridPointer,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
DeviceOp::DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
DeviceOp::EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
ComputePtrOffsetOfStridedBatch,
Block2ETileMap,
has_main_loop>;
return launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_ds_grid_,
arg.p_e_grid_,
arg.Batch_,
arg.a_element_op_,
arg.b_element_op_,
arg.cde_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.ds_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.e_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.compute_ptr_offset_of_batch_,
arg.block_2_etile_map_);
};
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
return launch_kernel(integral_constant<bool, true>{});
}
else
{
return launch_kernel(integral_constant<bool, false>{});
}
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static bool IsSupportedArgument(const Argument& arg)
{
if(!(ck::get_device_name() == "gfx908" || ck::get_device_name() == "gfx90a"))
{
return false;
}
return GridwiseGemm::CheckValidity(arg.a_grid_desc_m_k_,
arg.b_grid_desc_n_k_,
arg.ds_grid_desc_m_n_,
arg.e_grid_desc_m_n_,
arg.block_2_etile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const void* p_a,
const void* p_b,
const std::array<const void*, NumDTensor>& p_ds,
void* p_e,
index_t M,
index_t N,
index_t K,
index_t Batch,
index_t StrideA,
index_t StrideB,
const std::array<index_t, NumDTensor>& StrideDs,
index_t StrideE,
index_t BatchStrideA,
index_t BatchStrideB,
const std::array<ck::index_t, NumDTensor>& BatchStrideDs,
index_t BatchStrideE,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op)
{
return Argument{p_a,
p_b,
p_ds,
p_e,
M,
N,
K,
Batch,
StrideA,
StrideB,
StrideDs,
StrideE,
BatchStrideA,
BatchStrideB,
BatchStrideDs,
BatchStrideE,
a_element_op,
b_element_op,
cde_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_a,
const void* p_b,
const std::array<const void*, NumDTensor>& p_ds,
void* p_e,
index_t M,
index_t N,
index_t K,
index_t Batch,
index_t StrideA,
index_t StrideB,
const std::array<ck::index_t, NumDTensor>& StrideDs,
index_t StrideE,
index_t BatchStrideA,
index_t BatchStrideB,
const std::array<ck::index_t, NumDTensor>& BatchStrideDs,
index_t BatchStrideE,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op) override
{
return std::make_unique<Argument>(p_a,
p_b,
p_ds,
p_e,
M,
N,
K,
Batch,
StrideA,
StrideB,
StrideDs,
StrideE,
BatchStrideA,
BatchStrideB,
BatchStrideDs,
BatchStrideE,
a_element_op,
b_element_op,
cde_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceBatchedGemmMultiD_Xdl"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock
<< AK1 << ", "
<< BK1 << ", "
<< getGemmSpecializationString(GemmSpec)
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

951
include/ck/tensor_operation/gpu/device/impl/device_batched_gemm_multiple_d_gemm_multiple_d_xdl_cshuffle.hpp Normal file
View File

@@ -0,0 +1,951 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_batched_gemm_multiple_d_gemm_multiple_d.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_batched_gemm_multiple_d_gemm_multiple_d_xdl_cshuffle_v1.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <typename GridwiseGemm,
typename A0B0B1DataType,
typename D0sPointer,
typename D1sPointer,
typename E1DataType,
typename A0ElementwiseOperation,
typename B0ElementwiseOperation,
typename CDE0ElementwiseOperation,
typename B1ElementwiseOperation,
typename CDE1ElementwiseOperation,
typename A0GridDesc_AK0_M_AK1,
typename B0GridDesc_BK0_N_BK1,
typename D0sGridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5,
typename B1GridDesc_BK0_N_BK1,
typename D1sGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename E1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename Block2E1TileMap,
typename ComputeBasePtrOfStridedBatch,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_batched_gemm_gemm_xdl_cshuffle_v1(
const A0B0B1DataType* __restrict__ p_a0_grid,
const A0B0B1DataType* __restrict__ p_b0_grid,
D0sPointer p_d0s_grid,
const A0B0B1DataType* __restrict__ p_b1_grid,
D1sPointer p_d1s_grid,
E1DataType* __restrict__ p_e1_grid,
const A0ElementwiseOperation a0_element_op,
const B0ElementwiseOperation b0_element_op,
const CDE0ElementwiseOperation cde0_element_op,
const B1ElementwiseOperation b1_element_op,
const CDE1ElementwiseOperation cde1_element_op,
const A0GridDesc_AK0_M_AK1 a0_grid_desc_ak0_m_ak1,
const B0GridDesc_BK0_N_BK1 b0_grid_desc_bk0_n_bk1,
const D0sGridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5
d0s_griddesc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5,
const B1GridDesc_BK0_N_BK1 b1_grid_desc_bk0_n_bk1,
const D1sGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
d1s_grid_desc_mblock_mperblock_nblock_nperblock,
const E1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
e1_grid_desc_mblock_mperblock_nblock_nperblock,
const Block2E1TileMap block_2_e1tile_map,
const index_t batch_count,
const ComputeBasePtrOfStridedBatch compute_base_ptr_of_batch)
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__))
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
const index_t num_blocks_per_batch =
__builtin_amdgcn_readfirstlane(get_grid_size() / batch_count);
const index_t g_idx = __builtin_amdgcn_readfirstlane(get_block_1d_id() / num_blocks_per_batch);
const long_index_t a_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetABasePtr(g_idx)));
const long_index_t b_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetBBasePtr(g_idx)));
const long_index_t b1_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetB1BasePtr(g_idx)));
const long_index_t c_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetCBasePtr(g_idx)));
static_for<0, p_d0s_grid.Size(), 1>{}([&](auto In) {
const long_index_t d0_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetD0BasePtr(g_idx, In)));
p_d0s_grid(In) = p_d0s_grid(In) + d0_batch_offset;
});
static_for<0, p_d1s_grid.Size(), 1>{}([&](auto In) {
const long_index_t d1_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetD1BasePtr(g_idx, In)));
p_d1s_grid(In) = p_d1s_grid(In) + d1_batch_offset;
});
GridwiseGemm::template Run<HasMainKBlockLoop>(p_a0_grid + a_batch_offset,
p_b0_grid + b_batch_offset,
p_d0s_grid,
p_b1_grid + b1_batch_offset,
p_d1s_grid,
p_e1_grid + c_batch_offset,
p_shared,
a0_element_op,
b0_element_op,
cde0_element_op,
b1_element_op,
cde1_element_op,
a0_grid_desc_ak0_m_ak1,
b0_grid_desc_bk0_n_bk1,
d0s_griddesc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5,
b1_grid_desc_bk0_n_bk1,
d1s_grid_desc_mblock_mperblock_nblock_nperblock,
e1_grid_desc_mblock_mperblock_nblock_nperblock,
block_2_e1tile_map);
#else
ignore = p_a0_grid;
ignore = p_b0_grid;
ignore = p_d0s_grid;
ignore = p_b1_grid;
ignore = p_d1s_grid;
ignore = p_e1_grid;
ignore = a0_element_op;
ignore = b0_element_op;
ignore = cde0_element_op;
ignore = b1_element_op;
ignore = cde1_element_op;
ignore = a0_grid_desc_ak0_m_ak1;
ignore = b0_grid_desc_bk0_n_bk1;
ignore = d0s_griddesc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5;
ignore = b1_grid_desc_bk0_n_bk1;
ignore = d1s_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = e1_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = block_2_e1tile_map;
ignore = batch_count;
ignore = compute_base_ptr_of_batch;
#endif
}
// Computes C = A * B0 * B1
// ^^^^^^ (Acc0)
// ^^^^^^^^^^^ (Acc1)
template <typename A0Layout,
typename B0Layout, // B0Layout
typename D0sLayout,
typename B1Layout,
typename D1sLayout,
typename E1Layout,
typename A0DataType,
typename B0DataType,
typename Acc0DataType,
typename D0sDataType,
typename B1DataType,
typename Acc1DataType,
typename C1ShuffleDataType,
typename D1sDataType,
typename E1DataType,
typename A0ElementwiseOperation,
typename B0ElementwiseOperation,
typename CDE0ElementwiseOperation,
typename B1ElementwiseOperation,
typename CDE1ElementwiseOperation,
bool PadGemm0M,
bool PadGemm0N,
bool PadGemm0K,
bool PadGemm1N,
bool PadGemm1K,
index_t NumGemm0KPrefetchStage,
index_t BlockSize,
index_t Gemm0MPerBlock,
index_t Gemm0NPerBlock,
index_t Gemm0KPerBlock,
index_t Gemm1NPerBlock,
index_t Gemm1KPerBlock,
index_t A0K1,
index_t B0K1,
index_t B1K1,
index_t Gemm0MPerXdl,
index_t Gemm0NPerXdl,
index_t Gemm0MXdlPerWave,
index_t Gemm0NXdlPerWave,
index_t Gemm1NXdlPerWave,
typename A0BlockTransferThreadClusterLengths_AK0_M_AK1,
typename A0BlockTransferThreadClusterArrangeOrder,
typename A0BlockTransferSrcAccessOrder,
index_t A0BlockTransferSrcVectorDim,
index_t A0BlockTransferSrcScalarPerVector,
index_t A0BlockTransferDstScalarPerVector_AK1,
bool A0BlockLdsExtraM,
typename B0BlockTransferThreadClusterLengths_BK0_N_BK1,
typename B0BlockTransferThreadClusterArrangeOrder,
typename B0BlockTransferSrcAccessOrder,
index_t B0BlockTransferSrcVectorDim,
index_t B0BlockTransferSrcScalarPerVector,
index_t B0BlockTransferDstScalarPerVector_BK1,
bool B0BlockLdsExtraN,
typename B1BlockTransferThreadClusterLengths_BK0_N_BK1,
typename B1BlockTransferThreadClusterArrangeOrder,
typename B1BlockTransferSrcAccessOrder,
index_t B1BlockTransferSrcVectorDim,
index_t B1BlockTransferSrcScalarPerVector,
index_t B1BlockTransferDstScalarPerVector_BK1,
bool B1BlockLdsExtraN,
index_t C1ShuffleMXdlPerWavePerShuffle,
index_t C1ShuffleGemm0NXdlPerWavePerShuffle,
typename CDE1ShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CDE1ShuffleBlockTransferScalarPerVector_NPerBlock,
LoopScheduler LoopSched = LoopScheduler::Default>
struct DeviceBatchedGemmMultipleDGemmMultipleD_Xdl_CShuffle
: public DeviceBatchedGemmMultipleDGemmMultipleD<A0Layout,
B0Layout,
D0sLayout,
B1Layout,
D1sLayout,
E1Layout,
A0DataType,
B0DataType,
D0sDataType,
B1DataType,
D1sDataType,
E1DataType,
A0ElementwiseOperation,
B0ElementwiseOperation,
CDE0ElementwiseOperation,
B1ElementwiseOperation,
CDE1ElementwiseOperation>
{
using DeviceOp = DeviceBatchedGemmMultipleDGemmMultipleD_Xdl_CShuffle;
static constexpr index_t NumD0Tensor = D0sDataType::Size();
static constexpr index_t NumD1Tensor = D1sDataType::Size();
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr auto I4 = Number<4>{};
static constexpr auto I5 = Number<5>{};
static constexpr auto I6 = Number<6>{};
static constexpr auto I7 = Number<7>{};
static constexpr auto I8 = Number<8>{};
static constexpr auto I9 = Number<9>{};
static constexpr auto gemm0_padder =
GemmPadder_v2<PadGemm0M, PadGemm0N, PadGemm0K, index_t, index_t, index_t>{
Gemm0MPerBlock, Gemm0NPerBlock, Gemm0KPerBlock};
static constexpr auto gemm1_padder =
GemmPadder_v2<PadGemm0M, PadGemm1N, PadGemm1K, index_t, index_t, index_t>{
Gemm0MPerBlock, Gemm1NPerBlock, Gemm1KPerBlock};
// for Gemm0
static auto MakeA0GridDescriptor_M_K(index_t MRaw, index_t KRaw, index_t StrideA0)
{
const auto a0_grid_desc_mraw_kraw = [&]() {
if constexpr(is_same_v<tensor_layout::gemm::RowMajor, A0Layout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(StrideA0, I1));
}
else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, A0Layout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(I1, StrideA0));
}
}();
return gemm0_padder.PadADescriptor_M_K(a0_grid_desc_mraw_kraw);
}
// for Gemm0
static auto MakeB0GridDescriptor_N_K(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b0_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, B0Layout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, B0Layout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
return gemm0_padder.PadBDescriptor_N_K(b0_grid_desc_nraw_kraw);
}
// for Gemm0
template <typename DLay>
static auto MakeD0GridDescriptor_M_N(index_t MRaw, index_t NRaw, index_t StrideD0)
{
const auto d0_grid_desc_mraw_nraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, DLay>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(StrideD0, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, DLay>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(I1, StrideD0));
}
}();
return gemm0_padder.PadCDescriptor_M_N(d0_grid_desc_mraw_nraw);
}
// for Gemm1
static auto MakeB1GridDescriptor_N_K(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b1_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, B1Layout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, B1Layout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
return gemm1_padder.PadBDescriptor_N_K(b1_grid_desc_nraw_kraw);
}
// for Gemm1
template <typename ELay>
static auto MakeE1GridDescriptor_M_N(index_t MRaw, index_t NRaw, index_t StrideE1)
{
const auto e1_grid_desc_mraw_nraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, ELay>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(StrideE1, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, ELay>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(I1, StrideE1));
}
}();
return gemm1_padder.PadCDescriptor_M_N(e1_grid_desc_mraw_nraw);
}
static auto MakeD0sGridDescriptor_M_N(const std::array<index_t, NumD1Tensor>& MRaws,
const std::array<index_t, NumD1Tensor>& NRaws,
const std::array<index_t, NumD1Tensor>& DsStride)
{
return generate_tuple(
[&](auto i) {
using DLayout = remove_cvref_t<tuple_element_t<i.value, D0sLayout>>;
return DeviceOp::MakeD0GridDescriptor_M_N<DLayout>(MRaws[i], NRaws[i], DsStride[i]);
},
Number<NumD0Tensor>{});
}
static auto MakeD1sGridDescriptor_M_N(const std::array<index_t, NumD1Tensor>& MRaws,
const std::array<index_t, NumD1Tensor>& NRaws,
const std::array<index_t, NumD1Tensor>& DsStride)
{
return generate_tuple(
[&](auto i) {
using DLayout = remove_cvref_t<tuple_element_t<i.value, D1sLayout>>;
return DeviceOp::MakeE1GridDescriptor_M_N<DLayout>(MRaws[i], NRaws[i], DsStride[i]);
},
Number<NumD1Tensor>{});
}
struct ComputeBasePtrOfStridedBatch
{
ComputeBasePtrOfStridedBatch(index_t BatchStrideA0,
index_t BatchStrideB0,
std::array<index_t, NumD0Tensor> BatchStrideD0s,
index_t BatchStrideB1,
std::array<index_t, NumD1Tensor> BatchStrideD1s,
index_t BatchStrideE1)
: BatchStrideA0_(BatchStrideA0),
BatchStrideB0_(BatchStrideB0),
BatchStrideD0s_(BatchStrideD0s),
BatchStrideB1_(BatchStrideB1),
BatchStrideD1s_(BatchStrideD1s),
BatchStrideE1_(BatchStrideE1)
{
}
__host__ __device__ constexpr long_index_t GetABasePtr(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideA0_);
}
__host__ __device__ constexpr long_index_t GetBBasePtr(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideB0_);
}
template <index_t I>
__host__ __device__ constexpr long_index_t GetD0BasePtr(index_t g_idx,
Number<I> d1_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideD0s_[d1_idx]);
}
__host__ __device__ constexpr long_index_t GetB1BasePtr(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideB1_);
}
__host__ __device__ constexpr long_index_t GetCBasePtr(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideE1_);
}
template <index_t I>
__host__ __device__ constexpr auto GetD1BasePtr(index_t g_idx, Number<I> d1_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideD1s_[d1_idx]);
}
private:
index_t BatchStrideA0_;
index_t BatchStrideB0_;
std::array<index_t, NumD0Tensor> BatchStrideD0s_;
index_t BatchStrideB1_;
std::array<index_t, NumD1Tensor> BatchStrideD1s_;
index_t BatchStrideE1_;
};
using A0GridDesc_M_K = decltype(MakeA0GridDescriptor_M_K(1, 1, 1));
using B0GridDesc_N_K = decltype(MakeB0GridDescriptor_N_K(1, 1, 1));
using D0sGridDesc_M_N = remove_cvref_t<decltype(MakeD0sGridDescriptor_M_N({}, {}, {}))>;
using B1GridDesc_N_K = decltype(MakeB1GridDescriptor_N_K(1, 1, 1));
using D1sGridDesc_M_N = remove_cvref_t<decltype(MakeD1sGridDescriptor_M_N({}, {}, {}))>;
using E1GridDesc_M_N = decltype(MakeE1GridDescriptor_M_N<E1Layout>(1, 1, 1));
// GridwiseGemm
using GridwiseGemm = GridwiseBatchedGemmMultipleDGemmMultipleD_Xdl_CShuffle<
A0DataType, // TODO: distinguish A/B datatype
Acc0DataType,
D0sDataType,
Acc1DataType,
C1ShuffleDataType,
D1sDataType,
E1DataType,
A0ElementwiseOperation,
B0ElementwiseOperation,
CDE0ElementwiseOperation,
B1ElementwiseOperation,
CDE1ElementwiseOperation,
InMemoryDataOperationEnum::Set,
A0GridDesc_M_K,
B0GridDesc_N_K,
D0sGridDesc_M_N,
B1GridDesc_N_K,
D1sGridDesc_M_N,
E1GridDesc_M_N,
NumGemm0KPrefetchStage,
BlockSize,
Gemm0MPerBlock,
Gemm0NPerBlock,
Gemm0KPerBlock,
Gemm1NPerBlock,
Gemm1KPerBlock,
A0K1,
B0K1,
B1K1,
Gemm0MPerXdl,
Gemm0NPerXdl,
Gemm0MXdlPerWave,
Gemm0NXdlPerWave,
Gemm1NXdlPerWave,
A0BlockTransferThreadClusterLengths_AK0_M_AK1,
A0BlockTransferThreadClusterArrangeOrder,
A0BlockTransferSrcAccessOrder,
A0BlockTransferSrcVectorDim,
A0BlockTransferSrcScalarPerVector,
A0BlockTransferDstScalarPerVector_AK1,
true,
A0BlockLdsExtraM,
B0BlockTransferThreadClusterLengths_BK0_N_BK1,
B0BlockTransferThreadClusterArrangeOrder,
B0BlockTransferSrcAccessOrder,
B0BlockTransferSrcVectorDim,
B0BlockTransferSrcScalarPerVector,
B0BlockTransferDstScalarPerVector_BK1,
true,
B0BlockLdsExtraN,
B1BlockTransferThreadClusterLengths_BK0_N_BK1,
B1BlockTransferThreadClusterArrangeOrder,
B1BlockTransferSrcAccessOrder,
B1BlockTransferSrcVectorDim,
B1BlockTransferSrcScalarPerVector,
B1BlockTransferDstScalarPerVector_BK1,
false,
B1BlockLdsExtraN,
C1ShuffleMXdlPerWavePerShuffle,
C1ShuffleGemm0NXdlPerWavePerShuffle,
CDE1ShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CDE1ShuffleBlockTransferScalarPerVector_NPerBlock,
LoopSched>;
using A0GridDesc_AK0_M_AK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultA0GridDescriptor_AK0_M_AK1(A0GridDesc_M_K{}))>;
using B0GridDesc_BK0_N_BK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultB0GridDescriptor_BK0_N_BK1(B0GridDesc_N_K{}))>;
using B1GridDesc_BK0_N_BK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultB1GridDescriptor_BK0_N_BK1(B1GridDesc_N_K{}))>;
// Argument
struct Argument : public BaseArgument
{
Argument(const A0DataType* p_a0_grid,
const B0DataType* p_b0_grid,
std::array<const void*, NumD0Tensor> p_d0s_grid,
const B1DataType* p_b1_grid,
std::array<const void*, NumD1Tensor> p_d1s_grid,
E1DataType* p_e1_grid,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t Gemm1NRaw, // = ORaw
index_t Batch,
index_t StrideA0,
index_t StrideB0,
std::array<index_t, NumD0Tensor> StrideD0s,
index_t StrideB1,
std::array<index_t, NumD1Tensor> StrideD1s,
index_t StrideE1,
index_t BatchStrideA0,
index_t BatchStrideB0,
std::array<index_t, NumD0Tensor> BatchStrideD0s,
index_t BatchStrideB1,
std::array<index_t, NumD1Tensor> BatchStrideD1s,
index_t BatchStrideE1,
A0ElementwiseOperation a0_element_op,
B0ElementwiseOperation b0_element_op,
CDE0ElementwiseOperation cde0_element_op,
B1ElementwiseOperation b1_element_op,
CDE1ElementwiseOperation cde1_element_op)
: p_a0_grid_{p_a0_grid},
p_b0_grid_{p_b0_grid},
p_d0s_grid_{},
p_b1_grid_{p_b1_grid},
p_d1s_grid_{},
p_e1_grid_{p_e1_grid},
a0_grid_desc_m_k_{DeviceOp::MakeA0GridDescriptor_M_K(MRaw, KRaw, StrideA0)},
b0_grid_desc_n_k_{DeviceOp::MakeB0GridDescriptor_N_K(KRaw, NRaw, StrideB0)},
d0s_grid_desc_m_n_{},
b1_grid_desc_n_k_{DeviceOp::MakeB1GridDescriptor_N_K(NRaw, Gemm1NRaw, StrideB1)},
d1s_grid_desc_m_n_{},
e1_grid_desc_m_n_{
DeviceOp::MakeE1GridDescriptor_M_N<E1Layout>(MRaw, Gemm1NRaw, StrideE1)},
a0_grid_desc_ak0_m_ak1_{
GridwiseGemm::MakeDefaultA0GridDescriptor_AK0_M_AK1(a0_grid_desc_m_k_)},
b0_grid_desc_bk0_n_bk1_{
GridwiseGemm::MakeDefaultB0GridDescriptor_BK0_N_BK1(b0_grid_desc_n_k_)},
d0s_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_{},
b1_grid_desc_bk0_n_bk1_{
GridwiseGemm::MakeDefaultB1GridDescriptor_BK0_N_BK1(b1_grid_desc_n_k_)},
d1s_grid_desc_mblock_mperblock_nblock_nperblock_{},
e1_grid_desc_mblock_mperblock_nblock_nperblock_{},
block_2_e1tile_map_{GridwiseGemm::MakeDefaultBlock2E1TileMap(e1_grid_desc_m_n_)},
a0_element_op_{a0_element_op},
b0_element_op_{b0_element_op},
cde0_element_op_{cde0_element_op},
b1_element_op_{b1_element_op},
cde1_element_op_{cde1_element_op},
batch_count_(Batch),
compute_base_ptr_of_batch_{BatchStrideA0,
BatchStrideB0,
BatchStrideD0s,
BatchStrideB1,
BatchStrideD1s,
BatchStrideE1}
{
std::cout << "a0_grid_desc_m_k_{" << a0_grid_desc_m_k_.GetLength(I0) << ", "
<< a0_grid_desc_m_k_.GetLength(I1) << "}" << std::endl;
std::cout << "b0_grid_desc_n_k_{" << b0_grid_desc_n_k_.GetLength(I0) << ", "
<< b0_grid_desc_n_k_.GetLength(I1) << "}" << std::endl;
std::cout << "d0s_grid_desc_m_n_[I0]{" << d0s_grid_desc_m_n_[I0].GetLength(I0) << ", "
<< d0s_grid_desc_m_n_[I0].GetLength(I1) << "}" << std::endl;
std::cout << "b1_grid_desc_n_k_{" << b1_grid_desc_n_k_.GetLength(I0) << ", "
<< b1_grid_desc_n_k_.GetLength(I1) << "}" << std::endl;
std::cout << "d0s_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_{"
<< d0s_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_[I0].GetLength(I0) << ", "
<< d0s_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_[I0].GetLength(I1) << ", "
<< d0s_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_[I0].GetLength(I2) << ", "
<< d0s_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_[I0].GetLength(I3) << ", "
<< d0s_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_[I0].GetLength(I4) << ", "
<< d0s_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_[I0].GetLength(I5) << ", "
<< d0s_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_[I0].GetLength(I6) << ", "
<< d0s_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_[I0].GetLength(I7) << ", "
<< d0s_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_[I0].GetLength(I8) << ", "
<< d0s_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_[I0].GetLength(I9) << "}"
<< std::endl;
std::cout << "e1_grid_desc_m_n_{" << e1_grid_desc_m_n_.GetLength(I0) << ", "
<< e1_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
static_for<0, NumD0Tensor, 1>{}([&](auto i) {
using D0Layout = remove_cvref_t<tuple_element_t<i.value, D0sLayout>>;
using D0DataType = remove_cvref_t<tuple_element_t<i.value, D0sDataType>>;
// D0 pointer
p_d0s_grid_(i) = static_cast<const D0DataType*>(p_d0s_grid[i]);
// D0 desc
d0s_grid_desc_m_n_(i) =
DeviceOp::MakeD0GridDescriptor_M_N<D0Layout>(MRaw, NRaw, StrideD0s[i]);
});
static_for<0, NumD1Tensor, 1>{}([&](auto i) {
using D1Layout = remove_cvref_t<tuple_element_t<i.value, D1sLayout>>;
using D1DataType = remove_cvref_t<tuple_element_t<i.value, D1sDataType>>;
// D1 pointer
p_d1s_grid_(i) = static_cast<const D1DataType*>(p_d1s_grid[i]);
// D1 desc
d1s_grid_desc_m_n_(i) =
DeviceOp::MakeE1GridDescriptor_M_N<D1Layout>(MRaw, Gemm1NRaw, StrideD1s[i]);
});
if(GridwiseGemm::CheckValidity(a0_grid_desc_m_k_,
b0_grid_desc_n_k_,
b1_grid_desc_n_k_,
e1_grid_desc_m_n_,
block_2_e1tile_map_))
{
e1_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeE1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
e1_grid_desc_m_n_);
d0s_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_ =
GridwiseGemm::MakeD0sGridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5(
d0s_grid_desc_m_n_);
d1s_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeD1sGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
d1s_grid_desc_m_n_);
}
}
// private:
// pointers
const A0DataType* p_a0_grid_;
const B0DataType* p_b0_grid_;
typename GridwiseGemm::D0sGridPointer p_d0s_grid_;
const B1DataType* p_b1_grid_;
typename GridwiseGemm::D1sGridPointer p_d1s_grid_;
E1DataType* p_e1_grid_;
// tensor descriptors for problem definiton
A0GridDesc_M_K a0_grid_desc_m_k_;
B0GridDesc_N_K b0_grid_desc_n_k_;
D0sGridDesc_M_N d0s_grid_desc_m_n_;
B1GridDesc_N_K b1_grid_desc_n_k_;
D1sGridDesc_M_N d1s_grid_desc_m_n_;
E1GridDesc_M_N e1_grid_desc_m_n_;
// tensor descriptors for block/thread-wise copy
A0GridDesc_AK0_M_AK1 a0_grid_desc_ak0_m_ak1_;
B0GridDesc_BK0_N_BK1 b0_grid_desc_bk0_n_bk1_;
typename GridwiseGemm::D0sGridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5
d0s_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_;
B1GridDesc_BK0_N_BK1 b1_grid_desc_bk0_n_bk1_;
typename GridwiseGemm::D1sGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
d1s_grid_desc_mblock_mperblock_nblock_nperblock_;
typename GridwiseGemm::E1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
e1_grid_desc_mblock_mperblock_nblock_nperblock_;
// block-to-e1-tile map
typename GridwiseGemm::DefaultBlock2E1TileMap block_2_e1tile_map_;
// element-wise op
A0ElementwiseOperation a0_element_op_;
B0ElementwiseOperation b0_element_op_;
CDE0ElementwiseOperation cde0_element_op_;
B1ElementwiseOperation b1_element_op_;
CDE1ElementwiseOperation cde1_element_op_;
// batch
index_t batch_count_;
ComputeBasePtrOfStridedBatch compute_base_ptr_of_batch_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(!GridwiseGemm::CheckValidity(arg.a0_grid_desc_m_k_,
arg.b0_grid_desc_n_k_,
arg.b1_grid_desc_n_k_,
arg.e1_grid_desc_m_n_,
arg.block_2_e1tile_map_))
{
throw std::runtime_error("wrong! GridwiseGemm has invalid setting");
}
const index_t grid_size =
arg.block_2_e1tile_map_.CalculateGridSize(arg.e1_grid_desc_m_n_) * arg.batch_count_;
// Gemm0_K
const auto K = arg.a0_grid_desc_m_k_.GetLength(I1);
auto launch_kernel = [&](auto has_main_k_block_loop_) {
const auto kernel = kernel_batched_gemm_gemm_xdl_cshuffle_v1<
GridwiseGemm,
A0DataType, // TODO: distiguish A/B datatype
typename GridwiseGemm::D0sGridPointer,
typename GridwiseGemm::D1sGridPointer,
E1DataType,
A0ElementwiseOperation,
B0ElementwiseOperation,
CDE0ElementwiseOperation,
B1ElementwiseOperation,
CDE1ElementwiseOperation,
DeviceOp::A0GridDesc_AK0_M_AK1,
DeviceOp::B0GridDesc_BK0_N_BK1,
typename GridwiseGemm::D0sGridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5,
DeviceOp::B1GridDesc_BK0_N_BK1,
typename GridwiseGemm::D1sGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::E1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::DefaultBlock2E1TileMap,
ComputeBasePtrOfStridedBatch,
has_main_k_block_loop_>;
return launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a0_grid_,
arg.p_b0_grid_,
arg.p_d0s_grid_,
arg.p_b1_grid_,
arg.p_d1s_grid_,
arg.p_e1_grid_,
arg.a0_element_op_,
arg.b0_element_op_,
arg.cde0_element_op_,
arg.b1_element_op_,
arg.cde1_element_op_,
arg.a0_grid_desc_ak0_m_ak1_,
arg.b0_grid_desc_bk0_n_bk1_,
arg.d0s_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_,
arg.b1_grid_desc_bk0_n_bk1_,
arg.d1s_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.e1_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_e1tile_map_,
arg.batch_count_,
arg.compute_base_ptr_of_batch_);
};
// Gemm1_K is split into Gemm1_K0/K1 where K1 is known at compile time, so we only need
// to concern Gemm0's loop
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
return launch_kernel(integral_constant<bool, true>{});
}
else
{
return launch_kernel(integral_constant<bool, false>{});
}
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
if(!(ck::get_device_name() == "gfx908" || ck::get_device_name() == "gfx90a"))
{
return false;
}
return GridwiseGemm::CheckValidity(arg.a0_grid_desc_m_k_,
arg.b0_grid_desc_n_k_,
arg.b1_grid_desc_n_k_,
arg.e1_grid_desc_m_n_,
arg.block_2_e1tile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const A0DataType* p_a0,
const B0DataType* p_b0,
std::array<const void*, NumD0Tensor> p_d0s,
const B1DataType* p_b1,
std::array<const void*, NumD1Tensor> p_d1s,
E1DataType* p_e1,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t Gemm1NRaw,
index_t Batch,
index_t StrideA0,
index_t StrideB0,
std::array<index_t, NumD0Tensor> StrideD0s,
index_t StrideB1,
std::array<index_t, NumD1Tensor> StrideD1s,
index_t StrideE1,
index_t BatchStrideA0,
index_t BatchStrideB0,
std::array<index_t, NumD0Tensor> BatchStrideD0s,
index_t BatchStrideB1,
std::array<index_t, NumD1Tensor> BatchStrideD1s,
index_t BatchStrideE1,
A0ElementwiseOperation a0_element_op,
B0ElementwiseOperation b0_element_op,
CDE0ElementwiseOperation cde0_element_op,
B1ElementwiseOperation b1_element_op,
CDE1ElementwiseOperation cde1_element_op)
{
return Argument{p_a0, p_b0,
p_d0s, p_b1,
p_d1s, p_e1,
MRaw, NRaw,
KRaw, Gemm1NRaw,
Batch, StrideA0,
StrideB0, StrideD0s,
StrideB1, StrideD1s,
StrideE1, BatchStrideA0,
BatchStrideB0, BatchStrideD0s,
BatchStrideB1, BatchStrideD1s,
BatchStrideE1, a0_element_op,
b0_element_op, cde0_element_op,
b1_element_op, cde1_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_a0,
const void* p_b0,
std::array<const void*, NumD0Tensor> p_d0s,
const void* p_b1,
std::array<const void*, NumD1Tensor> p_d1s,
void* p_e1,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t Gemm1NRaw,
index_t Batch,
index_t StrideA0,
index_t StrideB0,
std::array<ck::index_t, NumD0Tensor> StrideD0s,
index_t StrideB1,
std::array<ck::index_t, NumD1Tensor> StrideD1s,
index_t StrideE1,
index_t BatchStrideA0,
index_t BatchStrideB0,
std::array<ck::index_t, NumD0Tensor> BatchStrideD0s,
index_t BatchStrideB1,
std::array<ck::index_t, NumD1Tensor> BatchStrideD1s,
index_t BatchStrideE1,
A0ElementwiseOperation a0_element_op,
B0ElementwiseOperation b0_element_op,
CDE0ElementwiseOperation cde0_element_op,
B1ElementwiseOperation b1_element_op,
CDE1ElementwiseOperation cde1_element_op) override
{
return std::make_unique<Argument>(static_cast<const A0DataType*>(p_a0),
static_cast<const B0DataType*>(p_b0),
p_d0s,
static_cast<const B1DataType*>(p_b1),
p_d1s,
static_cast<E1DataType*>(p_e1),
MRaw,
NRaw,
KRaw,
Gemm1NRaw,
Batch,
StrideA0,
StrideB0,
StrideD0s,
StrideB1,
StrideD1s,
StrideE1,
BatchStrideA0,
BatchStrideB0,
BatchStrideD0s,
BatchStrideB1,
BatchStrideD1s,
BatchStrideE1,
a0_element_op,
b0_element_op,
cde0_element_op,
b1_element_op,
cde1_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceBatchedGemmMultipleDGemmMultipleD_Xdl_CShuffle"
<< "<"
<< BlockSize << ", "
<< Gemm0MPerBlock << ", "
<< Gemm0NPerBlock << ", "
<< Gemm0KPerBlock << ", "
<< A0K1 << ", "
<< B0K1 << ", "
<< B1K1 << ", "
<< Gemm0MPerXdl << ", "
<< Gemm0NPerXdl << ", "
<< Gemm0MXdlPerWave << ", "
<< Gemm0NXdlPerWave << ", "
<< Gemm1NXdlPerWave << "> ";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

1001
include/ck/tensor_operation/gpu/device/impl/device_batched_gemm_reduce_xdl_cshuffle.hpp Normal file
View File

File diff suppressed because it is too large Load Diff

913
include/ck/tensor_operation/gpu/device/impl/device_batched_gemm_softmax_gemm_permute_xdl_cshuffle.hpp Normal file
View File

@@ -0,0 +1,913 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/device_batched_gemm_softmax_gemm_permute.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_batched_gemm_softmax_gemm_xdl_cshuffle_v1.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <typename GridwiseGemm,
typename FloatAB,
typename FloatC,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename AccElementwiseOperation,
typename B1ElementwiseOperation,
typename CElementwiseOperation,
typename AGridDesc_AK0_M_AK1,
typename BGridDesc_BK0_N_BK1,
typename B1GridDesc_BK0_N_BK1,
typename CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename Block2CTileMap,
typename ComputeBasePtrOfStridedBatch,
typename C0MatrixMask,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_batched_gemm_softmax_gemm_xdl_cshuffle_v1(
const FloatAB* __restrict__ p_a_grid,
const FloatAB* __restrict__ p_b_grid,
const FloatAB* __restrict__ p_b1_grid,
FloatC* __restrict__ p_c_grid,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const AccElementwiseOperation acc_element_op,
const B1ElementwiseOperation b1_element_op,
const CElementwiseOperation c_element_op,
const AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1,
const BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1,
const B1GridDesc_BK0_N_BK1 b1_grid_desc_bk0_n_bk1,
const CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
c_grid_desc_mblock_mperblock_nblock_nperblock,
const Block2CTileMap block_2_ctile_map,
const index_t batch_count,
const ComputeBasePtrOfStridedBatch compute_base_ptr_of_batch,
const C0MatrixMask c0_matrix_mask)
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__))
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
const index_t num_blocks_per_batch =
__builtin_amdgcn_readfirstlane(get_grid_size() / batch_count);
const index_t g_idx = __builtin_amdgcn_readfirstlane(get_block_1d_id() / num_blocks_per_batch);
const long_index_t a_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetABasePtr(g_idx)));
const long_index_t b_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetBBasePtr(g_idx)));
const long_index_t b1_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetB1BasePtr(g_idx)));
const long_index_t c_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetCBasePtr(g_idx)));
GridwiseGemm::template Run<HasMainKBlockLoop>(p_a_grid + a_batch_offset,
p_b_grid + b_batch_offset,
p_b1_grid + b1_batch_offset,
p_c_grid + c_batch_offset,
p_shared,
a_element_op,
b_element_op,
acc_element_op,
b1_element_op,
c_element_op,
a_grid_desc_ak0_m_ak1,
b_grid_desc_bk0_n_bk1,
b1_grid_desc_bk0_n_bk1,
c_grid_desc_mblock_mperblock_nblock_nperblock,
block_2_ctile_map,
c0_matrix_mask);
#else
ignore = p_a_grid;
ignore = p_b_grid;
ignore = p_b1_grid;
ignore = p_c_grid;
ignore = a_element_op;
ignore = b_element_op;
ignore = acc_element_op;
ignore = b1_element_op;
ignore = c_element_op;
ignore = a_grid_desc_ak0_m_ak1;
ignore = b_grid_desc_bk0_n_bk1;
ignore = b1_grid_desc_bk0_n_bk1;
ignore = c_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = block_2_ctile_map;
ignore = batch_count;
ignore = compute_base_ptr_of_batch;
ignore = c0_matrix_mask;
#endif // end of if (defined(__gfx908__) || defined(__gfx90a__))
}
// Computes C = A * B0 * B1
// ^^^^^^ (Acc0)
// ^^^^^^^^^^^ (Acc1)
template <typename ALayout,
typename BLayout, // B0Layout
typename B1Layout,
typename CPermuteNumDims_G_M_Gemm1N, // Sequence<NumDimG, NumDimM, NumDimGemm1N>
typename ADataType,
typename BDataType,
typename B1DataType,
typename CDataType,
typename GemmAccDataType,
typename CShuffleDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename AccElementwiseOperation,
typename B1ElementwiseOperation,
typename CElementwiseOperation,
GemmSpecialization GemmSpec,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock, // Gemm0NPerBlock
index_t KPerBlock, // Gemm0KPerBlock
index_t Gemm1NPerBlock,
index_t Gemm1KPerBlock,
index_t AK1,
index_t BK1,
index_t B1K1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
index_t Gemm1NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
bool ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
bool BBlockLdsExtraN,
typename B1BlockTransferThreadClusterLengths_BK0_N_BK1,
typename B1BlockTransferThreadClusterArrangeOrder,
typename B1BlockTransferSrcAccessOrder,
index_t B1BlockTransferSrcVectorDim,
index_t B1BlockTransferSrcScalarPerVector,
index_t B1BlockTransferDstScalarPerVector_BK1,
bool B1BlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CShuffleBlockTransferScalarPerVector_NPerBlock,
bool MaskOutUpperTriangle,
LoopScheduler LoopSched = LoopScheduler::Default>
struct DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle
: public DeviceBatchedGemmSoftmaxGemmPermute<ALayout,
BLayout,
B1Layout,
CPermuteNumDims_G_M_Gemm1N,
ADataType,
BDataType,
B1DataType,
CDataType,
AElementwiseOperation,
BElementwiseOperation,
AccElementwiseOperation,
B1ElementwiseOperation,
CElementwiseOperation>
{
using DeviceOp = DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto matrix_padder =
GemmGemmPadder<GemmSpec, index_t, index_t, index_t, index_t>{
MPerBlock, NPerBlock, KPerBlock, Gemm1NPerBlock};
static auto MakeAGridDescriptor_AK0_M_AK1(index_t MRaw, index_t KRaw, index_t StrideA)
{
const auto a_grid_desc_mraw_kraw = [&]() {
if constexpr(is_same_v<tensor_layout::gemm::RowMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(StrideA, I1));
}
else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(I1, StrideA));
}
}();
const auto a_grid_desc_m_k = matrix_padder.PadADescriptor_M_K(a_grid_desc_mraw_kraw);
const auto M = a_grid_desc_m_k.GetLength(I0);
const auto K = a_grid_desc_m_k.GetLength(I1);
const auto AK0 = K / AK1;
return transform_tensor_descriptor(a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_pass_through_transform(M)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
static auto MakeBGridDescriptor_BK0_N_BK1(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
const auto b_grid_desc_n_k = matrix_padder.PadBDescriptor_N_K(b_grid_desc_nraw_kraw);
const auto N = b_grid_desc_n_k.GetLength(I0);
const auto K = b_grid_desc_n_k.GetLength(I1);
const auto BK0 = K / BK1;
return transform_tensor_descriptor(b_grid_desc_n_k,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_pass_through_transform(N)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
// Args: Gemm1KRaw, Gemm1NRaw, StrideB1
static auto MakeB1GridDescriptor_BK0_N_BK1(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b1_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, B1Layout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, B1Layout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
const auto b1_grid_desc_n_k = matrix_padder.PadB1Descriptor_N_K(b1_grid_desc_nraw_kraw);
const auto N = b1_grid_desc_n_k.GetLength(I0);
const auto K = b1_grid_desc_n_k.GetLength(I1);
const auto B1K0 = K / B1K1;
return transform_tensor_descriptor(
b1_grid_desc_n_k,
make_tuple(make_unmerge_transform(make_tuple(B1K0, B1K1)),
make_pass_through_transform(N)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
// assume C[G0, G1, ..., M0, M1, M2, ..., N0, N1, N2...]
static auto MakeCGridDescriptor_M_N(const std::vector<index_t>& c_gs_ms_ns_lengths_vec,
const std::vector<index_t>& c_gs_ms_ns_strides_vec)
{
constexpr index_t NumDimG = CPermuteNumDims_G_M_Gemm1N::At(I0);
constexpr index_t NumDimM = CPermuteNumDims_G_M_Gemm1N::At(I1);
constexpr index_t NumDimN = CPermuteNumDims_G_M_Gemm1N::At(I2); // NumDimGemm1N
assert(c_gs_ms_ns_lengths_vec.size() == NumDimG + NumDimM + NumDimN &&
c_gs_ms_ns_strides_vec.size() == NumDimG + NumDimM + NumDimN);
const auto to_tuple = [&](auto& vec, auto start, auto end) {
return generate_tuple([&](auto i) { return vec[start + i]; }, Number<end - start>{});
};
const auto c_ms_ns_lengths = to_tuple(
c_gs_ms_ns_lengths_vec, Number<NumDimG>{}, Number<NumDimG + NumDimM + NumDimN>{});
const auto c_ms_ns_strides = to_tuple(
c_gs_ms_ns_strides_vec, Number<NumDimG>{}, Number<NumDimG + NumDimM + NumDimN>{});
// dimension Ids for M0, M1, ...
constexpr auto mDimIds = typename arithmetic_sequence_gen<0, NumDimM, 1>::type{};
// dimension Ids for N0, N1, ...
constexpr auto nDimIds =
typename arithmetic_sequence_gen<NumDimM, NumDimM + NumDimN, 1>::type{};
// lengths for M0, M1, ...
const auto mLengths = get_container_subset(c_ms_ns_lengths, mDimIds);
// lengths for K0, K1, ...
const auto nLengths = get_container_subset(c_ms_ns_lengths, nDimIds);
// naive tensor C[M0, M1, M2, ..., N0, N1, N2...]
const auto c_grid_desc_ms_ns =
make_naive_tensor_descriptor(c_ms_ns_lengths, c_ms_ns_strides);
// transformed tensor C[MRaw = M0 * M1 * M2 * ... , NRaw = N0 * N1 * N2 * ...]
const auto c_grid_desc_mraw_nraw = transform_tensor_descriptor(
c_grid_desc_ms_ns,
make_tuple(make_merge_transform(mLengths), make_merge_transform(nLengths)),
make_tuple(mDimIds, nDimIds),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return matrix_padder.PadCDescriptor_M_N(c_grid_desc_mraw_nraw);
}
// assume C[G0, G1, ..., M0, M1, M2, ..., N0, N1, N2...]
static auto MakeCGridDescriptor_G_M_N(const std::vector<index_t>& c_gs_ms_ns_lengths_vec,
const std::vector<index_t>& c_gs_ms_ns_strides_vec)
{
constexpr index_t NumDimG = CPermuteNumDims_G_M_Gemm1N::At(I0);
constexpr index_t NumDimM = CPermuteNumDims_G_M_Gemm1N::At(I1);
constexpr index_t NumDimN = CPermuteNumDims_G_M_Gemm1N::At(I2); // NumDimGemm1N
assert(c_gs_ms_ns_lengths_vec.size() == NumDimG + NumDimM + NumDimN &&
c_gs_ms_ns_strides_vec.size() == NumDimG + NumDimM + NumDimN);
const auto to_tuple = [&](auto& vec, auto start, auto end) {
return generate_tuple([&](auto i) { return vec[start + i]; }, Number<end - start>{});
};
const auto c_gs_ms_ns_lengths =
to_tuple(c_gs_ms_ns_lengths_vec, Number<0>{}, Number<NumDimG + NumDimM + NumDimN>{});
const auto c_gs_ms_ns_strides =
to_tuple(c_gs_ms_ns_strides_vec, Number<0>{}, Number<NumDimG + NumDimM + NumDimN>{});
// dimension Ids for G0, G1, ...
constexpr auto gDimIds = typename arithmetic_sequence_gen<0, NumDimG, 1>::type{};
// dimension Ids for M0, M1, ...
constexpr auto mDimIds =
typename arithmetic_sequence_gen<NumDimG, NumDimG + NumDimM, 1>::type{};
// dimension Ids for N0, N1, ...
constexpr auto nDimIds = typename arithmetic_sequence_gen<NumDimG + NumDimM,
NumDimG + NumDimM + NumDimN,
1>::type{};
// lengths for G0, G1, ...
const auto gLengths = get_container_subset(c_gs_ms_ns_lengths, gDimIds);
// lengths for M0, M1, ...
const auto mLengths = get_container_subset(c_gs_ms_ns_lengths, mDimIds);
// lengths for K0, K1, ...
const auto nLengths = get_container_subset(c_gs_ms_ns_lengths, nDimIds);
// naive tensor C[G0, G1, ..., M0, M1, M2, ..., N0, N1, N2...]
const auto c_grid_desc_gs_ms_ns =
make_naive_tensor_descriptor(c_gs_ms_ns_lengths, c_gs_ms_ns_strides);
// transformed tensor C[G = G0 * G1 * ..., MRaw = M0 * M1 * M2 * ... , NRaw = N0 * N1 *
// N2 * ...]
const auto c_grid_desc_g_mraw_nraw =
transform_tensor_descriptor(c_grid_desc_gs_ms_ns,
make_tuple(make_merge_transform(gLengths),
make_merge_transform(mLengths),
make_merge_transform(nLengths)),
make_tuple(gDimIds, mDimIds, nDimIds),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
// this desc is only for calculating batch offset so no padding needed
return c_grid_desc_g_mraw_nraw;
}
using AGridDesc_AK0_M_AK1 = decltype(MakeAGridDescriptor_AK0_M_AK1(1, 1, 1));
using BGridDesc_BK0_N_BK1 = decltype(MakeBGridDescriptor_BK0_N_BK1(1, 1, 1));
using B1GridDesc_BK0_N_BK1 = decltype(MakeB1GridDescriptor_BK0_N_BK1(1, 1, 1));
using CGridDesc_M_N = decltype(MakeCGridDescriptor_M_N({}, {}));
using CGridDesc_G_M_N = decltype(MakeCGridDescriptor_G_M_N({}, {}));
// to track the points which need to be set to -inf on C0
// Note: no need to reset M padding value, because they will not be stored out.
struct C0MatrixMask
{
C0MatrixMask(index_t NRaw) : NRaw_(NRaw) {}
__host__ __device__ bool IsUpperTriangle(index_t m, index_t n) const { return n > m; }
__host__ __device__ bool IsNOutOfBound(/*index_t m, */ index_t n) const
{
return n >= NRaw_;
}
__host__ __device__ bool IsMaskedElement(index_t m, index_t n) const
{
return IsUpperTriangle(m, n) || IsNOutOfBound(n);
}
private:
// index_t MRaw_;
index_t NRaw_;
};
struct ComputeBasePtrOfStridedBatch
{
ComputeBasePtrOfStridedBatch(index_t BatchStrideA,
index_t BatchStrideB,
index_t BatchStrideB1,
CGridDesc_G_M_N c_grid_desc_g_m_n)
: BatchStrideA_(BatchStrideA),
BatchStrideB_(BatchStrideB),
BatchStrideB1_(BatchStrideB1),
c_grid_desc_g_m_n_(c_grid_desc_g_m_n)
{
}
__host__ __device__ constexpr long_index_t GetABasePtr(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideA_);
}
__host__ __device__ constexpr long_index_t GetBBasePtr(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideB_);
}
__host__ __device__ constexpr long_index_t GetB1BasePtr(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideB1_);
}
__host__ __device__ constexpr long_index_t GetCBasePtr(index_t g_idx) const
{
return c_grid_desc_g_m_n_.CalculateOffset(make_multi_index(g_idx, 0, 0));
}
private:
index_t BatchStrideA_;
index_t BatchStrideB_;
index_t BatchStrideB1_;
CGridDesc_G_M_N c_grid_desc_g_m_n_;
};
// GridwiseGemm
using GridwiseGemm = GridwiseBatchedGemmSoftmaxGemm_Xdl_CShuffle<
ADataType, // TODO: distinguish A/B datatype
GemmAccDataType,
CShuffleDataType,
CDataType,
AElementwiseOperation,
BElementwiseOperation,
AccElementwiseOperation,
B1ElementwiseOperation,
CElementwiseOperation,
InMemoryDataOperationEnum::Set,
AGridDesc_AK0_M_AK1,
BGridDesc_BK0_N_BK1,
B1GridDesc_BK0_N_BK1,
CGridDesc_M_N,
NumGemmKPrefetchStage,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
Gemm1NPerBlock,
Gemm1KPerBlock,
AK1,
BK1,
B1K1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
Gemm1NXdlPerWave,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
true,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
true,
BBlockLdsExtraN,
B1BlockTransferThreadClusterLengths_BK0_N_BK1,
B1BlockTransferThreadClusterArrangeOrder,
B1BlockTransferSrcAccessOrder,
B1BlockTransferSrcVectorDim,
B1BlockTransferSrcScalarPerVector,
B1BlockTransferDstScalarPerVector_BK1,
false,
B1BlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CShuffleBlockTransferScalarPerVector_NPerBlock,
LoopSched,
matrix_padder.PadN,
MaskOutUpperTriangle>;
// Argument
// FIXME: constness
struct Argument : public BaseArgument
{
Argument(const ADataType* p_a_grid,
const BDataType* p_b_grid,
const B1DataType* p_b1_grid,
CDataType* p_c_grid,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t Gemm1NRaw, // = ORaw
index_t Batch,
std::vector<index_t> c_gs_ms_gemm1ns_lengths, // c_gs_ms_os_lengths
std::vector<index_t> c_gs_ms_gemm1ns_strides, // c_gs_ms_os_strides
index_t StrideA,
index_t StrideB,
index_t StrideB1,
index_t BatchStrideA,
index_t BatchStrideB,
index_t BatchStrideB1,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
AccElementwiseOperation acc_element_op,
B1ElementwiseOperation b1_element_op,
CElementwiseOperation c_element_op)
: p_a_grid_{p_a_grid},
p_b_grid_{p_b_grid},
p_b1_grid_{p_b1_grid},
p_c_grid_{p_c_grid},
a_grid_desc_ak0_m_ak1_{DeviceOp::MakeAGridDescriptor_AK0_M_AK1(MRaw, KRaw, StrideA)},
b_grid_desc_bk0_n_bk1_{DeviceOp::MakeBGridDescriptor_BK0_N_BK1(KRaw, NRaw, StrideB)},
b1_grid_desc_bk0_n_bk1_{
DeviceOp::MakeB1GridDescriptor_BK0_N_BK1(NRaw, Gemm1NRaw, StrideB1)},
c_grid_desc_m_n_{DeviceOp::MakeCGridDescriptor_M_N(c_gs_ms_gemm1ns_lengths,
c_gs_ms_gemm1ns_strides)},
c_grid_desc_g_m_n_{DeviceOp::MakeCGridDescriptor_G_M_N(c_gs_ms_gemm1ns_lengths,
c_gs_ms_gemm1ns_strides)},
c_grid_desc_mblock_mperblock_nblock_nperblock_{},
block_2_ctile_map_{GridwiseGemm::MakeDefaultBlock2CTileMap(c_grid_desc_m_n_)},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
acc_element_op_{acc_element_op},
b1_element_op_{b1_element_op},
c_element_op_{c_element_op},
batch_count_(Batch),
compute_base_ptr_of_batch_{
BatchStrideA, BatchStrideB, BatchStrideB1, c_grid_desc_g_m_n_},
c0_matrix_mask_{NRaw},
raw_lengths_m_n_k_o_{MRaw, NRaw, KRaw, Gemm1NRaw},
c_extent_lowest_{c_gs_ms_gemm1ns_lengths.back()},
c_stride_lowest_{c_gs_ms_gemm1ns_strides.back()}
{
if(GridwiseGemm::CheckValidity(a_grid_desc_ak0_m_ak1_,
b_grid_desc_bk0_n_bk1_,
b1_grid_desc_bk0_n_bk1_,
c_grid_desc_m_n_,
block_2_ctile_map_))
{
c_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeCGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
c_grid_desc_m_n_);
}
}
// private:
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
const B1DataType* p_b1_grid_;
CDataType* p_c_grid_;
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
B1GridDesc_BK0_N_BK1 b1_grid_desc_bk0_n_bk1_;
CGridDesc_M_N c_grid_desc_m_n_;
CGridDesc_G_M_N c_grid_desc_g_m_n_;
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
c_grid_desc_mblock_mperblock_nblock_nperblock_;
typename GridwiseGemm::DefaultBlock2CTileMap block_2_ctile_map_;
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
AccElementwiseOperation acc_element_op_;
B1ElementwiseOperation b1_element_op_;
CElementwiseOperation c_element_op_;
index_t batch_count_;
ComputeBasePtrOfStridedBatch compute_base_ptr_of_batch_;
// check C0 masking and padding
C0MatrixMask c0_matrix_mask_;
// For robust IsSupportedArgument() check
std::vector<index_t> raw_lengths_m_n_k_o_;
index_t c_extent_lowest_;
index_t c_stride_lowest_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.b1_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_))
{
throw std::runtime_error("wrong! GridwiseGemm has invalid setting");
}
const index_t grid_size =
arg.block_2_ctile_map_.CalculateGridSize(arg.c_grid_desc_m_n_) * arg.batch_count_;
// Gemm0_K
const auto K =
arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) * arg.a_grid_desc_ak0_m_ak1_.GetLength(I2);
float ave_time = 0;
auto launch_kernel = [&](auto has_main_k_block_loop_) {
const auto kernel = kernel_batched_gemm_softmax_gemm_xdl_cshuffle_v1<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
AElementwiseOperation,
BElementwiseOperation,
AccElementwiseOperation,
B1ElementwiseOperation,
CElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
DeviceOp::B1GridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::DefaultBlock2CTileMap,
ComputeBasePtrOfStridedBatch,
C0MatrixMask,
has_main_k_block_loop_>;
return launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_b1_grid_,
arg.p_c_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.acc_element_op_,
arg.b1_element_op_,
arg.c_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.b1_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_ctile_map_,
arg.batch_count_,
arg.compute_base_ptr_of_batch_,
arg.c0_matrix_mask_);
};
// Gemm1_K is split into Gemm1_K0/K1 where K1 is known at compile time, so we only need
// to concern Gemm0's loop
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
ave_time = launch_kernel(integral_constant<bool, true>{});
}
else
{
ave_time = launch_kernel(integral_constant<bool, false>{});
}
return ave_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
if(!(ck::get_device_name() == "gfx908" || ck::get_device_name() == "gfx90a"))
{
return false;
}
// Check if C permute dimension matches GEMM + GEMM shape
const index_t c_g = arg.c_grid_desc_g_m_n_.GetLength(I0); // unpadded
const index_t c_m = arg.c_grid_desc_m_n_.GetLength(I0);
const index_t c_gemm1n = arg.c_grid_desc_m_n_.GetLength(I1);
const index_t a_m = arg.a_grid_desc_ak0_m_ak1_.GetLength(I1);
const index_t b1_gemm1n = arg.b1_grid_desc_bk0_n_bk1_.GetLength(I1);
if(!(c_g == arg.batch_count_ && c_m == a_m && c_gemm1n == b1_gemm1n))
{
return false;
}
// Note: we need raw lengths since threadwise copy can not handle vector load when part of
// vector is out of bounds
const auto MRaw = arg.raw_lengths_m_n_k_o_[0];
const auto NRaw = arg.raw_lengths_m_n_k_o_[1];
const auto KRaw = arg.raw_lengths_m_n_k_o_[2];
const auto Gemm1NRaw = arg.raw_lengths_m_n_k_o_[3];
// Check scalar per vector requirement
const auto a_extent_lowest =
is_same_v<tensor_layout::gemm::RowMajor, ALayout> ? KRaw : MRaw;
const auto b_extent_lowest =
is_same_v<tensor_layout::gemm::RowMajor, BLayout> ? NRaw : KRaw;
const auto b1_extent_lowest =
is_same_v<tensor_layout::gemm::RowMajor, B1Layout> ? Gemm1NRaw : NRaw;
const auto c_extent_lowest = arg.c_extent_lowest_;
if(!(a_extent_lowest % ABlockTransferSrcScalarPerVector == 0 &&
b_extent_lowest % BBlockTransferSrcScalarPerVector == 0 &&
b1_extent_lowest % B1BlockTransferSrcScalarPerVector == 0 &&
c_extent_lowest % CShuffleBlockTransferScalarPerVector_NPerBlock == 0))
{
return false;
}
// Check vector store requirement; assumes last dimension in N to be contiguous
if(arg.c_stride_lowest_ != 1)
{
return false;
}
return GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.b1_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const ADataType* p_a,
const BDataType* p_b,
const B1DataType* p_b1,
CDataType* p_c,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t Gemm1NRaw,
index_t Batch,
std::vector<index_t> c_gs_ms_gemm1ns_lengths, // c_gs_ms_os_lengths
std::vector<index_t> c_gs_ms_gemm1ns_strides, // c_gs_ms_os_strides
index_t StrideA,
index_t StrideB,
index_t StrideB1,
index_t BatchStrideA,
index_t BatchStrideB,
index_t BatchStrideB1,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
AccElementwiseOperation acc_element_op,
B1ElementwiseOperation b1_element_op,
CElementwiseOperation c_element_op)
{
return Argument{p_a,
p_b,
p_b1,
p_c,
MRaw,
NRaw,
KRaw,
Gemm1NRaw,
Batch,
c_gs_ms_gemm1ns_lengths,
c_gs_ms_gemm1ns_strides,
StrideA,
StrideB,
StrideB1,
BatchStrideA,
BatchStrideB,
BatchStrideB1,
a_element_op,
b_element_op,
acc_element_op,
b1_element_op,
c_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
// FIXME: constness
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_a,
const void* p_b,
const void* p_b1,
void* p_c,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t Gemm1NRaw,
index_t Batch,
std::vector<index_t> c_gs_ms_gemm1ns_lengths, // c_gs_ms_os_lengths
std::vector<index_t> c_gs_ms_gemm1ns_strides, // c_gs_ms_os_strides
index_t StrideA,
index_t StrideB,
index_t StrideB1,
index_t BatchStrideA,
index_t BatchStrideB,
index_t BatchStrideB1,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
AccElementwiseOperation acc_element_op,
B1ElementwiseOperation b1_element_op,
CElementwiseOperation c_element_op) override
{
return std::make_unique<Argument>(static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
static_cast<const B1DataType*>(p_b1),
static_cast<CDataType*>(p_c),
MRaw,
NRaw,
KRaw,
Gemm1NRaw,
Batch,
c_gs_ms_gemm1ns_lengths,
c_gs_ms_gemm1ns_strides,
StrideA,
StrideB,
StrideB1,
BatchStrideA,
BatchStrideB,
BatchStrideB1,
a_element_op,
b_element_op,
acc_element_op,
b1_element_op,
c_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< AK1 << ", "
<< BK1 << ", "
<< MPerBlock << ", "
<< Gemm1NPerBlock << ", "
<< Gemm1KPerBlock << ", "
<< B1K1 << ", "
<< getGemmSpecializationString(GemmSpec) << ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

788
include/ck/tensor_operation/gpu/device/impl/device_batched_gemm_softmax_gemm_xdl_cshuffle.hpp Normal file
View File

@@ -0,0 +1,788 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_batched_gemm_softmax_gemm.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_batched_gemm_softmax_gemm_xdl_cshuffle_v1.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <typename GridwiseGemm,
typename FloatAB,
typename FloatC,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename AccElementwiseOperation,
typename B1ElementwiseOperation,
typename CElementwiseOperation,
typename AGridDesc_AK0_M_AK1,
typename BGridDesc_BK0_N_BK1,
typename B1GridDesc_BK0_N_BK1,
typename CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename Block2CTileMap,
typename ComputeBasePtrOfStridedBatch,
typename C0MatrixMask,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_batched_gemm_softmax_gemm_xdl_cshuffle_v1(
const FloatAB* __restrict__ p_a_grid,
const FloatAB* __restrict__ p_b_grid,
const FloatAB* __restrict__ p_b1_grid,
FloatC* __restrict__ p_c_grid,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const AccElementwiseOperation acc_element_op,
const B1ElementwiseOperation b1_element_op,
const CElementwiseOperation c_element_op,
const AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1,
const BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1,
const B1GridDesc_BK0_N_BK1 b1_grid_desc_bk0_n_bk1,
const CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
c_grid_desc_mblock_mperblock_nblock_nperblock,
const Block2CTileMap block_2_ctile_map,
const index_t batch_count,
const ComputeBasePtrOfStridedBatch compute_base_ptr_of_batch,
const C0MatrixMask c0_matrix_mask)
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__))
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
const index_t num_blocks_per_batch =
__builtin_amdgcn_readfirstlane(get_grid_size() / batch_count);
const index_t g_idx = __builtin_amdgcn_readfirstlane(get_block_1d_id() / num_blocks_per_batch);
const long_index_t a_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetABasePtr(g_idx)));
const long_index_t b_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetBBasePtr(g_idx)));
const long_index_t b1_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetB1BasePtr(g_idx)));
const long_index_t c_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetCBasePtr(g_idx)));
GridwiseGemm::template Run<HasMainKBlockLoop>(p_a_grid + a_batch_offset,
p_b_grid + b_batch_offset,
p_b1_grid + b1_batch_offset,
p_c_grid + c_batch_offset,
p_shared,
a_element_op,
b_element_op,
acc_element_op,
b1_element_op,
c_element_op,
a_grid_desc_ak0_m_ak1,
b_grid_desc_bk0_n_bk1,
b1_grid_desc_bk0_n_bk1,
c_grid_desc_mblock_mperblock_nblock_nperblock,
block_2_ctile_map,
c0_matrix_mask);
#else
ignore = p_a_grid;
ignore = p_b_grid;
ignore = p_b1_grid;
ignore = p_c_grid;
ignore = a_element_op;
ignore = b_element_op;
ignore = acc_element_op;
ignore = b1_element_op;
ignore = c_element_op;
ignore = a_grid_desc_ak0_m_ak1;
ignore = b_grid_desc_bk0_n_bk1;
ignore = b1_grid_desc_bk0_n_bk1;
ignore = c_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = block_2_ctile_map;
ignore = batch_count;
ignore = compute_base_ptr_of_batch;
ignore = c0_matrix_mask;
#endif // end of if (defined(__gfx908__) || defined(__gfx90a__))
}
// Computes C = A * B0 * B1
// ^^^^^^ (Acc0)
// ^^^^^^^^^^^ (Acc1)
// When using NPadding as GemmSpecialization, AccElementwiseOperation should be set to
// ScaleAndResetNaNToMinusInfinity.
// if !isNan(AccElement)
// AccElement *= scale
// else
// AccElement = -INFINITY
// Otherwise, result may be wrong.
template <typename ALayout,
typename BLayout, // B0Layout
typename B1Layout,
typename CLayout,
typename ADataType,
typename BDataType,
typename B1DataType,
typename CDataType,
typename GemmAccDataType,
typename CShuffleDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename AccElementwiseOperation,
typename B1ElementwiseOperation,
typename CElementwiseOperation,
GemmSpecialization GemmSpec,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock, // Gemm0NPerBlock
index_t KPerBlock, // Gemm0KPerBlock
index_t Gemm1NPerBlock,
index_t Gemm1KPerBlock,
index_t AK1,
index_t BK1,
index_t B1K1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
index_t Gemm1NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
bool ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
bool BBlockLdsExtraN,
typename B1BlockTransferThreadClusterLengths_BK0_N_BK1,
typename B1BlockTransferThreadClusterArrangeOrder,
typename B1BlockTransferSrcAccessOrder,
index_t B1BlockTransferSrcVectorDim,
index_t B1BlockTransferSrcScalarPerVector,
index_t B1BlockTransferDstScalarPerVector_BK1,
bool B1BlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CShuffleBlockTransferScalarPerVector_NPerBlock,
bool MaskOutUpperTriangle,
LoopScheduler LoopSched = LoopScheduler::Default>
struct DeviceBatchedGemmSoftmaxGemm_Xdl_CShuffle
: public DeviceBatchedGemmSoftmaxGemm<ALayout,
BLayout,
B1Layout,
CLayout,
ADataType,
BDataType,
B1DataType,
CDataType,
AElementwiseOperation,
BElementwiseOperation,
AccElementwiseOperation,
B1ElementwiseOperation,
CElementwiseOperation>
{
using DeviceOp = DeviceBatchedGemmSoftmaxGemm_Xdl_CShuffle;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto matrix_padder =
GemmGemmPadder<GemmSpec, index_t, index_t, index_t, index_t>{
MPerBlock, NPerBlock, KPerBlock, Gemm1NPerBlock};
static auto MakeAGridDescriptor_AK0_M_AK1(index_t MRaw, index_t KRaw, index_t StrideA)
{
const auto a_grid_desc_mraw_kraw = [&]() {
if constexpr(is_same_v<tensor_layout::gemm::RowMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(StrideA, I1));
}
else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(I1, StrideA));
}
}();
const auto a_grid_desc_m_k = matrix_padder.PadADescriptor_M_K(a_grid_desc_mraw_kraw);
const auto M = a_grid_desc_m_k.GetLength(I0);
const auto K = a_grid_desc_m_k.GetLength(I1);
const auto AK0 = K / AK1;
return transform_tensor_descriptor(a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_pass_through_transform(M)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
static auto MakeBGridDescriptor_BK0_N_BK1(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
const auto b_grid_desc_n_k = matrix_padder.PadBDescriptor_N_K(b_grid_desc_nraw_kraw);
const auto N = b_grid_desc_n_k.GetLength(I0);
const auto K = b_grid_desc_n_k.GetLength(I1);
const auto BK0 = K / BK1;
return transform_tensor_descriptor(b_grid_desc_n_k,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_pass_through_transform(N)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
// Args: Gemm1KRaw, Gemm1NRaw, StrideB1
static auto MakeB1GridDescriptor_BK0_N_BK1(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b1_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, B1Layout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, B1Layout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
const auto b1_grid_desc_n_k = matrix_padder.PadB1Descriptor_N_K(b1_grid_desc_nraw_kraw);
const auto N = b1_grid_desc_n_k.GetLength(I0);
const auto K = b1_grid_desc_n_k.GetLength(I1);
const auto B1K0 = K / B1K1;
return transform_tensor_descriptor(
b1_grid_desc_n_k,
make_tuple(make_unmerge_transform(make_tuple(B1K0, B1K1)),
make_pass_through_transform(N)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
static auto MakeCGridDescriptor_M_N(index_t MRaw, index_t NRaw, index_t StrideC)
{
const auto c_grid_desc_mraw_nraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(StrideC, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(I1, StrideC));
}
}();
return matrix_padder.PadCDescriptor_M_N(c_grid_desc_mraw_nraw);
}
// to track the points which need to be set to -inf on C0
// Note: no need to reset M padding value, because they will not be stored out.
struct C0MatrixMask
{
C0MatrixMask(index_t NRaw) : NRaw_(NRaw) {}
__host__ __device__ bool IsUpperTriangle(index_t m, index_t n) const { return n > m; }
__host__ __device__ bool IsNOutOfBound(/*index_t m, */ index_t n) const
{
return n >= NRaw_;
}
__host__ __device__ bool IsMaskedElement(index_t m, index_t n) const
{
return IsUpperTriangle(m, n) || IsNOutOfBound(n);
}
private:
// index_t MRaw_;
index_t NRaw_;
};
struct ComputeBasePtrOfStridedBatch
{
ComputeBasePtrOfStridedBatch(index_t BatchStrideA,
index_t BatchStrideB,
index_t BatchStrideB1,
index_t BatchStrideC)
: BatchStrideA_(BatchStrideA),
BatchStrideB_(BatchStrideB),
BatchStrideB1_(BatchStrideB1),
BatchStrideC_(BatchStrideC)
{
}
__host__ __device__ constexpr long_index_t GetABasePtr(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideA_);
}
__host__ __device__ constexpr long_index_t GetBBasePtr(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideB_);
}
__host__ __device__ constexpr long_index_t GetB1BasePtr(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideB1_);
}
__host__ __device__ constexpr long_index_t GetCBasePtr(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideC_);
}
private:
index_t BatchStrideA_;
index_t BatchStrideB_;
index_t BatchStrideB1_;
index_t BatchStrideC_;
};
using AGridDesc_AK0_M_AK1 = decltype(MakeAGridDescriptor_AK0_M_AK1(1, 1, 1));
using BGridDesc_BK0_N_BK1 = decltype(MakeBGridDescriptor_BK0_N_BK1(1, 1, 1));
using B1GridDesc_BK0_N_BK1 = decltype(MakeB1GridDescriptor_BK0_N_BK1(1, 1, 1));
using CGridDesc_M_N = decltype(MakeCGridDescriptor_M_N(1, 1, 1));
// GridwiseGemm
using GridwiseGemm = GridwiseBatchedGemmSoftmaxGemm_Xdl_CShuffle<
ADataType, // TODO: distinguish A/B datatype
GemmAccDataType,
CShuffleDataType,
CDataType,
AElementwiseOperation,
BElementwiseOperation,
AccElementwiseOperation,
B1ElementwiseOperation,
CElementwiseOperation,
InMemoryDataOperationEnum::Set,
AGridDesc_AK0_M_AK1,
BGridDesc_BK0_N_BK1,
B1GridDesc_BK0_N_BK1,
CGridDesc_M_N,
NumGemmKPrefetchStage,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
Gemm1NPerBlock,
Gemm1KPerBlock,
AK1,
BK1,
B1K1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
Gemm1NXdlPerWave,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
true,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
true,
BBlockLdsExtraN,
B1BlockTransferThreadClusterLengths_BK0_N_BK1,
B1BlockTransferThreadClusterArrangeOrder,
B1BlockTransferSrcAccessOrder,
B1BlockTransferSrcVectorDim,
B1BlockTransferSrcScalarPerVector,
B1BlockTransferDstScalarPerVector_BK1,
false,
B1BlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CShuffleBlockTransferScalarPerVector_NPerBlock,
LoopSched,
matrix_padder.PadN,
MaskOutUpperTriangle>;
// Argument
struct Argument : public BaseArgument
{
Argument(const ADataType* p_a_grid,
const BDataType* p_b_grid,
const B1DataType* p_b1_grid,
CDataType* p_c_grid,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t Gemm1NRaw, // = ORaw
index_t Batch,
index_t StrideA,
index_t StrideB,
index_t StrideB1,
index_t StrideC,
index_t BatchStrideA,
index_t BatchStrideB,
index_t BatchStrideB1,
index_t BatchStrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
AccElementwiseOperation acc_element_op,
B1ElementwiseOperation b1_element_op,
CElementwiseOperation c_element_op)
: p_a_grid_{p_a_grid},
p_b_grid_{p_b_grid},
p_b1_grid_{p_b1_grid},
p_c_grid_{p_c_grid},
a_grid_desc_ak0_m_ak1_{DeviceOp::MakeAGridDescriptor_AK0_M_AK1(MRaw, KRaw, StrideA)},
b_grid_desc_bk0_n_bk1_{DeviceOp::MakeBGridDescriptor_BK0_N_BK1(KRaw, NRaw, StrideB)},
b1_grid_desc_bk0_n_bk1_{
DeviceOp::MakeB1GridDescriptor_BK0_N_BK1(NRaw, Gemm1NRaw, StrideB1)},
c_grid_desc_m_n_{DeviceOp::MakeCGridDescriptor_M_N(MRaw, Gemm1NRaw, StrideC)},
c_grid_desc_mblock_mperblock_nblock_nperblock_{},
block_2_ctile_map_{GridwiseGemm::MakeDefaultBlock2CTileMap(c_grid_desc_m_n_)},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
acc_element_op_{acc_element_op},
b1_element_op_{b1_element_op},
c_element_op_{c_element_op},
batch_count_(Batch),
compute_base_ptr_of_batch_{BatchStrideA, BatchStrideB, BatchStrideB1, BatchStrideC},
c0_matrix_mask_{NRaw},
raw_lengths_m_n_k_o_{MRaw, NRaw, KRaw, Gemm1NRaw}
{
if(GridwiseGemm::CheckValidity(a_grid_desc_ak0_m_ak1_,
b_grid_desc_bk0_n_bk1_,
b1_grid_desc_bk0_n_bk1_,
c_grid_desc_m_n_,
block_2_ctile_map_))
{
c_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeCGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
c_grid_desc_m_n_);
}
}
// private:
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
const B1DataType* p_b1_grid_;
CDataType* p_c_grid_;
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
B1GridDesc_BK0_N_BK1 b1_grid_desc_bk0_n_bk1_;
CGridDesc_M_N c_grid_desc_m_n_;
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
c_grid_desc_mblock_mperblock_nblock_nperblock_;
typename GridwiseGemm::DefaultBlock2CTileMap block_2_ctile_map_;
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
AccElementwiseOperation acc_element_op_;
B1ElementwiseOperation b1_element_op_;
CElementwiseOperation c_element_op_;
index_t batch_count_;
ComputeBasePtrOfStridedBatch compute_base_ptr_of_batch_;
// check C0 masking and padding
C0MatrixMask c0_matrix_mask_;
// For robust IsSupportedArgument() check
std::vector<index_t> raw_lengths_m_n_k_o_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.b1_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_))
{
throw std::runtime_error("wrong! GridwiseGemm has invalid setting");
}
const index_t grid_size =
arg.block_2_ctile_map_.CalculateGridSize(arg.c_grid_desc_m_n_) * arg.batch_count_;
// Gemm0_K
const auto K =
arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) * arg.a_grid_desc_ak0_m_ak1_.GetLength(I2);
float ave_time = 0;
auto launch_kernel = [&](auto has_main_k_block_loop_) {
const auto kernel = kernel_batched_gemm_softmax_gemm_xdl_cshuffle_v1<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
AElementwiseOperation,
BElementwiseOperation,
AccElementwiseOperation,
B1ElementwiseOperation,
CElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
DeviceOp::B1GridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::DefaultBlock2CTileMap,
ComputeBasePtrOfStridedBatch,
C0MatrixMask,
has_main_k_block_loop_>;
return launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_b1_grid_,
arg.p_c_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.acc_element_op_,
arg.b1_element_op_,
arg.c_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.b1_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_ctile_map_,
arg.batch_count_,
arg.compute_base_ptr_of_batch_,
arg.c0_matrix_mask_);
};
// Gemm1_K is split into Gemm1_K0/K1 where K1 is known at compile time, so we only need
// to concern Gemm0's loop
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
ave_time = launch_kernel(integral_constant<bool, true>{});
}
else
{
ave_time = launch_kernel(integral_constant<bool, false>{});
}
return ave_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
if(!(ck::get_device_name() == "gfx908" || ck::get_device_name() == "gfx90a"))
{
return false;
}
// Note: we need raw lengths since threadwise copy can not handle vector load when part of
// vector is out of bounds
const auto MRaw = arg.raw_lengths_m_n_k_o_[0];
const auto NRaw = arg.raw_lengths_m_n_k_o_[1];
const auto KRaw = arg.raw_lengths_m_n_k_o_[2];
const auto Gemm1NRaw = arg.raw_lengths_m_n_k_o_[3];
// Check scalar per vector requirement
const auto a_extent_lowest =
is_same_v<tensor_layout::gemm::RowMajor, ALayout> ? KRaw : MRaw;
const auto b_extent_lowest =
is_same_v<tensor_layout::gemm::RowMajor, BLayout> ? NRaw : KRaw;
const auto b1_extent_lowest =
is_same_v<tensor_layout::gemm::RowMajor, B1Layout> ? Gemm1NRaw : NRaw;
const auto c_extent_lowest =
is_same_v<tensor_layout::gemm::RowMajor, CLayout> ? Gemm1NRaw : MRaw;
if(!(a_extent_lowest % ABlockTransferSrcScalarPerVector == 0 &&
b_extent_lowest % BBlockTransferSrcScalarPerVector == 0 &&
b1_extent_lowest % B1BlockTransferSrcScalarPerVector == 0 &&
c_extent_lowest % CShuffleBlockTransferScalarPerVector_NPerBlock == 0))
{
return false;
}
return GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.b1_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const ADataType* p_a,
const BDataType* p_b,
const B1DataType* p_b1,
CDataType* p_c,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t Gemm1NRaw,
index_t Batch,
index_t StrideA,
index_t StrideB,
index_t StrideB1,
index_t StrideC,
index_t BatchStrideA,
index_t BatchStrideB,
index_t BatchStrideB1,
index_t BatchStrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
AccElementwiseOperation acc_element_op,
B1ElementwiseOperation b1_element_op,
CElementwiseOperation c_element_op)
{
return Argument{p_a, p_b, p_b1, p_c, MRaw,
NRaw, KRaw, Gemm1NRaw, Batch, StrideA,
StrideB, StrideB1, StrideC, BatchStrideA, BatchStrideB,
BatchStrideB1, BatchStrideC, a_element_op, b_element_op, acc_element_op,
b1_element_op, c_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,
const void* p_b,
const void* p_b1,
void* p_c,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t Gemm1NRaw,
index_t Batch,
index_t StrideA,
index_t StrideB,
index_t StrideB1,
index_t StrideC,
index_t BatchStrideA,
index_t BatchStrideB,
index_t BatchStrideB1,
index_t BatchStrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
AccElementwiseOperation acc_element_op,
B1ElementwiseOperation b1_element_op,
CElementwiseOperation c_element_op) override
{
return std::make_unique<Argument>(static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
static_cast<const B1DataType*>(p_b1),
static_cast<CDataType*>(p_c),
MRaw,
NRaw,
KRaw,
Gemm1NRaw,
Batch,
StrideA,
StrideB,
StrideB1,
StrideC,
BatchStrideA,
BatchStrideB,
BatchStrideB1,
BatchStrideC,
a_element_op,
b_element_op,
acc_element_op,
b1_element_op,
c_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceBatchedGemmSoftmaxGemm_Xdl_CShuffle"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< AK1 << ", "
<< BK1 << ", "
<< MPerBlock << ", "
<< Gemm1NPerBlock << ", "
<< Gemm1KPerBlock << ", "
<< B1K1 << ", "
<< getGemmSpecializationString(GemmSpec) << ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

641
include/ck/tensor_operation/gpu/device/impl/device_batched_gemm_xdl.hpp Normal file
View File

@@ -0,0 +1,641 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_batched_gemm.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v2r3.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
/*
* \brief Wrapper function of GridwiseGemm::Run to realize BatchedGEMM.
*
* \tparam ComputePtrOffsetOfBatch Class that computes the base pointer offsets of A, B, C matrix
* given the batch. For example, ComputePtrOffsetOfStridedBatch() computes the offsets of evenly
* strided batched, but we can easily extend to other layouts. The returned offset can be either \p
* index_t or \p long_index_t. If it returns \p long_index_t, we are not subject to the 2GB
* limitations.
*
* \tparam Block2CTileMap Block2CTileMap::CalculateBottomIndex() takes in id of a workgroup and
* returns the 2D index of the tile that it computes. \see
* GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r3::Run().
*
* \note Using \p ComputePtrOffsetOfBatch gives us the flexibility that 2 workgroups can compute 2
* tiles from different matrices. Keep in mind that these 2 matrices can share the same grid
* descriptor (like in BatchedGEMM), or use their own grid descriptors (in GroupedGemm). \link
* device_conv3d_fwd_xdl_ndhwc_kzyxc_ndhwk.hpp kernel_gemm_xdlops_v2r3_for_conv3d \endlink for \link
* DeviceConv3d \endlink uses the same concept, but currently does NOT encapsulate the computing of
* pointer offset into \p ComputePtrOffsetOfStridedBatch.
*
* \note \p Block2CTileMap allows customized mapping between a workgroup and the C-tile it computes.
* Together with \p ComputePtrOffsetOfBatch, we can reuse GridwiseGemm (and GridwiseGemm fusion ) to
* realize BatchedGemm and GroupedGemm (and the corresponding GEMM fusion).
*
*/
template <typename GridwiseGemm,
typename FloatAB,
typename FloatC,
typename AGridDesc_K0_M_K1,
typename BGridDesc_K0_N_K1,
typename CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CElementwiseOperation,
typename ComputePtrOffsetOfBatch,
typename Block2CTileMap,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_batched_gemm_xdlops_v2r3(
const FloatAB* __restrict__ p_a_grid,
const FloatAB* __restrict__ p_b_grid,
FloatC* __restrict__ p_c_grid,
const index_t batch_count,
const AGridDesc_K0_M_K1 a_grid_desc_k0_m_k1,
const BGridDesc_K0_N_K1 b_grid_desc_k0_n_k1,
const CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2 c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const CElementwiseOperation c_element_op,
const ComputePtrOffsetOfBatch compute_ptr_offset_of_batch,
const Block2CTileMap block_2_ctile_map)
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__))
const index_t num_blocks_per_batch =
__builtin_amdgcn_readfirstlane(get_grid_size() / batch_count);
const index_t g_idx = __builtin_amdgcn_readfirstlane(get_block_1d_id() / num_blocks_per_batch);
const long_index_t a_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_ptr_offset_of_batch.GetAPtrOffset(g_idx)));
const long_index_t b_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_ptr_offset_of_batch.GetBPtrOffset(g_idx)));
const long_index_t c_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_ptr_offset_of_batch.GetCPtrOffset(g_idx)));
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
GridwiseGemm::template Run<HasMainKBlockLoop>(p_a_grid + a_batch_offset,
p_b_grid + b_batch_offset,
p_c_grid + c_batch_offset,
p_shared,
a_grid_desc_k0_m_k1,
b_grid_desc_k0_n_k1,
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2,
a_element_op,
b_element_op,
c_element_op,
block_2_ctile_map);
#else
ignore = p_a_grid;
ignore = p_b_grid;
ignore = p_c_grid;
ignore = batch_count;
ignore = a_grid_desc_k0_m_k1;
ignore = b_grid_desc_k0_n_k1;
ignore = c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2;
ignore = a_element_op;
ignore = b_element_op;
ignore = c_element_op;
ignore = compute_ptr_offset_of_batch;
ignore = block_2_ctile_map;
#endif
}
template <typename ADataType,
typename BDataType,
typename CDataType,
typename AccDataType,
typename ALayout,
typename BLayout,
typename CLayout,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CElementwiseOperation,
ck::index_t BlockSize,
ck::index_t MPerBlock,
ck::index_t NPerBlock,
ck::index_t K0PerBlock,
ck::index_t K1,
ck::index_t MPerXDL,
ck::index_t NPerXDL,
ck::index_t MXdlPerWave,
ck::index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_K0_M_K1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
ck::index_t ABlockTransferSrcVectorDim,
ck::index_t ABlockTransferSrcScalarPerVector,
ck::index_t ABlockTransferDstScalarPerVector_K1,
bool ABlockLdsAddExtraM,
typename BBlockTransferThreadClusterLengths_K0_N_K1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
ck::index_t BBlockTransferSrcVectorDim,
ck::index_t BBlockTransferSrcScalarPerVector,
ck::index_t BBlockTransferDstScalarPerVector_K1,
bool BBlockLdsAddExtraN,
ck::index_t CThreadTransferSrcDstVectorDim,
ck::index_t CThreadTransferDstScalarPerVector>
struct DeviceBatchedGemmXdl : public DeviceBatchedGemm<ALayout,
BLayout,
CLayout,
ADataType,
BDataType,
CDataType,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation>
{
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto K1Number = Number<K1>{};
static auto MakeAGridDescriptor_K0_M_K1(index_t M, index_t K, index_t StrideA)
{
assert(K % K1 == 0);
const index_t K0 = K / K1;
const auto a_grid_desc_m_k = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, ALayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, K), make_tuple(StrideA, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, ALayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, K), make_tuple(I1, StrideA));
}
}();
const auto PadM = (MPerBlock - M % MPerBlock) % MPerBlock;
const auto a_grid_desc_k0_mp_k1 =
transform_tensor_descriptor(a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(K0, K1Number)),
make_right_pad_transform(M, PadM)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_k0_mp_k1;
}
static auto MakeBGridDescriptor_K0_N_K1(index_t K, index_t N, index_t StrideB)
{
assert(K % K1 == 0);
const index_t K0 = K / K1;
const auto b_grid_desc_k_n = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(K, N), make_tuple(StrideB, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(K, N), make_tuple(I1, StrideB));
}
}();
const auto PadN = (NPerBlock - N % NPerBlock) % NPerBlock;
const auto b_grid_desc_k0_np_k1 =
transform_tensor_descriptor(b_grid_desc_k_n,
make_tuple(make_unmerge_transform(make_tuple(K0, K1Number)),
make_right_pad_transform(N, PadN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_k0_np_k1;
}
static auto MakeCGridDescriptor_M_N(index_t M, index_t N, index_t StrideC)
{
const auto c_grid_desc_m_n = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, N), make_tuple(StrideC, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, N), make_tuple(I1, StrideC));
}
}();
const auto PadM = (MPerBlock - M % MPerBlock) % MPerBlock;
const auto PadN = (NPerBlock - N % NPerBlock) % NPerBlock;
const auto c_grid_desc_mp_np = transform_tensor_descriptor(
c_grid_desc_m_n,
make_tuple(make_right_pad_transform(M, PadM), make_right_pad_transform(N, PadN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return c_grid_desc_mp_np;
}
using AGridDesc_K0_M_K1 = decltype(MakeAGridDescriptor_K0_M_K1(1, 1, 1));
using BGridDesc_K0_N_K1 = decltype(MakeBGridDescriptor_K0_N_K1(1, 1, 1));
using CGridDesc_M_N = decltype(MakeCGridDescriptor_M_N(1, 1, 1));
struct ComputePtrOffsetOfStridedBatch
{
ComputePtrOffsetOfStridedBatch(index_t BatchStrideA,
index_t BatchStrideB,
index_t BatchStrideC)
: BatchStrideA_(BatchStrideA), BatchStrideB_(BatchStrideB), BatchStrideC_(BatchStrideC)
{
}
__host__ __device__ constexpr long_index_t GetAPtrOffset(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideA_);
}
__host__ __device__ constexpr long_index_t GetBPtrOffset(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideB_);
}
__host__ __device__ constexpr long_index_t GetCPtrOffset(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideC_);
}
private:
index_t BatchStrideA_;
index_t BatchStrideB_;
index_t BatchStrideC_;
};
// GridwiseGemm
using GridwiseGemm =
GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r3<BlockSize,
ADataType, // TODO: distinguish A/B datatype
AccDataType,
CDataType,
InMemoryDataOperationEnum::Set,
AGridDesc_K0_M_K1,
BGridDesc_K0_N_K1,
CGridDesc_M_N,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
MPerBlock,
NPerBlock,
K0PerBlock,
MPerXDL,
NPerXDL,
K1,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_K0_M_K1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_K1,
false, // AThreadTransferSrcResetCoordinateAfterRun,
ABlockLdsAddExtraM,
BBlockTransferThreadClusterLengths_K0_N_K1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_K1,
false, // BThreadTransferSrcResetCoordinateAfterRun,
BBlockLdsAddExtraN,
Sequence<2, 3, 0, 1, 7, 5, 4, 6>,
CThreadTransferSrcDstVectorDim,
CThreadTransferDstScalarPerVector>;
using CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2 =
decltype(GridwiseGemm::MakeCGridDescriptor_M0_N0_M1_N1_M2_M3_M4_N2(CGridDesc_M_N{}));
using Block2CTileMap = typename GridwiseGemm::DefaultBlock2CTileMap;
// Argument
struct Argument : public BaseArgument
{
Argument(const ADataType* p_a_grid,
const BDataType* p_b_grid,
CDataType* p_c_grid,
index_t M,
index_t N,
index_t K,
index_t StrideA,
index_t StrideB,
index_t StrideC,
index_t BatchStrideA,
index_t BatchStrideB,
index_t BatchStrideC,
index_t Batch,
index_t M01,
index_t N01,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op)
: p_a_grid_{p_a_grid},
p_b_grid_{p_b_grid},
p_c_grid_{p_c_grid},
Batch_(Batch),
a_grid_desc_k0_m_k1_{
DeviceBatchedGemmXdl::MakeAGridDescriptor_K0_M_K1(M, K, StrideA)},
b_grid_desc_k0_n_k1_{
DeviceBatchedGemmXdl::MakeBGridDescriptor_K0_N_K1(K, N, StrideB)},
c_grid_desc_m_n_{DeviceBatchedGemmXdl::MakeCGridDescriptor_M_N(M, N, StrideC)},
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_{},
compute_ptr_offset_of_batch_{BatchStrideA, BatchStrideB, BatchStrideC},
block_2_ctile_map_{
GridwiseGemm::MakeDefaultBlock2CTileMap(c_grid_desc_m_n_, M01, N01)},
M01_{M01},
N01_{N01},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
c_element_op_{c_element_op}
{
if(GridwiseGemm::CheckValidity(a_grid_desc_k0_m_k1_,
b_grid_desc_k0_n_k1_,
c_grid_desc_m_n_,
block_2_ctile_map_))
{
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_ =
GridwiseGemm::MakeCGridDescriptor_M0_N0_M1_N1_M2_M3_M4_N2(c_grid_desc_m_n_);
}
}
// private:
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
CDataType* p_c_grid_;
index_t Batch_;
AGridDesc_K0_M_K1 a_grid_desc_k0_m_k1_;
BGridDesc_K0_N_K1 b_grid_desc_k0_n_k1_;
CGridDesc_M_N c_grid_desc_m_n_;
CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2 c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_;
ComputePtrOffsetOfStridedBatch compute_ptr_offset_of_batch_;
Block2CTileMap block_2_ctile_map_;
index_t M01_;
index_t N01_;
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CElementwiseOperation c_element_op_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceBatchedGemmXdl::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
{
std::cout << "arg.a_grid_desc_k0_m_k1_{" << arg.a_grid_desc_k0_m_k1_.GetLength(I0)
<< ", " << arg.a_grid_desc_k0_m_k1_.GetLength(I1) << ", "
<< arg.a_grid_desc_k0_m_k1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.b_grid_desc_k0_n_k1_{" << arg.b_grid_desc_k0_n_k1_.GetLength(I0)
<< ", " << arg.b_grid_desc_k0_n_k1_.GetLength(I1) << ", "
<< arg.b_grid_desc_k0_n_k1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.c_grid_desc_m_n_{" << arg.c_grid_desc_m_n_.GetLength(I0) << ", "
<< arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
}
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_))
{
throw std::runtime_error(
"wrong! GridwiseBatchedGemm_km_kn_m0m1n0n1_xdlops_v2r3 has invalid setting");
}
const index_t grid_size =
arg.block_2_ctile_map_.CalculateGridSize(arg.c_grid_desc_m_n_) * arg.Batch_;
const auto K =
arg.a_grid_desc_k0_m_k1_.GetLength(I0) * arg.a_grid_desc_k0_m_k1_.GetLength(I2);
float ave_time = 0;
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
const auto kernel = kernel_batched_gemm_xdlops_v2r3<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceBatchedGemmXdl::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceBatchedGemmXdl::BGridDesc_K0_N_K1>,
remove_reference_t<typename GridwiseGemm::CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2>,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
ComputePtrOffsetOfStridedBatch,
remove_reference_t<Block2CTileMap>,
true>;
ave_time = launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.Batch_,
arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.compute_ptr_offset_of_batch_,
arg.block_2_ctile_map_);
}
else
{
const auto kernel = kernel_batched_gemm_xdlops_v2r3<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceBatchedGemmXdl::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceBatchedGemmXdl::BGridDesc_K0_N_K1>,
remove_reference_t<typename GridwiseGemm::CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2>,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
ComputePtrOffsetOfStridedBatch,
remove_reference_t<Block2CTileMap>,
false>;
ave_time = launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.Batch_,
arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.compute_ptr_offset_of_batch_,
arg.block_2_ctile_map_);
}
return ave_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
return GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const ADataType* p_a,
const BDataType* p_b,
CDataType* p_c,
index_t M,
index_t N,
index_t K,
index_t StrideA,
index_t StrideB,
index_t StrideC,
index_t BatchStrideA,
index_t BatchStrideB,
index_t BatchStrideC,
index_t Batch,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op)
{
return Argument{p_a,
p_b,
p_c,
M,
N,
K,
StrideA,
StrideB,
StrideC,
BatchStrideA,
BatchStrideB,
BatchStrideC,
Batch,
1,
1,
a_element_op,
b_element_op,
c_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,
const void* p_b,
void* p_c,
index_t M,
index_t N,
index_t K,
index_t StrideA,
index_t StrideB,
index_t StrideC,
index_t BatchStrideA,
index_t BatchStrideB,
index_t BatchStrideC,
index_t Batch,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op) override
{
return std::make_unique<Argument>(static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
static_cast<CDataType*>(p_c),
M,
N,
K,
StrideA,
StrideB,
StrideC,
BatchStrideA,
BatchStrideB,
BatchStrideC,
Batch,
1,
1,
a_element_op,
b_element_op,
c_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceBatchedGemmXdl"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< K0PerBlock
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

948
include/ck/tensor_operation/gpu/device/impl/device_cgemm_4gemm_xdl_cshuffle.hpp Normal file
View File

@@ -0,0 +1,948 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_gemm.hpp"
#include "ck/tensor_operation/gpu/device/device_cgemm.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v1.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_elementwise_1d.hpp"
#include "ck/tensor_operation/gpu/element/binary_element_wise_operation.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <
typename ALayout,
typename BLayout,
typename CLayout,
typename ADataType,
typename BDataType,
typename CDataType,
typename GemmAccDataType,
typename CShuffleDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CElementwiseOperation,
GemmSpecialization GemmSpec,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock,
index_t KPerBlock,
index_t AK1,
index_t BK1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
bool ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
bool BBlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CShuffleBlockTransferScalarPerVector_NPerBlock,
LoopScheduler LoopSched = make_default_loop_scheduler(),
enable_if_t<
is_same_v<AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough> &&
is_same_v<BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough> &&
is_same_v<CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough>,
bool> = false>
struct DeviceCGemm_4Gemm_Xdl_CShuffle
: public DeviceCGemm<AElementwiseOperation, BElementwiseOperation, CElementwiseOperation>
{
using DeviceOp = DeviceCGemm_4Gemm_Xdl_CShuffle;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto MPerThread = Number<4>{};
static constexpr auto AScalarPerVector = Number<4>{};
static constexpr auto BScalarPerVector = Number<4>{};
static constexpr auto CScalarPerVector = Number<4>{};
template <typename Desc_M>
static auto PadDescriptor_M_1d(Desc_M desc_m, index_t gridSize, index_t blockSize)
{
const auto M = desc_m.GetLength(I0);
const index_t loop_step = gridSize * blockSize * MPerThread;
const auto pad = math::integer_least_multiple(M, loop_step) - M;
const auto desc_m_pad =
transform_tensor_descriptor(desc_m,
make_tuple(make_right_pad_transform(M, pad)),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0>{}));
return desc_m_pad;
}
static auto MakeDescriptor_M(const std::vector<index_t>& lengths,
const std::vector<index_t>& strides,
index_t gridSize,
index_t blockSize)
{
auto tupleOfShape = generate_tuple([&](auto I) { return lengths[I]; }, Number<2>{});
auto tupleOfStride = generate_tuple([&](auto I) { return strides[I]; }, Number<2>{});
// nd desc - [s0, s1, s2, ...]
const auto desc = make_naive_tensor_descriptor(tupleOfShape, tupleOfStride);
const auto desc_m = transform_tensor_descriptor(
desc,
make_tuple(make_merge_transform(tupleOfShape)),
make_tuple(generate_sequence_v2([&](auto I) { return I; }, Number<2>{})),
make_tuple(Sequence<0>{}));
return PadDescriptor_M_1d(desc_m, gridSize, blockSize);
}
static auto MakeAGridDescriptor_AK0_M_AK1(index_t MRaw, index_t KRaw, index_t StrideA)
{
const auto a_grid_desc_mraw_kraw = [&]() {
if constexpr(is_same_v<tensor_layout::gemm::RowMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(StrideA, I1));
}
else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(I1, StrideA));
}
}();
const auto M = math::integer_divide_ceil(MRaw, MPerBlock) * MPerBlock;
const auto K = math::integer_divide_ceil(KRaw, KPerBlock) * KPerBlock;
const auto MPad = M - MRaw;
const auto KPad = K - KRaw;
if constexpr(GemmSpec == GemmSpecialization::MKPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad both M and K
assert(K % AK1 == 0);
const auto AK0 = K / AK1;
const auto a_grid_desc_m_k =
transform_tensor_descriptor(a_grid_desc_mraw_kraw,
make_tuple(make_right_pad_transform(MRaw, MPad),
make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_pass_through_transform(M)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
else if constexpr(GemmSpec == GemmSpecialization::MPadding ||
GemmSpec == GemmSpecialization::MNPadding)
{
// pad M, but not K
assert(KRaw % AK1 == 0);
const auto AK0 = KRaw / AK1;
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_mraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_right_pad_transform(MRaw, MPad)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
else if constexpr(GemmSpec == GemmSpecialization::KPadding ||
GemmSpec == GemmSpecialization::NKPadding)
{
// pad K, but not M
assert(K % AK1 == 0);
const auto AK0 = K / AK1;
const auto a_grid_desc_m_k = transform_tensor_descriptor(
a_grid_desc_mraw_kraw,
make_tuple(make_pass_through_transform(MRaw), make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_pass_through_transform(MRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
else
{
// not pad M or K
assert(KRaw % AK1 == 0);
const auto AK0 = KRaw / AK1;
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_mraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_pass_through_transform(MRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
}
static auto MakeBGridDescriptor_BK0_N_BK1(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
const auto N = math::integer_divide_ceil(NRaw, NPerBlock) * NPerBlock;
const auto K = math::integer_divide_ceil(KRaw, KPerBlock) * KPerBlock;
const auto NPad = N - NRaw;
const auto KPad = K - KRaw;
if constexpr(GemmSpec == GemmSpecialization::NKPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad both N and K
assert(K % BK1 == 0);
const auto BK0 = K / BK1;
const auto b_grid_desc_n_k =
transform_tensor_descriptor(b_grid_desc_nraw_kraw,
make_tuple(make_right_pad_transform(NRaw, NPad),
make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_n_k,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_pass_through_transform(N)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
else if constexpr(GemmSpec == GemmSpecialization::NPadding ||
GemmSpec == GemmSpecialization::MNPadding)
{
// pad N, but not K
assert(KRaw % BK1 == 0);
const auto BK0 = KRaw / BK1;
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_nraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_right_pad_transform(NRaw, NPad)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
else if constexpr(GemmSpec == GemmSpecialization::KPadding ||
GemmSpec == GemmSpecialization::MKPadding)
{
// pad K, but not N
assert(K % BK1 == 0);
const auto BK0 = K / BK1;
const auto b_grid_desc_n_k = transform_tensor_descriptor(
b_grid_desc_nraw_kraw,
make_tuple(make_pass_through_transform(NRaw), make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_n_k,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_pass_through_transform(NRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
else
{
// not pad N or K
assert(KRaw % BK1 == 0);
const auto BK0 = KRaw / BK1;
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_nraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_pass_through_transform(NRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
}
static auto MakeCGridDescriptor_M_N(index_t MRaw, index_t NRaw, index_t StrideC)
{
const auto c_grid_desc_mraw_nraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(StrideC, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(I1, StrideC));
}
}();
const auto M = math::integer_divide_ceil(MRaw, MPerBlock) * MPerBlock;
const auto N = math::integer_divide_ceil(NRaw, NPerBlock) * NPerBlock;
const auto MPad = M - MRaw;
const auto NPad = N - NRaw;
if constexpr(GemmSpec == GemmSpecialization::MNPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad M and N
return transform_tensor_descriptor(c_grid_desc_mraw_nraw,
make_tuple(make_right_pad_transform(MRaw, MPad),
make_right_pad_transform(NRaw, NPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else if constexpr(GemmSpec == GemmSpecialization::MPadding ||
GemmSpec == GemmSpecialization::MKPadding)
{
// pad M, but not N
return transform_tensor_descriptor(
c_grid_desc_mraw_nraw,
make_tuple(make_right_pad_transform(MRaw, MPad), make_pass_through_transform(NRaw)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else if constexpr(GemmSpec == GemmSpecialization::NPadding ||
GemmSpec == GemmSpecialization::NKPadding)
{
// pad N, but not M
return transform_tensor_descriptor(
c_grid_desc_mraw_nraw,
make_tuple(make_pass_through_transform(MRaw), make_right_pad_transform(NRaw, NPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else
{
// not pad M or N
return c_grid_desc_mraw_nraw;
}
}
using AGridDesc_AK0_M_AK1 = decltype(MakeAGridDescriptor_AK0_M_AK1(1, 1, 1));
using BGridDesc_BK0_N_BK1 = decltype(MakeBGridDescriptor_BK0_N_BK1(1, 1, 1));
using CGridDesc_M_N = decltype(MakeCGridDescriptor_M_N(1, 1, 1));
using CGridDesc_M = decltype(MakeDescriptor_M({1, 1}, {1, 1}, 1, 1));
// GridwiseGemm
using GridwiseGemm = GridwiseGemm_k0mk1_k0nk1_mn_xdl_cshuffle_v1<
ADataType, // TODO: distinguish A/B datatype
GemmAccDataType,
CShuffleDataType,
CDataType,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
InMemoryDataOperationEnum::Set,
AGridDesc_AK0_M_AK1,
BGridDesc_BK0_N_BK1,
CGridDesc_M_N,
NumGemmKPrefetchStage,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
AK1,
BK1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
false,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
false,
BBlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CShuffleBlockTransferScalarPerVector_NPerBlock,
LoopSched>;
// Argument
struct Argument : public BaseArgument
{
Argument(const ADataType* p_a_grid_real,
const ADataType* p_a_grid_imag,
const BDataType* p_b_grid_real,
const BDataType* p_b_grid_imag,
CDataType* p_c_grid_real,
CDataType* p_c_grid_imag,
CDataType* p_workspace,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op)
: p_a_grid_real_{p_a_grid_real},
p_a_grid_imag_{p_a_grid_imag},
p_b_grid_real_{p_b_grid_real},
p_b_grid_imag_{p_b_grid_imag},
p_c_grid_real_{p_c_grid_real},
p_c_grid_imag_{p_c_grid_imag},
p_aux_grid_{p_workspace},
a_grid_desc_ak0_m_ak1_{DeviceOp::MakeAGridDescriptor_AK0_M_AK1(MRaw, KRaw, StrideA)},
b_grid_desc_bk0_n_bk1_{DeviceOp::MakeBGridDescriptor_BK0_N_BK1(KRaw, NRaw, StrideB)},
c_grid_desc_m_n_{DeviceOp::MakeCGridDescriptor_M_N(MRaw, NRaw, StrideC)},
c_grid_desc_mblock_mperblock_nblock_nperblock_{},
block_2_ctile_map_{GridwiseGemm::MakeDefaultBlock2CTileMap(c_grid_desc_m_n_)},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
c_element_op_{c_element_op}
{
if(GridwiseGemm::CheckValidity(a_grid_desc_ak0_m_ak1_,
b_grid_desc_bk0_n_bk1_,
c_grid_desc_m_n_,
block_2_ctile_map_))
{
c_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeCGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
c_grid_desc_m_n_);
}
const index_t grid_size = block_2_ctile_map_.CalculateGridSize(c_grid_desc_m_n_);
if constexpr(is_same<tensor_layout::gemm::RowMajor, CLayout>::value)
{
c_grid_desc_m_ =
DeviceOp::MakeDescriptor_M({MRaw, NRaw}, {StrideC, I1}, grid_size, BlockSize);
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, CLayout>::value)
{
c_grid_desc_m_ =
DeviceOp::MakeDescriptor_M({MRaw, NRaw}, {I1, StrideC}, grid_size, BlockSize);
}
p_aux_2_grid_ = p_workspace + c_grid_desc_m_n_.GetElementSpaceSize();
}
// private:
const ADataType* p_a_grid_real_;
const ADataType* p_a_grid_imag_;
const BDataType* p_b_grid_real_;
const BDataType* p_b_grid_imag_;
CDataType* p_c_grid_real_;
CDataType* p_c_grid_imag_;
CDataType* p_aux_grid_;
CDataType* p_aux_2_grid_;
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
CGridDesc_M_N c_grid_desc_m_n_;
CGridDesc_M c_grid_desc_m_;
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
c_grid_desc_mblock_mperblock_nblock_nperblock_;
typename GridwiseGemm::DefaultBlock2CTileMap block_2_ctile_map_;
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CElementwiseOperation c_element_op_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_))
{
throw std::runtime_error("wrong! GridwiseGemm has invalid setting");
}
const index_t grid_size =
arg.block_2_ctile_map_.CalculateGridSize(arg.c_grid_desc_m_n_);
const auto K =
arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) * arg.a_grid_desc_ak0_m_ak1_.GetLength(I2);
float ave_time = 0;
using Add = ck::tensor_operation::element_wise::Add;
using Subtract = ck::tensor_operation::element_wise::Subtract;
using GridwiseBinAdd =
GridwiseElementwise_1D<Tuple<CGridDesc_M, CGridDesc_M>,
Tuple<CGridDesc_M>,
Tuple<const CDataType*, const CDataType*>,
Tuple<CDataType*>,
Add,
MPerThread,
Sequence<AScalarPerVector, BScalarPerVector>,
Sequence<CScalarPerVector>>;
using GridwiseBinSubtract =
GridwiseElementwise_1D<Tuple<CGridDesc_M, CGridDesc_M>,
Tuple<CGridDesc_M>,
Tuple<const CDataType*, const CDataType*>,
Tuple<CDataType*>,
Subtract,
MPerThread,
Sequence<AScalarPerVector, BScalarPerVector>,
Sequence<CScalarPerVector>>;
const auto add_kernel = kernel_elementwise_1d<GridwiseBinAdd,
Tuple<CGridDesc_M, CGridDesc_M>,
Tuple<CGridDesc_M>,
Tuple<const CDataType*, const CDataType*>,
Tuple<CDataType*>,
Add>;
const auto subtract_kernel =
kernel_elementwise_1d<GridwiseBinSubtract,
Tuple<CGridDesc_M, CGridDesc_M>,
Tuple<CGridDesc_M>,
Tuple<const CDataType*, const CDataType*>,
Tuple<CDataType*>,
Subtract>;
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
const auto kernel = kernel_gemm_xdl_cshuffle_v1<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::DefaultBlock2CTileMap,
true>;
ave_time +=
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_real_,
arg.p_b_grid_real_,
arg.p_aux_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_ctile_map_);
ave_time +=
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_imag_,
arg.p_b_grid_imag_,
arg.p_aux_2_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_ctile_map_);
// c_real = aux - aux_2
ave_time += launch_and_time_kernel(
stream_config,
subtract_kernel,
dim3(grid_size),
dim3(BlockSize),
0,
make_tuple(arg.c_grid_desc_m_, arg.c_grid_desc_m_),
make_tuple(arg.c_grid_desc_m_),
make_tuple(const_cast<const CDataType*>(arg.p_aux_grid_),
const_cast<const CDataType*>(arg.p_aux_2_grid_)),
make_tuple(arg.p_c_grid_real_),
Subtract{});
ave_time +=
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_real_,
arg.p_b_grid_imag_,
arg.p_aux_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_ctile_map_);
ave_time +=
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_imag_,
arg.p_b_grid_real_,
arg.p_aux_2_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_ctile_map_);
// c_imag = aux + aux_2
ave_time += launch_and_time_kernel(
stream_config,
add_kernel,
dim3(grid_size),
dim3(BlockSize),
0,
make_tuple(arg.c_grid_desc_m_, arg.c_grid_desc_m_),
make_tuple(arg.c_grid_desc_m_),
make_tuple(const_cast<const CDataType*>(arg.p_aux_grid_),
const_cast<const CDataType*>(arg.p_aux_2_grid_)),
make_tuple(arg.p_c_grid_imag_),
Add{});
}
else
{
const auto kernel = kernel_gemm_xdl_cshuffle_v1<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::DefaultBlock2CTileMap,
false>;
ave_time +=
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_real_,
arg.p_b_grid_real_,
arg.p_aux_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_ctile_map_);
ave_time +=
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_imag_,
arg.p_b_grid_imag_,
arg.p_aux_2_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_ctile_map_);
// c_real = aux - aux_2
ave_time += launch_and_time_kernel(
stream_config,
subtract_kernel,
dim3(grid_size),
dim3(BlockSize),
0,
make_tuple(arg.c_grid_desc_m_, arg.c_grid_desc_m_),
make_tuple(arg.c_grid_desc_m_),
make_tuple(const_cast<const CDataType*>(arg.p_aux_grid_),
const_cast<const CDataType*>(arg.p_aux_2_grid_)),
make_tuple(arg.p_c_grid_real_),
Subtract{});
ave_time +=
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_real_,
arg.p_b_grid_imag_,
arg.p_aux_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_ctile_map_);
ave_time +=
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_imag_,
arg.p_b_grid_real_,
arg.p_aux_2_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_ctile_map_);
// c_imag = aux + aux_2
ave_time += launch_and_time_kernel(
stream_config,
add_kernel,
dim3(grid_size),
dim3(BlockSize),
0,
make_tuple(arg.c_grid_desc_m_, arg.c_grid_desc_m_),
make_tuple(arg.c_grid_desc_m_),
make_tuple(const_cast<const CDataType*>(arg.p_aux_grid_),
const_cast<const CDataType*>(arg.p_aux_2_grid_)),
make_tuple(arg.p_c_grid_imag_),
Add{});
}
return ave_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
return GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const ADataType* p_a_real,
const ADataType* p_a_imag,
const BDataType* p_b_real,
const BDataType* p_b_imag,
CDataType* p_c_real,
CDataType* p_c_imag,
CDataType* p_workspace,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op)
{
return Argument{p_a_real,
p_a_imag,
p_b_real,
p_b_imag,
p_c_real,
p_c_imag,
p_workspace,
MRaw,
NRaw,
KRaw,
StrideA,
StrideB,
StrideC,
a_element_op,
b_element_op,
c_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a_real,
const void* p_a_imag,
const void* p_b_real,
const void* p_b_imag,
void* p_c_real,
void* p_c_imag,
void* p_workspace,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op,
index_t /* KBatch */ = 1) override
{
return std::make_unique<Argument>(static_cast<const ADataType*>(p_a_real),
static_cast<const ADataType*>(p_a_imag),
static_cast<const BDataType*>(p_b_real),
static_cast<const BDataType*>(p_b_imag),
static_cast<CDataType*>(p_c_real),
static_cast<CDataType*>(p_c_imag),
static_cast<CDataType*>(p_workspace),
MRaw,
NRaw,
KRaw,
StrideA,
StrideB,
StrideC,
a_element_op,
b_element_op,
c_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceCGemm_4Gemm_Xdl_CShuffle"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< AK1 << ", "
<< BK1
<< ">";
// clang-format on
return str.str();
}
std::size_t GetWorkspaceSize(index_t MRaw,
index_t NRaw,
[[maybe_unused]] index_t KRaw,
[[maybe_unused]] index_t StrideA,
[[maybe_unused]] index_t StrideB,
index_t StrideC) override
{
const auto c_grid_desc_m_n = MakeCGridDescriptor_M_N(MRaw, NRaw, StrideC);
return 2 * sizeof(CDataType) * c_grid_desc_m_n.GetElementSpaceSize();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

779
include/ck/tensor_operation/gpu/device/impl/device_contraction_multiple_d_xdl_cshuffle.hpp Normal file
View File

@@ -0,0 +1,779 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
template <typename GridwiseGemm,
typename FloatAB,
typename FloatDsPointer,
typename FloatE,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
typename AGridDesc_AK0_M_AK1,
typename BGridDesc_BK0_N_BK1,
typename DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename Block2ETileMap,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_contraction_multiple_d_xdl_cshuffle(
const FloatAB* __restrict__ p_a_grid,
const FloatAB* __restrict__ p_b_grid,
FloatDsPointer p_ds_grid,
FloatE* __restrict__ p_e_grid,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const CDEElementwiseOperation cde_element_op,
const AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1,
const BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1,
const DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
ds_grid_desc_mblock_mperblock_nblock_nperblock,
const EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
e_grid_desc_mblock_mperblock_nblock_nperblock,
const Block2ETileMap block_2_etile_map)
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__))
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
GridwiseGemm::template Run<HasMainKBlockLoop>(p_a_grid,
p_b_grid,
p_ds_grid,
p_e_grid,
p_shared,
a_element_op,
b_element_op,
cde_element_op,
a_grid_desc_ak0_m_ak1,
b_grid_desc_bk0_n_bk1,
ds_grid_desc_mblock_mperblock_nblock_nperblock,
e_grid_desc_mblock_mperblock_nblock_nperblock,
block_2_etile_map);
#else
ignore = p_a_grid;
ignore = p_b_grid;
ignore = p_ds_grid;
ignore = p_e_grid;
ignore = a_element_op;
ignore = b_element_op;
ignore = cde_element_op;
ignore = a_grid_desc_ak0_m_ak1;
ignore = b_grid_desc_bk0_n_bk1;
ignore = ds_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = e_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = block_2_etile_map;
#endif
}
} // namespace ck
namespace ck {
namespace tensor_operation {
namespace device {
// Tensor Contraction:
// input : A
// input : B
// input : D0, D1, ...
// output : E
// C = a_op(A) * b_op(B)
// E = cde_op(C, D0, D1, ...)
// Assume:
// A[M0, M1, M2, ..., K0, K1, K2, ...]
// B[N0, N1, N2, ..., K0, K1, K2, ...]
// D[M0, M1, M2, ..., N0, N1, N2, ...]
// E[M0, M1, M2, ..., N0, N1, N2, ...]
template <index_t NumDimM,
index_t NumDimN,
index_t NumDimK,
typename ADataType,
typename BDataType,
typename AccDataType,
typename CShuffleDataType,
typename DsDataType,
typename EDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
GemmSpecialization GemmSpec,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock,
index_t KPerBlock,
index_t AK1,
index_t BK1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
bool ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
bool BBlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CDEBlockTransferScalarPerVector_NPerBlock,
LoopScheduler LoopSched = make_default_loop_scheduler()>
struct DeviceContractionMultipleD_Xdl_CShuffle
: public DeviceContractionMultipleD<NumDimM,
NumDimN,
NumDimK,
ADataType,
BDataType,
DsDataType,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation>
{
using DeviceOp = DeviceContractionMultipleD_Xdl_CShuffle;
static constexpr index_t NumDTensor = DsDataType::Size();
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr auto matrix_padder =
MatrixPadder<GemmSpec, index_t, index_t, index_t>{MPerBlock, NPerBlock, KPerBlock};
// Assume: A[M0, M1, M2, ..., K0, K1, K2, ...]
static auto MakeAGridDescriptor_M_K(const std::vector<index_t>& a_ms_ks_lengths_vec,
const std::vector<index_t>& a_ms_ks_strides_vec)
{
assert(a_ms_ks_lengths_vec.size() == NumDimM + NumDimK &&
a_ms_ks_strides_vec.size() == NumDimM + NumDimK);
const auto to_tuple = [&](auto& vec, auto num) {
return generate_tuple([&](auto i) { return vec[i]; }, num);
};
const auto a_ms_ns_lengths = to_tuple(a_ms_ks_lengths_vec, Number<NumDimM + NumDimK>{});
const auto a_ms_ks_strides = to_tuple(a_ms_ks_strides_vec, Number<NumDimM + NumDimK>{});
// dimension Ids for M0, M1, ...
constexpr auto mDimIds = typename arithmetic_sequence_gen<0, NumDimM, 1>::type{};
// dimension Ids for K0, K1, ...
constexpr auto kDimIds =
typename arithmetic_sequence_gen<NumDimM, NumDimM + NumDimK, 1>::type{};
// lengths for M0, M1, ...
const auto mLengths = get_container_subset(a_ms_ns_lengths, mDimIds);
// lengths for K0, K1, ...
const auto kLengths = get_container_subset(a_ms_ns_lengths, kDimIds);
// naive tensor A[M0, M1, M2, ..., K0, K1, K2...]
const auto a_grid_desc_ms_ks =
make_naive_tensor_descriptor(a_ms_ns_lengths, a_ms_ks_strides);
// transformed tensor A[MRaw = M0 * M1 * M2 * ... , KRaw = K0 * K1 * K2 * ...]
const auto a_grid_desc_mraw_kraw = transform_tensor_descriptor(
a_grid_desc_ms_ks,
make_tuple(make_merge_transform(mLengths), make_merge_transform(kLengths)),
make_tuple(mDimIds, kDimIds),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return matrix_padder.PadADescriptor_M_K(a_grid_desc_mraw_kraw);
}
// Assume: B[N0, N1, N2, ..., K0, K1, K2, ...]
static auto MakeBGridDescriptor_N_K(const std::vector<index_t>& b_ns_ks_lengths_vec,
const std::vector<index_t>& b_ns_ks_strides_vec)
{
assert(b_ns_ks_lengths_vec.size() == NumDimN + NumDimK &&
b_ns_ks_strides_vec.size() == NumDimN + NumDimK);
const auto to_tuple = [&](auto& vec, auto num) {
return generate_tuple([&](auto i) { return vec[i]; }, num);
};
const auto b_ns_ks_lengths = to_tuple(b_ns_ks_lengths_vec, Number<NumDimN + NumDimK>{});
const auto b_ns_ks_strides = to_tuple(b_ns_ks_strides_vec, Number<NumDimN + NumDimK>{});
// dimension Ids for N0, N1, ...
constexpr auto nDimIds = typename arithmetic_sequence_gen<0, NumDimN, 1>::type{};
// dimension Ids for K0, K1, ...
constexpr auto kDimIds =
typename arithmetic_sequence_gen<NumDimN, NumDimN + NumDimK, 1>::type{};
// lengths for K0, K1, ...
const auto kLengths = get_container_subset(b_ns_ks_lengths, kDimIds);
// lengths for N0, N1, ...
const auto nLengths = get_container_subset(b_ns_ks_lengths, nDimIds);
// naive tensor B[N0, N1, N2, ..., K0, K1, K2, ...]
const auto b_grid_desc_ns_ks =
make_naive_tensor_descriptor(b_ns_ks_lengths, b_ns_ks_strides);
// transformed tensor B[NRaw = N0 * N1 * N2 * ..., KRaw = K0 * K1 * K2 * ...]
const auto b_grid_desc_nraw_kraw = transform_tensor_descriptor(
b_grid_desc_ns_ks,
make_tuple(make_merge_transform(nLengths), make_merge_transform(kLengths)),
make_tuple(nDimIds, kDimIds),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return matrix_padder.PadBDescriptor_N_K(b_grid_desc_nraw_kraw);
}
// assume E[M0, M1, M2, ..., N0, N1, N2...]
static auto MakeEGridDescriptor_M_N(const std::vector<index_t>& e_ms_ns_lengths_vec,
const std::vector<index_t>& e_ms_ns_strides_vec)
{
assert(e_ms_ns_lengths_vec.size() == NumDimM + NumDimN &&
e_ms_ns_strides_vec.size() == NumDimM + NumDimN);
const auto to_tuple = [&](auto& vec, auto num) {
return generate_tuple([&](auto i) { return vec[i]; }, num);
};
const auto e_ms_ns_lengths = to_tuple(e_ms_ns_lengths_vec, Number<NumDimM + NumDimN>{});
const auto e_ms_ns_strides = to_tuple(e_ms_ns_strides_vec, Number<NumDimM + NumDimN>{});
// dimension Ids for M0, M1, ...
constexpr auto mDimIds = typename arithmetic_sequence_gen<0, NumDimM, 1>::type{};
// dimension Ids for N0, N1, ...
constexpr auto nDimIds =
typename arithmetic_sequence_gen<NumDimM, NumDimM + NumDimN, 1>::type{};
// lengths for M0, M1, ...
const auto mLengths = get_container_subset(e_ms_ns_lengths, mDimIds);
// lengths for K0, K1, ...
const auto nLengths = get_container_subset(e_ms_ns_lengths, nDimIds);
// naive tensor E[M0, M1, M2, ..., N0, N1, N2...]
const auto e_grid_desc_ms_ns =
make_naive_tensor_descriptor(e_ms_ns_lengths, e_ms_ns_strides);
// transformed tensor E[MRaw = M0 * M1 * M2 * ... , NRaw = N0 * N1 * N2 * ...]
const auto e_grid_desc_mraw_nraw = transform_tensor_descriptor(
e_grid_desc_ms_ns,
make_tuple(make_merge_transform(mLengths), make_merge_transform(nLengths)),
make_tuple(mDimIds, nDimIds),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return matrix_padder.PadCDescriptor_M_N(e_grid_desc_mraw_nraw);
}
static auto MakeDsGridDescriptor_M_N(
const std::array<std::vector<index_t>, NumDTensor>& ds_ms_ns_lengths_vec,
const std::array<std::vector<index_t>, NumDTensor>& ds_ms_ns_strides_vec)
{
return generate_tuple(
[&](auto i) {
return DeviceOp::MakeEGridDescriptor_M_N(ds_ms_ns_lengths_vec[i],
ds_ms_ns_strides_vec[i]);
},
Number<NumDTensor>{});
}
using AGridDesc_M_K = decltype(MakeAGridDescriptor_M_K({}, {}));
using BGridDesc_N_K = decltype(MakeBGridDescriptor_N_K({}, {}));
using DsGridDesc_M_N = remove_cvref_t<decltype(MakeDsGridDescriptor_M_N({{}}, {{}}))>;
using EGridDesc_M_N = decltype(MakeEGridDescriptor_M_N({}, {}));
// GridwiseGemm
using GridwiseGemm = GridwiseGemmMultipleD_xdl_cshuffle<
ADataType, // TODO: distinguish A/B datatype
AccDataType,
CShuffleDataType,
DsDataType,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
InMemoryDataOperationEnum::Set,
NumGemmKPrefetchStage,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
AK1,
BK1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
false,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
false,
BBlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CDEBlockTransferScalarPerVector_NPerBlock,
LoopSched>;
// desc for blockwise copy
using AGridDesc_AK0_M_AK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(AGridDesc_M_K{}))>;
using BGridDesc_BK0_N_BK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(BGridDesc_N_K{}))>;
using DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock = remove_cvref_t<decltype(
GridwiseGemm::MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(DsGridDesc_M_N{}))>;
using EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock = remove_cvref_t<decltype(
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(EGridDesc_M_N{}))>;
// block-to-e-tile map
using Block2ETileMap =
remove_cvref_t<decltype(GridwiseGemm::MakeDefaultBlock2ETileMap(EGridDesc_M_N{}))>;
// Argument
struct Argument : public BaseArgument
{
Argument(const void* p_a_grid,
const void* p_b_grid,
std::array<const void*, NumDTensor> p_ds_grid,
void* p_e_grid,
const std::vector<index_t>& a_ms_ns_lengths,
const std::vector<index_t>& a_ms_ks_strides,
const std::vector<index_t>& b_ns_ks_lengths,
const std::vector<index_t>& b_ns_ks_strides,
const std::array<std::vector<index_t>, NumDTensor>& ds_ms_ns_lengths,
const std::array<std::vector<index_t>, NumDTensor>& ds_ms_ns_strides,
const std::vector<index_t>& e_ms_ns_lengths,
const std::vector<index_t>& e_ms_ns_strides,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op)
: p_a_grid_{static_cast<const ADataType*>(p_a_grid)},
p_b_grid_{static_cast<const BDataType*>(p_b_grid)},
p_ds_grid_{},
p_e_grid_{static_cast<EDataType*>(p_e_grid)},
a_grid_desc_m_k_{DeviceOp::MakeAGridDescriptor_M_K(a_ms_ns_lengths, a_ms_ks_strides)},
b_grid_desc_n_k_{DeviceOp::MakeBGridDescriptor_N_K(b_ns_ks_lengths, b_ns_ks_strides)},
ds_grid_desc_m_n_{},
e_grid_desc_m_n_{DeviceOp::MakeEGridDescriptor_M_N(e_ms_ns_lengths, e_ms_ns_strides)},
a_grid_desc_ak0_m_ak1_{
GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(a_grid_desc_m_k_)},
b_grid_desc_bk0_n_bk1_{
GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(b_grid_desc_n_k_)},
ds_grid_desc_mblock_mperblock_nblock_nperblock_{},
e_grid_desc_mblock_mperblock_nblock_nperblock_{},
block_2_etile_map_{GridwiseGemm::MakeDefaultBlock2ETileMap(e_grid_desc_m_n_)},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
cde_element_op_{cde_element_op},
a_mz_stride_{},
a_kz_stride_{},
b_nz_stride_{},
b_kz_stride_{},
ds_nz_stride_{},
e_nz_stride_{}
{
// populate pointer, batch stride, desc for Ds
static_for<0, NumDTensor, 1>{}([&](auto i) {
using DDataType = remove_cvref_t<tuple_element_t<i.value, DsDataType>>;
// D pointer
p_ds_grid_(i) = static_cast<const DDataType*>(p_ds_grid[i]);
// D desc
ds_grid_desc_m_n_(i) =
DeviceOp::MakeEGridDescriptor_M_N(ds_ms_ns_lengths[i], ds_ms_ns_strides[i]);
});
// populate desc for Ds/E
if(GridwiseGemm::CheckValidity(a_grid_desc_m_k_,
b_grid_desc_n_k_,
ds_grid_desc_m_n_,
e_grid_desc_m_n_,
block_2_etile_map_))
{
e_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
e_grid_desc_m_n_);
ds_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
ds_grid_desc_m_n_);
}
// for sanity check of vector memory access
a_mz_stride_ = a_ms_ks_strides[NumDimM - 1];
a_kz_stride_ = a_ms_ks_strides[NumDimM + NumDimK - 1];
b_nz_stride_ = b_ns_ks_strides[NumDimN - 1];
b_kz_stride_ = b_ns_ks_strides[NumDimN + NumDimK - 1];
for(index_t i = 0; i < NumDTensor; ++i)
{
ds_nz_stride_[i] = ds_ms_ns_strides[i][NumDimM + NumDimN - 1];
}
e_nz_stride_ = e_ms_ns_strides[NumDimM + NumDimN - 1];
}
void Print() const
{
std::cout << "A[M, K]: " << a_grid_desc_m_k_ << std::endl;
std::cout << "B[N, K]: " << b_grid_desc_n_k_ << std::endl;
static_for<0, NumDTensor, 1>{}(
[&](auto i) { std::cout << "Ds[M, N]: " << ds_grid_desc_m_n_[i] << std::endl; });
std::cout << "E[M, N]: " << e_grid_desc_m_n_ << std::endl;
}
// private:
// pointers
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
typename GridwiseGemm::DsGridPointer p_ds_grid_;
EDataType* p_e_grid_;
// tensor descriptors for problem definiton
AGridDesc_M_K a_grid_desc_m_k_;
BGridDesc_N_K b_grid_desc_n_k_;
DsGridDesc_M_N ds_grid_desc_m_n_;
EGridDesc_M_N e_grid_desc_m_n_;
// tensor descriptors for block/thread-wise copy
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock
ds_grid_desc_mblock_mperblock_nblock_nperblock_;
EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock e_grid_desc_mblock_mperblock_nblock_nperblock_;
// block-to-e-tile map
Block2ETileMap block_2_etile_map_;
// element-wise op
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CDEElementwiseOperation cde_element_op_;
// Strides for the last M/N/K dimensions of A/B/Ds/E
// for sanity check of vector load/store
index_t a_mz_stride_;
index_t a_kz_stride_;
index_t b_nz_stride_;
index_t b_kz_stride_;
std::array<index_t, NumDTensor> ds_nz_stride_;
index_t e_mz_stride_;
index_t e_nz_stride_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_m_k_,
arg.b_grid_desc_n_k_,
arg.ds_grid_desc_m_n_,
arg.e_grid_desc_m_n_,
arg.block_2_etile_map_))
{
throw std::runtime_error(
"wrong! GridwiseGemmMultipleD_xdl_cshuffle has invalid setting");
}
const index_t grid_size =
arg.block_2_etile_map_.CalculateGridSize(arg.e_grid_desc_m_n_);
const auto K =
arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) * arg.a_grid_desc_ak0_m_ak1_.GetLength(I2);
auto launch_kernel = [&](auto has_main_k_block_loop) {
constexpr bool has_main_loop = has_main_k_block_loop.value;
const auto kernel = kernel_contraction_multiple_d_xdl_cshuffle<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
typename GridwiseGemm::DsGridPointer,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
DeviceOp::DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
DeviceOp::EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
DeviceOp::Block2ETileMap,
has_main_loop>;
return launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_ds_grid_,
arg.p_e_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.cde_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.ds_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.e_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_etile_map_);
};
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
return launch_kernel(integral_constant<bool, true>{});
}
else
{
return launch_kernel(integral_constant<bool, false>{});
}
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static bool IsSupportedArgument(const Argument& arg)
{
if(!(ck::get_device_name() == "gfx908" || ck::get_device_name() == "gfx90a"))
{
return false;
}
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_m_k_,
arg.b_grid_desc_n_k_,
arg.ds_grid_desc_m_n_,
arg.e_grid_desc_m_n_,
arg.block_2_etile_map_))
{
return false;
}
// check vector access
static_assert((ABlockTransferSrcVectorDim == 1 || ABlockTransferSrcVectorDim == 2) &&
(BBlockTransferSrcVectorDim == 1 || BBlockTransferSrcVectorDim == 2),
"wrong!");
// vector memory access of A: could be on M or AK1 dimension
if constexpr(ABlockTransferSrcVectorDim == 1)
{
if(!(arg.a_mz_stride_ == 1 &&
arg.a_grid_desc_ak0_m_ak1_.GetLength(I1) % ABlockTransferSrcScalarPerVector == 0))
{
return false;
}
}
else
{
if(!(arg.a_kz_stride_ == 1 &&
arg.a_grid_desc_ak0_m_ak1_.GetLength(I2) % ABlockTransferSrcScalarPerVector == 0))
{
return false;
}
}
// vector memory access of B: could be on N or BK1 dimension
if constexpr(BBlockTransferSrcVectorDim == 1)
{
if(!(arg.b_nz_stride_ == 1 &&
arg.b_grid_desc_bk0_n_bk1_.GetLength(I1) % BBlockTransferSrcScalarPerVector == 0))
{
return false;
}
}
else
{
if(!(arg.b_kz_stride_ == 1 &&
arg.b_grid_desc_bk0_n_bk1_.GetLength(I2) % BBlockTransferSrcScalarPerVector == 0))
{
return false;
}
}
// vector memory access of Ds: always on NPerBlock dimension
bool valid_d_access = true;
static_for<0, NumDTensor, 1>{}([&](auto i) {
if(!(arg.ds_nz_stride_[i] == 1 &&
arg.ds_grid_desc_mblock_mperblock_nblock_nperblock_[i].GetLength(I3) %
CDEBlockTransferScalarPerVector_NPerBlock ==
0))
{
valid_d_access = false;
}
});
if(valid_d_access == false)
{
return false;
}
// vector memory access of E: always on NPerBlock dimension
if(!(arg.e_nz_stride_ == 1 &&
arg.e_grid_desc_mblock_mperblock_nblock_nperblock_.GetLength(I3) %
CDEBlockTransferScalarPerVector_NPerBlock ==
0))
{
return false;
}
return true;
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const void* p_a,
const void* p_b,
std::array<const void*, NumDTensor> p_ds,
void* p_e,
const std::vector<index_t>& a_ms_ns_lengths,
const std::vector<index_t>& a_ms_ks_strides,
const std::vector<index_t>& b_ns_ks_lengths,
const std::vector<index_t>& b_ns_ks_strides,
const std::array<std::vector<index_t>, NumDTensor>& ds_ms_ns_lengths,
const std::array<std::vector<index_t>, NumDTensor>& ds_ms_ns_strides,
const std::vector<index_t>& e_ms_ns_lengths,
const std::vector<index_t>& e_ms_ns_strides,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op)
{
return Argument{p_a,
p_b,
p_ds,
p_e,
a_ms_ns_lengths,
a_ms_ks_strides,
b_ns_ks_lengths,
b_ns_ks_strides,
ds_ms_ns_lengths,
ds_ms_ns_strides,
e_ms_ns_lengths,
e_ms_ns_strides,
a_element_op,
b_element_op,
cde_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_a,
const void* p_b,
std::array<const void*, NumDTensor> p_ds,
void* p_e,
const std::vector<index_t>& a_ms_ns_lengths,
const std::vector<index_t>& a_ms_ks_strides,
const std::vector<index_t>& b_ns_ks_lengths,
const std::vector<index_t>& b_ns_ks_strides,
const std::array<std::vector<index_t>, NumDTensor>& ds_ms_ns_lengths,
const std::array<std::vector<index_t>, NumDTensor>& ds_ms_ns_strides,
const std::vector<index_t>& e_ms_ns_lengths,
const std::vector<index_t>& e_ms_ns_strides,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op) override
{
return std::make_unique<Argument>(p_a,
p_b,
p_ds,
p_e,
a_ms_ns_lengths,
a_ms_ks_strides,
b_ns_ks_lengths,
b_ns_ks_strides,
ds_ms_ns_lengths,
ds_ms_ns_strides,
e_ms_ns_lengths,
e_ms_ns_strides,
a_element_op,
b_element_op,
cde_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceContractionMultipleD_Xdl_CShuffle"
<< "<"
<< NumDimM << ", "
<< NumDimN << ", "
<< NumDimK << ", "
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< AK1 << ", "
<< BK1 << ", "
<< ABlockTransferSrcVectorDim << ", "
<< BBlockTransferSrcVectorDim
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

785
include/ck/tensor_operation/gpu/device/impl/device_conv2d_backward_weight_xdl_c_shuffle_nhwc_kyxc_nhwk.hpp Normal file
View File

@@ -0,0 +1,785 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_conv_bwd_weight.hpp"
#include "ck/tensor_operation/gpu/device/convolution_backward_weight_specialization.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_bwd_weight.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
// out[N, Ho, Wo, K] = in[N, Hi, Wi, C] * wei[K, Y, X, C]
template <typename InDataType,
typename WeiDataType,
typename OutDataType,
typename AccDataType,
typename InElementwiseOperation,
typename WeiElementwiseOperation,
typename OutElementwiseOperation,
ck::index_t BlockSize,
ck::index_t MPerBlock,
ck::index_t NPerBlock,
ck::index_t K0PerBlock,
ck::index_t K1,
ck::index_t MPerXdl,
ck::index_t NPerXdl,
ck::index_t MXdlPerWave,
ck::index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_K0_M_K1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
ck::index_t ABlockTransferSrcVectorDim,
ck::index_t ABlockTransferSrcScalarPerVector,
ck::index_t ABlockTransferDstScalarPerVector_K1,
bool ABlockLdsAddExtraM,
typename BBlockTransferThreadClusterLengths_K0_N_K1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
ck::index_t BBlockTransferSrcVectorDim,
ck::index_t BBlockTransferSrcScalarPerVector,
ck::index_t BBlockTransferDstScalarPerVector_K1,
bool BBlockLdsAddExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CBlockTransferScalarPerVector_NWaveNPerXdl>
struct DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
: public DeviceConvBwdWeight<2,
ck::tensor_layout::convolution::NHWC,
ck::tensor_layout::convolution::KYXC,
ck::tensor_layout::convolution::NHWK,
InDataType,
WeiDataType,
OutDataType,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation>
{
using DeviceOp =
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K;
using ADataType = OutDataType;
using BDataType = InDataType;
using CDataType = WeiDataType;
using AElementwiseOperation = OutElementwiseOperation;
using BElementwiseOperation = InElementwiseOperation;
using CElementwiseOperation = WeiElementwiseOperation;
// TODO make A/B datatype different
using ABDataType = InDataType;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr auto I4 = Number<4>{};
static constexpr auto I5 = Number<5>{};
static constexpr auto K1Number = Number<K1>{};
static constexpr auto GemmK1Number = K1Number;
static constexpr auto N1Number = K1Number;
// Bytes per 32 lds bank: 32 * 4 bytes
static constexpr auto BankLength = 128;
static constexpr auto ElePerBank = BankLength / sizeof(ADataType);
// M1 & M0
static constexpr auto ABlockLdsM1PerBlock = ElePerBank / K1;
static constexpr auto ABlockLdsM0PerBlock = MPerBlock / ABlockLdsM1PerBlock;
static constexpr auto ABlockLdsM1Padding = 4;
// N1 & N0
static constexpr auto BBlockLdsN1PerBlock = ElePerBank / K1;
static constexpr auto BBlockLdsN0PerBlock = NPerBlock / BBlockLdsN1PerBlock;
static constexpr auto BBlockLdsN1Padding = 4;
static auto
MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
ck::index_t batch_k)
{
using namespace ck;
const index_t Hi = input_spatial_lengths[0];
const index_t Wi = input_spatial_lengths[1];
const index_t Ho = output_spatial_lengths[0];
const index_t Wo = output_spatial_lengths[1];
const index_t Y = filter_spatial_lengths[0];
const index_t X = filter_spatial_lengths[1];
const index_t ConvStrideH = conv_filter_strides[0];
const index_t ConvStrideW = conv_filter_strides[1];
const index_t ConvDilationH = conv_filter_dilations[0];
const index_t ConvDilationW = conv_filter_dilations[1];
const index_t InLeftPadH = input_left_pads[0];
const index_t InLeftPadW = input_left_pads[1];
const index_t InRightPadH = input_right_pads[0];
const index_t InRightPadW = input_right_pads[1];
const index_t GemmKTotal = N * Ho * Wo;
const index_t GemmM = K;
const index_t GemmN = C * X * Y;
const index_t GemmKBatch = batch_k;
const index_t GemmK0 =
math::integer_divide_ceil(GemmKTotal, GemmK1Number * K0PerBlock * GemmKBatch) *
K0PerBlock;
const auto in_n_hi_wi_c_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N, Hi, Wi, C));
// A: output tensor
const index_t N0 = N / N1Number;
const index_t GemmK0Total = N0 * Ho * Wo;
const index_t GemmK0S =
math::integer_divide_ceil(GemmK0Total, K0PerBlock * GemmKBatch) * K0PerBlock;
const index_t GemmK0Pad = GemmKBatch * GemmK0S;
const auto out_n_ho_wo_k_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N, Ho * Wo, K));
const auto out_n0_ho_wo_k_n1_grid_desc =
transform_tensor_descriptor(out_n_ho_wo_k_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(N0, N1Number)),
make_pass_through_transform(Ho * Wo),
make_pass_through_transform(K)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}),
make_tuple(Sequence<0, 3>{}, Sequence<1>{}, Sequence<2>{}));
const auto out_gemmk0total_gemmm_gemmk1_grid_desc =
transform_tensor_descriptor(out_n0_ho_wo_k_n1_grid_desc,
make_tuple(make_merge_transform(make_tuple(N0, Ho * Wo)),
make_pass_through_transform(K),
make_pass_through_transform(N1Number)),
make_tuple(Sequence<0, 1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
const auto out_gemmk0pad_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
out_gemmk0total_gemmm_gemmk1_grid_desc,
make_tuple(make_right_pad_transform(GemmK0Total, GemmK0Pad - GemmK0Total),
make_pass_through_transform(GemmM),
make_pass_through_transform(N1Number)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
const auto out_gemmkbatch_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
out_gemmk0pad_gemmm_gemmk1_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmKBatch, GemmK0)),
make_pass_through_transform(GemmM),
make_pass_through_transform(N1Number)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}),
make_tuple(Sequence<0, 1>{}, Sequence<2>{}, Sequence<3>{}));
// B: input tensor
const auto in_n_hip_wip_c_grid_desc = transform_tensor_descriptor(
in_n_hi_wi_c_grid_desc,
make_tuple(make_pass_through_transform(N),
make_pad_transform(Hi, InLeftPadH, InRightPadH),
make_pad_transform(Wi, InLeftPadW, InRightPadW),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
const auto in_n_y_ho_x_wo_c_grid_desc = transform_tensor_descriptor(
in_n_hip_wip_c_grid_desc,
make_tuple(
make_pass_through_transform(N),
make_embed_transform(make_tuple(Y, Ho), make_tuple(ConvDilationH, ConvStrideH)),
make_embed_transform(make_tuple(X, Wo), make_tuple(ConvDilationW, ConvStrideW)),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1, 2>{}, Sequence<3, 4>{}, Sequence<5>{}));
const auto in_n0_y_ho_x_wo_c_n1_grid_desc =
transform_tensor_descriptor(in_n_y_ho_x_wo_c_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(N0, N1Number)),
make_pass_through_transform(Y),
make_pass_through_transform(Ho),
make_pass_through_transform(X),
make_pass_through_transform(Wo),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{},
Sequence<1>{},
Sequence<2>{},
Sequence<3>{},
Sequence<4>{},
Sequence<5>{}),
make_tuple(Sequence<0, 6>{},
Sequence<1>{},
Sequence<2>{},
Sequence<3>{},
Sequence<4>{},
Sequence<5>{}));
const auto in_gemmk0total_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
in_n0_y_ho_x_wo_c_n1_grid_desc,
make_tuple(make_merge_transform(make_tuple(N0, Ho, Wo)),
make_merge_transform(make_tuple(Y, X, C)),
make_pass_through_transform(N1Number)),
make_tuple(Sequence<0, 2, 4>{}, Sequence<1, 3, 5>{}, Sequence<6>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
const auto in_gemmk0pad_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmk0total_gemmn_gemmk1_grid_desc,
make_tuple(make_right_pad_transform(GemmK0Total, GemmK0Pad - GemmK0Total),
make_pass_through_transform(GemmN),
make_pass_through_transform(N1Number)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
const auto in_gemmkbatch_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmk0pad_gemmn_gemmk1_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmKBatch, GemmK0)),
make_pass_through_transform(GemmN),
make_pass_through_transform(N1Number)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}),
make_tuple(Sequence<0, 1>{}, Sequence<2>{}, Sequence<3>{}));
// C: weight tensor
const auto wei_gemmm_gemmn_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(K, Y * X * C));
return make_tuple(out_gemmkbatch_gemmk0_gemmm_gemmk1_grid_desc,
in_gemmkbatch_gemmk0_gemmn_gemmk1_grid_desc,
wei_gemmm_gemmn_grid_desc);
}
using ABCGridDescs = decltype(MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(
1, 1, 1, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}, 1));
using AGridDesc_K0_M_K1 = remove_cvref_t<decltype(ABCGridDescs{}[I0])>;
using BGridDesc_K0_N_K1 = remove_cvref_t<decltype(ABCGridDescs{}[I1])>;
using CGridDesc_M_N = remove_cvref_t<decltype(ABCGridDescs{}[I2])>;
// GridwiseGemm
using GridwiseGemm = GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_bwd_weight<
BlockSize,
ADataType, // TODO: distinguish A/B datatype
AccDataType,
CDataType,
InMemoryDataOperationEnum::Set,
AGridDesc_K0_M_K1,
BGridDesc_K0_N_K1,
CGridDesc_M_N,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
MPerBlock,
NPerBlock,
K0PerBlock,
MPerXdl,
NPerXdl,
K1,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_K0_M_K1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_K1,
false, // AThreadTransferSrcResetCoordinateAfterRun,
ABlockLdsAddExtraM,
ABlockLdsM1PerBlock,
ABlockLdsM0PerBlock,
ABlockLdsM1Padding,
BBlockTransferThreadClusterLengths_K0_N_K1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_K1,
false, // BThreadTransferSrcResetCoordinateAfterRun,
BBlockLdsAddExtraN,
BBlockLdsN1PerBlock,
BBlockLdsN0PerBlock,
BBlockLdsN1Padding,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CBlockTransferScalarPerVector_NWaveNPerXdl,
CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
true,
true>;
using GridwiseGemmAtomicAdd = GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_bwd_weight<
BlockSize,
ADataType, // TODO: distinguish A/B datatype
AccDataType,
CDataType,
InMemoryDataOperationEnum::AtomicAdd,
AGridDesc_K0_M_K1,
BGridDesc_K0_N_K1,
CGridDesc_M_N,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
MPerBlock,
NPerBlock,
K0PerBlock,
MPerXdl,
NPerXdl,
K1,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_K0_M_K1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_K1,
false, // AThreadTransferSrcResetCoordinateAfterRun,
ABlockLdsAddExtraM,
ABlockLdsM1PerBlock,
ABlockLdsM0PerBlock,
ABlockLdsM1Padding,
BBlockTransferThreadClusterLengths_K0_N_K1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_K1,
false, // BThreadTransferSrcResetCoordinateAfterRun,
BBlockLdsAddExtraN,
BBlockLdsN1PerBlock,
BBlockLdsN0PerBlock,
BBlockLdsN1Padding,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CBlockTransferScalarPerVector_NWaveNPerXdl,
CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
true,
true>;
// Argument
using CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock =
decltype(GridwiseGemm::MakeCGridDesc_MBlock_MPerBlock_NBlock_NPerBlock(CGridDesc_M_N{}));
using Block2CTileMap =
decltype(GridwiseGemm::MakeCBlockClusterAdaptor(CGridDesc_M_N{}, 1, 1, 1));
struct Argument : public BaseArgument
{
Argument(const InDataType* p_in_grid,
WeiDataType* p_wei_grid,
const OutDataType* p_out_grid,
ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
ck::index_t M01,
ck::index_t N01,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op,
ck::index_t split_k)
: p_a_grid_{p_out_grid},
p_b_grid_{p_in_grid},
p_c_grid_{p_wei_grid},
a_grid_desc_kbatch_k0_m_k1_{},
b_grid_desc_kbatch_k0_n_k1_{},
c_grid_desc_m_n_{},
c_grid_desc_mblock_mperblock_nblock_nperblock_{},
block_2_ctile_map_{},
M01_{M01},
N01_{N01},
a_element_op_{out_element_op},
b_element_op_{in_element_op},
c_element_op_{wei_element_op},
Conv_N_{N},
Conv_K_{K},
Conv_C_{C},
output_spatial_lengths_{output_spatial_lengths},
filter_spatial_lengths_{filter_spatial_lengths},
conv_filter_strides_{conv_filter_strides},
input_left_pads_{input_left_pads},
input_right_pads_{input_right_pads},
k_batch_{split_k}
{
const auto descs =
DeviceOp::MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
k_batch_);
a_grid_desc_kbatch_k0_m_k1_ = descs[I0];
b_grid_desc_kbatch_k0_n_k1_ = descs[I1];
c_grid_desc_m_n_ = descs[I2];
block_2_ctile_map_ =
GridwiseGemm::MakeCBlockClusterAdaptor(c_grid_desc_m_n_, M01, N01, k_batch_);
if(GridwiseGemm::CheckValidity(a_grid_desc_kbatch_k0_m_k1_,
b_grid_desc_kbatch_k0_n_k1_,
c_grid_desc_m_n_,
block_2_ctile_map_))
{
c_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeCGridDesc_MBlock_MPerBlock_NBlock_NPerBlock(c_grid_desc_m_n_);
}
}
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
CDataType* p_c_grid_;
AGridDesc_K0_M_K1 a_grid_desc_kbatch_k0_m_k1_;
BGridDesc_K0_N_K1 b_grid_desc_kbatch_k0_n_k1_;
CGridDesc_M_N c_grid_desc_m_n_;
CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock c_grid_desc_mblock_mperblock_nblock_nperblock_;
Block2CTileMap block_2_ctile_map_;
index_t M01_;
index_t N01_;
InElementwiseOperation a_element_op_;
OutElementwiseOperation b_element_op_;
WeiElementwiseOperation c_element_op_;
// for checking IsSupportedArgument()
index_t Conv_N_;
index_t Conv_K_;
index_t Conv_C_;
std::vector<index_t> output_spatial_lengths_;
std::vector<index_t> filter_spatial_lengths_;
std::vector<index_t> conv_filter_strides_;
std::vector<index_t> input_left_pads_;
std::vector<index_t> input_right_pads_;
index_t k_batch_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
void ShowInfo(const Argument& arg)
{
std::cout << "arg.a_grid_desc_kbatch_k0_m_k1_{"
<< arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I0) << ", "
<< arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I1) << ", "
<< arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I2) << ", "
<< arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I3) << "}" << std::endl;
std::cout << "arg.b_grid_desc_kbatch_k0_n_k1_{"
<< arg.b_grid_desc_kbatch_k0_n_k1_.GetLength(I0) << ", "
<< arg.b_grid_desc_kbatch_k0_n_k1_.GetLength(I1) << ", "
<< arg.b_grid_desc_kbatch_k0_n_k1_.GetLength(I2) << ", "
<< arg.b_grid_desc_kbatch_k0_n_k1_.GetLength(I3) << "}" << std::endl;
std::cout << "arg.c_grid_desc_m_n_{ " << arg.c_grid_desc_m_n_.GetLength(I0) << ", "
<< arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
}
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
ShowInfo(arg);
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_kbatch_k0_m_k1_,
arg.b_grid_desc_kbatch_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_))
{
throw std::runtime_error(
"wrong! GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_bwd_weight has invalid setting");
}
const auto kbatch = arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I0);
const index_t grid_size =
arg.block_2_ctile_map_.CalculateGridSize(arg.c_grid_desc_m_n_);
const auto K0 = arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I1);
const bool has_main_k0_block_loop = GridwiseGemm::CalculateHasMainK0BlockLoop(K0);
float ave_time = 0;
const auto Run = [&](const auto& kernel) {
hipGetErrorString(hipMemset(
arg.p_c_grid_,
0,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_.GetElementSpaceSize() *
sizeof(CDataType)));
ave_time =
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_grid_desc_kbatch_k0_m_k1_,
arg.b_grid_desc_kbatch_k0_n_k1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.block_2_ctile_map_);
};
if(has_main_k0_block_loop)
{
if(kbatch == 1)
{
const auto kernel = kernel_gemm_xdlops_bwd_weight<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
remove_reference_t<DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
OutElementwiseOperation,
InElementwiseOperation,
WeiElementwiseOperation,
remove_reference_t<DeviceOp::Block2CTileMap>,
true>;
Run(kernel);
}
else
{
const auto kernel = kernel_gemm_xdlops_bwd_weight<
GridwiseGemmAtomicAdd,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
remove_reference_t<DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
OutElementwiseOperation,
InElementwiseOperation,
WeiElementwiseOperation,
remove_reference_t<DeviceOp::Block2CTileMap>,
true>;
Run(kernel);
}
}
else
{
if(kbatch == 1)
{
const auto kernel = kernel_gemm_xdlops_bwd_weight<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
remove_reference_t<DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
OutElementwiseOperation,
InElementwiseOperation,
WeiElementwiseOperation,
remove_reference_t<DeviceOp::Block2CTileMap>,
false>;
Run(kernel);
}
else
{
const auto kernel = kernel_gemm_xdlops_bwd_weight<
GridwiseGemmAtomicAdd,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
remove_reference_t<DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
OutElementwiseOperation,
InElementwiseOperation,
WeiElementwiseOperation,
remove_reference_t<DeviceOp::Block2CTileMap>,
false>;
Run(kernel);
}
}
return ave_time;
}
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
// vector load A/B matrix from global memory
if(!(ABlockTransferSrcVectorDim == 2 && BBlockTransferSrcVectorDim == 2 &&
arg.Conv_K_ % ABlockTransferSrcScalarPerVector == 0 &&
arg.Conv_C_ % BBlockTransferSrcScalarPerVector == 0))
{
return false;
}
// unmerge N to N0 and N1, where N1 equals to K1
if(!(arg.Conv_N_ % K1 == 0))
{
return false;
}
// vector store C matrix into global memory
if(!(arg.Conv_C_ % CBlockTransferScalarPerVector_NWaveNPerXdl == 0))
{
return false;
}
// Gridwise GEMM size
return GridwiseGemm::CheckValidity(arg.a_grid_desc_kbatch_k0_m_k1_,
arg.b_grid_desc_kbatch_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_);
}
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const InDataType* p_in_grid,
WeiDataType* p_wei_grid,
const OutDataType* p_out_grid,
ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op,
ck::index_t split_k)
{
return Argument{p_in_grid,
p_wei_grid,
p_out_grid,
N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
1,
1,
in_element_op,
wei_element_op,
out_element_op,
split_k};
}
static auto MakeInvoker() { return Invoker{}; }
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_in_grid,
void* p_wei_grid,
const void* p_out_grid,
ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op,
ck::index_t split_k) override
{
return std::make_unique<Argument>(static_cast<const InDataType*>(p_in_grid),
static_cast<WeiDataType*>(p_wei_grid),
static_cast<const OutDataType*>(p_out_grid),
N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
1,
1,
in_element_op,
wei_element_op,
out_element_op,
split_k);
}
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< K0PerBlock
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

833
include/ck/tensor_operation/gpu/device/impl/device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk.hpp Normal file
View File

@@ -0,0 +1,833 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_conv_bwd_data.hpp"
#include "ck/tensor_operation/gpu/device/convolution_backward_data_specialization.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v2r3.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
// out[N, Ho, Wo, K] = in[N, Hi, Wi, C] * wei[K, Y, X, C]
template <typename InDataType,
typename WeiDataType,
typename OutDataType,
typename AccDataType,
typename InElementwiseOperation,
typename WeiElementwiseOperation,
typename OutElementwiseOperation,
ConvolutionBackwardDataSpecialization ConvBackwardDataSpecialization,
ck::index_t BlockSize,
ck::index_t MPerBlock,
ck::index_t NPerBlock,
ck::index_t K0PerBlock,
ck::index_t K1,
ck::index_t MPerXdl,
ck::index_t NPerXdl,
ck::index_t MXdlPerWave,
ck::index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_K0_M_K1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
ck::index_t ABlockTransferSrcVectorDim,
ck::index_t ABlockTransferSrcScalarPerVector,
ck::index_t ABlockTransferDstScalarPerVector_K1,
bool ABlockLdsAddExtraM,
typename BBlockTransferThreadClusterLengths_K0_N_K1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
ck::index_t BBlockTransferSrcVectorDim,
ck::index_t BBlockTransferSrcScalarPerVector,
ck::index_t BBlockTransferDstScalarPerVector_K1,
bool BBlockLdsAddExtraN,
ck::index_t CThreadTransferSrcDstVectorDim,
ck::index_t CThreadTransferDstScalarPerVector>
struct DeviceConv2dBwdDataXdl_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
: public DeviceConvBwdData<2,
ck::tensor_layout::convolution::NHWC,
ck::tensor_layout::convolution::KYXC,
ck::tensor_layout::convolution::NHWK,
InDataType,
WeiDataType,
OutDataType,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation>
{
using DeviceOp = DeviceConv2dBwdDataXdl_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K;
using ADataType = OutDataType;
using BDataType = WeiDataType;
using CDataType = InDataType;
// TODO make A/B datatype different
using ABDataType = InDataType;
static constexpr index_t NDimSpatial = 2;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr auto I4 = Number<4>{};
static constexpr auto I5 = Number<5>{};
static_assert((K1 % ABlockTransferThreadClusterLengths_K0_M_K1{}[I2]) %
ABlockTransferSrcScalarPerVector ==
0);
static_assert((NPerBlock / BBlockTransferThreadClusterLengths_K0_N_K1{}[I1]) %
BBlockTransferSrcScalarPerVector ==
0);
static constexpr auto K1Number = Number<K1>{};
static constexpr auto GemmK1Number = K1Number;
static auto
MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
index_t i_ytilde,
index_t i_xtilde)
{
using namespace ck;
const index_t Hi = input_spatial_lengths[0];
const index_t Wi = input_spatial_lengths[1];
const index_t Ho = output_spatial_lengths[0];
const index_t Wo = output_spatial_lengths[1];
const index_t Y = filter_spatial_lengths[0];
const index_t X = filter_spatial_lengths[1];
const index_t InLeftPadH = input_left_pads[0];
const index_t InLeftPadW = input_left_pads[1];
const index_t InRightPadH = input_right_pads[0];
const index_t InRightPadW = input_right_pads[1];
const index_t ConvStrideH = conv_filter_strides[0];
const index_t ConvStrideW = conv_filter_strides[1];
const index_t ConvDilationH = conv_filter_dilations[0];
const index_t ConvDilationW = conv_filter_dilations[1];
const auto K0 = K / K1;
const auto out_n_ho_wo_k_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N, Ho, Wo, K));
const auto wei_k_y_x_c_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(K, Y, X, C));
const auto in_n_hi_wi_c_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N, Hi, Wi, C));
if constexpr(ConvBackwardDataSpecialization ==
ConvolutionBackwardDataSpecialization::Filter1x1Stride1Pad0)
{
// A: output tensor
const auto out_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, K)),
make_tuple(make_pass_through_transform(N * Ho * Wo),
make_unmerge_transform(make_tuple(K0, K1))),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<1>{}, Sequence<0, 2>{}));
// B: weight tensor
const auto wei_gemmk0_gemmn_gemmk1_grid_desc =
transform_tensor_descriptor(make_naive_tensor_descriptor_packed(make_tuple(K, C)),
make_tuple(make_unmerge_transform(make_tuple(K0, K1)),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// C: input tensor
const auto in_n_y_ho_x_wo_c_grid_desc = transform_tensor_descriptor(
in_n_hi_wi_c_grid_desc,
make_tuple(make_pass_through_transform(N),
make_embed_transform(make_tuple(I1, Ho), make_tuple(I1, ConvStrideH)),
make_embed_transform(make_tuple(I1, Wo), make_tuple(I1, ConvStrideW)),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1, 2>{}, Sequence<3, 4>{}, Sequence<5>{}));
const auto in_gemmm_gemmn_grid_desc = transform_tensor_descriptor(
in_n_y_ho_x_wo_c_grid_desc,
make_tuple(make_freeze_transform(I0),
make_freeze_transform(I0),
make_merge_transform(make_tuple(N, Ho, Wo)),
make_pass_through_transform(C)),
make_tuple(Sequence<1>{}, Sequence<3>{}, Sequence<0, 2, 4>{}, Sequence<5>{}),
make_tuple(Sequence<>{}, Sequence<>{}, Sequence<0>{}, Sequence<1>{}));
return make_tuple(out_gemmk0_gemmm_gemmk1_grid_desc,
wei_gemmk0_gemmn_gemmk1_grid_desc,
in_gemmm_gemmn_grid_desc);
}
else
{
const auto GcdStrideDilationH = math::gcd(ConvStrideH, ConvDilationH);
const auto GcdStrideDilationW = math::gcd(ConvStrideW, ConvDilationW);
const auto YTilde = ConvStrideH / GcdStrideDilationH;
const auto XTilde = ConvStrideW / GcdStrideDilationW;
const auto YDot = math::integer_divide_ceil(Y, YTilde);
const auto XDot = math::integer_divide_ceil(X, XTilde);
const auto HTilde =
Ho + math::integer_divide_ceil(ConvDilationH * (Y - I1), ConvStrideH);
const auto WTilde =
Wo + math::integer_divide_ceil(ConvDilationW * (X - I1), ConvStrideW);
// only work on HTilde and WTilde that contribute to non-padding area of input tensor
const auto IHTildeSliceBegin = math::integer_divide_floor(
math::max(I0, InLeftPadH - ConvDilationH * (YTilde - I1)), ConvStrideH);
const auto IWTildeSliceBegin = math::integer_divide_floor(
math::max(I0, InLeftPadW - ConvDilationW * (XTilde - I1)), ConvStrideW);
const auto IHTildeSliceEnd = math::min(
HTilde, math::integer_divide_ceil(InLeftPadH + Hi - I1, ConvStrideH) + I1);
const auto IWTildeSliceEnd = math::min(
WTilde, math::integer_divide_ceil(InLeftPadW + Wi - I1, ConvStrideW) + I1);
const auto HTildeSlice = IHTildeSliceEnd - IHTildeSliceBegin;
const auto WTildeSlice = IWTildeSliceEnd - IWTildeSliceBegin;
// GemmK is different for each GEMM
const auto YDotSlice = math::integer_divide_ceil(Y - i_ytilde, YTilde);
const auto XDotSlice = math::integer_divide_ceil(X - i_xtilde, XTilde);
// A: output tensor
const auto out_n_hop_wop_k_grid_desc = transform_tensor_descriptor(
out_n_ho_wo_k_grid_desc,
make_tuple(make_pass_through_transform(N),
make_pad_transform(Ho, I0, I0),
make_pad_transform(Wo, I0, I0),
make_pass_through_transform(K)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
const auto out_n_ydot_htilde_xdot_wtilde_k_grid_desc = transform_tensor_descriptor(
out_n_hop_wop_k_grid_desc,
make_tuple(
make_pass_through_transform(N),
make_embed_transform(make_tuple(YDot, HTilde),
make_tuple(-ConvDilationH / GcdStrideDilationH, I1)),
make_embed_transform(make_tuple(XDot, WTilde),
make_tuple(-ConvDilationW / GcdStrideDilationW, I1)),
make_pass_through_transform(K)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1, 2>{}, Sequence<3, 4>{}, Sequence<5>{}));
const auto out_n_ydotslice_htildeslice_xdotslice_wtildeslice_k0_k1_grid_desc =
transform_tensor_descriptor(
out_n_ydot_htilde_xdot_wtilde_k_grid_desc,
make_tuple(make_pass_through_transform(N),
make_slice_transform(YDot, I0, YDotSlice),
make_slice_transform(HTilde, IHTildeSliceBegin, HTildeSlice),
make_slice_transform(XDot, I0, XDotSlice),
make_slice_transform(WTilde, IWTildeSliceBegin, WTildeSlice),
make_unmerge_transform(make_tuple(K0, K1))),
make_tuple(Sequence<0>{},
Sequence<1>{},
Sequence<2>{},
Sequence<3>{},
Sequence<4>{},
Sequence<5>{}),
make_tuple(Sequence<0>{},
Sequence<1>{},
Sequence<2>{},
Sequence<3>{},
Sequence<4>{},
Sequence<5, 6>{}));
const auto out_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
out_n_ydotslice_htildeslice_xdotslice_wtildeslice_k0_k1_grid_desc,
make_tuple(make_merge_transform(make_tuple(YDotSlice, XDotSlice, K0)),
make_merge_transform(make_tuple(N, HTildeSlice, WTildeSlice)),
make_pass_through_transform(K1)),
make_tuple(Sequence<1, 3, 5>{}, Sequence<0, 2, 4>{}, Sequence<6>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
// B weight tensor
const auto wei_k_ydot_ytilde_xdot_xtilde_c_grid_desc = transform_tensor_descriptor(
wei_k_y_x_c_grid_desc,
make_tuple(make_pass_through_transform(K),
make_embed_transform(make_tuple(YDot, YTilde),
make_tuple(ConvStrideH / GcdStrideDilationH, I1)),
make_embed_transform(make_tuple(XDot, XTilde),
make_tuple(ConvStrideW / GcdStrideDilationW, I1)),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1, 2>{}, Sequence<3, 4>{}, Sequence<5>{}));
const auto wei_k0_k1_ydotslice_xdotslice_c_grid_desc =
transform_tensor_descriptor(wei_k_ydot_ytilde_xdot_xtilde_c_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(K0, K1)),
make_slice_transform(YDot, I0, YDotSlice),
make_slice_transform(XDot, I0, XDotSlice),
make_freeze_transform(i_ytilde),
make_freeze_transform(i_xtilde),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{},
Sequence<1>{},
Sequence<3>{},
Sequence<2>{},
Sequence<4>{},
Sequence<5>{}),
make_tuple(Sequence<0, 1>{},
Sequence<2>{},
Sequence<3>{},
Sequence<>{},
Sequence<>{},
Sequence<4>{}));
const auto wei_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
wei_k0_k1_ydotslice_xdotslice_c_grid_desc,
make_tuple(make_merge_transform(make_tuple(YDotSlice, XDotSlice, K0)),
make_pass_through_transform(C),
make_pass_through_transform(K1)),
make_tuple(Sequence<2, 3, 0>{}, Sequence<4>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
// C: input tensor
const auto in_n_hip_wip_c_grid_desc = transform_tensor_descriptor(
in_n_hi_wi_c_grid_desc,
make_tuple(make_pass_through_transform(N),
make_pad_transform(Hi, InLeftPadH, InRightPadH),
make_pad_transform(Wi, InLeftPadW, InRightPadW),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
const auto in_n_ytilde_htilde_xtilde_wtilde_c_grid_desc = transform_tensor_descriptor(
in_n_hip_wip_c_grid_desc,
make_tuple(make_pass_through_transform(N),
make_embed_transform(make_tuple(YTilde, HTilde),
make_tuple(ConvDilationH, ConvStrideH)),
make_embed_transform(make_tuple(XTilde, WTilde),
make_tuple(ConvDilationW, ConvStrideW)),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1, 2>{}, Sequence<3, 4>{}, Sequence<5>{}));
const auto in_n_htildeslice_wtildeslice_c_grid_desc = transform_tensor_descriptor(
in_n_ytilde_htilde_xtilde_wtilde_c_grid_desc,
make_tuple(make_pass_through_transform(N),
make_freeze_transform(i_ytilde),
make_slice_transform(HTilde, IHTildeSliceBegin, HTildeSlice),
make_freeze_transform(i_xtilde),
make_slice_transform(WTilde, IWTildeSliceBegin, WTildeSlice),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{},
Sequence<1>{},
Sequence<2>{},
Sequence<3>{},
Sequence<4>{},
Sequence<5>{}),
make_tuple(Sequence<0>{},
Sequence<>{},
Sequence<1>{},
Sequence<>{},
Sequence<2>{},
Sequence<3>{}));
const auto in_gemmm_gemmn_grid_desc = transform_tensor_descriptor(
in_n_htildeslice_wtildeslice_c_grid_desc,
make_tuple(make_merge_transform(make_tuple(N, HTildeSlice, WTildeSlice)),
make_pass_through_transform(C)),
make_tuple(Sequence<0, 1, 2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return make_tuple(out_gemmk0_gemmm_gemmk1_grid_desc,
wei_gemmk0_gemmn_gemmk1_grid_desc,
in_gemmm_gemmn_grid_desc);
}
} // function end
using ABCGridDescs = decltype(MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(
1, 1, 1, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}, 0, 0));
using AGridDesc_K0_M_K1 = remove_cvref_t<decltype(ABCGridDescs{}[I0])>;
using BGridDesc_K0_N_K1 = remove_cvref_t<decltype(ABCGridDescs{}[I1])>;
using CGridDesc_M_N = remove_cvref_t<decltype(ABCGridDescs{}[I2])>;
// GridwiseGemm
using GridwiseGemm = GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r3<
BlockSize,
ABDataType, // TODO: distinguish A/B datatype
AccDataType,
CDataType,
InMemoryDataOperationEnum::Set,
AGridDesc_K0_M_K1,
BGridDesc_K0_N_K1,
CGridDesc_M_N,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation,
MPerBlock,
NPerBlock,
K0PerBlock,
MPerXdl,
NPerXdl,
K1,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_K0_M_K1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_K1,
false, // AThreadTransferSrcResetCoordinateAfterRun,
ABlockLdsAddExtraM,
BBlockTransferThreadClusterLengths_K0_N_K1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_K1,
false, // BThreadTransferSrcResetCoordinateAfterRun,
BBlockLdsAddExtraN,
Sequence<2, 3, 0, 1, 7, 5, 4, 6>, // CThreadTransferSrcDstAccessOrder,
7, // CThreadTransferSrcDstVectorDim,
CThreadTransferDstScalarPerVector>;
// Argument
struct Argument : public BaseArgument
{
Argument(InDataType* p_in_grid,
const WeiDataType* p_wei_grid,
const OutDataType* p_out_grid,
ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
ck::index_t M01,
ck::index_t N01,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op)
: p_a_grid_{p_out_grid},
p_b_grid_{p_wei_grid},
p_c_grid_{p_in_grid},
M01_{M01},
N01_{N01},
a_element_op_{out_element_op},
b_element_op_{wei_element_op},
c_element_op_{in_element_op},
Conv_N_{N},
Conv_K_{K},
Conv_C_{C},
input_spatial_lengths_{input_spatial_lengths},
filter_spatial_lengths_{filter_spatial_lengths},
output_spatial_lengths_{output_spatial_lengths},
conv_filter_strides_{conv_filter_strides},
conv_filter_dilations_{conv_filter_dilations},
input_left_pads_{input_left_pads},
input_right_pads_{input_right_pads}
{
const index_t ConvStrideH = conv_filter_strides[0];
const index_t ConvStrideW = conv_filter_strides[1];
const index_t ConvDilationH = conv_filter_dilations[0];
const index_t ConvDilationW = conv_filter_dilations[1];
const auto GcdStrideDilationH = math::gcd(ConvStrideH, ConvDilationH);
const auto GcdStrideDilationW = math::gcd(ConvStrideW, ConvDilationW);
const auto YTilde = ConvStrideH / GcdStrideDilationH;
const auto XTilde = ConvStrideW / GcdStrideDilationW;
for(index_t i_ytilde = 0; i_ytilde < YTilde; ++i_ytilde)
{
for(index_t i_xtilde = 0; i_xtilde < XTilde; ++i_xtilde)
{
// check slice is valid
const index_t Y = filter_spatial_lengths_[0];
const index_t X = filter_spatial_lengths_[1];
const auto YDotSlice = math::integer_divide_ceil(Y - i_ytilde, YTilde);
const auto XDotSlice = math::integer_divide_ceil(X - i_xtilde, XTilde);
if(YDotSlice * XDotSlice <= 0)
{
continue;
}
const auto descs = DeviceOp::MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(
N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
i_ytilde,
i_xtilde);
a_grid_desc_k0_m_k1_container_.push_back(descs[I0]);
b_grid_desc_k0_n_k1_container_.push_back(descs[I1]);
c_grid_desc_m_n_container_.push_back(descs[I2]);
auto block_2_ctile_map =
GridwiseGemm::MakeDefaultBlock2CTileMap(descs[I2], M01, N01);
if(GridwiseGemm::CheckValidity(
descs[I0], descs[I1], descs[I2], block_2_ctile_map))
{
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_container_.push_back(
GridwiseGemm::MakeCGridDescriptor_M0_N0_M1_N1_M2_M3_M4_N2(descs[I2]));
block_2_ctile_map_container_.push_back(block_2_ctile_map);
}
}
}
}
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
CDataType* p_c_grid_;
std::vector<AGridDesc_K0_M_K1> a_grid_desc_k0_m_k1_container_;
std::vector<BGridDesc_K0_N_K1> b_grid_desc_k0_n_k1_container_;
std::vector<CGridDesc_M_N> c_grid_desc_m_n_container_;
std::vector<typename GridwiseGemm::CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2>
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_container_;
std::vector<typename GridwiseGemm::DefaultBlock2CTileMap> block_2_ctile_map_container_;
index_t M01_;
index_t N01_;
OutElementwiseOperation a_element_op_;
WeiElementwiseOperation b_element_op_;
InElementwiseOperation c_element_op_;
// for checking IsSupportedArgument()
index_t Conv_N_;
index_t Conv_K_;
index_t Conv_C_;
std::vector<ck::index_t> input_spatial_lengths_;
std::vector<ck::index_t> filter_spatial_lengths_;
std::vector<ck::index_t> output_spatial_lengths_;
std::vector<ck::index_t> conv_filter_strides_;
std::vector<ck::index_t> conv_filter_dilations_;
std::vector<ck::index_t> input_left_pads_;
std::vector<ck::index_t> input_right_pads_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
float ave_time = 0;
for(size_t i = 0; i < arg.a_grid_desc_k0_m_k1_container_.size(); i++)
{
{
std::cout << "arg.a_grid_desc_k0_m_k1_container_{"
<< arg.a_grid_desc_k0_m_k1_container_[i].GetLength(I0) << ", "
<< arg.a_grid_desc_k0_m_k1_container_[i].GetLength(I1) << ", "
<< arg.a_grid_desc_k0_m_k1_container_[i].GetLength(I2) << "}"
<< std::endl;
std::cout << "arg.b_grid_desc_k0_n_k1_container_{"
<< arg.b_grid_desc_k0_n_k1_container_[i].GetLength(I0) << ", "
<< arg.b_grid_desc_k0_n_k1_container_[i].GetLength(I1) << ", "
<< arg.b_grid_desc_k0_n_k1_container_[i].GetLength(I2) << "}"
<< std::endl;
std::cout << "arg.c_grid_desc_m_n_container_{ "
<< arg.c_grid_desc_m_n_container_[i].GetLength(I0) << ", "
<< arg.c_grid_desc_m_n_container_[i].GetLength(I1) << "}"
<< std::endl;
std::cout << "arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_container_( "
<< arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_container_[i].GetLength(I0)
<< ", "
<< arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_container_[i].GetLength(I1)
<< ", "
<< arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_container_[i].GetLength(I2)
<< ", "
<< arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_container_[i].GetLength(I3)
<< ", "
<< arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_container_[i].GetLength(I4)
<< ", "
<< arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_container_[i].GetLength(I5)
<< " ) " << std::endl;
}
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_container_[i],
arg.b_grid_desc_k0_n_k1_container_[i],
arg.c_grid_desc_m_n_container_[i],
arg.block_2_ctile_map_container_[i]))
{
throw std::runtime_error(
"wrong! GridwiseGemm_km_kn_m0m1n0n1_xdlops_v3r1 has invalid setting");
}
const index_t grid_size = arg.block_2_ctile_map_container_[i].CalculateGridSize(
arg.c_grid_desc_m_n_container_[i]);
const auto K = arg.a_grid_desc_k0_m_k1_container_[i].GetLength(I0) *
arg.a_grid_desc_k0_m_k1_container_[i].GetLength(I2);
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
const auto kernel = kernel_gemm_xdlops_v2r3<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
remove_reference_t<
typename GridwiseGemm::CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2>,
OutElementwiseOperation,
WeiElementwiseOperation,
InElementwiseOperation,
remove_reference_t<typename GridwiseGemm::DefaultBlock2CTileMap>,
true>;
ave_time += launch_and_time_kernel(
stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_grid_desc_k0_m_k1_container_[i],
arg.b_grid_desc_k0_n_k1_container_[i],
arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_container_[i],
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.block_2_ctile_map_container_[i]);
}
else
{
const auto kernel = kernel_gemm_xdlops_v2r3<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
remove_reference_t<
typename GridwiseGemm::CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2>,
OutElementwiseOperation,
WeiElementwiseOperation,
InElementwiseOperation,
remove_reference_t<typename GridwiseGemm::DefaultBlock2CTileMap>,
false>;
ave_time += launch_and_time_kernel(
stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_grid_desc_k0_m_k1_container_[i],
arg.b_grid_desc_k0_n_k1_container_[i],
arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_container_[i],
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.block_2_ctile_map_container_[i]);
}
}
return ave_time;
}
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
if constexpr(ConvBackwardDataSpecialization ==
ConvolutionBackwardDataSpecialization::Filter1x1Stride1Pad0)
{
// check if it's 1x1, stride=1 pad = 0 conv
if(!(arg.filter_spatial_lengths_[0] == 1 && arg.filter_spatial_lengths_[1] == 1 &&
arg.conv_filter_strides_[0] == 1 && arg.conv_filter_strides_[1] == 1 &&
arg.input_left_pads_[0] == 0 && arg.input_left_pads_[1] == 0 &&
arg.input_right_pads_[0] == 0 && arg.input_right_pads_[1] == 0))
{
return false;
}
}
// vector load A/B matrix from global memory
if(!(ABlockTransferSrcVectorDim == 2 && BBlockTransferSrcVectorDim == 1 &&
arg.Conv_K_ % ABlockTransferSrcScalarPerVector == 0 &&
arg.Conv_C_ % BBlockTransferSrcScalarPerVector == 0))
{
return false;
}
// vector store C matrix into global memory
if(!(arg.Conv_C_ % CThreadTransferDstScalarPerVector == 0))
{
return false;
}
// Gridwise GEMM size
for(std::size_t i = 0; i < arg.a_grid_desc_k0_m_k1_container_.size(); i++)
{
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_container_[i],
arg.b_grid_desc_k0_n_k1_container_[i],
arg.c_grid_desc_m_n_container_[i],
arg.block_2_ctile_map_container_[i]))
{
return false;
}
}
return true;
}
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(InDataType* p_in_grid,
const WeiDataType* p_wei_grid,
const OutDataType* p_out_grid,
ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op)
{
return Argument{p_in_grid,
p_wei_grid,
p_out_grid,
N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
1,
1,
in_element_op,
wei_element_op,
out_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
std::unique_ptr<BaseArgument>
MakeArgumentPointer(void* p_in_grid,
const void* p_wei_grid,
const void* p_out_grid,
ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op) override
{
return std::make_unique<Argument>(static_cast<InDataType*>(p_in_grid),
static_cast<const WeiDataType*>(p_wei_grid),
static_cast<const OutDataType*>(p_out_grid),
N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
1,
1,
in_element_op,
wei_element_op,
out_element_op);
}
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceConv2dBwdDataXdl_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< K0PerBlock
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

968
include/ck/tensor_operation/gpu/device/impl/device_conv2d_fwd_xdl_c_shuffle_bias_activation_add_nhwc_kyxc_nhwk.hpp Normal file
View File

@@ -0,0 +1,968 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_conv_fwd_bias_activation_add.hpp"
#include "ck/tensor_operation/gpu/device/convolution_forward_specialization.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v3r3.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
// out[N, Ho, Wo, K] =
// activate(in[N, Hi, Wi, C] * wei[K, Y, X, C] + bias[K]) + residual[N, Ho, Wo, K]
template <
typename InDataType,
typename WeiDataType,
typename OutDataType,
typename AccDataType,
typename InElementwiseOperation,
typename WeiElementwiseOperation,
typename OutElementwiseOperation,
ConvolutionForwardSpecialization ConvForwardSpecialization,
ck::index_t BlockSize,
ck::index_t MPerBlock,
ck::index_t NPerBlock,
ck::index_t K0PerBlock,
ck::index_t K1,
ck::index_t MPerXDL,
ck::index_t NPerXDL,
ck::index_t MXdlPerWave,
ck::index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_K0_M_K1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
ck::index_t ABlockTransferSrcVectorDim,
ck::index_t ABlockTransferSrcScalarPerVector,
ck::index_t ABlockTransferDstScalarPerVector_K1,
bool ABlockLdsAddExtraM,
typename BBlockTransferThreadClusterLengths_K0_N_K1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
ck::index_t BBlockTransferSrcVectorDim,
ck::index_t BBlockTransferSrcScalarPerVector,
ck::index_t BBlockTransferDstScalarPerVector_K1,
bool BBlockLdsAddExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl,
index_t CBlockTransferScalarPerVector_NWaveNPerXdl>
struct
DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
: public DeviceConvFwdBiasActivationAdd<InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation>
{
using DeviceOp =
DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K;
using ADataType = InDataType;
using BDataType = WeiDataType;
using CDataType = OutDataType;
// TODO make A/B datatype different
using ABDataType = InDataType;
// TODO make it support any # of spatial dimensions
static constexpr index_t NDimSpatial = 2;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr auto I4 = Number<4>{};
static constexpr auto K1Number = Number<K1>{};
static constexpr auto GemmK1Number = K1Number;
static auto
MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads)
{
using namespace ck;
const index_t Hi = input_spatial_lengths[0];
const index_t Wi = input_spatial_lengths[1];
const index_t Ho = output_spatial_lengths[0];
const index_t Wo = output_spatial_lengths[1];
const index_t Y = filter_spatial_lengths[0];
const index_t X = filter_spatial_lengths[1];
const index_t ConvStrideH = conv_filter_strides[0];
const index_t ConvStrideW = conv_filter_strides[1];
const index_t ConvDilationH = conv_filter_dilations[0];
const index_t ConvDilationW = conv_filter_dilations[1];
const index_t InLeftPadH = input_left_pads[0];
const index_t InLeftPadW = input_left_pads[1];
const index_t InRightPadH = input_right_pads[0];
const index_t InRightPadW = input_right_pads[1];
const index_t GemmMRaw = N * Ho * Wo;
const index_t GemmN = K;
const auto GemmM = math::integer_least_multiple(GemmMRaw, MPerBlock);
const auto GemmMPad = GemmM - GemmMRaw;
if constexpr(ConvForwardSpecialization ==
ConvolutionForwardSpecialization::Filter1x1Stride1Pad0)
{ // 1x1, stride=1, pad=0
const index_t GemmK = Y * X * C;
assert(GemmK % GemmK1Number == 0);
const index_t GemmK0 = GemmK / GemmK1Number;
// A: input tensor
const auto in_gemmmraw_gemmk_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, C));
const auto in_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmmraw_gemmk_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_right_pad_transform(GemmMRaw, GemmMPad)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// B: weight tensor
const auto wei_gemmn_gemmk_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(K, C));
const auto wei_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
wei_gemmn_gemmk_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// C: output tensor
const auto out_gemmmraw_gemmn_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, K));
const auto out_gemmm_gemmn_grid_desc =
transform_tensor_descriptor(out_gemmmraw_gemmn_grid_desc,
make_tuple(make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
// C0: bias tensor: assume a contiguous vector
const auto bias_grid_desc_gemmm_gemmn =
make_naive_tensor_descriptor(make_tuple(GemmM, GemmN), make_tuple(I0, I1));
// C1: residual tensor: assume same layout as output tensor
const auto resi_grid_desc_gemmm_gemmn = out_gemmm_gemmn_grid_desc;
return make_tuple(in_gemmk0_gemmm_gemmk1_grid_desc,
wei_gemmk0_gemmn_gemmk1_grid_desc,
out_gemmm_gemmn_grid_desc,
bias_grid_desc_gemmm_gemmn,
resi_grid_desc_gemmm_gemmn);
}
else if constexpr(ConvForwardSpecialization ==
ConvolutionForwardSpecialization::Filter1x1Pad0)
{ // 1x1, pad=0
const index_t GemmK = Y * X * C;
assert(GemmK % GemmK1Number == 0);
const index_t GemmK0 = GemmK / GemmK1Number;
// A: input tensor
const auto in_n_hi_wi_c_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N, Hi, Wi, C));
const auto in_n_ho_wo_c_grid_desc = transform_tensor_descriptor(
in_n_hi_wi_c_grid_desc,
make_tuple(make_pass_through_transform(N),
make_embed_transform(make_tuple(Ho), make_tuple(ConvStrideH)),
make_embed_transform(make_tuple(Wo), make_tuple(ConvStrideW)),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
const auto in_gemmk0_gemmmraw_gemmk1_grid_desc = transform_tensor_descriptor(
in_n_ho_wo_c_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_merge_transform(make_tuple(N, Ho, Wo))),
make_tuple(Sequence<3>{}, Sequence<0, 1, 2>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
const auto in_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmk0_gemmmraw_gemmk1_grid_desc,
make_tuple(make_pass_through_transform(GemmK0),
make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmK1Number)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
// B: weight tensor
const auto wei_gemmn_gemmk_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(K, C));
const auto wei_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
wei_gemmn_gemmk_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// C: output tensor
const auto out_gemmmraw_gemmn_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, K));
const auto out_gemmm_gemmn_grid_desc =
transform_tensor_descriptor(out_gemmmraw_gemmn_grid_desc,
make_tuple(make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
// C0: bias tensor: assume a contiguous vector
const auto bias_grid_desc_gemmm_gemmn =
make_naive_tensor_descriptor(make_tuple(GemmM, GemmN), make_tuple(I0, I1));
// C1: residual tensor: assume same layout as output tensor
const auto resi_grid_desc_gemmm_gemmn = out_gemmm_gemmn_grid_desc;
return make_tuple(in_gemmk0_gemmm_gemmk1_grid_desc,
wei_gemmk0_gemmn_gemmk1_grid_desc,
out_gemmm_gemmn_grid_desc,
bias_grid_desc_gemmm_gemmn,
resi_grid_desc_gemmm_gemmn);
}
else if constexpr(ConvForwardSpecialization == ConvolutionForwardSpecialization::OddC)
{ // C = odd value
const index_t GemmKRaw = Y * X * C;
const index_t GemmK = math::integer_least_multiple(GemmKRaw, K0PerBlock * GemmK1Number);
const index_t GemmKPad = GemmK - GemmKRaw;
const index_t GemmK0 = GemmK / GemmK1Number;
// A: input tensor
const auto in_n_hi_wi_c_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N, Hi, Wi, C));
const auto in_n_hip_wip_c_grid_desc = transform_tensor_descriptor(
in_n_hi_wi_c_grid_desc,
make_tuple(make_pass_through_transform(N),
make_pad_transform(Hi, InLeftPadH, InRightPadH),
make_pad_transform(Wi, InLeftPadW, InRightPadW),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
const auto in_n_y_ho_x_wo_c_grid_desc = transform_tensor_descriptor(
in_n_hip_wip_c_grid_desc,
make_tuple(
make_pass_through_transform(N),
make_embed_transform(make_tuple(Y, Ho), make_tuple(ConvDilationH, ConvStrideH)),
make_embed_transform(make_tuple(X, Wo), make_tuple(ConvDilationW, ConvStrideW)),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1, 2>{}, Sequence<3, 4>{}, Sequence<5>{}));
const auto in_gemmkraw_gemmmraw_grid_desc =
transform_tensor_descriptor(in_n_y_ho_x_wo_c_grid_desc,
make_tuple(make_merge_transform(make_tuple(Y, X, C)),
make_merge_transform(make_tuple(N, Ho, Wo))),
make_tuple(Sequence<1, 3, 5>{}, Sequence<0, 2, 4>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto in_gemmk_gemmm_grid_desc = transform_tensor_descriptor(
in_gemmkraw_gemmmraw_grid_desc,
make_tuple(make_right_pad_transform(GemmKRaw, GemmKPad),
make_right_pad_transform(GemmMRaw, GemmMPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto in_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmk_gemmm_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmM)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// B: weight tensor
const auto wei_k_yxc_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(K, Y * X * C));
const auto wei_gemmk_gemmn_grid_desc = transform_tensor_descriptor(
wei_k_yxc_grid_desc,
make_tuple(make_pass_through_transform(K),
make_right_pad_transform(GemmKRaw, GemmKPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<1>{}, Sequence<0>{}));
const auto wei_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
wei_gemmk_gemmn_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// C: output tensor
const auto out_nhowo_k_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, K));
const auto out_gemmmraw_gemmn_grid_desc =
transform_tensor_descriptor(out_nhowo_k_grid_desc,
make_tuple(make_pass_through_transform(N * Ho * Wo),
make_pass_through_transform(K)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto out_gemmm_gemmn_grid_desc =
transform_tensor_descriptor(out_gemmmraw_gemmn_grid_desc,
make_tuple(make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
// C0: bias tensor: assume a contiguous vector
const auto bias_grid_desc_gemmm_gemmn =
make_naive_tensor_descriptor(make_tuple(GemmM, GemmN), make_tuple(I0, I1));
// C1: residual tensor: assume same layout as output tensor
const auto resi_grid_desc_gemmm_gemmn = out_gemmm_gemmn_grid_desc;
return make_tuple(in_gemmk0_gemmm_gemmk1_grid_desc,
wei_gemmk0_gemmn_gemmk1_grid_desc,
out_gemmm_gemmn_grid_desc,
bias_grid_desc_gemmm_gemmn,
resi_grid_desc_gemmm_gemmn);
}
else
{
const index_t GemmK = Y * X * C;
assert(GemmK % GemmK1Number == 0);
const index_t GemmK0 = GemmK / GemmK1Number;
// A: input tensor
const auto in_n_hi_wi_c_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N, Hi, Wi, C));
const auto in_n_hip_wip_c_grid_desc = transform_tensor_descriptor(
in_n_hi_wi_c_grid_desc,
make_tuple(make_pass_through_transform(N),
make_pad_transform(Hi, InLeftPadH, InRightPadH),
make_pad_transform(Wi, InLeftPadW, InRightPadW),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
const auto in_n_y_ho_x_wo_c_grid_desc = transform_tensor_descriptor(
in_n_hip_wip_c_grid_desc,
make_tuple(
make_pass_through_transform(N),
make_embed_transform(make_tuple(Y, Ho), make_tuple(ConvDilationH, ConvStrideH)),
make_embed_transform(make_tuple(X, Wo), make_tuple(ConvDilationW, ConvStrideW)),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1, 2>{}, Sequence<3, 4>{}, Sequence<5>{}));
const auto in_gemmk_gemmmraw_grid_desc =
transform_tensor_descriptor(in_n_y_ho_x_wo_c_grid_desc,
make_tuple(make_merge_transform(make_tuple(Y, X, C)),
make_merge_transform(make_tuple(N, Ho, Wo))),
make_tuple(Sequence<1, 3, 5>{}, Sequence<0, 2, 4>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto in_gemmk0_gemmmraw_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmk_gemmmraw_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmMRaw)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
const auto in_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmk0_gemmmraw_gemmk1_grid_desc,
make_tuple(make_pass_through_transform(GemmK0),
make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmK1Number)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
// B: weight tensor
const auto wei_k_yxc_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(K, Y * X * C));
const auto wei_gemmk_gemmn_grid_desc = transform_tensor_descriptor(
wei_k_yxc_grid_desc,
make_tuple(make_pass_through_transform(K), make_pass_through_transform(Y * X * C)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<1>{}, Sequence<0>{}));
const auto wei_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
wei_gemmk_gemmn_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// C: output tensor
const auto out_nhowo_k_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, K));
const auto out_gemmmraw_gemmn_grid_desc =
transform_tensor_descriptor(out_nhowo_k_grid_desc,
make_tuple(make_pass_through_transform(N * Ho * Wo),
make_pass_through_transform(K)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto out_gemmm_gemmn_grid_desc =
transform_tensor_descriptor(out_gemmmraw_gemmn_grid_desc,
make_tuple(make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
// C0: bias tensor: assume a contiguous vector
const auto bias_grid_desc_gemmm_gemmn =
make_naive_tensor_descriptor(make_tuple(GemmM, GemmN), make_tuple(I0, I1));
// C1: residual tensor: assume same layout as output tensor
const auto resi_grid_desc_gemmm_gemmn = out_gemmm_gemmn_grid_desc;
return make_tuple(in_gemmk0_gemmm_gemmk1_grid_desc,
wei_gemmk0_gemmn_gemmk1_grid_desc,
out_gemmm_gemmn_grid_desc,
bias_grid_desc_gemmm_gemmn,
resi_grid_desc_gemmm_gemmn);
}
}
using GridDescs = decltype(MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(
1, 1, 1, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}));
using AGridDesc_K0_M_K1 = remove_cvref_t<decltype(GridDescs{}[I0])>;
using BGridDesc_K0_N_K1 = remove_cvref_t<decltype(GridDescs{}[I1])>;
using CGridDesc_M_N = remove_cvref_t<decltype(GridDescs{}[I2])>;
using C0GridDesc_M_N = remove_cvref_t<decltype(GridDescs{}[I3])>;
using C1GridDesc_M_N = remove_cvref_t<decltype(GridDescs{}[I4])>;
using Block2CTileMap = BlockToCTileMap_M00_N0_M01<MPerBlock, NPerBlock, CGridDesc_M_N>;
// GridwiseGemm
using GridwiseGemm = GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v3r3<
BlockSize,
ABDataType, // TODO: distinguish A/B datatype
AccDataType,
CDataType,
InMemoryDataOperationEnum::Set,
AGridDesc_K0_M_K1,
BGridDesc_K0_N_K1,
CGridDesc_M_N,
C0GridDesc_M_N,
C1GridDesc_M_N,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation,
MPerBlock,
NPerBlock,
K0PerBlock,
MPerXDL,
NPerXDL,
K1,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_K0_M_K1,
Sequence<1, 0, 2>, // ABlockTransferThreadClusterArrangeOrder,
Sequence<1, 0, 2>, // ABlockTransferSrcAccessOrder,
2, // ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_K1,
false, // AThreadTransferSrcResetCoordinateAfterRun,
ABlockLdsAddExtraM,
BBlockTransferThreadClusterLengths_K0_N_K1,
Sequence<1, 0, 2>, // BBlockTransferThreadClusterArrangeOrder,
Sequence<1, 0, 2>, // BBlockTransferSrcAccessOrder,
2, // BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_K1,
false, // BThreadTransferSrcResetCoordinateAfterRun,
BBlockLdsAddExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl,
CBlockTransferScalarPerVector_NWaveNPerXdl>;
// Argument
struct Argument : public BaseArgument
{
Argument(const InDataType* p_in_grid,
const WeiDataType* p_wei_grid,
OutDataType* p_out_grid,
const OutDataType* p_bias_grid,
const OutDataType* p_resi_grid,
ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op)
: p_a_grid_{p_in_grid},
p_b_grid_{p_wei_grid},
p_c_grid_{p_out_grid},
p_c0_grid_{p_bias_grid},
p_c1_grid_{p_resi_grid},
a_grid_desc_k0_m_k1_{},
b_grid_desc_k0_n_k1_{},
c_grid_desc_m_n_{},
c0_grid_desc_m_n_{},
c1_grid_desc_m_n_{},
c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_{},
c0_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_{},
c1_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_{},
block_2_ctile_map_{},
in_element_op_{in_element_op},
wei_element_op_{wei_element_op},
out_element_op_{out_element_op},
Conv_N_{N},
Conv_K_{K},
Conv_C_{C},
input_spatial_lengths_{input_spatial_lengths},
filter_spatial_lengths_{filter_spatial_lengths},
output_spatial_lengths_{output_spatial_lengths},
conv_filter_strides_{conv_filter_strides},
conv_filter_dilations_{conv_filter_dilations},
input_left_pads_{input_left_pads},
input_right_pads_{input_right_pads}
{
const auto descs =
DeviceOp::MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads);
a_grid_desc_k0_m_k1_ = descs[I0];
b_grid_desc_k0_n_k1_ = descs[I1];
c_grid_desc_m_n_ = descs[I2];
c0_grid_desc_m_n_ = descs[I3];
c1_grid_desc_m_n_ = descs[I4];
block_2_ctile_map_ = Block2CTileMap{c_grid_desc_m_n_};
if(GridwiseGemm::CheckValidity(a_grid_desc_k0_m_k1_,
b_grid_desc_k0_n_k1_,
c_grid_desc_m_n_,
block_2_ctile_map_))
{
c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_ =
GridwiseGemm::
MakeCGridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl(
c_grid_desc_m_n_);
c0_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_ =
GridwiseGemm::
MakeCGridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl(
c0_grid_desc_m_n_);
c1_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_ =
GridwiseGemm::
MakeCGridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl(
c1_grid_desc_m_n_);
}
}
// private:
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
CDataType* p_c_grid_;
const CDataType* p_c0_grid_;
const CDataType* p_c1_grid_;
AGridDesc_K0_M_K1 a_grid_desc_k0_m_k1_;
BGridDesc_K0_N_K1 b_grid_desc_k0_n_k1_;
CGridDesc_M_N c_grid_desc_m_n_;
C0GridDesc_M_N c0_grid_desc_m_n_;
C1GridDesc_M_N c1_grid_desc_m_n_;
typename GridwiseGemm::
CGridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl
c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_;
typename GridwiseGemm::
C0GridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl
c0_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_;
typename GridwiseGemm::
C1GridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl
c1_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_;
Block2CTileMap block_2_ctile_map_;
InElementwiseOperation in_element_op_;
WeiElementwiseOperation wei_element_op_;
OutElementwiseOperation out_element_op_;
// for checking IsSupportedArgument()
index_t Conv_N_;
index_t Conv_K_;
index_t Conv_C_;
std::vector<index_t> input_spatial_lengths_;
std::vector<index_t> filter_spatial_lengths_;
std::vector<index_t> output_spatial_lengths_;
std::vector<index_t> conv_filter_strides_;
std::vector<index_t> conv_filter_dilations_;
std::vector<index_t> input_left_pads_;
std::vector<index_t> input_right_pads_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
#if 0
{
std::cout << DeviceOp{}.GetTypeString() << std::endl;
std::cout << "N " << arg.Conv_N_ << ", "
<< "K " << arg.Conv_K_ << ", "
<< "C " << arg.Conv_C_ << ", " << std::endl;
std::cout << "Y X " << arg.filter_spatial_lengths_[0] << ", "
<< arg.filter_spatial_lengths_[1] << ", " << std::endl;
std::cout << "Hi Wi " << arg.input_spatial_lengths_[0] << ", "
<< arg.input_spatial_lengths_[1] << ", " << std::endl;
std::cout << "Ho Wo " << arg.output_spatial_lengths_[0] << ", "
<< arg.output_spatial_lengths_[1] << ", " << std::endl;
std::cout << "Strides " << arg.conv_filter_strides_[0] << ", "
<< arg.conv_filter_strides_[1] << ", " << std::endl;
std::cout << "Dilations " << arg.conv_filter_dilations_[0] << ", "
<< arg.conv_filter_dilations_[1] << ", " << std::endl;
std::cout << "InLeftPads " << arg.input_left_pads_[0] << ", "
<< arg.input_left_pads_[1] << ", " << std::endl;
std::cout << "InLeftPads " << arg.input_right_pads_[0] << ", "
<< arg.input_right_pads_[1] << ", " << std::endl;
}
{
std::cout << "arg.a_grid_desc_k0_m_k1_{" << arg.a_grid_desc_k0_m_k1_.GetLength(I0)
<< ", " << arg.a_grid_desc_k0_m_k1_.GetLength(I1) << ", "
<< arg.a_grid_desc_k0_m_k1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.b_grid_desc_k0_n_k1_{" << arg.b_grid_desc_k0_n_k1_.GetLength(I0)
<< ", " << arg.b_grid_desc_k0_n_k1_.GetLength(I1) << ", "
<< arg.b_grid_desc_k0_n_k1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.c_grid_desc_m_n_{ " << arg.c_grid_desc_m_n_.GetLength(I0) << ", "
<< arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
std::cout << "arg.c0_grid_desc_m_n_{ " << arg.c0_grid_desc_m_n_.GetLength(I0)
<< ", " << arg.c0_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
std::cout << "arg.c1_grid_desc_m_n_{ " << arg.c1_grid_desc_m_n_.GetLength(I0)
<< ", " << arg.c1_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
}
#endif
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_))
{
throw std::runtime_error(
"wrong! GridwiseGemm_km_kn_m0m1n0n1_xdlops_v3r3 has invalid setting");
}
const index_t grid_size =
arg.block_2_ctile_map_.CalculateGridSize(arg.c_grid_desc_m_n_);
const auto K =
arg.a_grid_desc_k0_m_k1_.GetLength(I0) * arg.a_grid_desc_k0_m_k1_.GetLength(I2);
float ave_time = 0;
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
const auto kernel = kernel_gemm_xdlops_v3r3<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
remove_reference_t<
typename GridwiseGemm::
CGridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl>,
remove_reference_t<
typename GridwiseGemm::
C0GridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl>,
remove_reference_t<
typename GridwiseGemm::
C1GridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl>,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation,
Block2CTileMap,
true>;
ave_time = launch_and_time_kernel(
stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.p_c0_grid_,
arg.p_c1_grid_,
arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_,
arg.c0_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_,
arg.c1_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_,
arg.in_element_op_,
arg.wei_element_op_,
arg.out_element_op_,
arg.block_2_ctile_map_);
}
else
{
const auto kernel = kernel_gemm_xdlops_v3r3<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
remove_reference_t<
typename GridwiseGemm::
CGridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl>,
remove_reference_t<
typename GridwiseGemm::
C0GridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl>,
remove_reference_t<
typename GridwiseGemm::
C1GridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl>,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation,
Block2CTileMap,
false>;
ave_time = launch_and_time_kernel(
stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.p_c0_grid_,
arg.p_c1_grid_,
arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_,
arg.c0_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_,
arg.c1_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_,
arg.in_element_op_,
arg.wei_element_op_,
arg.out_element_op_,
arg.block_2_ctile_map_);
}
return ave_time;
}
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
if constexpr(ConvForwardSpecialization ==
ConvolutionForwardSpecialization::Filter1x1Stride1Pad0)
{
// check if it's 1x1, stride=1 conv
if(!(arg.filter_spatial_lengths_[0] == 1 && arg.filter_spatial_lengths_[1] == 1 &&
arg.conv_filter_strides_[0] == 1 && arg.conv_filter_strides_[1] == 1 &&
arg.input_left_pads_[0] == 0 && arg.input_left_pads_[1] == 0 &&
arg.input_right_pads_[0] == 0 && arg.input_right_pads_[1] == 0))
{
return false;
}
}
else if constexpr(ConvForwardSpecialization ==
ConvolutionForwardSpecialization::Filter1x1Pad0)
{
// check if it's 1x1 conv
if(!(arg.filter_spatial_lengths_[0] == 1 && arg.filter_spatial_lengths_[1] == 1 &&
arg.input_left_pads_[0] == 0 && arg.input_left_pads_[1] == 0 &&
arg.input_right_pads_[0] == 0 && arg.input_right_pads_[1] == 0))
{
return false;
}
}
// vector load A/B matrix from global memory
if(!(ABlockTransferSrcVectorDim == 2 && BBlockTransferSrcVectorDim == 2 &&
arg.Conv_C_ % ABlockTransferSrcScalarPerVector == 0 &&
arg.Conv_C_ % BBlockTransferSrcScalarPerVector == 0))
{
return false;
}
// vector store C matrix into global memory
if(!(arg.Conv_K_ % CBlockTransferScalarPerVector_NWaveNPerXdl == 0))
{
return false;
}
// Gridwise GEMM size
return GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_);
}
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const InDataType* p_in_grid,
const WeiDataType* p_wei_grid,
OutDataType* p_out_grid,
const OutDataType* p_bias_grid,
const OutDataType* p_resi_grid,
ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op)
{
return Argument{p_in_grid,
p_wei_grid,
p_out_grid,
p_bias_grid,
p_resi_grid,
N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
in_element_op,
wei_element_op,
out_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_in_grid,
const void* p_wei_grid,
void* p_out_grid,
const void* p_bias_grid,
const void* p_resi_grid,
ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op) override
{
return std::make_unique<Argument>(static_cast<const InDataType*>(p_in_grid),
static_cast<const WeiDataType*>(p_wei_grid),
static_cast<OutDataType*>(p_out_grid),
static_cast<const OutDataType*>(p_bias_grid),
static_cast<const OutDataType*>(p_resi_grid),
N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
in_element_op,
wei_element_op,
out_element_op);
}
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Add_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< K0PerBlock
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

925
include/ck/tensor_operation/gpu/device/impl/device_conv2d_fwd_xdl_c_shuffle_bias_activation_nhwc_kyxc_nhwk.hpp Normal file
View File

@@ -0,0 +1,925 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include <vector>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_conv_fwd_bias_activation.hpp"
#include "ck/tensor_operation/gpu/device/convolution_forward_specialization.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v3r2.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
// out[N, Ho, Wo, K] =
// activate(in[N, Hi, Wi, C] * wei[K, Y, X, C] + bias[K])
template <
typename InDataType,
typename WeiDataType,
typename OutDataType,
typename AccDataType,
typename InElementwiseOperation,
typename WeiElementwiseOperation,
typename OutElementwiseOperation,
InMemoryDataOperationEnum OutGlobalMemoryDataOperation,
ConvolutionForwardSpecialization ConvForwardSpecialization,
ck::index_t BlockSize,
ck::index_t MPerBlock,
ck::index_t NPerBlock,
ck::index_t K0PerBlock,
ck::index_t K1,
ck::index_t MPerXDL,
ck::index_t NPerXDL,
ck::index_t MXdlPerWave,
ck::index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_K0_M_K1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
ck::index_t ABlockTransferSrcVectorDim,
ck::index_t ABlockTransferSrcScalarPerVector,
ck::index_t ABlockTransferDstScalarPerVector_K1,
bool ABlockLdsAddExtraM,
typename BBlockTransferThreadClusterLengths_K0_N_K1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
ck::index_t BBlockTransferSrcVectorDim,
ck::index_t BBlockTransferSrcScalarPerVector,
ck::index_t BBlockTransferDstScalarPerVector_K1,
bool BBlockLdsAddExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl,
index_t CBlockTransferScalarPerVector_NWaveNPerXdl>
struct DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
: public DeviceConvFwdBiasActivation<InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation>
{
using DeviceOp =
DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K;
using ADataType = InDataType;
using BDataType = WeiDataType;
using CDataType = OutDataType;
// TODO make A/B datatype different
using ABDataType = InDataType;
// TODO make it support any # of spatial dimensions
static constexpr index_t NDimSpatial = 2;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr auto K1Number = Number<K1>{};
static constexpr auto GemmK1Number = K1Number;
static auto
MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads)
{
using namespace ck;
const index_t Hi = input_spatial_lengths[0];
const index_t Wi = input_spatial_lengths[1];
const index_t Ho = output_spatial_lengths[0];
const index_t Wo = output_spatial_lengths[1];
const index_t Y = filter_spatial_lengths[0];
const index_t X = filter_spatial_lengths[1];
const index_t ConvStrideH = conv_filter_strides[0];
const index_t ConvStrideW = conv_filter_strides[1];
const index_t ConvDilationH = conv_filter_dilations[0];
const index_t ConvDilationW = conv_filter_dilations[1];
const index_t InLeftPadH = input_left_pads[0];
const index_t InLeftPadW = input_left_pads[1];
const index_t InRightPadH = input_right_pads[0];
const index_t InRightPadW = input_right_pads[1];
const index_t GemmMRaw = N * Ho * Wo;
const index_t GemmN = K;
const auto GemmM = math::integer_least_multiple(GemmMRaw, MPerBlock);
const auto GemmMPad = GemmM - GemmMRaw;
if constexpr(ConvForwardSpecialization ==
ConvolutionForwardSpecialization::Filter1x1Stride1Pad0)
{ // 1x1, stride=1, pad=0
const index_t GemmK = Y * X * C;
assert(GemmK % GemmK1Number == 0);
const index_t GemmK0 = GemmK / GemmK1Number;
// A: input tensor
const auto in_gemmmraw_gemmk_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, C));
const auto in_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmmraw_gemmk_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_right_pad_transform(GemmMRaw, GemmMPad)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// B: weight tensor
const auto wei_gemmn_gemmk_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(K, C));
const auto wei_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
wei_gemmn_gemmk_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// C: output tensor
const auto out_gemmmraw_gemmn_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, K));
const auto out_gemmm_gemmn_grid_desc =
transform_tensor_descriptor(out_gemmmraw_gemmn_grid_desc,
make_tuple(make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
// C0: bias tensor: assume a contiguous vector
const auto bias_grid_desc_gemmm_gemmn =
make_naive_tensor_descriptor(make_tuple(GemmM, GemmN), make_tuple(I0, I1));
return make_tuple(in_gemmk0_gemmm_gemmk1_grid_desc,
wei_gemmk0_gemmn_gemmk1_grid_desc,
out_gemmm_gemmn_grid_desc,
bias_grid_desc_gemmm_gemmn);
}
else if constexpr(ConvForwardSpecialization ==
ConvolutionForwardSpecialization::Filter1x1Pad0)
{ // 1x1, pad=0
const index_t GemmK = Y * X * C;
assert(GemmK % GemmK1Number == 0);
const index_t GemmK0 = GemmK / GemmK1Number;
// A: input tensor
const auto in_n_hi_wi_c_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N, Hi, Wi, C));
const auto in_n_ho_wo_c_grid_desc = transform_tensor_descriptor(
in_n_hi_wi_c_grid_desc,
make_tuple(make_pass_through_transform(N),
make_embed_transform(make_tuple(Ho), make_tuple(ConvStrideH)),
make_embed_transform(make_tuple(Wo), make_tuple(ConvStrideW)),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
const auto in_gemmk0_gemmmraw_gemmk1_grid_desc = transform_tensor_descriptor(
in_n_ho_wo_c_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_merge_transform(make_tuple(N, Ho, Wo))),
make_tuple(Sequence<3>{}, Sequence<0, 1, 2>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
const auto in_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmk0_gemmmraw_gemmk1_grid_desc,
make_tuple(make_pass_through_transform(GemmK0),
make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmK1Number)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
// B: weight tensor
const auto wei_gemmn_gemmk_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(K, C));
const auto wei_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
wei_gemmn_gemmk_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// C: output tensor
const auto out_gemmmraw_gemmn_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, K));
const auto out_gemmm_gemmn_grid_desc =
transform_tensor_descriptor(out_gemmmraw_gemmn_grid_desc,
make_tuple(make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
// C0: bias tensor: assume a contiguous vector
const auto bias_grid_desc_gemmm_gemmn =
make_naive_tensor_descriptor(make_tuple(GemmM, GemmN), make_tuple(I0, I1));
return make_tuple(in_gemmk0_gemmm_gemmk1_grid_desc,
wei_gemmk0_gemmn_gemmk1_grid_desc,
out_gemmm_gemmn_grid_desc,
bias_grid_desc_gemmm_gemmn);
}
else if constexpr(ConvForwardSpecialization == ConvolutionForwardSpecialization::OddC)
{ // C = odd value
const index_t GemmKRaw = Y * X * C;
const index_t GemmK = math::integer_least_multiple(GemmKRaw, K0PerBlock * GemmK1Number);
const index_t GemmKPad = GemmK - GemmKRaw;
const index_t GemmK0 = GemmK / GemmK1Number;
// A: input tensor
const auto in_n_hi_wi_c_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N, Hi, Wi, C));
const auto in_n_hip_wip_c_grid_desc = transform_tensor_descriptor(
in_n_hi_wi_c_grid_desc,
make_tuple(make_pass_through_transform(N),
make_pad_transform(Hi, InLeftPadH, InRightPadH),
make_pad_transform(Wi, InLeftPadW, InRightPadW),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
const auto in_n_y_ho_x_wo_c_grid_desc = transform_tensor_descriptor(
in_n_hip_wip_c_grid_desc,
make_tuple(
make_pass_through_transform(N),
make_embed_transform(make_tuple(Y, Ho), make_tuple(ConvDilationH, ConvStrideH)),
make_embed_transform(make_tuple(X, Wo), make_tuple(ConvDilationW, ConvStrideW)),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1, 2>{}, Sequence<3, 4>{}, Sequence<5>{}));
const auto in_gemmkraw_gemmmraw_grid_desc =
transform_tensor_descriptor(in_n_y_ho_x_wo_c_grid_desc,
make_tuple(make_merge_transform(make_tuple(Y, X, C)),
make_merge_transform(make_tuple(N, Ho, Wo))),
make_tuple(Sequence<1, 3, 5>{}, Sequence<0, 2, 4>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto in_gemmk_gemmm_grid_desc = transform_tensor_descriptor(
in_gemmkraw_gemmmraw_grid_desc,
make_tuple(make_right_pad_transform(GemmKRaw, GemmKPad),
make_right_pad_transform(GemmMRaw, GemmMPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto in_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmk_gemmm_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmM)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// B: weight tensor
const auto wei_k_yxc_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(K, Y * X * C));
const auto wei_gemmk_gemmn_grid_desc = transform_tensor_descriptor(
wei_k_yxc_grid_desc,
make_tuple(make_pass_through_transform(K),
make_right_pad_transform(GemmKRaw, GemmKPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<1>{}, Sequence<0>{}));
const auto wei_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
wei_gemmk_gemmn_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// C: output tensor
const auto out_nhowo_k_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, K));
const auto out_gemmmraw_gemmn_grid_desc =
transform_tensor_descriptor(out_nhowo_k_grid_desc,
make_tuple(make_pass_through_transform(N * Ho * Wo),
make_pass_through_transform(K)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto out_gemmm_gemmn_grid_desc =
transform_tensor_descriptor(out_gemmmraw_gemmn_grid_desc,
make_tuple(make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
// C0: bias tensor: assume a contiguous vector
const auto bias_grid_desc_gemmm_gemmn =
make_naive_tensor_descriptor(make_tuple(GemmM, GemmN), make_tuple(I0, I1));
return make_tuple(in_gemmk0_gemmm_gemmk1_grid_desc,
wei_gemmk0_gemmn_gemmk1_grid_desc,
out_gemmm_gemmn_grid_desc,
bias_grid_desc_gemmm_gemmn);
}
else
{
const index_t GemmK = Y * X * C;
assert(GemmK % GemmK1Number == 0);
const index_t GemmK0 = GemmK / GemmK1Number;
// A: input tensor
const auto in_n_hi_wi_c_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N, Hi, Wi, C));
const auto in_n_hip_wip_c_grid_desc = transform_tensor_descriptor(
in_n_hi_wi_c_grid_desc,
make_tuple(make_pass_through_transform(N),
make_pad_transform(Hi, InLeftPadH, InRightPadH),
make_pad_transform(Wi, InLeftPadW, InRightPadW),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
const auto in_n_y_ho_x_wo_c_grid_desc = transform_tensor_descriptor(
in_n_hip_wip_c_grid_desc,
make_tuple(
make_pass_through_transform(N),
make_embed_transform(make_tuple(Y, Ho), make_tuple(ConvDilationH, ConvStrideH)),
make_embed_transform(make_tuple(X, Wo), make_tuple(ConvDilationW, ConvStrideW)),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1, 2>{}, Sequence<3, 4>{}, Sequence<5>{}));
const auto in_gemmk_gemmmraw_grid_desc =
transform_tensor_descriptor(in_n_y_ho_x_wo_c_grid_desc,
make_tuple(make_merge_transform(make_tuple(Y, X, C)),
make_merge_transform(make_tuple(N, Ho, Wo))),
make_tuple(Sequence<1, 3, 5>{}, Sequence<0, 2, 4>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto in_gemmk0_gemmmraw_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmk_gemmmraw_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmMRaw)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
const auto in_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmk0_gemmmraw_gemmk1_grid_desc,
make_tuple(make_pass_through_transform(GemmK0),
make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmK1Number)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
// B: weight tensor
const auto wei_k_yxc_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(K, Y * X * C));
const auto wei_gemmk_gemmn_grid_desc = transform_tensor_descriptor(
wei_k_yxc_grid_desc,
make_tuple(make_pass_through_transform(K), make_pass_through_transform(Y * X * C)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<1>{}, Sequence<0>{}));
const auto wei_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
wei_gemmk_gemmn_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// C: output tensor
const auto out_nhowo_k_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, K));
const auto out_gemmmraw_gemmn_grid_desc =
transform_tensor_descriptor(out_nhowo_k_grid_desc,
make_tuple(make_pass_through_transform(N * Ho * Wo),
make_pass_through_transform(K)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto out_gemmm_gemmn_grid_desc =
transform_tensor_descriptor(out_gemmmraw_gemmn_grid_desc,
make_tuple(make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
// C0: bias tensor: assume a contiguous vector
const auto bias_grid_desc_gemmm_gemmn =
make_naive_tensor_descriptor(make_tuple(GemmM, GemmN), make_tuple(I0, I1));
return make_tuple(in_gemmk0_gemmm_gemmk1_grid_desc,
wei_gemmk0_gemmn_gemmk1_grid_desc,
out_gemmm_gemmn_grid_desc,
bias_grid_desc_gemmm_gemmn);
}
}
using ABCGridDescs = decltype(MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(
1, 1, 1, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}));
using AGridDesc_K0_M_K1 = remove_cvref_t<decltype(ABCGridDescs{}[I0])>;
using BGridDesc_K0_N_K1 = remove_cvref_t<decltype(ABCGridDescs{}[I1])>;
using CGridDesc_M_N = remove_cvref_t<decltype(ABCGridDescs{}[I2])>;
using C0GridDesc_M_N = remove_cvref_t<decltype(ABCGridDescs{}[I3])>;
// GridwiseGemm
using GridwiseGemm = GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v3r2<
BlockSize,
ABDataType, // TODO: distinguish A/B datatype
AccDataType,
CDataType,
OutGlobalMemoryDataOperation,
AGridDesc_K0_M_K1,
BGridDesc_K0_N_K1,
CGridDesc_M_N,
C0GridDesc_M_N,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation,
MPerBlock,
NPerBlock,
K0PerBlock,
MPerXDL,
NPerXDL,
K1,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_K0_M_K1,
Sequence<1, 0, 2>, // ABlockTransferThreadClusterArrangeOrder,
Sequence<1, 0, 2>, // ABlockTransferSrcAccessOrder,
2, // ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_K1,
false, // AThreadTransferSrcResetCoordinateAfterRun,
ABlockLdsAddExtraM,
BBlockTransferThreadClusterLengths_K0_N_K1,
Sequence<1, 0, 2>, // BBlockTransferThreadClusterArrangeOrder,
Sequence<1, 0, 2>, // BBlockTransferSrcAccessOrder,
2, // BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_K1,
false, // BThreadTransferSrcResetCoordinateAfterRun,
BBlockLdsAddExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl,
CBlockTransferScalarPerVector_NWaveNPerXdl>;
// Argument
struct Argument : public BaseArgument
{
Argument(const InDataType* p_in_grid,
const WeiDataType* p_wei_grid,
OutDataType* p_out_grid,
const OutDataType* p_bias_grid,
ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
ck::index_t M01,
ck::index_t N01,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op)
: p_a_grid_{p_in_grid},
p_b_grid_{p_wei_grid},
p_c_grid_{p_out_grid},
p_c0_grid_{p_bias_grid},
a_grid_desc_k0_m_k1_{},
b_grid_desc_k0_n_k1_{},
c_grid_desc_m_n_{},
c0_grid_desc_m_n_{},
c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_{},
c0_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_{},
block_2_ctile_map_{},
M01_{M01},
N01_{N01},
in_element_op_{in_element_op},
wei_element_op_{wei_element_op},
out_element_op_{out_element_op},
Conv_N_{N},
Conv_K_{K},
Conv_C_{C},
input_spatial_lengths_{input_spatial_lengths},
filter_spatial_lengths_{filter_spatial_lengths},
output_spatial_lengths_{output_spatial_lengths},
conv_filter_strides_{conv_filter_strides},
conv_filter_dilations_{conv_filter_dilations},
input_left_pads_{input_left_pads},
input_right_pads_{input_right_pads}
{
const auto descs =
DeviceOp::MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads);
a_grid_desc_k0_m_k1_ = descs[I0];
b_grid_desc_k0_n_k1_ = descs[I1];
c_grid_desc_m_n_ = descs[I2];
c0_grid_desc_m_n_ = descs[I3];
block_2_ctile_map_ =
GridwiseGemm::MakeDefaultBlock2CTileMap(c_grid_desc_m_n_, M01, N01);
if(GridwiseGemm::CheckValidity(a_grid_desc_k0_m_k1_,
b_grid_desc_k0_n_k1_,
c_grid_desc_m_n_,
block_2_ctile_map_))
{
c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_ =
GridwiseGemm::
MakeCGridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl(
c_grid_desc_m_n_);
c0_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_ =
GridwiseGemm::
MakeCGridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl(
c0_grid_desc_m_n_);
}
}
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
CDataType* p_c_grid_;
const CDataType* p_c0_grid_;
AGridDesc_K0_M_K1 a_grid_desc_k0_m_k1_;
BGridDesc_K0_N_K1 b_grid_desc_k0_n_k1_;
CGridDesc_M_N c_grid_desc_m_n_;
C0GridDesc_M_N c0_grid_desc_m_n_;
typename GridwiseGemm::
CGridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl
c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_;
typename GridwiseGemm::
C0GridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl
c0_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_;
typename GridwiseGemm::DefaultBlock2CTileMap block_2_ctile_map_;
index_t M01_;
index_t N01_;
InElementwiseOperation in_element_op_;
WeiElementwiseOperation wei_element_op_;
OutElementwiseOperation out_element_op_;
// for checking IsSupportedArgument()
index_t Conv_N_;
index_t Conv_K_;
index_t Conv_C_;
std::vector<index_t> input_spatial_lengths_;
std::vector<index_t> filter_spatial_lengths_;
std::vector<index_t> output_spatial_lengths_;
std::vector<index_t> conv_filter_strides_;
std::vector<index_t> conv_filter_dilations_;
std::vector<index_t> input_left_pads_;
std::vector<index_t> input_right_pads_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
#if 0
{
std::cout << DeviceOp{}.GetTypeString() << std::endl;
std::cout << "N " << arg.Conv_N_ << ", "
<< "K " << arg.Conv_K_ << ", "
<< "C " << arg.Conv_C_ << ", " << std::endl;
std::cout << "Y X " << arg.filter_spatial_lengths_[0] << ", "
<< arg.filter_spatial_lengths_[1] << ", " << std::endl;
std::cout << "Hi Wi " << arg.input_spatial_lengths_[0] << ", "
<< arg.input_spatial_lengths_[1] << ", " << std::endl;
std::cout << "Ho Wo " << arg.output_spatial_lengths_[0] << ", "
<< arg.output_spatial_lengths_[1] << ", " << std::endl;
std::cout << "Strides " << arg.conv_filter_strides_[0] << ", "
<< arg.conv_filter_strides_[1] << ", " << std::endl;
std::cout << "Dilations " << arg.conv_filter_dilations_[0] << ", "
<< arg.conv_filter_dilations_[1] << ", " << std::endl;
std::cout << "InLeftPads " << arg.input_left_pads_[0] << ", "
<< arg.input_left_pads_[1] << ", " << std::endl;
std::cout << "InLeftPads " << arg.input_right_pads_[0] << ", "
<< arg.input_right_pads_[1] << ", " << std::endl;
}
{
std::cout << "arg.a_grid_desc_k0_m_k1_{" << arg.a_grid_desc_k0_m_k1_.GetLength(I0)
<< ", " << arg.a_grid_desc_k0_m_k1_.GetLength(I1) << ", "
<< arg.a_grid_desc_k0_m_k1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.b_grid_desc_k0_n_k1_{" << arg.b_grid_desc_k0_n_k1_.GetLength(I0)
<< ", " << arg.b_grid_desc_k0_n_k1_.GetLength(I1) << ", "
<< arg.b_grid_desc_k0_n_k1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.c_grid_desc_m_n_{ " << arg.c_grid_desc_m_n_.GetLength(I0) << ", "
<< arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
std::cout << "arg.c0_grid_desc_m_n_{ " << arg.c0_grid_desc_m_n_.GetLength(I0)
<< ", " << arg.c0_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
}
#endif
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_))
{
throw std::runtime_error(
"wrong! GridwiseGemm_km_kn_m0m1n0n1_xdlops_v3r2 has invalid setting");
}
const index_t grid_size =
arg.block_2_ctile_map_.CalculateGridSize(arg.c_grid_desc_m_n_);
const auto K =
arg.a_grid_desc_k0_m_k1_.GetLength(I0) * arg.a_grid_desc_k0_m_k1_.GetLength(I2);
float ave_time = 0;
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
const auto kernel = kernel_gemm_xdlops_v3r2<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
remove_reference_t<
typename GridwiseGemm::
CGridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl>,
remove_reference_t<
typename GridwiseGemm::
C0GridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl>,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation,
remove_reference_t<typename GridwiseGemm::DefaultBlock2CTileMap>,
true>;
ave_time = launch_and_time_kernel(
stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.p_c0_grid_,
arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_,
arg.c0_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_,
arg.in_element_op_,
arg.wei_element_op_,
arg.out_element_op_,
arg.block_2_ctile_map_);
}
else
{
const auto kernel = kernel_gemm_xdlops_v3r2<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
remove_reference_t<
typename GridwiseGemm::
CGridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl>,
remove_reference_t<
typename GridwiseGemm::
C0GridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl>,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation,
remove_reference_t<typename GridwiseGemm::DefaultBlock2CTileMap>,
false>;
ave_time = launch_and_time_kernel(
stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.p_c0_grid_,
arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_,
arg.c0_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_,
arg.in_element_op_,
arg.wei_element_op_,
arg.out_element_op_,
arg.block_2_ctile_map_);
}
return ave_time;
}
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
if constexpr(ConvForwardSpecialization ==
ConvolutionForwardSpecialization::Filter1x1Stride1Pad0)
{
// check if it's 1x1, stride=1 conv
if(!(arg.filter_spatial_lengths_[0] == 1 && arg.filter_spatial_lengths_[1] == 1 &&
arg.conv_filter_strides_[0] == 1 && arg.conv_filter_strides_[1] == 1 &&
arg.input_left_pads_[0] == 0 && arg.input_left_pads_[1] == 0 &&
arg.input_right_pads_[0] == 0 && arg.input_right_pads_[1] == 0))
{
return false;
}
}
else if constexpr(ConvForwardSpecialization ==
ConvolutionForwardSpecialization::Filter1x1Pad0)
{
// check if it's 1x1 conv
if(!(arg.filter_spatial_lengths_[0] == 1 && arg.filter_spatial_lengths_[1] == 1 &&
arg.input_left_pads_[0] == 0 && arg.input_left_pads_[1] == 0 &&
arg.input_right_pads_[0] == 0 && arg.input_right_pads_[1] == 0))
{
return false;
}
}
// vector load A/B matrix from global memory
if(!(ABlockTransferSrcVectorDim == 2 && BBlockTransferSrcVectorDim == 2 &&
arg.Conv_C_ % ABlockTransferSrcScalarPerVector == 0 &&
arg.Conv_C_ % BBlockTransferSrcScalarPerVector == 0))
{
return false;
}
// vector store C matrix into global memory
if(!(arg.Conv_K_ % CBlockTransferScalarPerVector_NWaveNPerXdl == 0))
{
return false;
}
// Gridwise GEMM size
return GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_);
}
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const InDataType* p_in_grid,
const WeiDataType* p_wei_grid,
OutDataType* p_out_grid,
const OutDataType* p_bias_grid,
ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op)
{
return Argument{p_in_grid,
p_wei_grid,
p_out_grid,
p_bias_grid,
N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
1,
1,
in_element_op,
wei_element_op,
out_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_in_grid,
const void* p_wei_grid,
void* p_out_grid,
const void* p_bias_grid,
ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op) override
{
return std::make_unique<Argument>(static_cast<const InDataType*>(p_in_grid),
static_cast<const WeiDataType*>(p_wei_grid),
static_cast<OutDataType*>(p_out_grid),
static_cast<const OutDataType*>(p_bias_grid),
N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
1,
1,
in_element_op,
wei_element_op,
out_element_op);
}
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceConv2dFwdXdl_C_Shuffle_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< K0PerBlock
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

893
include/ck/tensor_operation/gpu/device/impl/device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk.hpp Normal file
View File

@@ -0,0 +1,893 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_conv_fwd.hpp"
#include "ck/tensor_operation/gpu/device/convolution_forward_specialization.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v3r1.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
// out[N, Ho, Wo, K] = in[N, Hi, Wi, C] * wei[K, Y, X, C]
template <
typename InDataType,
typename WeiDataType,
typename OutDataType,
typename AccDataType,
typename InElementwiseOperation,
typename WeiElementwiseOperation,
typename OutElementwiseOperation,
ConvolutionForwardSpecialization ConvForwardSpecialization,
ck::index_t BlockSize,
ck::index_t MPerBlock,
ck::index_t NPerBlock,
ck::index_t K0PerBlock,
ck::index_t K1,
ck::index_t MPerXdl,
ck::index_t NPerXdl,
ck::index_t MXdlPerWave,
ck::index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_K0_M_K1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
ck::index_t ABlockTransferSrcVectorDim,
ck::index_t ABlockTransferSrcScalarPerVector,
ck::index_t ABlockTransferDstScalarPerVector_K1,
bool ABlockLdsAddExtraM,
typename BBlockTransferThreadClusterLengths_K0_N_K1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
ck::index_t BBlockTransferSrcVectorDim,
ck::index_t BBlockTransferSrcScalarPerVector,
ck::index_t BBlockTransferDstScalarPerVector_K1,
bool BBlockLdsAddExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl,
index_t CBlockTransferScalarPerVector_NWaveNPerXdl>
struct DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
: public DeviceConvFwd<2,
ck::tensor_layout::convolution::NHWC,
ck::tensor_layout::convolution::KYXC,
ck::tensor_layout::convolution::NHWK,
InDataType,
WeiDataType,
OutDataType,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation>
{
using DeviceOp = DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K;
using ADataType = InDataType;
using BDataType = WeiDataType;
using CDataType = OutDataType;
// TODO make A/B datatype different
using ABDataType = InDataType;
static constexpr index_t NDimSpatial = 2;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr auto I4 = Number<4>{};
static constexpr auto I5 = Number<5>{};
static constexpr auto K1Number = Number<K1>{};
static constexpr auto GemmK1Number = K1Number;
static auto
MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads)
{
using namespace ck;
const index_t Hi = input_spatial_lengths[0];
const index_t Wi = input_spatial_lengths[1];
const index_t Ho = output_spatial_lengths[0];
const index_t Wo = output_spatial_lengths[1];
const index_t Y = filter_spatial_lengths[0];
const index_t X = filter_spatial_lengths[1];
const index_t ConvStrideH = conv_filter_strides[0];
const index_t ConvStrideW = conv_filter_strides[1];
const index_t ConvDilationH = conv_filter_dilations[0];
const index_t ConvDilationW = conv_filter_dilations[1];
const index_t InLeftPadH = input_left_pads[0];
const index_t InLeftPadW = input_left_pads[1];
const index_t InRightPadH = input_right_pads[0];
const index_t InRightPadW = input_right_pads[1];
const index_t GemmMRaw = N * Ho * Wo;
const index_t GemmN = K;
const auto GemmM = math::integer_least_multiple(GemmMRaw, MPerBlock);
const auto GemmMPad = GemmM - GemmMRaw;
if constexpr(ConvForwardSpecialization ==
ConvolutionForwardSpecialization::Filter1x1Stride1Pad0)
{ // 1x1, stride=1, pad=0
const index_t GemmK = Y * X * C;
assert(GemmK % GemmK1Number == 0);
const index_t GemmK0 = GemmK / GemmK1Number;
// A: input tensor
const auto in_gemmmraw_gemmk_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, C));
const auto in_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmmraw_gemmk_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_right_pad_transform(GemmMRaw, GemmMPad)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// B: weight tensor
const auto wei_gemmn_gemmk_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(K, C));
const auto wei_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
wei_gemmn_gemmk_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// C: output tensor
const auto out_gemmmraw_gemmn_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, K));
const auto out_gemmm_gemmn_grid_desc =
transform_tensor_descriptor(out_gemmmraw_gemmn_grid_desc,
make_tuple(make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return make_tuple(in_gemmk0_gemmm_gemmk1_grid_desc,
wei_gemmk0_gemmn_gemmk1_grid_desc,
out_gemmm_gemmn_grid_desc);
}
else if constexpr(ConvForwardSpecialization ==
ConvolutionForwardSpecialization::Filter1x1Pad0)
{ // 1x1, pad=0
const index_t GemmK = Y * X * C;
assert(GemmK % GemmK1Number == 0);
const index_t GemmK0 = GemmK / GemmK1Number;
// A: input tensor
const auto in_n_hi_wi_c_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N, Hi, Wi, C));
const auto in_n_ho_wo_c_grid_desc = transform_tensor_descriptor(
in_n_hi_wi_c_grid_desc,
make_tuple(make_pass_through_transform(N),
make_embed_transform(make_tuple(Ho), make_tuple(ConvStrideH)),
make_embed_transform(make_tuple(Wo), make_tuple(ConvStrideW)),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
const auto in_gemmk0_gemmmraw_gemmk1_grid_desc = transform_tensor_descriptor(
in_n_ho_wo_c_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_merge_transform(make_tuple(N, Ho, Wo))),
make_tuple(Sequence<3>{}, Sequence<0, 1, 2>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
const auto in_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmk0_gemmmraw_gemmk1_grid_desc,
make_tuple(make_pass_through_transform(GemmK0),
make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmK1Number)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
// B: weight tensor
const auto wei_gemmn_gemmk_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(K, C));
const auto wei_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
wei_gemmn_gemmk_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// C: output tensor
const auto out_gemmmraw_gemmn_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, K));
const auto out_gemmm_gemmn_grid_desc =
transform_tensor_descriptor(out_gemmmraw_gemmn_grid_desc,
make_tuple(make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return make_tuple(in_gemmk0_gemmm_gemmk1_grid_desc,
wei_gemmk0_gemmn_gemmk1_grid_desc,
out_gemmm_gemmn_grid_desc);
}
else if constexpr(ConvForwardSpecialization == ConvolutionForwardSpecialization::OddC)
{ // C = odd value
const index_t GemmKRaw = Y * X * C;
const index_t GemmK = math::integer_least_multiple(GemmKRaw, K0PerBlock * GemmK1Number);
const index_t GemmKPad = GemmK - GemmKRaw;
const index_t GemmK0 = GemmK / GemmK1Number;
// A: input tensor
const auto in_n_hi_wi_c_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N, Hi, Wi, C));
const auto in_n_hip_wip_c_grid_desc = transform_tensor_descriptor(
in_n_hi_wi_c_grid_desc,
make_tuple(make_pass_through_transform(N),
make_pad_transform(Hi, InLeftPadH, InRightPadH),
make_pad_transform(Wi, InLeftPadW, InRightPadW),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
const auto in_n_y_ho_x_wo_c_grid_desc = transform_tensor_descriptor(
in_n_hip_wip_c_grid_desc,
make_tuple(
make_pass_through_transform(N),
make_embed_transform(make_tuple(Y, Ho), make_tuple(ConvDilationH, ConvStrideH)),
make_embed_transform(make_tuple(X, Wo), make_tuple(ConvDilationW, ConvStrideW)),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1, 2>{}, Sequence<3, 4>{}, Sequence<5>{}));
const auto in_gemmkraw_gemmmraw_grid_desc =
transform_tensor_descriptor(in_n_y_ho_x_wo_c_grid_desc,
make_tuple(make_merge_transform(make_tuple(Y, X, C)),
make_merge_transform(make_tuple(N, Ho, Wo))),
make_tuple(Sequence<1, 3, 5>{}, Sequence<0, 2, 4>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto in_gemmk_gemmm_grid_desc = transform_tensor_descriptor(
in_gemmkraw_gemmmraw_grid_desc,
make_tuple(make_right_pad_transform(GemmKRaw, GemmKPad),
make_right_pad_transform(GemmMRaw, GemmMPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto in_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmk_gemmm_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmM)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// B: weight tensor
const auto wei_k_yxc_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(K, Y * X * C));
const auto wei_gemmk_gemmn_grid_desc = transform_tensor_descriptor(
wei_k_yxc_grid_desc,
make_tuple(make_pass_through_transform(K),
make_right_pad_transform(GemmKRaw, GemmKPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<1>{}, Sequence<0>{}));
const auto wei_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
wei_gemmk_gemmn_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// C: output tensor
const auto out_nhowo_k_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, K));
const auto out_gemmmraw_gemmn_grid_desc =
transform_tensor_descriptor(out_nhowo_k_grid_desc,
make_tuple(make_pass_through_transform(N * Ho * Wo),
make_pass_through_transform(K)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto out_gemmm_gemmn_grid_desc =
transform_tensor_descriptor(out_gemmmraw_gemmn_grid_desc,
make_tuple(make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return make_tuple(in_gemmk0_gemmm_gemmk1_grid_desc,
wei_gemmk0_gemmn_gemmk1_grid_desc,
out_gemmm_gemmn_grid_desc);
}
else
{
const index_t GemmK = Y * X * C;
assert(GemmK % GemmK1Number == 0);
const index_t GemmK0 = GemmK / GemmK1Number;
// A: input tensor
const auto in_n_hi_wi_c_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N, Hi, Wi, C));
const auto in_n_hip_wip_c_grid_desc = transform_tensor_descriptor(
in_n_hi_wi_c_grid_desc,
make_tuple(make_pass_through_transform(N),
make_pad_transform(Hi, InLeftPadH, InRightPadH),
make_pad_transform(Wi, InLeftPadW, InRightPadW),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
const auto in_n_y_ho_x_wo_c_grid_desc = transform_tensor_descriptor(
in_n_hip_wip_c_grid_desc,
make_tuple(
make_pass_through_transform(N),
make_embed_transform(make_tuple(Y, Ho), make_tuple(ConvDilationH, ConvStrideH)),
make_embed_transform(make_tuple(X, Wo), make_tuple(ConvDilationW, ConvStrideW)),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1, 2>{}, Sequence<3, 4>{}, Sequence<5>{}));
const auto in_gemmk_gemmmraw_grid_desc =
transform_tensor_descriptor(in_n_y_ho_x_wo_c_grid_desc,
make_tuple(make_merge_transform(make_tuple(Y, X, C)),
make_merge_transform(make_tuple(N, Ho, Wo))),
make_tuple(Sequence<1, 3, 5>{}, Sequence<0, 2, 4>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto in_gemmk0_gemmmraw_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmk_gemmmraw_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmMRaw)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
const auto in_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmk0_gemmmraw_gemmk1_grid_desc,
make_tuple(make_pass_through_transform(GemmK0),
make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmK1Number)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
// B: weight tensor
const auto wei_k_yxc_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(K, Y * X * C));
const auto wei_gemmk_gemmn_grid_desc = transform_tensor_descriptor(
wei_k_yxc_grid_desc,
make_tuple(make_pass_through_transform(K), make_pass_through_transform(Y * X * C)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<1>{}, Sequence<0>{}));
const auto wei_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
wei_gemmk_gemmn_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// C: output tensor
const auto out_nhowo_k_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, K));
const auto out_gemmmraw_gemmn_grid_desc =
transform_tensor_descriptor(out_nhowo_k_grid_desc,
make_tuple(make_pass_through_transform(N * Ho * Wo),
make_pass_through_transform(K)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto out_gemmm_gemmn_grid_desc =
transform_tensor_descriptor(out_gemmmraw_gemmn_grid_desc,
make_tuple(make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return make_tuple(in_gemmk0_gemmm_gemmk1_grid_desc,
wei_gemmk0_gemmn_gemmk1_grid_desc,
out_gemmm_gemmn_grid_desc);
}
}
using ABCGridDescs = decltype(MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(
1, 1, 1, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}));
using AGridDesc_K0_M_K1 = remove_cvref_t<decltype(ABCGridDescs{}[I0])>;
using BGridDesc_K0_N_K1 = remove_cvref_t<decltype(ABCGridDescs{}[I1])>;
using CGridDesc_M_N = remove_cvref_t<decltype(ABCGridDescs{}[I2])>;
using Block2CTileMap = BlockToCTileMap_M00_N0_M01<MPerBlock, NPerBlock, CGridDesc_M_N>;
// GridwiseGemm
using GridwiseGemm = GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v3r1<
BlockSize,
ABDataType, // TODO: distinguish A/B datatype
AccDataType,
CDataType, // TODO: Add ShuffleType for DeviceConv2d
CDataType,
InMemoryDataOperationEnum::Set,
AGridDesc_K0_M_K1,
BGridDesc_K0_N_K1,
CGridDesc_M_N,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation,
MPerBlock,
NPerBlock,
K0PerBlock * K1,
K1, // AK1
K1, // BK1
MPerXdl,
NPerXdl,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_K0_M_K1,
Sequence<1, 0, 2>, // ABlockTransferThreadClusterArrangeOrder,
Sequence<1, 0, 2>, // ABlockTransferSrcAccessOrder,
2, // ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_K1,
false, // AThreadTransferSrcResetCoordinateAfterRun,
ABlockLdsAddExtraM,
BBlockTransferThreadClusterLengths_K0_N_K1,
Sequence<1, 0, 2>, // BBlockTransferThreadClusterArrangeOrder,
Sequence<1, 0, 2>, // BBlockTransferSrcAccessOrder,
2, // BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_K1,
false, // BThreadTransferSrcResetCoordinateAfterRun,
BBlockLdsAddExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl,
CBlockTransferScalarPerVector_NWaveNPerXdl>;
// Argument
struct Argument : public BaseArgument
{
Argument(const InDataType* p_in_grid,
const WeiDataType* p_wei_grid,
OutDataType* p_out_grid,
ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op)
: p_a_grid_{p_in_grid},
p_b_grid_{p_wei_grid},
p_c_grid_{p_out_grid},
a_grid_desc_k0_m_k1_{},
b_grid_desc_k0_n_k1_{},
c_grid_desc_m_n_{},
c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_{},
block_2_ctile_map_{},
in_element_op_{in_element_op},
wei_element_op_{wei_element_op},
out_element_op_{out_element_op},
Conv_N_{N},
Conv_K_{K},
Conv_C_{C},
input_spatial_lengths_{input_spatial_lengths},
filter_spatial_lengths_{filter_spatial_lengths},
output_spatial_lengths_{output_spatial_lengths},
conv_filter_strides_{conv_filter_strides},
conv_filter_dilations_{conv_filter_dilations},
input_left_pads_{input_left_pads},
input_right_pads_{input_right_pads}
{
const auto descs =
DeviceOp::MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads);
a_grid_desc_k0_m_k1_ = descs[I0];
b_grid_desc_k0_n_k1_ = descs[I1];
c_grid_desc_m_n_ = descs[I2];
block_2_ctile_map_ = Block2CTileMap{c_grid_desc_m_n_};
if(GridwiseGemm::CheckValidity(a_grid_desc_k0_m_k1_,
b_grid_desc_k0_n_k1_,
c_grid_desc_m_n_,
block_2_ctile_map_))
{
c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_ =
GridwiseGemm::
MakeCGridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl(
c_grid_desc_m_n_);
}
}
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
CDataType* p_c_grid_;
AGridDesc_K0_M_K1 a_grid_desc_k0_m_k1_;
BGridDesc_K0_N_K1 b_grid_desc_k0_n_k1_;
CGridDesc_M_N c_grid_desc_m_n_;
typename GridwiseGemm::
CGridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl
c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_;
Block2CTileMap block_2_ctile_map_;
InElementwiseOperation in_element_op_;
WeiElementwiseOperation wei_element_op_;
OutElementwiseOperation out_element_op_;
// for checking IsSupportedArgument()
index_t Conv_N_;
index_t Conv_K_;
index_t Conv_C_;
std::vector<index_t> input_spatial_lengths_;
std::vector<index_t> filter_spatial_lengths_;
std::vector<index_t> output_spatial_lengths_;
std::vector<index_t> conv_filter_strides_;
std::vector<index_t> conv_filter_dilations_;
std::vector<index_t> input_left_pads_;
std::vector<index_t> input_right_pads_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
#if 0
{
std::cout << DeviceOp{}.GetTypeString() << std::endl;
std::cout << "N " << arg.Conv_N_ << ", "
<< "K " << arg.Conv_K_ << ", "
<< "C " << arg.Conv_C_ << ", " << std::endl;
std::cout << "Y X " << arg.filter_spatial_lengths_[0] << ", "
<< arg.filter_spatial_lengths_[1] << ", " << std::endl;
std::cout << "Hi Wi " << arg.input_spatial_lengths_[0] << ", "
<< arg.input_spatial_lengths_[1] << ", " << std::endl;
std::cout << "Ho Wo " << arg.output_spatial_lengths_[0] << ", "
<< arg.output_spatial_lengths_[1] << ", " << std::endl;
std::cout << "Strides " << arg.conv_filter_strides_[0] << ", "
<< arg.conv_filter_strides_[1] << ", " << std::endl;
std::cout << "Dilations " << arg.conv_filter_dilations_[0] << ", "
<< arg.conv_filter_dilations_[1] << ", " << std::endl;
std::cout << "InLeftPads " << arg.input_left_pads_[0] << ", "
<< arg.input_left_pads_[1] << ", " << std::endl;
std::cout << "InLeftPads " << arg.input_right_pads_[0] << ", "
<< arg.input_right_pads_[1] << ", " << std::endl;
}
{
std::cout << "arg.a_grid_desc_k0_m_k1_{" << arg.a_grid_desc_k0_m_k1_.GetLength(I0)
<< ", " << arg.a_grid_desc_k0_m_k1_.GetLength(I1) << ", "
<< arg.a_grid_desc_k0_m_k1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.b_grid_desc_k0_n_k1_{" << arg.b_grid_desc_k0_n_k1_.GetLength(I0)
<< ", " << arg.b_grid_desc_k0_n_k1_.GetLength(I1) << ", "
<< arg.b_grid_desc_k0_n_k1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.c_grid_desc_m_n_{ " << arg.c_grid_desc_m_n_.GetLength(I0) << ", "
<< arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
std::cout
<< "arg.c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_"
"nwavenperxdl_{ "
<< arg.c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_
.GetLength(I0)
<< ", "
<< arg.c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_
.GetLength(I1)
<< ", "
<< arg.c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_
.GetLength(I2)
<< ", "
<< arg.c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_
.GetLength(I3)
<< ", "
<< arg.c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_
.GetLength(I4)
<< ", "
<< arg.c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_
.GetLength(I5)
<< "}" << std::endl;
}
#endif
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_))
{
throw std::runtime_error(
"wrong! GridwiseGemm_km_kn_m0m1n0n1_xdlops_v3r1 has invalid setting");
}
const index_t grid_size =
arg.block_2_ctile_map_.CalculateGridSize(arg.c_grid_desc_m_n_);
const auto K =
arg.a_grid_desc_k0_m_k1_.GetLength(I0) * arg.a_grid_desc_k0_m_k1_.GetLength(I2);
float ave_time = 0;
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
const auto kernel = kernel_gemm_xdlops_v3r1<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
remove_reference_t<
typename GridwiseGemm::
CGridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl>,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation,
Block2CTileMap,
true>;
ave_time = launch_and_time_kernel(
stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_,
arg.in_element_op_,
arg.wei_element_op_,
arg.out_element_op_,
arg.block_2_ctile_map_);
}
else
{
const auto kernel = kernel_gemm_xdlops_v3r1<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
remove_reference_t<
typename GridwiseGemm::
CGridDescriptor_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl>,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation,
Block2CTileMap,
false>;
ave_time = launch_and_time_kernel(
stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_mblock_mxdlperwave_mwavemperxdl_nblock_nxdlperwave_nwavenperxdl_,
arg.in_element_op_,
arg.wei_element_op_,
arg.out_element_op_,
arg.block_2_ctile_map_);
}
return ave_time;
}
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
if constexpr(ConvForwardSpecialization ==
ConvolutionForwardSpecialization::Filter1x1Stride1Pad0)
{
// check if it's 1x1, stride=1 conv
if(!(arg.filter_spatial_lengths_[0] == 1 && arg.filter_spatial_lengths_[1] == 1 &&
arg.conv_filter_strides_[0] == 1 && arg.conv_filter_strides_[1] == 1 &&
arg.input_left_pads_[0] == 0 && arg.input_left_pads_[1] == 0 &&
arg.input_right_pads_[0] == 0 && arg.input_right_pads_[1] == 0))
{
return false;
}
}
else if constexpr(ConvForwardSpecialization ==
ConvolutionForwardSpecialization::Filter1x1Pad0)
{
// check if it's 1x1 conv
if(!(arg.filter_spatial_lengths_[0] == 1 && arg.filter_spatial_lengths_[1] == 1 &&
arg.input_left_pads_[0] == 0 && arg.input_left_pads_[1] == 0 &&
arg.input_right_pads_[0] == 0 && arg.input_right_pads_[1] == 0))
{
return false;
}
}
// vector load A/B matrix from global memory
if(!(ABlockTransferSrcVectorDim == 2 && BBlockTransferSrcVectorDim == 2 &&
arg.Conv_C_ % ABlockTransferSrcScalarPerVector == 0 &&
arg.Conv_C_ % BBlockTransferSrcScalarPerVector == 0))
{
return false;
}
// vector store C matrix into global memory
if(!(arg.Conv_K_ % CBlockTransferScalarPerVector_NWaveNPerXdl == 0))
{
return false;
}
// Gridwise GEMM size
return GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_);
}
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const InDataType* p_in_grid,
const WeiDataType* p_wei_grid,
OutDataType* p_out_grid,
ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op)
{
return Argument{p_in_grid,
p_wei_grid,
p_out_grid,
N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
in_element_op,
wei_element_op,
out_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_in_grid,
const void* p_wei_grid,
void* p_out_grid,
ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op) override
{
return std::make_unique<Argument>(static_cast<const InDataType*>(p_in_grid),
static_cast<const WeiDataType*>(p_wei_grid),
static_cast<OutDataType*>(p_out_grid),
N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
in_element_op,
wei_element_op,
out_element_op);
}
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< K0PerBlock << ", "
<< getConvForwardSpecializationString(ConvForwardSpecialization)
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

733
include/ck/tensor_operation/gpu/device/impl/device_conv2d_fwd_xdl_nhwc_kyxc_nhwk.hpp Normal file
View File

@@ -0,0 +1,733 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_conv_fwd.hpp"
#include "ck/tensor_operation/gpu/device/convolution_forward_specialization.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v2r3.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
// out[N, Ho, Wo, K] = in[N, Hi, Wi, C] * wei[K, Y, X, C]
template <typename InDataType,
typename WeiDataType,
typename OutDataType,
typename AccDataType,
typename InElementwiseOperation,
typename WeiElementwiseOperation,
typename OutElementwiseOperation,
ConvolutionForwardSpecialization ConvForwardSpecialization,
ck::index_t BlockSize,
ck::index_t MPerBlock,
ck::index_t NPerBlock,
ck::index_t K0PerBlock,
ck::index_t K1,
ck::index_t MPerXDL,
ck::index_t NPerXDL,
ck::index_t MXdlPerWave,
ck::index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_K0_M_K1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
ck::index_t ABlockTransferSrcVectorDim,
ck::index_t ABlockTransferSrcScalarPerVector,
ck::index_t ABlockTransferDstScalarPerVector_K1,
bool ABlockLdsAddExtraM,
typename BBlockTransferThreadClusterLengths_K0_N_K1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
ck::index_t BBlockTransferSrcVectorDim,
ck::index_t BBlockTransferSrcScalarPerVector,
ck::index_t BBlockTransferDstScalarPerVector_K1,
bool BBlockLdsAddExtraN,
ck::index_t CThreadTransferSrcDstVectorDim,
ck::index_t CThreadTransferDstScalarPerVector>
struct DeviceConv2dFwdXdl_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
: public DeviceConvFwd<2,
ck::tensor_layout::convolution::NHWC,
ck::tensor_layout::convolution::KYXC,
ck::tensor_layout::convolution::NHWK,
InDataType,
WeiDataType,
OutDataType,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation>
{
using DeviceOp = DeviceConv2dFwdXdl_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K;
using ADataType = InDataType;
using BDataType = WeiDataType;
using CDataType = OutDataType;
// TODO make A/B datatype different
using ABDataType = InDataType;
static constexpr index_t NDimSpatial = 2;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr auto K1Number = Number<K1>{};
static constexpr auto GemmK1Number = K1Number;
static auto
MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads)
{
using namespace ck;
const index_t Hi = input_spatial_lengths[0];
const index_t Wi = input_spatial_lengths[1];
const index_t Ho = output_spatial_lengths[0];
const index_t Wo = output_spatial_lengths[1];
const index_t Y = filter_spatial_lengths[0];
const index_t X = filter_spatial_lengths[1];
const index_t ConvStrideH = conv_filter_strides[0];
const index_t ConvStrideW = conv_filter_strides[1];
const index_t ConvDilationH = conv_filter_dilations[0];
const index_t ConvDilationW = conv_filter_dilations[1];
const index_t InLeftPadH = input_left_pads[0];
const index_t InLeftPadW = input_left_pads[1];
const index_t InRightPadH = input_right_pads[0];
const index_t InRightPadW = input_right_pads[1];
const index_t GemmMRaw = N * Ho * Wo;
const index_t GemmN = K;
const index_t GemmK = Y * X * C;
const auto GemmMPad = math::integer_least_multiple(GemmMRaw, MPerBlock) - GemmMRaw;
assert(GemmK % GemmK1Number == 0);
const index_t GemmK0 = GemmK / GemmK1Number;
if constexpr(ConvForwardSpecialization ==
ConvolutionForwardSpecialization::Filter1x1Stride1Pad0)
{
// A: input tensor
const auto in_gemmmraw_gemmk_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, C));
const auto in_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmmraw_gemmk_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_right_pad_transform(GemmMRaw, GemmMPad)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// B: weight tensor
const auto wei_gemmn_gemmk_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(K, C));
const auto wei_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
wei_gemmn_gemmk_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// C: output tensor
const auto out_gemmmraw_gemmn_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, K));
const auto out_gemmm_gemmn_grid_desc =
transform_tensor_descriptor(out_gemmmraw_gemmn_grid_desc,
make_tuple(make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return make_tuple(in_gemmk0_gemmm_gemmk1_grid_desc,
wei_gemmk0_gemmn_gemmk1_grid_desc,
out_gemmm_gemmn_grid_desc);
}
else if constexpr(ConvForwardSpecialization ==
ConvolutionForwardSpecialization::Filter1x1Pad0)
{
// A: input tensor
const auto in_n_hi_wi_c_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N, Hi, Wi, C));
const auto in_n_ho_wo_c_grid_desc = transform_tensor_descriptor(
in_n_hi_wi_c_grid_desc,
make_tuple(make_pass_through_transform(N),
make_embed_transform(make_tuple(Ho), make_tuple(ConvStrideH)),
make_embed_transform(make_tuple(Wo), make_tuple(ConvStrideW)),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
const auto in_gemmk0_gemmmraw_gemmk1_grid_desc = transform_tensor_descriptor(
in_n_ho_wo_c_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_merge_transform(make_tuple(N, Ho, Wo))),
make_tuple(Sequence<3>{}, Sequence<0, 1, 2>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
const auto in_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmk0_gemmmraw_gemmk1_grid_desc,
make_tuple(make_pass_through_transform(GemmK0),
make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmK1Number)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
// B: weight tensor
const auto wei_gemmn_gemmk_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(K, C));
const auto wei_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
wei_gemmn_gemmk_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// C: output tensor
const auto out_gemmmraw_gemmn_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, K));
const auto out_gemmm_gemmn_grid_desc =
transform_tensor_descriptor(out_gemmmraw_gemmn_grid_desc,
make_tuple(make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return make_tuple(in_gemmk0_gemmm_gemmk1_grid_desc,
wei_gemmk0_gemmn_gemmk1_grid_desc,
out_gemmm_gemmn_grid_desc);
}
else
{
// A: input tensor
const auto in_n_hi_wi_c_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N, Hi, Wi, C));
const auto in_n_hip_wip_c_grid_desc = transform_tensor_descriptor(
in_n_hi_wi_c_grid_desc,
make_tuple(make_pass_through_transform(N),
make_pad_transform(Hi, InLeftPadH, InRightPadH),
make_pad_transform(Wi, InLeftPadW, InRightPadW),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
const auto in_n_y_ho_x_wo_c_grid_desc = transform_tensor_descriptor(
in_n_hip_wip_c_grid_desc,
make_tuple(
make_pass_through_transform(N),
make_embed_transform(make_tuple(Y, Ho), make_tuple(ConvDilationH, ConvStrideH)),
make_embed_transform(make_tuple(X, Wo), make_tuple(ConvDilationW, ConvStrideW)),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1, 2>{}, Sequence<3, 4>{}, Sequence<5>{}));
const auto in_gemmk_gemmmraw_grid_desc =
transform_tensor_descriptor(in_n_y_ho_x_wo_c_grid_desc,
make_tuple(make_merge_transform(make_tuple(Y, X, C)),
make_merge_transform(make_tuple(N, Ho, Wo))),
make_tuple(Sequence<1, 3, 5>{}, Sequence<0, 2, 4>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto in_gemmk0_gemmmraw_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmk_gemmmraw_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmMRaw)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
const auto in_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
in_gemmk0_gemmmraw_gemmk1_grid_desc,
make_tuple(make_pass_through_transform(GemmK0),
make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmK1Number)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
// B: weight tensor
const auto wei_k_yxc_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(K, Y * X * C));
const auto wei_gemmk_gemmn_grid_desc = transform_tensor_descriptor(
wei_k_yxc_grid_desc,
make_tuple(make_pass_through_transform(K), make_pass_through_transform(Y * X * C)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<1>{}, Sequence<0>{}));
const auto wei_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
wei_gemmk_gemmn_grid_desc,
make_tuple(make_unmerge_transform(make_tuple(GemmK0, GemmK1Number)),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
// C: output tensor
const auto out_nhowo_k_grid_desc =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo, K));
const auto out_gemmmraw_gemmn_grid_desc =
transform_tensor_descriptor(out_nhowo_k_grid_desc,
make_tuple(make_pass_through_transform(N * Ho * Wo),
make_pass_through_transform(K)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto out_gemmm_gemmn_grid_desc =
transform_tensor_descriptor(out_gemmmraw_gemmn_grid_desc,
make_tuple(make_right_pad_transform(GemmMRaw, GemmMPad),
make_pass_through_transform(GemmN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return make_tuple(in_gemmk0_gemmm_gemmk1_grid_desc,
wei_gemmk0_gemmn_gemmk1_grid_desc,
out_gemmm_gemmn_grid_desc);
}
}
using ABCGridDescs = decltype(MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(
1, 1, 1, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}));
using AGridDesc_K0_M_K1 = remove_cvref_t<decltype(ABCGridDescs{}[I0])>;
using BGridDesc_K0_N_K1 = remove_cvref_t<decltype(ABCGridDescs{}[I1])>;
using CGridDesc_M_N = remove_cvref_t<decltype(ABCGridDescs{}[I2])>;
// GridwiseGemm
using GridwiseGemm = GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r3<
BlockSize,
ABDataType, // TODO: distinguish A/B datatype
AccDataType,
CDataType,
InMemoryDataOperationEnum::Set,
AGridDesc_K0_M_K1,
BGridDesc_K0_N_K1,
CGridDesc_M_N,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation,
MPerBlock,
NPerBlock,
K0PerBlock,
MPerXDL,
NPerXDL,
K1,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_K0_M_K1,
Sequence<1, 0, 2>, // ABlockTransferThreadClusterArrangeOrder,
Sequence<1, 0, 2>, // ABlockTransferSrcAccessOrder,
2, // ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_K1,
false, // AThreadTransferSrcResetCoordinateAfterRun,
ABlockLdsAddExtraM,
BBlockTransferThreadClusterLengths_K0_N_K1,
Sequence<1, 0, 2>, // BBlockTransferThreadClusterArrangeOrder,
Sequence<1, 0, 2>, // BBlockTransferSrcAccessOrder,
2, // BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_K1,
false, // BThreadTransferSrcResetCoordinateAfterRun,
BBlockLdsAddExtraN,
Sequence<2, 3, 0, 1, 7, 5, 4, 6>, // CThreadTransferSrcDstAccessOrder,
7, // CThreadTransferSrcDstVectorDim,
CThreadTransferDstScalarPerVector>;
// Argument
struct Argument : public BaseArgument
{
Argument(const InDataType* p_in_grid,
const WeiDataType* p_wei_grid,
OutDataType* p_out_grid,
ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
ck::index_t M01,
ck::index_t N01,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op)
: p_a_grid_{p_in_grid},
p_b_grid_{p_wei_grid},
p_c_grid_{p_out_grid},
a_grid_desc_k0_m_k1_{},
b_grid_desc_k0_n_k1_{},
c_grid_desc_m_n_{},
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_{},
block_2_ctile_map_{},
M01_{M01},
N01_{N01},
in_element_op_{in_element_op},
wei_element_op_{wei_element_op},
out_element_op_{out_element_op},
Conv_N_{N},
Conv_K_{K},
Conv_C_{C},
filter_spatial_lengths_{filter_spatial_lengths},
conv_filter_strides_{conv_filter_strides},
input_left_pads_{input_left_pads},
input_right_pads_{input_right_pads}
{
const auto descs =
DeviceOp::MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads);
a_grid_desc_k0_m_k1_ = descs[I0];
b_grid_desc_k0_n_k1_ = descs[I1];
c_grid_desc_m_n_ = descs[I2];
block_2_ctile_map_ =
GridwiseGemm::MakeDefaultBlock2CTileMap(c_grid_desc_m_n_, M01, N01);
if(GridwiseGemm::CheckValidity(a_grid_desc_k0_m_k1_,
b_grid_desc_k0_n_k1_,
c_grid_desc_m_n_,
block_2_ctile_map_))
{
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_ =
GridwiseGemm::MakeCGridDescriptor_M0_N0_M1_N1_M2_M3_M4_N2(c_grid_desc_m_n_);
}
}
// private:
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
CDataType* p_c_grid_;
AGridDesc_K0_M_K1 a_grid_desc_k0_m_k1_;
BGridDesc_K0_N_K1 b_grid_desc_k0_n_k1_;
CGridDesc_M_N c_grid_desc_m_n_;
typename GridwiseGemm::CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_;
typename GridwiseGemm::DefaultBlock2CTileMap block_2_ctile_map_;
index_t M01_;
index_t N01_;
InElementwiseOperation in_element_op_;
WeiElementwiseOperation wei_element_op_;
OutElementwiseOperation out_element_op_;
// for checking IsSupportedArgument()
index_t Conv_N_;
index_t Conv_K_;
index_t Conv_C_;
std::vector<index_t> filter_spatial_lengths_;
std::vector<index_t> conv_filter_strides_;
std::vector<index_t> input_left_pads_;
std::vector<index_t> input_right_pads_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
#if 0
{
std::cout << "arg.a_grid_desc_k0_m_k1_{" << arg.a_grid_desc_k0_m_k1_.GetLength(I0)
<< ", " << arg.a_grid_desc_k0_m_k1_.GetLength(I1) << ", "
<< arg.a_grid_desc_k0_m_k1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.b_grid_desc_k0_n_k1_{" << arg.b_grid_desc_k0_n_k1_.GetLength(I0)
<< ", " << arg.b_grid_desc_k0_n_k1_.GetLength(I1) << ", "
<< arg.b_grid_desc_k0_n_k1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.c_grid_desc_m_n_{ " << arg.c_grid_desc_m_n_.GetLength(I0) << ", "
<< arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
}
#endif
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_))
{
throw std::runtime_error(
"wrong! GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r3 has invalid setting");
}
const index_t grid_size =
arg.block_2_ctile_map_.CalculateGridSize(arg.c_grid_desc_m_n_);
const auto K =
arg.a_grid_desc_k0_m_k1_.GetLength(I0) * arg.a_grid_desc_k0_m_k1_.GetLength(I2);
float ave_time = 0;
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
const auto kernel = kernel_gemm_xdlops_v2r3<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
remove_reference_t<typename GridwiseGemm::CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2>,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation,
remove_reference_t<typename GridwiseGemm::DefaultBlock2CTileMap>,
true>;
ave_time = launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_,
arg.in_element_op_,
arg.wei_element_op_,
arg.out_element_op_,
arg.block_2_ctile_map_);
}
else
{
const auto kernel = kernel_gemm_xdlops_v2r3<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
remove_reference_t<typename GridwiseGemm::CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2>,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation,
remove_reference_t<typename GridwiseGemm::DefaultBlock2CTileMap>,
false>;
ave_time = launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_,
arg.in_element_op_,
arg.wei_element_op_,
arg.out_element_op_,
arg.block_2_ctile_map_);
}
return ave_time;
}
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
if constexpr(ConvForwardSpecialization ==
ConvolutionForwardSpecialization::Filter1x1Stride1Pad0)
{
// check if it's 1x1, stride=1 conv
if(!(arg.filter_spatial_lengths_[0] == 1 && arg.filter_spatial_lengths_[1] == 1 &&
arg.conv_filter_strides_[0] == 1 && arg.conv_filter_strides_[1] == 1 &&
arg.input_left_pads_[0] == 0 && arg.input_left_pads_[1] == 0 &&
arg.input_right_pads_[0] == 0 && arg.input_right_pads_[1] == 0))
{
return false;
}
}
else if constexpr(ConvForwardSpecialization ==
ConvolutionForwardSpecialization::Filter1x1Pad0)
{
// check if it's 1x1 conv
if(!(arg.filter_spatial_lengths_[0] == 1 && arg.filter_spatial_lengths_[1] == 1 &&
arg.input_left_pads_[0] == 0 && arg.input_left_pads_[1] == 0 &&
arg.input_right_pads_[0] == 0 && arg.input_right_pads_[1] == 0))
{
return false;
}
}
// vector load A/B matrix from global memory
if(!(ABlockTransferSrcVectorDim == 2 && BBlockTransferSrcVectorDim == 2 &&
arg.Conv_C_ % ABlockTransferSrcScalarPerVector == 0 &&
arg.Conv_C_ % BBlockTransferSrcScalarPerVector == 0))
{
return false;
}
// vector store C matrix into global memory
if(!(arg.Conv_K_ % CThreadTransferDstScalarPerVector == 0))
{
return false;
}
// Gridwise GEMM size
return GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_);
}
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const InDataType* p_in_grid,
const WeiDataType* p_wei_grid,
OutDataType* p_out_grid,
ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op)
{
return Argument{p_in_grid,
p_wei_grid,
p_out_grid,
N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
1,
1,
in_element_op,
wei_element_op,
out_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_in_grid,
const void* p_wei_grid,
void* p_out_grid,
ck::index_t N,
ck::index_t K,
ck::index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op) override
{
return std::make_unique<Argument>(static_cast<const InDataType*>(p_in_grid),
static_cast<const WeiDataType*>(p_wei_grid),
static_cast<OutDataType*>(p_out_grid),
N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
1,
1,
in_element_op,
wei_element_op,
out_element_op);
}
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceConv2dFwdXdl_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< K0PerBlock << ", "
<< getConvForwardSpecializationString(ConvForwardSpecialization)
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

268
include/ck/tensor_operation/gpu/device/impl/device_conv3d_fwd_naive_ndhwc_kzyxc_ndhwk.hpp Normal file
View File

@@ -0,0 +1,268 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#ifndef DEVICE_CONV3D_FWD_NAIVE_HPP
#define DEVICE_CONV3D_FWD_NAIVE_HPP
#include <iostream>
#include <memory>
#include <sstream>
#include "conv_util.hpp"
#include "device.hpp"
#include "device_conv_fwd.hpp"
#include "common_header.hpp"
#include "naive_conv_fwd.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
// specialization for #D conv: in[n, di, hi, wi, c] * wei[k, z, y, x, c] = out[n, do, ho, wo, k]
template <typename InDataType,
typename WeiDataType, // WeiDataType must be the same as InDataType
typename OutDataType,
typename AccDataType,
typename InElementwiseOperation,
typename WeiElementwiseOperation,
typename OutElementwiseOperation>
struct DeviceConv3dFwdNaive_Input_N_Di_Hi_Wi_C_Weight_K_Z_Y_X_C_Output_N_Do_Ho_Wo_K
: public DeviceConvFwd<InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation>
{
using DeviceOp = DeviceConv3dFwdNaive_Input_N_Di_Hi_Wi_C_Weight_K_Z_Y_X_C_Output_N_Do_Ho_Wo_K;
using ADataType = InDataType;
using BDataType = WeiDataType;
using CDataType = OutDataType;
// TODO make A/B datatype different
using ABDataType = InDataType;
// Argument
struct Argument : public BaseArgument
{
Argument(const InDataType* p_in,
const WeiDataType* p_wei,
OutDataType* p_out,
const index_t N,
const index_t K,
const index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op)
: params_{3,
N,
K,
C,
filter_spatial_lengths,
input_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads},
out_spatial_lengths_{output_spatial_lengths},
p_in_{p_in},
p_wei_{p_wei},
p_out_{p_out},
in_element_op_{in_element_op},
wei_element_op_{wei_element_op},
out_element_op_{out_element_op}
{
}
// private:
utils::conv::ConvParams params_;
std::vector<index_t> out_spatial_lengths_;
const InDataType* p_in_;
const WeiDataType* p_wei_;
OutDataType* p_out_;
InElementwiseOperation in_element_op_;
WeiElementwiseOperation wei_element_op_;
OutElementwiseOperation out_element_op_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
const auto naive_conv3d_fwd =
ref::naive_conv_fwd_ndhwc_kzyxc_ndhwk<InDataType,
WeiDataType,
OutDataType,
AccDataType,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation>;
float ave_time = launch_and_time_kernel(stream_config,
naive_conv3d_fwd,
dim3(256),
dim3(256),
0,
arg.p_in_,
arg.p_wei_,
arg.p_out_,
arg.N_,
arg.K_,
arg.C_,
arg.in_spatial_lengths_[0],
arg.in_spatial_lengths_[1],
arg.in_spatial_lengths_[2],
arg.filter_spatial_lengths_[0],
arg.filter_spatial_lengths_[1],
arg.filter_spatial_lengths_[2],
arg.out_spatial_lengths_[0],
arg.out_spatial_lengths_[1],
arg.out_spatial_lengths_[2],
arg.conv_filter_strides_[0],
arg.conv_filter_strides_[1],
arg.conv_filter_strides_[2],
arg.conv_filter_dilations_[0],
arg.conv_filter_dilations_[1],
arg.conv_filter_dilations_[2],
arg.in_left_pads_[0],
arg.in_left_pads_[1],
arg.in_left_pads_[2]);
return ave_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
std::vector<index_t> out_spatial_lengths = arg.params_.GetOutputSpatialLengths();
bool out_lengths_are_consistent = out_spatial_lengths[0] == arg.out_spatial_lengths_[0] &&
out_spatial_lengths[1] == arg.out_spatial_lengths_[1] &&
out_spatial_lengths[2] == arg.out_spatial_lengths_[2];
return out_lengths_are_consistent;
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const InDataType* p_in,
const WeiDataType* p_wei,
OutDataType* p_out,
const index_t N,
const index_t K,
const index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op)
{
return Argument{p_in,
p_wei,
p_out,
N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
in_element_op,
wei_element_op,
out_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_in,
const void* p_wei,
void* p_out,
const index_t N,
const index_t K,
const index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op) override
{
return std::make_unique<Argument>(static_cast<const InDataType*>(p_in),
static_cast<const WeiDataType*>(p_wei),
static_cast<OutDataType*>(p_out),
N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
in_element_op,
wei_element_op,
out_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceConv3dFwdNaive_Input_N_Di_Hi_Wi_C_Weight_K_Z_Y_X_C_Output_N_Do_Ho_Wo_K<>";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck
#endif

642
include/ck/tensor_operation/gpu/device/impl/device_conv3d_fwd_xdl_ndhwc_kzyxc_ndhwk.hpp Normal file
View File

@@ -0,0 +1,642 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#ifndef DEVICE_CONV3D_FWD_XDL_HPP
#define DEVICE_CONV3D_FWD_XDL_HPP
#include <iostream>
#include <memory>
#include <sstream>
#include "device.hpp"
#include "device_conv_fwd.hpp"
#include "common_header.hpp"
#include "tensor_layout.hpp"
#include "convolution_forward_specialization.hpp"
#include "tensor_descriptor.hpp"
#include "tensor_descriptor_helper.hpp"
#include "transform_forward_convolution3d_into_gemm_v4r4r4_ndhwc_kzyxc_ndhwk.hpp"
#include "gridwise_gemm_xdlops_v2r3.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
/*
* \see \link impl/device_batched_gemm_xdl.hpp kernel_batched_gemm_xdlops_v2r3() \endlink.
*/
template <typename GridwiseGemm,
typename FloatAB,
typename FloatC,
typename AGridDesc_K0_M_K1,
typename BGridDesc_K0_N_K1,
typename CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CElementwiseOperation,
typename Block2CTileMap,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_gemm_xdlops_v2r3_for_conv3d(
const FloatAB* __restrict__ p_a_grid,
const FloatAB* __restrict__ p_b_grid,
FloatC* __restrict__ p_c_grid,
const index_t num_batches,
const index_t a_batch_stride,
const index_t b_batch_stride,
const index_t c_batch_stride,
const AGridDesc_K0_M_K1 a_grid_desc_k0_m_k1,
const BGridDesc_K0_N_K1 b_grid_desc_k0_n_k1,
const CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2 c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const CElementwiseOperation c_element_op,
const Block2CTileMap block_2_ctile_map)
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__))
const index_t num_blocks_per_batch =
__builtin_amdgcn_readfirstlane(get_grid_size() / num_batches);
const index_t g_idx = __builtin_amdgcn_readfirstlane(get_block_1d_id() / num_blocks_per_batch);
const long_index_t a_batch_offset =
__builtin_amdgcn_readfirstlane(static_cast<long_index_t>(a_batch_stride) * g_idx);
const long_index_t b_batch_offset =
__builtin_amdgcn_readfirstlane(static_cast<long_index_t>(b_batch_stride) * g_idx);
const long_index_t c_batch_offset =
__builtin_amdgcn_readfirstlane(static_cast<long_index_t>(c_batch_stride) * g_idx);
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
GridwiseGemm::template Run<HasMainKBlockLoop>(p_a_grid + a_batch_offset,
p_b_grid + b_batch_offset,
p_c_grid + c_batch_offset,
p_shared,
a_grid_desc_k0_m_k1,
b_grid_desc_k0_n_k1,
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2,
a_element_op,
b_element_op,
c_element_op,
block_2_ctile_map);
#else
ignore = p_a_grid;
ignore = p_b_grid;
ignore = p_c_grid;
ignore = num_batches;
ignore = a_batch_stride;
ignore = b_batch_stride;
ignore = c_batch_stride;
ignore = a_grid_desc_k0_m_k1;
ignore = b_grid_desc_k0_n_k1;
ignore = c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2;
ignore = a_element_op;
ignore = b_element_op;
ignore = c_element_op;
ignore = block_2_ctile_map;
#endif // end of if (defined(__gfx908__) || defined(__gfx90a__))
}
// specialization for #D conv: in[n, di, hi, wi, c] * wei[k, z, y, x, c] = out[n, do, ho, wo, k]
template <typename InDataType,
typename WeiDataType, // WeiDataType must be the same as InDataType
typename OutDataType,
typename AccDataType,
typename InElementwiseOperation,
typename WeiElementwiseOperation,
typename OutElementwiseOperation,
ConvolutionForwardSpecialization ConvForwardSpecialization,
ck::index_t BlockSize,
ck::index_t MPerBlock,
ck::index_t NPerBlock,
ck::index_t K0PerBlock,
ck::index_t K1,
ck::index_t MPerXDL,
ck::index_t NPerXDL,
ck::index_t MXdlPerWave,
ck::index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_K0_M_K1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
ck::index_t ABlockTransferSrcVectorDim,
ck::index_t ABlockTransferSrcScalarPerVector,
ck::index_t ABlockTransferDstScalarPerVector_K1,
bool ABlockLdsAddExtraM,
typename BBlockTransferThreadClusterLengths_K0_N_K1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
ck::index_t BBlockTransferSrcVectorDim,
ck::index_t BBlockTransferSrcScalarPerVector,
ck::index_t BBlockTransferDstScalarPerVector_K1,
bool BBlockLdsAddExtraN,
ck::index_t CThreadTransferSrcDstVectorDim,
ck::index_t CThreadTransferDstScalarPerVector>
struct DeviceConv3dFwdXdl_Input_N_Di_Hi_Wi_C_Weight_K_Z_Y_X_C_Output_N_Do_Ho_Wo_K
: public DeviceConvFwd<InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation>
{
using DeviceOp = DeviceConv3dFwdXdl_Input_N_Di_Hi_Wi_C_Weight_K_Z_Y_X_C_Output_N_Do_Ho_Wo_K;
using ADataType = InDataType;
using BDataType = WeiDataType;
using CDataType = OutDataType;
// TODO make A/B datatype different
using ABDataType = InDataType;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
/*
* \brief Split the number of batches, \p N, into N = B * N1, such that the memory
* space of input and output tensors stays with the value range of index_t, and each subbatch
* can be dealed with GridwiseGemm.
*/
static index_t GetMaxAllowableSubBatchSize(const index_t N,
const index_t K,
const index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths)
{
const index_t Di = input_spatial_lengths[0];
const index_t Hi = input_spatial_lengths[1];
const index_t Wi = input_spatial_lengths[2];
const index_t Do = output_spatial_lengths[0];
const index_t Ho = output_spatial_lengths[1];
const index_t Wo = output_spatial_lengths[2];
// N1 should satisfy that
// 1) N % N1 = 0;
// 2) N1 * (Do * Ho * Wo * K) < (2^31 - 1)
// 3) N1 * (Di * Hi * Wi * C) < (2^31 - 1)
//
// Do NOT confuse (B, N1) in this function with (B, N1) in gridewise GEMM.
auto N1 = N + 1;
const auto stride =
math::max(long_index_t(Do) * Ho * Wo * K, long_index_t(Di) * Hi * Wi * C);
const index_t max_stride = NumericLimits<index_t>::Max();
for(index_t n0 = 1; n0 <= N; ++n0)
{
index_t n1 = N / n0;
if(n0 * n1 == N && long_index_t(n1) * long_index_t(stride) < max_stride)
{
N1 = n1;
break;
}
}
const auto B = N / N1;
if(B * N1 != N)
{
throw std::runtime_error(__func__ +
std::string(": failed to find num_subbatches for conv3d.\n"));
}
return N1;
}
static auto
MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(const index_t N,
const index_t K,
const index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads)
{
assert(input_spatial_lengths.size() > 2);
assert(filter_spatial_lengths.size() > 2);
assert(conv_filter_strides.size() > 2);
assert(conv_filter_dilations.size() > 2);
assert(input_left_pads.size() > 2);
assert(input_right_pads.size() > 2);
const index_t Di = input_spatial_lengths[0];
const index_t Hi = input_spatial_lengths[1];
const index_t Wi = input_spatial_lengths[2];
const index_t Z = filter_spatial_lengths[0];
const index_t Y = filter_spatial_lengths[1];
const index_t X = filter_spatial_lengths[2];
const index_t Do = output_spatial_lengths[0];
const index_t Ho = output_spatial_lengths[1];
const index_t Wo = output_spatial_lengths[2];
static_assert(ConvForwardSpecialization == ConvolutionForwardSpecialization::Default,
"Wrong! This specialization not implemented!");
const auto in_desc_n_di_hi_wi_c =
make_naive_tensor_descriptor_packed(make_tuple(N, Di, Hi, Wi, C));
const auto wei_desc_k_z_y_x_c =
make_naive_tensor_descriptor_packed(make_tuple(K, Z, Y, X, C));
const auto out_desc_n_do_ho_wo_k =
make_naive_tensor_descriptor_packed(make_tuple(N, Do, Ho, Wo, K));
const auto descs = transform_forward_convolution3d_into_gemm_v4r4r4_ndhwc_kzyxc_ndhwk_pad(
in_desc_n_di_hi_wi_c,
wei_desc_k_z_y_x_c,
out_desc_n_do_ho_wo_k,
make_tuple(conv_filter_strides[0], conv_filter_strides[1], conv_filter_strides[2]),
make_tuple(
conv_filter_dilations[0], conv_filter_dilations[1], conv_filter_dilations[2]),
make_tuple(input_left_pads[0], input_left_pads[1], input_left_pads[2]),
make_tuple(input_right_pads[0], input_right_pads[1], input_right_pads[2]),
Number<K1>{});
return descs;
}
using ABCGridDescs = remove_cvref_t<decltype(MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(
1, 1, 1, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}))>;
using AGridDesc_K0_M_K1 = remove_cvref_t<decltype(ABCGridDescs{}[I0])>;
using BGridDesc_K0_N_K1 = remove_cvref_t<decltype(ABCGridDescs{}[I1])>;
using CGridDesc_M_N = remove_cvref_t<decltype(ABCGridDescs{}[I2])>;
using GridwiseGemm = GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r3<
BlockSize,
InDataType,
AccDataType,
OutDataType,
InMemoryDataOperationEnum::Set,
AGridDesc_K0_M_K1,
BGridDesc_K0_N_K1,
CGridDesc_M_N,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation,
MPerBlock,
NPerBlock,
K0PerBlock,
MPerXDL,
NPerXDL,
K1,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_K0_M_K1,
Sequence<1, 0, 2>, // ABlockTransferThreadClusterArrangeOrder,
Sequence<1, 0, 2>, // ABlockTransferSrcAccessOrder,
2,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_K1,
false, // AThreadTransferSrcResetCoordinateAfterRun,
ABlockLdsAddExtraM,
BBlockTransferThreadClusterLengths_K0_N_K1,
Sequence<1, 0, 2>, // ABlockTransferThreadClusterArrangeOrder,
Sequence<1, 0, 2>, // ABlockTransferSrcAccessOrder,
2,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_K1,
false, // BThreadTransferSrcResetCoordinateAfterRun,
BBlockLdsAddExtraN,
Sequence<2, 3, 0, 1, 7, 5, 4, 6>,
7,
CThreadTransferDstScalarPerVector>;
using CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2 =
decltype(GridwiseGemm::MakeCGridDescriptor_M0_N0_M1_N1_M2_M3_M4_N2(CGridDesc_M_N{}));
using Block2CTileMap = typename GridwiseGemm::DefaultBlock2CTileMap;
// Argument
struct Argument : public BaseArgument
{
Argument(const InDataType* p_in,
const WeiDataType* p_wei,
OutDataType* p_out,
const index_t N,
const index_t K,
const index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
index_t M01,
index_t N01,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op)
: p_a_grid_{p_in},
p_b_grid_{p_wei},
p_c_grid_{p_out},
M01_{M01},
N01_{N01},
in_element_op_{in_element_op},
wei_element_op_{wei_element_op},
out_element_op_{out_element_op}
{
const index_t subbatch_size =
GetMaxAllowableSubBatchSize(N, K, C, input_spatial_lengths, output_spatial_lengths);
num_subbatches_ = N / subbatch_size;
const auto descs =
MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(subbatch_size,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads);
a_grid_desc_k0_m_k1_ = descs[I0];
b_grid_desc_k0_n_k1_ = descs[I1];
c_grid_desc_m_n_ = descs[I2];
block_2_ctile_map_ =
GridwiseGemm::MakeDefaultBlock2CTileMap(c_grid_desc_m_n_, M01, N01);
a_batch_stride_ = a_grid_desc_k0_m_k1_.GetElementSpaceSize();
b_batch_stride_ = 0;
c_batch_stride_ = c_grid_desc_m_n_.GetElementSpaceSize();
if(GridwiseGemm::CheckValidity(a_grid_desc_k0_m_k1_,
b_grid_desc_k0_n_k1_,
c_grid_desc_m_n_,
block_2_ctile_map_))
{
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_ =
GridwiseGemm::MakeCGridDescriptor_M0_N0_M1_N1_M2_M3_M4_N2(c_grid_desc_m_n_);
}
}
// private:
const InDataType* p_a_grid_;
const WeiDataType* p_b_grid_;
OutDataType* p_c_grid_;
index_t num_subbatches_;
index_t a_batch_stride_;
index_t b_batch_stride_;
index_t c_batch_stride_;
AGridDesc_K0_M_K1 a_grid_desc_k0_m_k1_;
BGridDesc_K0_N_K1 b_grid_desc_k0_n_k1_;
CGridDesc_M_N c_grid_desc_m_n_;
CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2 c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_;
Block2CTileMap block_2_ctile_map_;
index_t M01_;
index_t N01_;
InElementwiseOperation in_element_op_;
WeiElementwiseOperation wei_element_op_;
OutElementwiseOperation out_element_op_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
{
std::cout << "num_batches_of_GEMM = " << arg.num_subbatches_ << std::endl;
std::cout << "a_grid_desc_k0_m_k1{" << arg.a_grid_desc_k0_m_k1_.GetLength(I0)
<< ", " << arg.a_grid_desc_k0_m_k1_.GetLength(I1) << ", "
<< arg.a_grid_desc_k0_m_k1_.GetLength(I2) << "}" << std::endl;
std::cout << "b_grid_desc_k0_n_k1{" << arg.b_grid_desc_k0_n_k1_.GetLength(I0)
<< ", " << arg.b_grid_desc_k0_n_k1_.GetLength(I1) << ", "
<< arg.b_grid_desc_k0_n_k1_.GetLength(I2) << "}" << std::endl;
std::cout << "c_grid_desc_m_n{ " << arg.c_grid_desc_m_n_.GetLength(I0) << ", "
<< arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
}
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_))
{
throw std::runtime_error(
"wrong! GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r3 has invalid setting");
}
const index_t grid_size =
arg.block_2_ctile_map_.CalculateGridSize(arg.c_grid_desc_m_n_) *
arg.num_subbatches_;
const auto K0 = arg.a_grid_desc_k0_m_k1_.GetLength(I0);
const bool has_main_k0_block_loop = GridwiseGemm::CalculateHasMainK0BlockLoop(K0);
float ave_time = 0;
if(has_main_k0_block_loop)
{
const auto kernel = kernel_gemm_xdlops_v2r3_for_conv3d<
GridwiseGemm,
InDataType,
OutDataType,
remove_reference_t<AGridDesc_K0_M_K1>,
remove_reference_t<BGridDesc_K0_N_K1>,
remove_reference_t<CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2>,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation,
remove_reference_t<Block2CTileMap>,
true>;
ave_time = launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.num_subbatches_,
arg.a_batch_stride_,
arg.b_batch_stride_,
arg.c_batch_stride_,
arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_,
arg.in_element_op_,
arg.wei_element_op_,
arg.out_element_op_,
arg.block_2_ctile_map_);
}
else
{
const auto kernel = kernel_gemm_xdlops_v2r3_for_conv3d<
GridwiseGemm,
InDataType,
OutDataType,
remove_reference_t<AGridDesc_K0_M_K1>,
remove_reference_t<BGridDesc_K0_N_K1>,
remove_reference_t<CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2>,
InElementwiseOperation,
WeiElementwiseOperation,
OutElementwiseOperation,
remove_reference_t<Block2CTileMap>,
false>;
ave_time = launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.num_subbatches_,
arg.a_batch_stride_,
arg.b_batch_stride_,
arg.c_batch_stride_,
arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_,
arg.in_element_op_,
arg.wei_element_op_,
arg.out_element_op_,
arg.block_2_ctile_map_);
}
return ave_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
return GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const InDataType* p_in,
const WeiDataType* p_wei,
OutDataType* p_out,
const index_t N,
const index_t K,
const index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op)
{
return Argument{p_in,
p_wei,
p_out,
N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
1,
1,
in_element_op,
wei_element_op,
out_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_in,
const void* p_wei,
void* p_out,
const index_t N,
const index_t K,
const index_t C,
std::vector<ck::index_t> input_spatial_lengths,
std::vector<ck::index_t> filter_spatial_lengths,
std::vector<ck::index_t> output_spatial_lengths,
std::vector<ck::index_t> conv_filter_strides,
std::vector<ck::index_t> conv_filter_dilations,
std::vector<ck::index_t> input_left_pads,
std::vector<ck::index_t> input_right_pads,
InElementwiseOperation in_element_op,
WeiElementwiseOperation wei_element_op,
OutElementwiseOperation out_element_op) override
{
return std::make_unique<Argument>(static_cast<const InDataType*>(p_in),
static_cast<const WeiDataType*>(p_wei),
static_cast<OutDataType*>(p_out),
N,
K,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
1,
1,
in_element_op,
wei_element_op,
out_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceConv3dFwdXdl_Input_N_Di_Hi_Wi_C_Weight_K_Z_Y_X_C_Output_N_Do_Ho_Wo_K"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< K0PerBlock
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck
#endif

1568
include/ck/tensor_operation/gpu/device/impl/device_convnd_bwd_data_nwc_kxc_nwk_xdl.hpp Normal file
View File

File diff suppressed because it is too large Load Diff

1105
include/ck/tensor_operation/gpu/device/impl/device_convnd_bwd_weight_nwc_kxc_nwk_xdl_cshuffle.hpp Normal file
View File

File diff suppressed because it is too large Load Diff

304
include/ck/tensor_operation/gpu/device/impl/device_elementwise.hpp Normal file
View File

@@ -0,0 +1,304 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/math.hpp"
#include "ck/utility/sequence.hpp"
#include "ck/tensor_operation/gpu/device/device_elementwise_base.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_elementwise_1d.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <typename InDataTypeTuple,
typename OutDataTypeTuple,
typename ElementwiseOperation,
index_t NumDim,
index_t MPerThread,
typename InScalarPerVectorSeq,
typename OutScalarPerVectorSeq>
struct DeviceElementwise
: public DeviceElementwiseBase<InDataTypeTuple, OutDataTypeTuple, ElementwiseOperation, NumDim>
{
static constexpr int NumInput = InDataTypeTuple::Size();
static constexpr int NumOutput = OutDataTypeTuple::Size();
static_assert(NumInput == InScalarPerVectorSeq::Size() &&
NumOutput == OutScalarPerVectorSeq::Size(),
"Tuple size is inconsistent with the number of in/out!");
static auto GenerateInDataTypePointerTuple()
{
return generate_tuple(
[&](auto I) {
using DataType = remove_cvref_t<decltype(InDataTypeTuple{}[I])>;
return static_cast<const DataType*>(nullptr);
},
Number<NumInput>{});
};
static auto GenerateOutDataTypePointerTuple()
{
return generate_tuple(
[&](auto I) {
using DataType = remove_cvref_t<decltype(OutDataTypeTuple{}[I])>;
return static_cast<DataType*>(nullptr);
},
Number<NumOutput>{});
};
using InDataTypePointerTuple = decltype(GenerateInDataTypePointerTuple());
using OutDataTypePointerTuple = decltype(GenerateOutDataTypePointerTuple());
template <typename Desc_M>
static auto PadDescriptor_M_1d(Desc_M desc_m, index_t gridSize, index_t blockSize)
{
constexpr auto I0 = Number<0>{};
const auto m = desc_m.GetLength(I0);
const index_t loop_step = gridSize * blockSize * MPerThread;
const auto pad = math::integer_least_multiple(m, loop_step) - m;
const auto desc_m_pad =
transform_tensor_descriptor(desc_m,
make_tuple(make_right_pad_transform(m, pad)),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0>{}));
return desc_m_pad;
}
static auto MakeDescriptor_M(const std::array<index_t, NumDim>& lengths,
const std::array<index_t, NumDim>& stride,
index_t gridSize,
index_t blockSize)
{
auto tupleOfShape = generate_tuple([&](auto I) { return lengths[I]; }, Number<NumDim>{});
auto tupleOfStride = generate_tuple([&](auto I) { return stride[I]; }, Number<NumDim>{});
// nd desc - [s0, s1, s2, ...]
const auto desc = make_naive_tensor_descriptor(tupleOfShape, tupleOfStride);
// merge nd to 1d desc - [s0 * s1 * ...]
if constexpr(NumDim > 1)
{
const auto desc_m = transform_tensor_descriptor(
desc,
make_tuple(make_merge_transform(tupleOfShape)),
make_tuple(generate_sequence_v2([&](auto I) { return I; }, Number<NumDim>{})),
make_tuple(Sequence<0>{}));
return PadDescriptor_M_1d(desc_m, gridSize, blockSize);
}
else
return PadDescriptor_M_1d(desc, gridSize, blockSize);
}
template <index_t TupleSize>
static auto GenerateInOutGrid1dDescTuple(Number<TupleSize>)
{
return generate_tuple(
[&](auto) {
if constexpr(NumDim > 1)
{
return MakeDescriptor_M({1, 1}, {1, 1}, 1, 1);
}
else
{
return MakeDescriptor_M({1}, {1}, 1, 1);
};
},
Number<TupleSize>{});
};
using InGrid1dDescTuple = decltype(GenerateInOutGrid1dDescTuple(Number<NumInput>{}));
using OutGrid1dDescTuple = decltype(GenerateInOutGrid1dDescTuple(Number<NumOutput>{}));
using GridwiseElementwise = GridwiseElementwise_1D<InGrid1dDescTuple,
OutGrid1dDescTuple,
InDataTypePointerTuple,
OutDataTypePointerTuple,
ElementwiseOperation,
MPerThread,
InScalarPerVectorSeq,
OutScalarPerVectorSeq>;
struct Argument : public BaseArgument
{
Argument(const std::array<index_t, NumDim> lengths,
const std::array<std::array<index_t, NumDim>, NumInput> inStridesArray,
const std::array<std::array<index_t, NumDim>, NumOutput> outStridesArray,
const std::array<const void*, NumInput> in_dev_buffers,
const std::array<void*, NumOutput> out_dev_buffers,
ElementwiseOperation elementwise_op)
: lengths_(lengths),
inStridesArray_(inStridesArray),
outStridesArray_(outStridesArray),
elementwise_op_(elementwise_op),
blockSize_(256),
gridSize_(120) // FIXME - Calculate the grid size by number of CU in the future
{
in_dev_buffers_ = generate_tuple(
[&](auto I) {
using DataType = remove_cvref_t<decltype(InDataTypeTuple{}[I])>;
return static_cast<const DataType*>(in_dev_buffers[I.value]);
},
Number<NumInput>{});
out_dev_buffers_ = generate_tuple(
[&](auto I) {
using DataType = remove_cvref_t<decltype(OutDataTypeTuple{}[I])>;
return static_cast<DataType*>(out_dev_buffers[I.value]);
},
Number<NumOutput>{});
in_grid_1d_desc_tuple_ = generate_tuple(
[&](auto I) {
return MakeDescriptor_M(
lengths, inStridesArray[I.value], gridSize_, blockSize_);
},
Number<NumInput>{});
out_grid_1d_desc_tuple_ = generate_tuple(
[&](auto I) {
return MakeDescriptor_M(
lengths, outStridesArray[I.value], gridSize_, blockSize_);
},
Number<NumOutput>{});
}
InDataTypePointerTuple in_dev_buffers_;
OutDataTypePointerTuple out_dev_buffers_;
InGrid1dDescTuple in_grid_1d_desc_tuple_;
OutGrid1dDescTuple out_grid_1d_desc_tuple_;
std::array<index_t, NumDim> lengths_;
std::array<std::array<index_t, NumDim>, NumInput> inStridesArray_;
std::array<std::array<index_t, NumDim>, NumOutput> outStridesArray_;
ElementwiseOperation elementwise_op_;
index_t blockSize_;
index_t gridSize_;
};
struct Invoker : public BaseInvoker
{
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
const auto kernel = kernel_elementwise_1d<GridwiseElementwise,
InGrid1dDescTuple,
OutGrid1dDescTuple,
InDataTypePointerTuple,
OutDataTypePointerTuple,
ElementwiseOperation>;
float elapsed_time = launch_and_time_kernel(stream_config,
kernel,
dim3(arg.gridSize_),
dim3(arg.blockSize_),
0,
arg.in_grid_1d_desc_tuple_,
arg.out_grid_1d_desc_tuple_,
arg.in_dev_buffers_,
arg.out_dev_buffers_,
arg.elementwise_op_);
return elapsed_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static bool IsSupportedArgument(const Argument& arg)
{
if(arg.lengths_.back() % MPerThread != 0)
return false;
auto IsScalarPerVectorValid = [&](const std::array<index_t, NumDim>& lengths,
const std::array<index_t, NumDim>& strides,
index_t scalarPerVector) {
if(strides.back() == 1 && lengths.back() % scalarPerVector == 0)
return true;
if(strides.back() != 1 && scalarPerVector == 1)
return true;
return false;
};
bool valid = true;
static_for<0, NumInput, 1>{}([&](auto I) {
if(!IsScalarPerVectorValid(
arg.lengths_, arg.inStridesArray_[I.value], InScalarPerVectorSeq::At(I)))
valid = false;
});
static_for<0, NumOutput, 1>{}([&](auto I) {
if(!IsScalarPerVectorValid(
arg.lengths_, arg.outStridesArray_[I.value], OutScalarPerVectorSeq::At(I)))
valid = false;
});
return valid;
};
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto
MakeArgument(const std::array<index_t, NumDim> lengths,
const std::array<std::array<index_t, NumDim>, NumInput> inStridesArray,
const std::array<std::array<index_t, NumDim>, NumOutput> outStridesArray,
const std::array<const void*, NumInput> in_dev_buffers,
const std::array<void*, NumOutput> out_dev_buffers,
ElementwiseOperation elementwise_op)
{
return Argument{lengths,
inStridesArray,
outStridesArray,
in_dev_buffers,
out_dev_buffers,
elementwise_op};
}
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const std::array<index_t, NumDim> lengths,
const std::array<std::array<index_t, NumDim>, NumInput> inStridesArray,
const std::array<std::array<index_t, NumDim>, NumOutput> outStridesArray,
const std::array<const void*, NumInput> in_dev_buffers,
const std::array<void*, NumOutput> out_dev_buffers,
ElementwiseOperation elementwise_op) override
{
return std::make_unique<Argument>(lengths,
inStridesArray,
outStridesArray,
in_dev_buffers,
out_dev_buffers,
elementwise_op);
}
static auto MakeInvoker() { return Invoker{}; }
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>();
};
}; // namespace device
} // namespace device
} // namespace tensor_operation
} // namespace ck

875
include/ck/tensor_operation/gpu/device/impl/device_gemm_bias_add_reduce_xdl_cshuffle.hpp Normal file
View File

@@ -0,0 +1,875 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_gemm_reduce.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_bias_add_reduce_xdl_cshuffle_v1.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
// Note: inter-wave loop scheduler is rolled out to c-shuffle version first. Becuase non c-shuffle
// version currently has compiler issues with register spill which further causes validation
// failures.
template <typename ALayout,
typename BLayout,
typename CLayout,
typename ADataType,
typename BDataType,
typename CDataType,
typename BiasDataType,
typename D0DataType,
typename GemmAccDataType,
typename CShuffleDataType,
typename ReduceAccDataType,
typename ReducePtrsGlobal,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CElementwiseOperation,
typename D0ElementwiseOperation,
typename ReduceOperations,
typename ReduceInElementwiseOperations,
typename ReduceAccElementwiseOperations,
typename ReduceGlobalMemoryDataOperation,
GemmSpecialization GemmSpec,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock,
index_t KPerBlock,
index_t AK1,
index_t BK1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
bool ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
bool BBlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CShuffleBlockTransferScalarPerVector_NPerBlock,
typename CReduceThreadClusterLengths_MPerBlock_NPerBlock,
index_t CReduceThreadLds2VGprCopySrcDstScalarPerVector_NPerBlock,
index_t CReduceThreadVgpr2GlobalCopySrcDstScalarPerVector_MPerBlock,
LoopScheduler LoopSched = make_default_loop_scheduler()>
struct DeviceGemmBiasAddReduce_Xdl_CShuffle : public DeviceGemmReduce<1, ReduceOperations::Size()>
{
using DeviceOp = DeviceGemmBiasAddReduce_Xdl_CShuffle;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static auto MakeAGridDescriptor_AK0_M_AK1(index_t MRaw, index_t KRaw, index_t StrideA)
{
const auto a_grid_desc_mraw_kraw = [&]() {
if constexpr(is_same_v<tensor_layout::gemm::RowMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(StrideA, I1));
}
else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(I1, StrideA));
}
}();
const auto M = math::integer_divide_ceil(MRaw, MPerBlock) * MPerBlock;
const auto K = math::integer_divide_ceil(KRaw, KPerBlock) * KPerBlock;
const auto MPad = M - MRaw;
const auto KPad = K - KRaw;
if constexpr(GemmSpec == GemmSpecialization::MKPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad both M and K
assert(K % AK1 == 0);
const auto AK0 = K / AK1;
const auto a_grid_desc_m_k =
transform_tensor_descriptor(a_grid_desc_mraw_kraw,
make_tuple(make_right_pad_transform(MRaw, MPad),
make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_pass_through_transform(M)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
else if constexpr(GemmSpec == GemmSpecialization::MPadding ||
GemmSpec == GemmSpecialization::MNPadding)
{
// pad M, but not K
assert(KRaw % AK1 == 0);
const auto AK0 = KRaw / AK1;
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_mraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_right_pad_transform(MRaw, MPad)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
else if constexpr(GemmSpec == GemmSpecialization::KPadding ||
GemmSpec == GemmSpecialization::NKPadding)
{
// pad K, but not M
assert(K % AK1 == 0);
const auto AK0 = K / AK1;
const auto a_grid_desc_m_k = transform_tensor_descriptor(
a_grid_desc_mraw_kraw,
make_tuple(make_pass_through_transform(MRaw), make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_pass_through_transform(MRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
else
{
// not pad M or K
assert(KRaw % AK1 == 0);
const auto AK0 = KRaw / AK1;
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_mraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_pass_through_transform(MRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
}
static auto MakeBGridDescriptor_BK0_N_BK1(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
const auto N = math::integer_divide_ceil(NRaw, NPerBlock) * NPerBlock;
const auto K = math::integer_divide_ceil(KRaw, KPerBlock) * KPerBlock;
const auto NPad = N - NRaw;
const auto KPad = K - KRaw;
if constexpr(GemmSpec == GemmSpecialization::NKPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad both N and K
assert(K % BK1 == 0);
const auto BK0 = K / BK1;
const auto b_grid_desc_n_k =
transform_tensor_descriptor(b_grid_desc_nraw_kraw,
make_tuple(make_right_pad_transform(NRaw, NPad),
make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_n_k,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_pass_through_transform(N)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
else if constexpr(GemmSpec == GemmSpecialization::NPadding ||
GemmSpec == GemmSpecialization::MNPadding)
{
// pad N, but not K
assert(KRaw % BK1 == 0);
const auto BK0 = KRaw / BK1;
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_nraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_right_pad_transform(NRaw, NPad)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
else if constexpr(GemmSpec == GemmSpecialization::KPadding ||
GemmSpec == GemmSpecialization::MKPadding)
{
// pad K, but not N
assert(K % BK1 == 0);
const auto BK0 = K / BK1;
const auto b_grid_desc_n_k = transform_tensor_descriptor(
b_grid_desc_nraw_kraw,
make_tuple(make_pass_through_transform(NRaw), make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_n_k,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_pass_through_transform(NRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
else
{
// not pad N or K
assert(KRaw % BK1 == 0);
const auto BK0 = KRaw / BK1;
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_nraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_pass_through_transform(NRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
}
static auto MakeCGridDescriptor_M_N(index_t MRaw, index_t NRaw, index_t StrideC)
{
const auto c_grid_desc_mraw_nraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(StrideC, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(I1, StrideC));
}
}();
const auto M = math::integer_divide_ceil(MRaw, MPerBlock) * MPerBlock;
const auto N = math::integer_divide_ceil(NRaw, NPerBlock) * NPerBlock;
const auto MPad = M - MRaw;
const auto NPad = N - NRaw;
if constexpr(GemmSpec == GemmSpecialization::MNPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad M and N
return transform_tensor_descriptor(c_grid_desc_mraw_nraw,
make_tuple(make_right_pad_transform(MRaw, MPad),
make_right_pad_transform(NRaw, NPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else if constexpr(GemmSpec == GemmSpecialization::MPadding ||
GemmSpec == GemmSpecialization::MKPadding)
{
// pad M, but not N
return transform_tensor_descriptor(
c_grid_desc_mraw_nraw,
make_tuple(make_right_pad_transform(MRaw, MPad), make_pass_through_transform(NRaw)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else if constexpr(GemmSpec == GemmSpecialization::NPadding ||
GemmSpec == GemmSpecialization::NKPadding)
{
// pad N, but not M
return transform_tensor_descriptor(
c_grid_desc_mraw_nraw,
make_tuple(make_pass_through_transform(MRaw), make_right_pad_transform(NRaw, NPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else
{
// not pad M or N
return c_grid_desc_mraw_nraw;
}
}
// assume D is packed tensor
static auto MakeReduceGridDescriptor_M(index_t MRaw)
{
const auto d_grid_desc_mraw = make_naive_tensor_descriptor_packed(make_tuple(MRaw));
const auto M = math::integer_divide_ceil(MRaw, MPerBlock) * MPerBlock;
const auto MPad = M - MRaw;
if constexpr(GemmSpec == GemmSpecialization::MPadding ||
GemmSpec == GemmSpecialization::MNPadding ||
GemmSpec == GemmSpecialization::MKPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad M
return transform_tensor_descriptor(d_grid_desc_mraw,
make_tuple(make_right_pad_transform(MRaw, MPad)),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0>{}));
}
else
{
// not pad M
return d_grid_desc_mraw;
}
}
using AGridDesc_AK0_M_AK1 = decltype(MakeAGridDescriptor_AK0_M_AK1(1, 1, 1));
using BGridDesc_BK0_N_BK1 = decltype(MakeBGridDescriptor_BK0_N_BK1(1, 1, 1));
using CGridDesc_M_N = decltype(MakeCGridDescriptor_M_N(1, 1, 1));
using C0GridDesc_M_N = decltype(MakeCGridDescriptor_M_N(1, 1, 0));
using C1GridDesc_M_N = decltype(MakeCGridDescriptor_M_N(1, 1, 1));
using ReduceGridDesc_M = decltype(MakeReduceGridDescriptor_M(1));
// GridwiseGemm
using GridwiseGemm = GridwiseGemmBiasAddReduce_k0mk1_k0nk1_mn_xdl_cshuffle_v1<
ADataType, // TODO: distinguish A/B datatype
GemmAccDataType,
CShuffleDataType,
CDataType,
BiasDataType,
D0DataType,
ReduceAccDataType,
ReducePtrsGlobal,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
D0ElementwiseOperation,
ReduceOperations,
ReduceInElementwiseOperations,
ReduceAccElementwiseOperations,
InMemoryDataOperationEnum::Set,
ReduceGlobalMemoryDataOperation,
AGridDesc_AK0_M_AK1,
BGridDesc_BK0_N_BK1,
CGridDesc_M_N,
C0GridDesc_M_N,
C1GridDesc_M_N,
ReduceGridDesc_M,
NumGemmKPrefetchStage,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
AK1,
BK1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
false,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
false,
BBlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CShuffleBlockTransferScalarPerVector_NPerBlock,
CReduceThreadClusterLengths_MPerBlock_NPerBlock,
CReduceThreadLds2VGprCopySrcDstScalarPerVector_NPerBlock,
CReduceThreadVgpr2GlobalCopySrcDstScalarPerVector_MPerBlock,
LoopSched>;
// Argument
struct Argument : public BaseArgument
{
Argument(const ADataType* p_a_grid,
const BDataType* p_b_grid,
CDataType* p_c_grid,
const BiasDataType* p_bias_grid,
const D0DataType* p_d0_grid,
ReducePtrsGlobal p_reduces_grid,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
index_t StrideC,
index_t StrideC1,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op,
D0ElementwiseOperation d0_element_op,
ReduceInElementwiseOperations reduce_in_element_ops,
ReduceAccElementwiseOperations reduce_out_element_ops)
: p_a_grid_{p_a_grid},
p_b_grid_{p_b_grid},
p_c_grid_{p_c_grid},
p_bias_grid_{p_bias_grid},
p_d0_grid_{p_d0_grid},
p_reduces_grid_{p_reduces_grid},
a_grid_desc_ak0_m_ak1_{DeviceOp::MakeAGridDescriptor_AK0_M_AK1(MRaw, KRaw, StrideA)},
b_grid_desc_bk0_n_bk1_{DeviceOp::MakeBGridDescriptor_BK0_N_BK1(KRaw, NRaw, StrideB)},
c_grid_desc_m_n_{DeviceOp::MakeCGridDescriptor_M_N(MRaw, NRaw, StrideC)},
c0_grid_desc_m_n_{DeviceOp::MakeCGridDescriptor_M_N(MRaw, NRaw, 0)},
c1_grid_desc_m_n_{DeviceOp::MakeCGridDescriptor_M_N(MRaw, NRaw, StrideC1)},
reduce_grid_desc_m_{DeviceOp::MakeReduceGridDescriptor_M(MRaw)},
c_grid_desc_mblock_mperblock_nblock_nperblock_{},
c0_grid_desc_mblock_mperblock_nblock_nperblock_{},
c1_grid_desc_mblock_mperblock_nblock_nperblock_{},
reduce_grid_desc_mblock_mperblock_{},
block_2_ctile_map_{GridwiseGemm::MakeDefaultBlock2CTileMap(c_grid_desc_m_n_)},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
c_element_op_{c_element_op},
d0_element_op_{d0_element_op},
reduce_in_element_ops_{reduce_in_element_ops},
reduce_out_element_ops_{reduce_out_element_ops}
{
if(GridwiseGemm::CheckValidity(a_grid_desc_ak0_m_ak1_,
b_grid_desc_bk0_n_bk1_,
c_grid_desc_m_n_,
block_2_ctile_map_))
{
c_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeCGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
c_grid_desc_m_n_);
c0_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeCGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
c0_grid_desc_m_n_);
c1_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeCGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
c1_grid_desc_m_n_);
reduce_grid_desc_mblock_mperblock_ =
GridwiseGemm::MakeReduceGridDescriptor_MBlock_MPerBlock(reduce_grid_desc_m_);
}
}
// private:
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
CDataType* p_c_grid_;
const BiasDataType* p_bias_grid_;
const D0DataType* p_d0_grid_;
ReducePtrsGlobal p_reduces_grid_;
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
CGridDesc_M_N c_grid_desc_m_n_;
C0GridDesc_M_N c0_grid_desc_m_n_;
C1GridDesc_M_N c1_grid_desc_m_n_;
ReduceGridDesc_M reduce_grid_desc_m_;
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
c_grid_desc_mblock_mperblock_nblock_nperblock_;
typename GridwiseGemm::C0GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
c0_grid_desc_mblock_mperblock_nblock_nperblock_;
typename GridwiseGemm::C1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
c1_grid_desc_mblock_mperblock_nblock_nperblock_;
typename GridwiseGemm::ReduceGridDescriptor_MBlock_MPerBlock
reduce_grid_desc_mblock_mperblock_;
typename GridwiseGemm::DefaultBlock2CTileMap block_2_ctile_map_;
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CElementwiseOperation c_element_op_;
D0ElementwiseOperation d0_element_op_;
ReduceInElementwiseOperations reduce_in_element_ops_;
ReduceAccElementwiseOperations reduce_out_element_ops_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_))
{
throw std::runtime_error("wrong! GridwiseGemm has invalid setting");
}
const index_t grid_size =
arg.block_2_ctile_map_.CalculateGridSize(arg.c_grid_desc_m_n_);
const auto K =
arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) * arg.a_grid_desc_ak0_m_ak1_.GetLength(I2);
float elapsed_time = 0.0f;
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
const auto kernel = kernel_gemm_bias_add_reduce_xdl_cshuffle_v1<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
BiasDataType,
D0DataType,
ReducePtrsGlobal,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
D0ElementwiseOperation,
ReduceInElementwiseOperations,
ReduceAccElementwiseOperations,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::C0GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::C1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::ReduceGridDescriptor_MBlock_MPerBlock,
typename GridwiseGemm::DefaultBlock2CTileMap,
true>;
elapsed_time =
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.p_bias_grid_,
arg.p_d0_grid_,
arg.p_reduces_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.d0_element_op_,
arg.reduce_in_element_ops_,
arg.reduce_out_element_ops_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.c0_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.c1_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.reduce_grid_desc_mblock_mperblock_,
arg.block_2_ctile_map_);
}
else
{
const auto kernel = kernel_gemm_bias_add_reduce_xdl_cshuffle_v1<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
BiasDataType,
D0DataType,
ReducePtrsGlobal,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
D0ElementwiseOperation,
ReduceInElementwiseOperations,
ReduceAccElementwiseOperations,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::C0GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::C1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::ReduceGridDescriptor_MBlock_MPerBlock,
typename GridwiseGemm::DefaultBlock2CTileMap,
false>;
elapsed_time =
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.p_bias_grid_,
arg.p_d0_grid_,
arg.p_reduces_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.d0_element_op_,
arg.reduce_in_element_ops_,
arg.reduce_out_element_ops_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.c0_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.c1_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.reduce_grid_desc_mblock_mperblock_,
arg.block_2_ctile_map_);
}
return elapsed_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
return GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static constexpr int NumReduce = ReduceOperations::Size();
static auto MakeArgument(const void* p_a,
const void* p_b,
const void* p_bias,
std::array<const void*, 1> p_ds,
void* p_c,
std::array<void*, NumReduce> p_reduces,
ck::index_t M,
ck::index_t N,
ck::index_t K,
ck::index_t StrideA,
ck::index_t StrideB,
ck::index_t StrideC,
std::array<ck::index_t, 1> StrideDs,
std::array<void*, 3> gemm_element_ops,
std::array<void*, 1> d_element_ops,
std::array<void*, NumReduce> reduce_in_element_op,
std::array<void*, NumReduce> reduce_out_element_op)
{
ReducePtrsGlobal reduce_tuple = generate_tuple(
[&](auto I) {
auto tmp = ReducePtrsGlobal{}[I];
using T = remove_pointer_t<decltype(tmp)>;
return static_cast<T*>(p_reduces[I]);
},
Number<NumReduce>{});
ReduceInElementwiseOperations reduce_in_element_ops = generate_tuple(
[&](auto I) {
auto tmp = ReduceInElementwiseOperations{}[I];
using T = remove_pointer_t<decltype(tmp)>;
return *(static_cast<T*>(reduce_in_element_op[I]));
},
Number<NumReduce>{});
ReduceAccElementwiseOperations reduce_out_element_ops = generate_tuple(
[&](auto I) {
auto tmp = ReduceAccElementwiseOperations{}[I];
using T = remove_pointer_t<decltype(tmp)>;
return *(static_cast<T*>(reduce_out_element_op[I]));
},
Number<NumReduce>{});
AElementwiseOperation a_element_op =
*(static_cast<AElementwiseOperation*>(gemm_element_ops[0]));
BElementwiseOperation b_element_op =
*(static_cast<BElementwiseOperation*>(gemm_element_ops[1]));
CElementwiseOperation c_element_op =
*(static_cast<CElementwiseOperation*>(gemm_element_ops[2]));
D0ElementwiseOperation d_element_op =
*(static_cast<D0ElementwiseOperation*>(d_element_ops[0]));
return Argument{static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
static_cast<CDataType*>(p_c),
static_cast<const BiasDataType*>(p_bias),
static_cast<const D0DataType*>(p_ds[0]),
reduce_tuple,
M,
N,
K,
StrideA,
StrideB,
StrideC,
StrideDs[0],
a_element_op,
b_element_op,
c_element_op,
d_element_op,
reduce_in_element_ops,
reduce_out_element_ops};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_a,
const void* p_b,
const void* p_bias,
std::array<const void*, 1> p_ds,
void* p_c,
std::array<void*, NumReduce> p_reduces,
ck::index_t M,
ck::index_t N,
ck::index_t K,
ck::index_t StrideA,
ck::index_t StrideB,
ck::index_t StrideC,
std::array<ck::index_t, 1> StrideDs,
std::array<void*, 3> gemm_element_ops,
std::array<void*, 1> d_element_ops,
std::array<void*, NumReduce> reduce_in_element_op,
std::array<void*, NumReduce> reduce_out_element_op,
index_t /* KBatch */ = 1) override
{
ReducePtrsGlobal reduce_tuple = generate_tuple(
[&](auto I) {
auto tmp = ReducePtrsGlobal{}[I];
using T = remove_pointer_t<decltype(tmp)>;
return static_cast<T*>(p_reduces[I]);
},
Number<NumReduce>{});
ReduceInElementwiseOperations reduce_in_element_ops = generate_tuple(
[&](auto I) {
auto tmp = ReduceInElementwiseOperations{}[I];
using T = remove_pointer_t<decltype(tmp)>;
return *(static_cast<T*>(reduce_in_element_op[I]));
},
Number<NumReduce>{});
ReduceAccElementwiseOperations reduce_out_element_ops = generate_tuple(
[&](auto I) {
auto tmp = ReduceAccElementwiseOperations{}[I];
using T = remove_pointer_t<decltype(tmp)>;
return *(static_cast<T*>(reduce_out_element_op[I]));
},
Number<NumReduce>{});
AElementwiseOperation a_element_op =
*(static_cast<AElementwiseOperation*>(gemm_element_ops[0]));
BElementwiseOperation b_element_op =
*(static_cast<BElementwiseOperation*>(gemm_element_ops[1]));
CElementwiseOperation c_element_op =
*(static_cast<CElementwiseOperation*>(gemm_element_ops[2]));
D0ElementwiseOperation d_element_op =
*(static_cast<D0ElementwiseOperation*>(d_element_ops[0]));
return std::make_unique<Argument>(static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
static_cast<CDataType*>(p_c),
static_cast<const BiasDataType*>(p_bias),
static_cast<const D0DataType*>(p_ds[0]),
reduce_tuple,
M,
N,
K,
StrideA,
StrideB,
StrideC,
StrideDs[0],
a_element_op,
b_element_op,
c_element_op,
d_element_op,
reduce_in_element_ops,
reduce_out_element_ops);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceGemmBiasAddReduce_Xdl_CShuffle"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< AK1 << ", "
<< BK1
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

572
include/ck/tensor_operation/gpu/device/impl/device_gemm_bias_e_permute_xdl.hpp Normal file
View File

@@ -0,0 +1,572 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_gemm_bias_e_permute.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
template <typename GridwiseGemm,
typename FloatAB,
typename FloatDsPointer,
typename FloatE,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
typename AGridDesc_AK0_M_AK1,
typename BGridDesc_BK0_N_BK1,
typename DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename Block2ETileMap,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_gemm_bias_e_permute(const FloatAB* __restrict__ p_a_grid,
const FloatAB* __restrict__ p_b_grid,
FloatDsPointer p_ds_grid,
FloatE* __restrict__ p_e_grid,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const CDEElementwiseOperation cde_element_op,
const AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1,
const BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1,
const DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
ds_grid_desc_mblock_mperblock_nblock_nperblock,
const EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
e_grid_desc_mblock_mperblock_nblock_nperblock,
const Block2ETileMap block_2_etile_map)
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__))
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
GridwiseGemm::template Run<HasMainKBlockLoop>(p_a_grid,
p_b_grid,
p_ds_grid,
p_e_grid,
p_shared,
a_element_op,
b_element_op,
cde_element_op,
a_grid_desc_ak0_m_ak1,
b_grid_desc_bk0_n_bk1,
ds_grid_desc_mblock_mperblock_nblock_nperblock,
e_grid_desc_mblock_mperblock_nblock_nperblock,
block_2_etile_map);
#else
ignore = p_a_grid;
ignore = p_b_grid;
ignore = p_ds_grid;
ignore = p_e_grid;
ignore = a_element_op;
ignore = b_element_op;
ignore = cde_element_op;
ignore = a_grid_desc_ak0_m_ak1;
ignore = b_grid_desc_bk0_n_bk1;
ignore = ds_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = e_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = block_2_etile_map;
#endif
}
} // namespace ck
namespace ck {
namespace tensor_operation {
namespace device {
// input : A[M, K], or A[K, N]
// input : B[K, N], or A[N, K]
// input : D0[M, N], D1[M, N], ...
// output : E[M, N]
// C = a_op(A) * b_op(B)
// E = cde_op(C, D0, D1, ...)
template <typename ALayout,
typename BLayout,
typename CDELayout,
typename ADataType,
typename BDataType,
typename AccDataType,
typename CShuffleDataType,
typename DDataType,
typename EDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
GemmSpecialization GemmSpec,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock,
index_t KPerBlock,
index_t AK1,
index_t BK1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
index_t ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
index_t BBlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CDEBlockTransferScalarPerVector_NPerBlock,
LoopScheduler LoopSched = make_default_loop_scheduler()>
struct DeviceGemmBiasEPermute_Xdl : public DeviceGemmBiasCPermute<AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation>
{
using DeviceOp = DeviceGemmBiasEPermute_Xdl;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr auto matrix_padder =
MatrixPadder<GemmSpec, index_t, index_t, index_t>{MPerBlock, NPerBlock, KPerBlock};
static constexpr index_t NumDTensor = 1;
static auto MakeAGridDescriptor_M_K(index_t MRaw, index_t KRaw, index_t StrideA)
{
const auto a_grid_desc_mraw_kraw = [&]() {
if constexpr(is_same_v<tensor_layout::gemm::RowMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(StrideA, I1));
}
else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(I1, StrideA));
}
}();
return matrix_padder.PadADescriptor_M_K(a_grid_desc_mraw_kraw);
}
static auto MakeBGridDescriptor_N_K(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
return matrix_padder.PadBDescriptor_N_K(b_grid_desc_nraw_kraw);
}
static auto MakeEGridDescriptor_M_N(DEGridDesc_M0_M1_M2_N0_N1 d_e_grid_desc)
{
index_t M0 = d_e_grid_desc.M0_;
index_t M1 = d_e_grid_desc.M1_;
index_t M2 = d_e_grid_desc.M2_;
index_t N0 = d_e_grid_desc.N0_;
index_t N1 = d_e_grid_desc.N1_;
index_t stride_M0 = d_e_grid_desc.stride_M0_;
index_t stride_M1 = d_e_grid_desc.stride_M1_;
index_t stride_M2 = d_e_grid_desc.stride_M2_;
index_t stride_N0 = d_e_grid_desc.stride_N0_;
index_t stride_N1 = d_e_grid_desc.stride_N1_;
const auto e_grid_desc_mraw_nraw = [&]() {
const auto e_grid_desc_m0_m1_m2_n0_n1 = make_naive_tensor_descriptor(
make_tuple(M0, M1, M2, N0, N1),
make_tuple(stride_M0, stride_M1, stride_M2, stride_N0, stride_N1));
return transform_tensor_descriptor(
e_grid_desc_m0_m1_m2_n0_n1,
make_tuple(make_merge_transform(make_tuple(M0, M1, M2)),
make_merge_transform(make_tuple(N0, N1))),
make_tuple(Sequence<0, 1, 2>{}, Sequence<3, 4>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}();
return matrix_padder.PadCDescriptor_M_N(e_grid_desc_mraw_nraw);
}
using AGridDesc_M_K = decltype(MakeAGridDescriptor_M_K(1, 1, 1));
using BGridDesc_N_K = decltype(MakeBGridDescriptor_N_K(1, 1, 1));
using EGridDesc_M_N = decltype(MakeEGridDescriptor_M_N(DEGridDesc_M0_M1_M2_N0_N1{}));
using DsGridDesc_M_N = Tuple<EGridDesc_M_N>;
// GridwiseGemm
using GridwiseGemm = GridwiseGemmMultipleD_xdl_cshuffle<
ADataType, // TODO: distinguish A/B datatype
AccDataType,
CShuffleDataType,
ck::Tuple<DDataType>,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
InMemoryDataOperationEnum::Set,
NumGemmKPrefetchStage,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
AK1,
BK1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
false,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
false,
BBlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CDEBlockTransferScalarPerVector_NPerBlock,
LoopSched>;
using AGridDesc_AK0_M_AK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(AGridDesc_M_K{}))>;
using BGridDesc_BK0_N_BK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(BGridDesc_N_K{}))>;
using Block2ETileMap = typename GridwiseGemm::DefaultBlock2ETileMap;
// Argument
struct Argument : public BaseArgument
{
Argument(const void* p_a_grid,
const void* p_b_grid,
const void* p_d_grid,
void* p_e_grid,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
DEGridDesc_M0_M1_M2_N0_N1 d_grid_desc,
DEGridDesc_M0_M1_M2_N0_N1 e_grid_desc,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op)
: p_a_grid_{static_cast<const ADataType*>(p_a_grid)},
p_b_grid_{static_cast<const BDataType*>(p_b_grid)},
p_ds_grid_{},
p_e_grid_{static_cast<EDataType*>(p_e_grid)},
a_grid_desc_m_k_{DeviceOp::MakeAGridDescriptor_M_K(MRaw, KRaw, StrideA)},
b_grid_desc_n_k_{DeviceOp::MakeBGridDescriptor_N_K(KRaw, NRaw, StrideB)},
ds_grid_desc_m_n_{},
e_grid_desc_m_n_{DeviceOp::MakeEGridDescriptor_M_N(e_grid_desc)},
a_grid_desc_ak0_m_ak1_{
GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(a_grid_desc_m_k_)},
b_grid_desc_bk0_n_bk1_{
GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(b_grid_desc_n_k_)},
ds_grid_desc_mblock_mperblock_nblock_nperblock_{},
e_grid_desc_mblock_mperblock_nblock_nperblock_{},
block_2_etile_map_{GridwiseGemm::MakeDefaultBlock2ETileMap(e_grid_desc_m_n_)},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
cde_element_op_{cde_element_op}
{
if(MRaw != d_grid_desc.M0_ * d_grid_desc.M1_ * d_grid_desc.M2_)
{
throw std::runtime_error("wrong! GridwiseGemm has invalid setting");
}
if(NRaw != d_grid_desc.N0_ * d_grid_desc.N1_)
{
throw std::runtime_error("wrong! GridwiseGemm has invalid setting");
}
// populate pointer, desc for Ds
// D pointer
p_ds_grid_(I0) = static_cast<const DDataType*>(p_d_grid);
// D desc
ds_grid_desc_m_n_(I0) = DeviceOp::MakeEGridDescriptor_M_N(d_grid_desc);
if(GridwiseGemm::CheckValidity(a_grid_desc_m_k_,
b_grid_desc_n_k_,
ds_grid_desc_m_n_,
e_grid_desc_m_n_,
block_2_etile_map_))
{
e_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
e_grid_desc_m_n_);
ds_grid_desc_mblock_mperblock_nblock_nperblock_(I0) =
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
ds_grid_desc_m_n_[I0]);
}
}
// private:
// pointers
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
typename GridwiseGemm::DsGridPointer p_ds_grid_;
EDataType* p_e_grid_;
// tensor descriptors for problem definiton
AGridDesc_M_K a_grid_desc_m_k_;
BGridDesc_N_K b_grid_desc_n_k_;
DsGridDesc_M_N ds_grid_desc_m_n_;
EGridDesc_M_N e_grid_desc_m_n_;
// tensor descriptors for block/thread-wise copy
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
typename GridwiseGemm::DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
ds_grid_desc_mblock_mperblock_nblock_nperblock_;
typename GridwiseGemm::EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
e_grid_desc_mblock_mperblock_nblock_nperblock_;
// block-to-e-tile map
Block2ETileMap block_2_etile_map_;
// element-wise op
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CDEElementwiseOperation cde_element_op_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_m_k_,
arg.b_grid_desc_n_k_,
arg.ds_grid_desc_m_n_,
arg.e_grid_desc_m_n_,
arg.block_2_etile_map_))
{
throw std::runtime_error("wrong! GridwiseGemm has invalid setting");
}
const index_t grid_size =
arg.block_2_etile_map_.CalculateGridSize(arg.e_grid_desc_m_n_);
const auto K =
arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) * arg.a_grid_desc_ak0_m_ak1_.GetLength(I2);
auto launch_kernel = [&](auto has_main_k_block_loop) {
constexpr bool has_main_loop = has_main_k_block_loop.value;
const auto kernel = kernel_gemm_bias_e_permute<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
typename GridwiseGemm::DsGridPointer,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
typename GridwiseGemm::DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::DefaultBlock2ETileMap,
has_main_loop>;
return launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_ds_grid_,
arg.p_e_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.cde_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.ds_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.e_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_etile_map_);
};
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
return launch_kernel(integral_constant<bool, true>{});
}
else
{
return launch_kernel(integral_constant<bool, false>{});
}
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static bool IsSupportedArgument(const Argument& arg)
{
if(!(ck::get_device_name() == "gfx908" || ck::get_device_name() == "gfx90a"))
{
return false;
}
return GridwiseGemm::CheckValidity(arg.a_grid_desc_m_k_,
arg.b_grid_desc_n_k_,
arg.ds_grid_desc_m_n_,
arg.e_grid_desc_m_n_,
arg.block_2_etile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const void* p_a,
const void* p_b,
const void* p_d,
void* p_e,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
DEGridDesc_M0_M1_M2_N0_N1 d_grid_desc,
DEGridDesc_M0_M1_M2_N0_N1 e_grid_desc,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op)
{
return Argument{p_a,
p_b,
p_d,
p_e,
MRaw,
NRaw,
KRaw,
StrideA,
StrideB,
d_grid_desc,
e_grid_desc,
a_element_op,
b_element_op,
cde_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_a,
const void* p_b,
const void* p_d,
void* p_e,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
DEGridDesc_M0_M1_M2_N0_N1 d_grid_desc,
DEGridDesc_M0_M1_M2_N0_N1 e_grid_desc,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op) override
{
return std::make_unique<Argument>(p_a,
p_b,
p_d,
p_e,
MRaw,
NRaw,
KRaw,
StrideA,
StrideB,
d_grid_desc,
e_grid_desc,
a_element_op,
b_element_op,
cde_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceGemmBiasEPermute_Xdl"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< AK1 << ", "
<< BK1
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

594
include/ck/tensor_operation/gpu/device/impl/device_gemm_dl.hpp Normal file
View File

@@ -0,0 +1,594 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_gemm.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_dl_v1r3.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <
typename ADataType,
typename BDataType,
typename CDataType,
typename AccDataType,
typename ALayout,
typename BLayout,
typename CLayout,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CElementwiseOperation,
GemmSpecialization GemmSpec,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock,
index_t K0PerBlock,
index_t K1,
index_t M1PerThread,
index_t N1PerThread,
index_t KPerThread,
typename M1N1ThreadClusterM1Xs,
typename M1N1ThreadClusterN1Xs,
typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1,
typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1,
typename ABlockTransferSrcVectorTensorContiguousDimOrder,
typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1,
typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1,
typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1,
typename BBlockTransferSrcVectorTensorContiguousDimOrder,
typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1,
typename CThreadTransferSrcDstAccessOrder,
index_t CThreadTransferSrcDstVectorDim,
index_t CThreadTransferDstScalarPerVector,
enable_if_t<
is_same_v<AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough> &&
is_same_v<BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough> &&
is_same_v<CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough>,
bool> = false>
struct DeviceGemmDl : public DeviceGemm<ALayout,
BLayout,
CLayout,
ADataType,
BDataType,
CDataType,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation>
{
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr auto I4 = Number<4>{};
static constexpr auto I5 = Number<5>{};
static constexpr auto K1Number = Number<K1>{};
static auto MakeAGridDescriptor_K0_M_K1(index_t M, index_t K, index_t StrideA)
{
assert(K % K1 == 0);
const index_t K0 = K / K1;
const auto a_grid_desc_m_k = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, ALayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, K), make_tuple(StrideA, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, ALayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, K), make_tuple(I1, StrideA));
}
}();
if constexpr(GemmSpec == GemmSpecialization::MNPadding)
{
const auto PadM = (MPerBlock - M % MPerBlock) % MPerBlock;
return transform_tensor_descriptor(
a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(K0, K1Number)),
make_right_pad_transform(M, PadM)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
else
{
return transform_tensor_descriptor(
a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(K0, K1Number)),
make_pass_through_transform(M)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
}
static auto MakeBGridDescriptor_K0_N_K1(index_t K, index_t N, index_t StrideB)
{
assert(K % K1 == 0);
const index_t K0 = K / K1;
const auto b_grid_desc_k_n = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(K, N), make_tuple(StrideB, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(K, N), make_tuple(I1, StrideB));
}
}();
if constexpr(GemmSpec == GemmSpecialization::MNPadding)
{
const auto PadN = (NPerBlock - N % NPerBlock) % NPerBlock;
return transform_tensor_descriptor(
b_grid_desc_k_n,
make_tuple(make_unmerge_transform(make_tuple(K0, K1Number)),
make_right_pad_transform(N, PadN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
else
{
return transform_tensor_descriptor(
b_grid_desc_k_n,
make_tuple(make_unmerge_transform(make_tuple(K0, K1Number)),
make_pass_through_transform(N)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
}
static auto MakeCGridDescriptor_M_N(index_t M, index_t N, index_t StrideC)
{
const auto c_grid_desc_m_n = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, N), make_tuple(StrideC, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, N), make_tuple(I1, StrideC));
}
}();
if constexpr(GemmSpec == GemmSpecialization::MNPadding)
{
const auto PadM = (MPerBlock - M % MPerBlock) % MPerBlock;
const auto PadN = (NPerBlock - N % NPerBlock) % NPerBlock;
return transform_tensor_descriptor(
c_grid_desc_m_n,
make_tuple(make_right_pad_transform(M, PadM), make_right_pad_transform(N, PadN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else
{
return transform_tensor_descriptor(
c_grid_desc_m_n,
make_tuple(make_pass_through_transform(M), make_pass_through_transform(N)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
}
using AGridDesc_K0_M_K1 = decltype(MakeAGridDescriptor_K0_M_K1(1, 1, 1));
using BGridDesc_K0_N_K1 = decltype(MakeBGridDescriptor_K0_N_K1(1, 1, 1));
using CGridDesc_M_N = decltype(MakeCGridDescriptor_M_N(1, 1, 1));
// GridwiseGemm
using GridwiseGemm =
GridwiseGemmDl_km_kn_mn_v1r3<BlockSize,
ADataType,
AccDataType,
CDataType,
InMemoryDataOperationEnum::Set,
AGridDesc_K0_M_K1,
BGridDesc_K0_N_K1,
CGridDesc_M_N,
MPerBlock,
NPerBlock,
K0PerBlock,
M1PerThread,
N1PerThread,
KPerThread,
M1N1ThreadClusterM1Xs,
M1N1ThreadClusterN1Xs,
ABlockTransferThreadSliceLengths_K0_M0_M1_K1,
ABlockTransferThreadClusterLengths_K0_M0_M1_K1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1,
ABlockTransferSrcVectorTensorContiguousDimOrder,
ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1,
BBlockTransferThreadSliceLengths_K0_N0_N1_K1,
BBlockTransferThreadClusterLengths_K0_N0_N1_K1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1,
BBlockTransferSrcVectorTensorContiguousDimOrder,
BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1,
CThreadTransferSrcDstAccessOrder,
CThreadTransferSrcDstVectorDim,
CThreadTransferDstScalarPerVector>;
using AGridDesc_K0_M0_M1_K1 =
decltype(GridwiseGemm::MakeAGridDescriptor_K0_M0_M1_K1(AGridDesc_K0_M_K1{}));
using BGridDesc_K0_N0_N1_K1 =
decltype(GridwiseGemm::MakeBGridDescriptor_K0_N0_N1_K1(BGridDesc_K0_N_K1{}));
using CGridDesc_M0_M10_M11_N0_N10_N11 =
decltype(GridwiseGemm::MakeCGridDescriptor_M0_M10_M11_N0_N10_N11(CGridDesc_M_N{}));
using DefaultBlock2CTileMap =
decltype(GridwiseGemm::MakeDefaultBlock2CTileMap(CGridDesc_M_N{}));
// Argument
struct Argument : public BaseArgument
{
Argument(const ADataType* p_a_grid,
const BDataType* p_b_grid,
CDataType* p_c_grid,
index_t M,
index_t N,
index_t K,
index_t StrideA,
index_t StrideB,
index_t StrideC,
index_t M01,
index_t N01,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op)
: p_a_grid_{p_a_grid},
p_b_grid_{p_b_grid},
p_c_grid_{p_c_grid},
a_grid_desc_k0_m0_m1_k1_{},
b_grid_desc_k0_n0_n1_k1_{},
c_grid_desc_m0_m10_m11_n0_n10_n11_{},
block_2_ctile_map_{},
M01_{M01},
N01_{N01},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
c_element_op_{c_element_op}
{
a_grid_desc_k0_m_k1_ = DeviceGemmDl::MakeAGridDescriptor_K0_M_K1(M, K, StrideA);
b_grid_desc_k0_n_k1_ = DeviceGemmDl::MakeBGridDescriptor_K0_N_K1(K, N, StrideB);
c_grid_desc_m_n_ = DeviceGemmDl::MakeCGridDescriptor_M_N(M, N, StrideC);
if(GridwiseGemm::CheckValidity(
a_grid_desc_k0_m_k1_, b_grid_desc_k0_n_k1_, c_grid_desc_m_n_))
{
a_grid_desc_k0_m0_m1_k1_ =
GridwiseGemm::MakeAGridDescriptor_K0_M0_M1_K1(a_grid_desc_k0_m_k1_);
b_grid_desc_k0_n0_n1_k1_ =
GridwiseGemm::MakeBGridDescriptor_K0_N0_N1_K1(b_grid_desc_k0_n_k1_);
c_grid_desc_m0_m10_m11_n0_n10_n11_ =
GridwiseGemm::MakeCGridDescriptor_M0_M10_M11_N0_N10_N11(c_grid_desc_m_n_);
block_2_ctile_map_ = GridwiseGemm::MakeDefaultBlock2CTileMap(c_grid_desc_m_n_);
}
}
// private:
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
CDataType* p_c_grid_;
AGridDesc_K0_M_K1 a_grid_desc_k0_m_k1_;
BGridDesc_K0_N_K1 b_grid_desc_k0_n_k1_;
CGridDesc_M_N c_grid_desc_m_n_;
AGridDesc_K0_M0_M1_K1 a_grid_desc_k0_m0_m1_k1_;
BGridDesc_K0_N0_N1_K1 b_grid_desc_k0_n0_n1_k1_;
CGridDesc_M0_M10_M11_N0_N10_N11 c_grid_desc_m0_m10_m11_n0_n10_n11_;
DefaultBlock2CTileMap block_2_ctile_map_;
// TODO: unused, but may be useful in future.
index_t M01_;
index_t N01_;
// TODO: unused since gridwise_gemm_dl_v1r3 does NOT support prologue for the time being.
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CElementwiseOperation c_element_op_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceGemmDl::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
{
std::cout << "arg.a_grid_desc_k0_m0_m1_k1_{"
<< arg.a_grid_desc_k0_m_k1_.GetLength(I0) << ", "
<< arg.a_grid_desc_k0_m_k1_.GetLength(I1) << ", "
<< arg.a_grid_desc_k0_m_k1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.b_grid_desc_k0_n0_n1_k1_{"
<< arg.b_grid_desc_k0_n_k1_.GetLength(I0) << ", "
<< arg.b_grid_desc_k0_n_k1_.GetLength(I1) << ", "
<< arg.b_grid_desc_k0_n_k1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.c_grid_desc_m_n_{ " << arg.c_grid_desc_m_n_.GetLength(I0) << ", "
<< arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
}
if(!GridwiseGemm::CheckValidity(
arg.a_grid_desc_k0_m_k1_, arg.b_grid_desc_k0_n_k1_, arg.c_grid_desc_m_n_))
{
throw std::runtime_error(
"wrong! GridwiseGemm_k0mk1_k0nk1_mn_xdl_v2r3 has invalid setting");
}
const index_t grid_size = GridwiseGemm::CalculateGridSize(
arg.c_grid_desc_m_n_.GetLength(I0), arg.c_grid_desc_m_n_.GetLength(I1));
const auto K0 = arg.a_grid_desc_k0_m0_m1_k1_.GetLength(I0);
const bool has_main_k_block_loop = GridwiseGemm::CalculateHasMainKBlockLoop(K0);
const bool has_double_tail_k_block_loop =
GridwiseGemm::CalculateHasDoubleTailKBlockLoop(K0);
float ave_time = 0;
if(has_main_k_block_loop && has_double_tail_k_block_loop)
{
const auto kernel =
kernel_gemm_dl_v1r3<GridwiseGemm,
ADataType,
CDataType,
remove_reference_t<AGridDesc_K0_M0_M1_K1>,
remove_reference_t<BGridDesc_K0_N0_N1_K1>,
remove_reference_t<CGridDesc_M0_M10_M11_N0_N10_N11>,
remove_reference_t<DefaultBlock2CTileMap>,
true,
true>;
ave_time = launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_grid_desc_k0_m0_m1_k1_,
arg.b_grid_desc_k0_n0_n1_k1_,
arg.c_grid_desc_m0_m10_m11_n0_n10_n11_,
arg.block_2_ctile_map_);
}
else if(has_main_k_block_loop && !has_double_tail_k_block_loop)
{
const auto kernel =
kernel_gemm_dl_v1r3<GridwiseGemm,
ADataType,
CDataType,
remove_reference_t<AGridDesc_K0_M0_M1_K1>,
remove_reference_t<BGridDesc_K0_N0_N1_K1>,
remove_reference_t<CGridDesc_M0_M10_M11_N0_N10_N11>,
remove_reference_t<DefaultBlock2CTileMap>,
true,
false>;
ave_time = launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_grid_desc_k0_m0_m1_k1_,
arg.b_grid_desc_k0_n0_n1_k1_,
arg.c_grid_desc_m0_m10_m11_n0_n10_n11_,
arg.block_2_ctile_map_);
}
else if(!has_main_k_block_loop && has_double_tail_k_block_loop)
{
const auto kernel =
kernel_gemm_dl_v1r3<GridwiseGemm,
ADataType,
CDataType,
remove_reference_t<AGridDesc_K0_M0_M1_K1>,
remove_reference_t<BGridDesc_K0_N0_N1_K1>,
remove_reference_t<CGridDesc_M0_M10_M11_N0_N10_N11>,
remove_reference_t<DefaultBlock2CTileMap>,
false,
true>;
ave_time = launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_grid_desc_k0_m0_m1_k1_,
arg.b_grid_desc_k0_n0_n1_k1_,
arg.c_grid_desc_m0_m10_m11_n0_n10_n11_,
arg.block_2_ctile_map_);
}
else
{
const auto kernel =
kernel_gemm_dl_v1r3<GridwiseGemm,
ADataType,
CDataType,
remove_reference_t<AGridDesc_K0_M0_M1_K1>,
remove_reference_t<BGridDesc_K0_N0_N1_K1>,
remove_reference_t<CGridDesc_M0_M10_M11_N0_N10_N11>,
remove_reference_t<DefaultBlock2CTileMap>,
false,
false>;
ave_time = launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_grid_desc_k0_m0_m1_k1_,
arg.b_grid_desc_k0_n0_n1_k1_,
arg.c_grid_desc_m0_m10_m11_n0_n10_n11_,
arg.block_2_ctile_map_);
}
return ave_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
if(ck::get_device_name() == "gfx906" || ck::get_device_name() == "gfx1030")
{
return GridwiseGemm::CheckValidity(
arg.a_grid_desc_k0_m_k1_, arg.b_grid_desc_k0_n_k1_, arg.c_grid_desc_m_n_);
}
else
{
return false;
}
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const ADataType* p_a,
const BDataType* p_b,
CDataType* p_c,
index_t M,
index_t N,
index_t K,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op)
{
return Argument{p_a,
p_b,
p_c,
M,
N,
K,
StrideA,
StrideB,
StrideC,
1,
1,
a_element_op,
b_element_op,
c_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,
const void* p_b,
void* p_c,
index_t M,
index_t N,
index_t K,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op) override
{
return std::make_unique<Argument>(static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
static_cast<CDataType*>(p_c),
M,
N,
K,
StrideA,
StrideB,
StrideC,
1,
1,
a_element_op,
b_element_op,
c_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceGemmDl"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< K0PerBlock << ", "
<< K1 << ", "
<< M1PerThread << ", "
<< N1PerThread << ", "
<< KPerThread
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

682
include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp Normal file
View File

@@ -0,0 +1,682 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_gemm_multiple_d_multiple_r.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_multiple_r_xdl_cshuffle.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
template <typename GridwiseGemm,
typename FloatAB,
typename FloatDsPointer,
typename FloatE,
typename FloatRsPointer,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
typename QsElementwiseOperation,
typename RsElementwiseOperation,
typename AGridDesc_AK0_M_AK1,
typename BGridDesc_BK0_N_BK1,
typename DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename RsGridDescriptor_MBlock_MPerBlock,
typename Block2ETileMap,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_gemm_multiple_d_multiple_r_xdl_cshuffle(
const FloatAB* __restrict__ p_a_grid,
const FloatAB* __restrict__ p_b_grid,
FloatDsPointer p_ds_grid,
FloatE* __restrict__ p_e_grid,
FloatRsPointer p_rs_grid,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const CDEElementwiseOperation cde_element_op,
const QsElementwiseOperation qs_element_op,
const RsElementwiseOperation rs_element_op,
const AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1,
const BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1,
const DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
ds_grid_desc_mblock_mperblock_nblock_nperblock,
const EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
e_grid_desc_mblock_mperblock_nblock_nperblock,
const RsGridDescriptor_MBlock_MPerBlock rs_grid_desc_mblock_mperblock,
const Block2ETileMap block_2_etile_map)
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__))
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
GridwiseGemm::template Run<HasMainKBlockLoop>(p_a_grid,
p_b_grid,
p_ds_grid,
p_e_grid,
p_rs_grid,
p_shared,
a_element_op,
b_element_op,
cde_element_op,
qs_element_op,
rs_element_op,
a_grid_desc_ak0_m_ak1,
b_grid_desc_bk0_n_bk1,
ds_grid_desc_mblock_mperblock_nblock_nperblock,
e_grid_desc_mblock_mperblock_nblock_nperblock,
rs_grid_desc_mblock_mperblock,
block_2_etile_map);
#else
ignore = p_a_grid;
ignore = p_b_grid;
ignore = p_ds_grid;
ignore = p_e_grid;
ignore = p_rs_grid;
ignore = a_element_op;
ignore = b_element_op;
ignore = cde_element_op;
ignore = qs_element_op;
ignore = rs_element_op;
ignore = a_grid_desc_ak0_m_ak1;
ignore = b_grid_desc_bk0_n_bk1;
ignore = ds_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = e_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = rs_grid_desc_mblock_mperblock;
ignore = block_2_etile_map;
#endif
}
} // namespace ck
namespace ck {
namespace tensor_operation {
namespace device {
// GEMM:
// input : A[AK0, M, AK1]
// input : B[AK0, N, AK1]
// input : D0[M, N], D1[M, N], ...
// output : E[M, N]
// output : R0[M], R1[M], ...
// C = a_op(A) * b_op(B)
// E = cde_op(C, D0, D1, ...)
// Q0 = reduce0(q_op0(E)), Q1 = reduce1(q_op0(E)), ...
// R0 = r_op0(Q0), R1 = r_op1(Q1), ...
// Assume:
// D0, D1, ... and E have the same layout
template <typename ALayout,
typename BLayout,
typename DELayout,
typename ADataType,
typename BDataType,
typename GemmAccDataType,
typename CShuffleDataType,
typename DsDataType,
typename EDataType,
typename ReduceAccDataType,
typename RsDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
typename QsElementwiseOperation,
typename RsElementwiseOperation,
typename ThreadReduceOperations,
typename RsGlobalMemoryDataOperation,
GemmSpecialization GemmSpec,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock,
index_t KPerBlock,
index_t AK1,
index_t BK1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
bool ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
bool BBlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CDRThreadTransferClusterLengths_MPerBlock_NPerBlock,
index_t CDEReduceThreadTransferScalarPerVector_NPerBlock,
index_t RThreadTransferDstScalarPerVector_MPerBlock,
LoopScheduler LoopSched = make_default_loop_scheduler()>
struct DeviceGemmMultipleDMultipleR_Xdl_CShuffle
: public DeviceGemmMultipleDMultipleR<ALayout,
BLayout,
DELayout,
ADataType,
BDataType,
DsDataType,
EDataType,
RsDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
QsElementwiseOperation,
RsElementwiseOperation>
{
using DeviceOp = DeviceGemmMultipleDMultipleR_Xdl_CShuffle;
static constexpr index_t NumDTensor = DsDataType::Size();
static constexpr index_t NumRTensor = RsDataType::Size();
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr auto matrix_padder =
MatrixPadder<GemmSpec, index_t, index_t, index_t>{MPerBlock, NPerBlock, KPerBlock};
static auto MakeAGridDescriptor_M_K(index_t MRaw, index_t KRaw, index_t StrideA)
{
const auto a_grid_desc_mraw_kraw = [&]() {
if constexpr(is_same_v<tensor_layout::gemm::RowMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(StrideA, I1));
}
else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(I1, StrideA));
}
}();
return matrix_padder.PadADescriptor_M_K(a_grid_desc_mraw_kraw);
}
static auto MakeBGridDescriptor_N_K(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
return matrix_padder.PadBDescriptor_N_K(b_grid_desc_nraw_kraw);
}
static auto MakeEGridDescriptor_M_N(index_t MRaw, index_t NRaw, index_t StrideE)
{
const auto e_grid_desc_mraw_nraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, DELayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(StrideE, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, DELayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(I1, StrideE));
}
}();
return matrix_padder.PadCDescriptor_M_N(e_grid_desc_mraw_nraw);
}
// assume D is packed tensor
static auto MakeRGridDescriptor_M(index_t MRaw)
{
const auto r_grid_desc_mraw = make_naive_tensor_descriptor_packed(make_tuple(MRaw));
const auto M = math::integer_divide_ceil(MRaw, MPerBlock) * MPerBlock;
const auto MPad = M - MRaw;
if constexpr(GemmSpec == GemmSpecialization::MPadding ||
GemmSpec == GemmSpecialization::MNPadding ||
GemmSpec == GemmSpecialization::MKPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad M
return transform_tensor_descriptor(r_grid_desc_mraw,
make_tuple(make_right_pad_transform(MRaw, MPad)),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0>{}));
}
else
{
// not pad M
return r_grid_desc_mraw;
}
}
using AGridDesc_M_K = decltype(MakeAGridDescriptor_M_K(1, 1, 1));
using BGridDesc_N_K = decltype(MakeBGridDescriptor_N_K(1, 1, 1));
using EGridDesc_M_N = decltype(MakeEGridDescriptor_M_N(1, 1, 1));
using RGridDesc_M = decltype(MakeRGridDescriptor_M(1));
// GridwiseGemm
using GridwiseGemm = GridwiseGemmMultipleDMultipleR_k0mk1_k0nk1_mn_xdl_cshuffle_v1<
ADataType, // TODO: distinguish A/B datatype
GemmAccDataType,
CShuffleDataType,
DsDataType,
EDataType,
ReduceAccDataType,
RsDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
QsElementwiseOperation,
RsElementwiseOperation,
ThreadReduceOperations,
InMemoryDataOperationEnum::Set,
RsGlobalMemoryDataOperation,
AGridDesc_M_K,
BGridDesc_N_K,
EGridDesc_M_N,
RGridDesc_M,
NumGemmKPrefetchStage,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
AK1,
BK1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
false,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
false,
BBlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CDRThreadTransferClusterLengths_MPerBlock_NPerBlock,
CDEReduceThreadTransferScalarPerVector_NPerBlock,
RThreadTransferDstScalarPerVector_MPerBlock,
LoopSched>;
using AGridDesc_AK0_M_AK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(AGridDesc_M_K{}))>;
using BGridDesc_BK0_N_BK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(BGridDesc_N_K{}))>;
using Block2ETileMap = typename GridwiseGemm::DefaultBlock2ETileMap;
// Argument
struct Argument : public BaseArgument
{
Argument(const void* p_a_grid,
const void* p_b_grid,
std::array<const void*, NumDTensor> p_ds_grid,
void* p_e_grid,
std::array<void*, NumRTensor> p_rs_grid,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
std::array<index_t, NumDTensor> StrideDs,
index_t StrideE,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op,
QsElementwiseOperation qs_element_op,
RsElementwiseOperation rs_element_op)
: p_a_grid_{static_cast<const ADataType*>(p_a_grid)},
p_b_grid_{static_cast<const BDataType*>(p_b_grid)},
p_ds_grid_{}, // FIXME
p_e_grid_{static_cast<EDataType*>(p_e_grid)},
p_rs_grid_{}, // FIXME
a_grid_desc_m_k_{DeviceOp::MakeAGridDescriptor_M_K(MRaw, KRaw, StrideA)},
b_grid_desc_n_k_{DeviceOp::MakeBGridDescriptor_N_K(KRaw, NRaw, StrideB)},
e_grid_desc_m_n_{DeviceOp::MakeEGridDescriptor_M_N(MRaw, NRaw, StrideE)},
r_grid_desc_m_{DeviceOp::MakeRGridDescriptor_M(MRaw)},
a_grid_desc_ak0_m_ak1_{
GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(a_grid_desc_m_k_)},
b_grid_desc_bk0_n_bk1_{
GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(b_grid_desc_n_k_)},
ds_grid_desc_mblock_mperblock_nblock_nperblock_{},
e_grid_desc_mblock_mperblock_nblock_nperblock_{},
rs_grid_desc_mblock_mperblock_{},
block_2_etile_map_{GridwiseGemm::MakeDefaultBlock2ETileMap(e_grid_desc_m_n_)},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
cde_element_op_{cde_element_op},
qs_element_op_{qs_element_op},
rs_element_op_{rs_element_op}
{
if(GridwiseGemm::CheckValidity(a_grid_desc_m_k_,
b_grid_desc_n_k_,
e_grid_desc_m_n_,
r_grid_desc_m_,
block_2_etile_map_))
{
e_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
e_grid_desc_m_n_);
static_for<0, NumDTensor, 1>{}([&](auto i) {
using DDataType = remove_cvref_t<tuple_element_t<i.value, DsDataType>>;
p_ds_grid_(i) = static_cast<const DDataType*>(p_ds_grid[i]);
const auto d_grid_desc_m_n =
DeviceOp::MakeEGridDescriptor_M_N(MRaw, NRaw, StrideDs[i]);
ds_grid_desc_mblock_mperblock_nblock_nperblock_(i) =
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
d_grid_desc_m_n);
});
static_for<0, NumRTensor, 1>{}([&](auto i) {
using RDataType = remove_cvref_t<tuple_element_t<i.value, RsDataType>>;
p_rs_grid_(i) = static_cast<RDataType*>(p_rs_grid[i]);
rs_grid_desc_mblock_mperblock_(i) =
GridwiseGemm::MakeRGridDescriptor_MBlock_MPerBlock(r_grid_desc_m_);
});
}
}
// private:
// pointers
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
typename GridwiseGemm::DsGridPointer p_ds_grid_;
EDataType* p_e_grid_;
typename GridwiseGemm::RsGridPointer p_rs_grid_;
// tensor descriptors
AGridDesc_M_K a_grid_desc_m_k_;
BGridDesc_N_K b_grid_desc_n_k_;
EGridDesc_M_N e_grid_desc_m_n_;
RGridDesc_M r_grid_desc_m_;
// tensor descriptors for block/thread-wise copy
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
StaticallyIndexedArray<
typename GridwiseGemm::EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
NumDTensor>
ds_grid_desc_mblock_mperblock_nblock_nperblock_; // FIXME: Ds desc may be of different
// type from E
typename GridwiseGemm::EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
e_grid_desc_mblock_mperblock_nblock_nperblock_;
StaticallyIndexedArray<typename GridwiseGemm::RGridDescriptor_MBlock_MPerBlock, NumRTensor>
rs_grid_desc_mblock_mperblock_;
// block-to-e-tile map
Block2ETileMap block_2_etile_map_;
// element-wise op
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CDEElementwiseOperation cde_element_op_;
QsElementwiseOperation qs_element_op_;
RsElementwiseOperation rs_element_op_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_m_k_,
arg.b_grid_desc_n_k_,
arg.e_grid_desc_m_n_,
arg.r_grid_desc_m_,
arg.block_2_etile_map_))
{
throw std::runtime_error("wrong! GridwiseGemm has invalid setting");
}
const index_t grid_size =
arg.block_2_etile_map_.CalculateGridSize(arg.e_grid_desc_m_n_);
const auto K =
arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) * arg.a_grid_desc_ak0_m_ak1_.GetLength(I2);
auto launch_kernel = [&](auto has_main_k_block_loop) {
constexpr bool has_main_loop = has_main_k_block_loop.value;
const auto kernel = kernel_gemm_multiple_d_multiple_r_xdl_cshuffle<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
typename GridwiseGemm::DsGridPointer,
EDataType,
typename GridwiseGemm::RsGridPointer,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
QsElementwiseOperation,
RsElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
ck::StaticallyIndexedArray<
typename GridwiseGemm::EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
NumDTensor>,
typename GridwiseGemm::EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
ck::StaticallyIndexedArray<
typename GridwiseGemm::RGridDescriptor_MBlock_MPerBlock,
NumRTensor>,
typename GridwiseGemm::DefaultBlock2ETileMap,
has_main_loop>;
return launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_ds_grid_,
arg.p_e_grid_,
arg.p_rs_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.cde_element_op_,
arg.qs_element_op_,
arg.rs_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.ds_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.e_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.rs_grid_desc_mblock_mperblock_,
arg.block_2_etile_map_);
};
float ave_time = 0;
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
ave_time = launch_kernel(integral_constant<bool, true>{});
}
else
{
ave_time = launch_kernel(integral_constant<bool, false>{});
}
return ave_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static bool IsSupportedArgument(const Argument& arg)
{
if(!(ck::get_device_name() == "gfx908" || ck::get_device_name() == "gfx90a"))
{
return false;
}
return GridwiseGemm::CheckValidity(arg.a_grid_desc_m_k_,
arg.b_grid_desc_n_k_,
arg.e_grid_desc_m_n_,
arg.r_grid_desc_m_,
arg.block_2_etile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const void* p_a,
const void* p_b,
std::array<const void*, NumDTensor> p_ds,
void* p_e,
std::array<void*, NumRTensor> p_rs,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
std::array<index_t, NumDTensor> StrideDs,
index_t StrideE,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op,
QsElementwiseOperation qs_element_op,
RsElementwiseOperation rs_element_op)
{
return Argument{p_a,
p_b,
p_ds,
p_e,
p_rs,
MRaw,
NRaw,
KRaw,
StrideA,
StrideB,
StrideDs,
StrideE,
a_element_op,
b_element_op,
cde_element_op,
qs_element_op,
rs_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,
const void* p_b,
std::array<const void*, NumDTensor> p_ds,
void* p_e,
std::array<void*, NumRTensor> p_rs,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
std::array<index_t, NumDTensor> StrideDs,
index_t StrideE,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op,
QsElementwiseOperation qs_element_op,
RsElementwiseOperation rs_element_op) override
{
return std::make_unique<Argument>(p_a,
p_b,
p_ds,
p_e,
p_rs,
MRaw,
NRaw,
KRaw,
StrideA,
StrideB,
StrideDs,
StrideE,
a_element_op,
b_element_op,
cde_element_op,
qs_element_op,
rs_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceGemmMultipleDMultipleR_Xdl_CShuffle"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< AK1 << ", "
<< BK1 << ", "
<< getGemmSpecializationString(GemmSpec)
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

686
include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle.hpp Normal file
View File

@@ -0,0 +1,686 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_gemm_multiple_d.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
template <typename GridwiseGemm,
typename ABDataType,
typename DsPointer,
typename EDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
typename AGridDesc_AK0_M_AK1,
typename BGridDesc_BK0_N_BK1,
typename DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename Block2ETileMap,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_gemm_multiple_d_xdl_cshuffle(const ABDataType* __restrict__ p_a_grid,
const ABDataType* __restrict__ p_b_grid,
DsPointer p_ds_grid,
EDataType* __restrict__ p_e_grid,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const CDEElementwiseOperation cde_element_op,
const AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1,
const BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1,
const DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
ds_grid_desc_mblock_mperblock_nblock_nperblock,
const EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
e_grid_desc_mblock_mperblock_nblock_nperblock,
const Block2ETileMap block_2_etile_map)
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__))
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
GridwiseGemm::template Run<HasMainKBlockLoop>(p_a_grid,
p_b_grid,
p_ds_grid,
p_e_grid,
p_shared,
a_element_op,
b_element_op,
cde_element_op,
a_grid_desc_ak0_m_ak1,
b_grid_desc_bk0_n_bk1,
ds_grid_desc_mblock_mperblock_nblock_nperblock,
e_grid_desc_mblock_mperblock_nblock_nperblock,
block_2_etile_map);
#else
ignore = p_a_grid;
ignore = p_b_grid;
ignore = p_ds_grid;
ignore = p_e_grid;
ignore = a_element_op;
ignore = b_element_op;
ignore = cde_element_op;
ignore = a_grid_desc_ak0_m_ak1;
ignore = b_grid_desc_bk0_n_bk1;
ignore = ds_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = e_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = block_2_etile_map;
#endif
}
} // namespace ck
namespace ck {
namespace tensor_operation {
namespace device {
// GEMM:
// input : A[M, K]
// input : B[N, K]
// input : D0[M, N], D1[M, N], ...
// output : E[M, N]
// C = a_op(A) * b_op(B)
// E = cde_op(C, D0, D1, ...)
// Assume:
// D0, D1, ... and E have the same layout
template <typename ALayout,
typename BLayout,
typename DsLayout,
typename ELayout,
typename ADataType,
typename BDataType,
typename AccDataType,
typename CShuffleDataType,
typename DsDataType,
typename EDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
GemmSpecialization GemmSpec,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock,
index_t KPerBlock,
index_t AK1,
index_t BK1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
index_t ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
index_t BBlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CDEBlockTransferScalarPerVector_NPerBlock,
LoopScheduler LoopSched = make_default_loop_scheduler()>
struct DeviceGemmMultipleD_Xdl_CShuffle : public DeviceGemmMultipleD<ALayout,
BLayout,
DsLayout,
ELayout,
ADataType,
BDataType,
DsDataType,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation>
{
using DeviceOp = DeviceGemmMultipleD_Xdl_CShuffle;
static constexpr index_t NumDTensor = DsDataType::Size();
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr auto matrix_padder =
MatrixPadder<GemmSpec, index_t, index_t, index_t>{MPerBlock, NPerBlock, KPerBlock};
static auto MakeAGridDescriptor_M_K(index_t MRaw, index_t KRaw, index_t StrideA)
{
const auto a_grid_desc_mraw_kraw = [&]() {
if constexpr(is_same_v<tensor_layout::gemm::RowMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(StrideA, I1));
}
else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(I1, StrideA));
}
}();
return matrix_padder.PadADescriptor_M_K(a_grid_desc_mraw_kraw);
}
static auto MakeBGridDescriptor_N_K(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
return matrix_padder.PadBDescriptor_N_K(b_grid_desc_nraw_kraw);
}
template <typename ELay>
static auto MakeEGridDescriptor_M_N(index_t MRaw, index_t NRaw, index_t StrideE)
{
const auto e_grid_desc_mraw_nraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, ELay>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(StrideE, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, ELay>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(I1, StrideE));
}
}();
return matrix_padder.PadCDescriptor_M_N(e_grid_desc_mraw_nraw);
}
static auto MakeDsGridDescriptor_M_N(const std::array<index_t, NumDTensor>& MRaws,
const std::array<index_t, NumDTensor>& NRaws,
const std::array<index_t, NumDTensor>& DsStride)
{
return generate_tuple(
[&](auto i) {
using DLayout = remove_cvref_t<tuple_element_t<i.value, DsLayout>>;
return DeviceOp::MakeEGridDescriptor_M_N<DLayout>(MRaws[i], NRaws[i], DsStride[i]);
},
Number<NumDTensor>{});
}
// desc for problem definition
using AGridDesc_M_K = decltype(MakeAGridDescriptor_M_K(1, 1, 1));
using BGridDesc_N_K = decltype(MakeBGridDescriptor_N_K(1, 1, 1));
using DsGridDesc_M_N = remove_cvref_t<decltype(MakeDsGridDescriptor_M_N({}, {}, {}))>;
using EGridDesc_M_N = decltype(MakeEGridDescriptor_M_N<ELayout>(1, 1, 1));
// GridwiseGemm
using GridwiseGemm = GridwiseGemmMultipleD_xdl_cshuffle<
ADataType, // TODO: distinguish A/B datatype
AccDataType,
CShuffleDataType,
DsDataType,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
InMemoryDataOperationEnum::Set,
NumGemmKPrefetchStage,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
AK1,
BK1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
false,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
false,
BBlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CDEBlockTransferScalarPerVector_NPerBlock,
LoopSched>;
// desc for blockwise copy
using AGridDesc_AK0_M_AK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(AGridDesc_M_K{}))>;
using BGridDesc_BK0_N_BK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(BGridDesc_N_K{}))>;
using DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock = remove_cvref_t<decltype(
GridwiseGemm::MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(DsGridDesc_M_N{}))>;
using EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock = remove_cvref_t<decltype(
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(EGridDesc_M_N{}))>;
// block-to-e-tile map
using Block2ETileMap =
remove_cvref_t<decltype(GridwiseGemm::MakeDefaultBlock2ETileMap(EGridDesc_M_N{}))>;
// Argument
struct Argument : public BaseArgument
{
Argument(const void* p_a_grid,
const void* p_b_grid,
std::array<const void*, NumDTensor> p_ds_grid,
void* p_e_grid,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
std::array<index_t, NumDTensor> StrideDs,
index_t StrideE,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op)
: p_a_grid_{static_cast<const ADataType*>(p_a_grid)},
p_b_grid_{static_cast<const BDataType*>(p_b_grid)},
p_ds_grid_{},
p_e_grid_{static_cast<EDataType*>(p_e_grid)},
a_grid_desc_m_k_{DeviceOp::MakeAGridDescriptor_M_K(MRaw, KRaw, StrideA)},
b_grid_desc_n_k_{DeviceOp::MakeBGridDescriptor_N_K(KRaw, NRaw, StrideB)},
ds_grid_desc_m_n_{},
e_grid_desc_m_n_{DeviceOp::MakeEGridDescriptor_M_N<ELayout>(MRaw, NRaw, StrideE)},
a_grid_desc_ak0_m_ak1_{
GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(a_grid_desc_m_k_)},
b_grid_desc_bk0_n_bk1_{
GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(b_grid_desc_n_k_)},
ds_grid_desc_mblock_mperblock_nblock_nperblock_{},
e_grid_desc_mblock_mperblock_nblock_nperblock_{},
block_2_etile_map_{GridwiseGemm::MakeDefaultBlock2ETileMap(e_grid_desc_m_n_)},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
cde_element_op_{cde_element_op},
MRaw_{MRaw},
NRaw_{NRaw},
KRaw_{KRaw}
{
// populate pointer, desc for Ds
static_for<0, NumDTensor, 1>{}([&](auto i) {
using DLayout = remove_cvref_t<tuple_element_t<i.value, DsLayout>>;
using DDataType = remove_cvref_t<tuple_element_t<i.value, DsDataType>>;
// D pointer
p_ds_grid_(i) = static_cast<const DDataType*>(p_ds_grid[i]);
// D desc
ds_grid_desc_m_n_(i) =
DeviceOp::MakeEGridDescriptor_M_N<DLayout>(MRaw, NRaw, StrideDs[i]);
});
// populate desc for Ds/E
if(GridwiseGemm::CheckValidity(a_grid_desc_m_k_,
b_grid_desc_n_k_,
ds_grid_desc_m_n_,
e_grid_desc_m_n_,
block_2_etile_map_))
{
ds_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
ds_grid_desc_m_n_);
e_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
e_grid_desc_m_n_);
}
}
void Print() const
{
std::cout << "A[M, K]: " << a_grid_desc_m_k_ << std::endl;
std::cout << "B[N, K]: " << b_grid_desc_n_k_ << std::endl;
static_for<0, NumDTensor, 1>{}(
[&](auto i) { std::cout << "Ds[M, N]: " << ds_grid_desc_m_n_[i] << std::endl; });
std::cout << "E[M, N]: " << e_grid_desc_m_n_ << std::endl;
}
// private:
// pointers
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
typename GridwiseGemm::DsGridPointer p_ds_grid_;
EDataType* p_e_grid_;
// tensor descriptors for problem definiton
AGridDesc_M_K a_grid_desc_m_k_;
BGridDesc_N_K b_grid_desc_n_k_;
DsGridDesc_M_N ds_grid_desc_m_n_;
EGridDesc_M_N e_grid_desc_m_n_;
// tensor descriptors for block/thread-wise copy
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock
ds_grid_desc_mblock_mperblock_nblock_nperblock_;
EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock e_grid_desc_mblock_mperblock_nblock_nperblock_;
// block-to-e-tile map
Block2ETileMap block_2_etile_map_;
// element-wise op
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CDEElementwiseOperation cde_element_op_;
// for checking vector load/store
index_t MRaw_;
index_t NRaw_;
index_t KRaw_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_m_k_,
arg.b_grid_desc_n_k_,
arg.ds_grid_desc_m_n_,
arg.e_grid_desc_m_n_,
arg.block_2_etile_map_))
{
throw std::runtime_error("wrong! GridwiseGemm has invalid setting");
}
const index_t grid_size =
arg.block_2_etile_map_.CalculateGridSize(arg.e_grid_desc_m_n_);
const auto K =
arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) * arg.a_grid_desc_ak0_m_ak1_.GetLength(I2);
auto launch_kernel = [&](auto has_main_k_block_loop) {
constexpr bool has_main_loop = has_main_k_block_loop.value;
const auto kernel = kernel_gemm_multiple_d_xdl_cshuffle<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
typename GridwiseGemm::DsGridPointer,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
DeviceOp::DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
DeviceOp::EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
DeviceOp::Block2ETileMap,
has_main_loop>;
return launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_ds_grid_,
arg.p_e_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.cde_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.ds_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.e_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_etile_map_);
};
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
return launch_kernel(integral_constant<bool, true>{});
}
else
{
return launch_kernel(integral_constant<bool, false>{});
}
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static bool IsSupportedArgument(const Argument& arg)
{
if(!(ck::get_device_name() == "gfx908" || ck::get_device_name() == "gfx90a"))
{
return false;
}
// check vector load/store
{
using Row = ck::tensor_layout::gemm::RowMajor;
using Col = ck::tensor_layout::gemm::ColumnMajor;
// check vector load of A
if constexpr(is_same_v<ALayout, Row> && ABlockTransferSrcVectorDim == 2)
{
if(arg.KRaw_ % ABlockTransferSrcScalarPerVector != 0)
{
return false;
}
}
else if constexpr(is_same_v<ALayout, Col> && ABlockTransferSrcVectorDim == 1)
{
// FIXME: not rigorous
if(arg.MRaw_ % ABlockTransferSrcScalarPerVector != 0)
{
return false;
}
}
else
{
return false;
}
// check vector laod of B
if constexpr(is_same_v<BLayout, Col> && BBlockTransferSrcVectorDim == 2)
{
if(arg.KRaw_ % BBlockTransferSrcScalarPerVector != 0)
{
return false;
}
}
else if constexpr(is_same_v<BLayout, Row> && BBlockTransferSrcVectorDim == 1)
{
// FIXME: not rigorous
if(arg.NRaw_ % BBlockTransferSrcScalarPerVector != 0)
{
return false;
}
}
else
{
return false;
}
// check vector load of Ds
// only support RowMajor for now
bool all_valid = true;
static_for<0, NumDTensor, 1>{}([&](auto i) {
using DLayout = remove_cvref_t<tuple_element_t<i.value, DsLayout>>;
if constexpr(!is_same_v<DLayout, Row>)
{
all_valid = false;
}
});
if(!all_valid)
{
return false;
}
// check vector store of E
// only support RowMajor for now
if constexpr(is_same_v<ELayout, Row>)
{
if(arg.NRaw_ % CDEBlockTransferScalarPerVector_NPerBlock != 0)
{
return false;
}
}
else
{
return false;
}
}
return GridwiseGemm::CheckValidity(arg.a_grid_desc_m_k_,
arg.b_grid_desc_n_k_,
arg.ds_grid_desc_m_n_,
arg.e_grid_desc_m_n_,
arg.block_2_etile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const void* p_a,
const void* p_b,
std::array<const void*, NumDTensor> p_ds,
void* p_e,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
std::array<index_t, NumDTensor> StrideDs,
index_t StrideE,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op)
{
return Argument{p_a,
p_b,
p_ds,
p_e,
MRaw,
NRaw,
KRaw,
StrideA,
StrideB,
StrideDs,
StrideE,
a_element_op,
b_element_op,
cde_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_a,
const void* p_b,
std::array<const void*, NumDTensor> p_ds,
void* p_e,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
std::array<ck::index_t, NumDTensor> StrideDs,
index_t StrideE,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op) override
{
return std::make_unique<Argument>(p_a,
p_b,
p_ds,
p_e,
MRaw,
NRaw,
KRaw,
StrideA,
StrideB,
StrideDs,
StrideE,
a_element_op,
b_element_op,
cde_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceGemmMultipleD_Xdl_CShuffle"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< AK1 << ", "
<< BK1 << ", "
<< getGemmSpecializationString(GemmSpec)
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

835
include/ck/tensor_operation/gpu/device/impl/device_gemm_reduce_xdl_cshuffle.hpp Normal file
View File

@@ -0,0 +1,835 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_gemm_reduce.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_reduce_xdl_cshuffle_v1.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
// Note: inter-wave loop scheduler is rolled out to c-shuffle version first. Becuase non c-shuffle
// version currently has compiler issues with register spill which further causes validation
// failures.
template <typename ALayout,
typename BLayout,
typename CLayout,
typename ADataType,
typename BDataType,
typename CDataType,
typename GemmAccDataType,
typename CShuffleDataType,
typename ReduceAccDataType,
typename ReducePtrsGlobal,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CElementwiseOperation,
typename ReduceOperations,
typename ReduceInElementwiseOperations,
typename ReduceAccElementwiseOperations,
typename ReduceGlobalMemoryDataOperation,
GemmSpecialization GemmSpec,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock,
index_t KPerBlock,
index_t AK1,
index_t BK1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
bool ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
bool BBlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CShuffleBlockTransferScalarPerVector_NPerBlock,
typename CReduceThreadClusterLengths_MPerBlock_NPerBlock,
index_t CReduceThreadLds2VGprCopySrcDstScalarPerVector_NPerBlock,
index_t CReduceThreadVgpr2GlobalCopySrcDstScalarPerVector_MPerBlock,
LoopScheduler LoopSched = make_default_loop_scheduler()>
struct DeviceGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<0, ReduceOperations::Size()>
{
using DeviceOp = DeviceGemmReduce_Xdl_CShuffle;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static auto MakeAGridDescriptor_AK0_M_AK1(index_t MRaw, index_t KRaw, index_t StrideA)
{
const auto a_grid_desc_mraw_kraw = [&]() {
if constexpr(is_same_v<tensor_layout::gemm::RowMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(StrideA, I1));
}
else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(I1, StrideA));
}
}();
const auto M = math::integer_divide_ceil(MRaw, MPerBlock) * MPerBlock;
const auto K = math::integer_divide_ceil(KRaw, KPerBlock) * KPerBlock;
const auto MPad = M - MRaw;
const auto KPad = K - KRaw;
if constexpr(GemmSpec == GemmSpecialization::MKPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad both M and K
assert(K % AK1 == 0);
const auto AK0 = K / AK1;
const auto a_grid_desc_m_k =
transform_tensor_descriptor(a_grid_desc_mraw_kraw,
make_tuple(make_right_pad_transform(MRaw, MPad),
make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_pass_through_transform(M)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
else if constexpr(GemmSpec == GemmSpecialization::MPadding ||
GemmSpec == GemmSpecialization::MNPadding)
{
// pad M, but not K
assert(KRaw % AK1 == 0);
const auto AK0 = KRaw / AK1;
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_mraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_right_pad_transform(MRaw, MPad)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
else if constexpr(GemmSpec == GemmSpecialization::KPadding ||
GemmSpec == GemmSpecialization::NKPadding)
{
// pad K, but not M
assert(K % AK1 == 0);
const auto AK0 = K / AK1;
const auto a_grid_desc_m_k = transform_tensor_descriptor(
a_grid_desc_mraw_kraw,
make_tuple(make_pass_through_transform(MRaw), make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_pass_through_transform(MRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
else
{
// not pad M or K
assert(KRaw % AK1 == 0);
const auto AK0 = KRaw / AK1;
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_mraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_pass_through_transform(MRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
}
static auto MakeBGridDescriptor_BK0_N_BK1(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
const auto N = math::integer_divide_ceil(NRaw, NPerBlock) * NPerBlock;
const auto K = math::integer_divide_ceil(KRaw, KPerBlock) * KPerBlock;
const auto NPad = N - NRaw;
const auto KPad = K - KRaw;
if constexpr(GemmSpec == GemmSpecialization::NKPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad both N and K
assert(K % BK1 == 0);
const auto BK0 = K / BK1;
const auto b_grid_desc_n_k =
transform_tensor_descriptor(b_grid_desc_nraw_kraw,
make_tuple(make_right_pad_transform(NRaw, NPad),
make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_n_k,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_pass_through_transform(N)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
else if constexpr(GemmSpec == GemmSpecialization::NPadding ||
GemmSpec == GemmSpecialization::MNPadding)
{
// pad N, but not K
assert(KRaw % BK1 == 0);
const auto BK0 = KRaw / BK1;
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_nraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_right_pad_transform(NRaw, NPad)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
else if constexpr(GemmSpec == GemmSpecialization::KPadding ||
GemmSpec == GemmSpecialization::MKPadding)
{
// pad K, but not N
assert(K % BK1 == 0);
const auto BK0 = K / BK1;
const auto b_grid_desc_n_k = transform_tensor_descriptor(
b_grid_desc_nraw_kraw,
make_tuple(make_pass_through_transform(NRaw), make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_n_k,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_pass_through_transform(NRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
else
{
// not pad N or K
assert(KRaw % BK1 == 0);
const auto BK0 = KRaw / BK1;
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_nraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_pass_through_transform(NRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
}
static auto MakeCGridDescriptor_M_N(index_t MRaw, index_t NRaw, index_t StrideC)
{
const auto c_grid_desc_mraw_nraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(StrideC, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(I1, StrideC));
}
}();
const auto M = math::integer_divide_ceil(MRaw, MPerBlock) * MPerBlock;
const auto N = math::integer_divide_ceil(NRaw, NPerBlock) * NPerBlock;
const auto MPad = M - MRaw;
const auto NPad = N - NRaw;
if constexpr(GemmSpec == GemmSpecialization::MNPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad M and N
return transform_tensor_descriptor(c_grid_desc_mraw_nraw,
make_tuple(make_right_pad_transform(MRaw, MPad),
make_right_pad_transform(NRaw, NPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else if constexpr(GemmSpec == GemmSpecialization::MPadding ||
GemmSpec == GemmSpecialization::MKPadding)
{
// pad M, but not N
return transform_tensor_descriptor(
c_grid_desc_mraw_nraw,
make_tuple(make_right_pad_transform(MRaw, MPad), make_pass_through_transform(NRaw)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else if constexpr(GemmSpec == GemmSpecialization::NPadding ||
GemmSpec == GemmSpecialization::NKPadding)
{
// pad N, but not M
return transform_tensor_descriptor(
c_grid_desc_mraw_nraw,
make_tuple(make_pass_through_transform(MRaw), make_right_pad_transform(NRaw, NPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else
{
// not pad M or N
return c_grid_desc_mraw_nraw;
}
}
// assume Reduce is packed tensor
static auto MakeReduceGridDescriptor_M(index_t MRaw)
{
const auto d_grid_desc_mraw = make_naive_tensor_descriptor_packed(make_tuple(MRaw));
const auto M = math::integer_divide_ceil(MRaw, MPerBlock) * MPerBlock;
const auto MPad = M - MRaw;
if constexpr(GemmSpec == GemmSpecialization::MPadding ||
GemmSpec == GemmSpecialization::MNPadding ||
GemmSpec == GemmSpecialization::MKPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad M
return transform_tensor_descriptor(d_grid_desc_mraw,
make_tuple(make_right_pad_transform(MRaw, MPad)),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0>{}));
}
else
{
// not pad M
return d_grid_desc_mraw;
}
}
using AGridDesc_AK0_M_AK1 = decltype(MakeAGridDescriptor_AK0_M_AK1(1, 1, 1));
using BGridDesc_BK0_N_BK1 = decltype(MakeBGridDescriptor_BK0_N_BK1(1, 1, 1));
using CGridDesc_M_N = decltype(MakeCGridDescriptor_M_N(1, 1, 1));
using ReduceGridDesc_M = decltype(MakeReduceGridDescriptor_M(1));
// GridwiseGemm
using GridwiseGemm = GridwiseGemmReduce_k0mk1_k0nk1_mn_xdl_cshuffle_v1<
ADataType, // TODO: distinguish A/B datatype
GemmAccDataType,
CShuffleDataType,
CDataType,
ReduceAccDataType,
ReducePtrsGlobal,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
ReduceOperations,
ReduceInElementwiseOperations,
ReduceAccElementwiseOperations,
InMemoryDataOperationEnum::Set,
ReduceGlobalMemoryDataOperation,
AGridDesc_AK0_M_AK1,
BGridDesc_BK0_N_BK1,
CGridDesc_M_N,
ReduceGridDesc_M,
NumGemmKPrefetchStage,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
AK1,
BK1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
false,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
false,
BBlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CShuffleBlockTransferScalarPerVector_NPerBlock,
CReduceThreadClusterLengths_MPerBlock_NPerBlock,
CReduceThreadLds2VGprCopySrcDstScalarPerVector_NPerBlock,
CReduceThreadVgpr2GlobalCopySrcDstScalarPerVector_MPerBlock,
LoopSched>;
// Argument
struct Argument : public BaseArgument
{
Argument(const ADataType* p_a_grid,
const BDataType* p_b_grid,
CDataType* p_c_grid,
ReducePtrsGlobal p_reduces_grid,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op,
ReduceInElementwiseOperations reduce_in_element_ops,
ReduceAccElementwiseOperations reduce_out_element_ops)
: p_a_grid_{p_a_grid},
p_b_grid_{p_b_grid},
p_c_grid_{p_c_grid},
p_reduces_grid_{p_reduces_grid},
a_grid_desc_ak0_m_ak1_{DeviceOp::MakeAGridDescriptor_AK0_M_AK1(MRaw, KRaw, StrideA)},
b_grid_desc_bk0_n_bk1_{DeviceOp::MakeBGridDescriptor_BK0_N_BK1(KRaw, NRaw, StrideB)},
c_grid_desc_m_n_{DeviceOp::MakeCGridDescriptor_M_N(MRaw, NRaw, StrideC)},
reduce_grid_desc_m_{DeviceOp::MakeReduceGridDescriptor_M(MRaw)},
c_grid_desc_mblock_mperblock_nblock_nperblock_{},
reduce_grid_desc_mblock_mperblock_{},
block_2_ctile_map_{GridwiseGemm::MakeDefaultBlock2CTileMap(c_grid_desc_m_n_)},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
c_element_op_{c_element_op},
reduce_in_element_ops_{reduce_in_element_ops},
reduce_out_element_ops_{reduce_out_element_ops}
{
if(GridwiseGemm::CheckValidity(a_grid_desc_ak0_m_ak1_,
b_grid_desc_bk0_n_bk1_,
c_grid_desc_m_n_,
block_2_ctile_map_))
{
c_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeCGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
c_grid_desc_m_n_);
reduce_grid_desc_mblock_mperblock_ =
GridwiseGemm::MakeReduceGridDescriptor_MBlock_MPerBlock(reduce_grid_desc_m_);
}
}
// private:
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
CDataType* p_c_grid_;
ReducePtrsGlobal p_reduces_grid_;
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
CGridDesc_M_N c_grid_desc_m_n_;
ReduceGridDesc_M reduce_grid_desc_m_;
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
c_grid_desc_mblock_mperblock_nblock_nperblock_;
typename GridwiseGemm::ReduceGridDescriptor_MBlock_MPerBlock
reduce_grid_desc_mblock_mperblock_;
typename GridwiseGemm::DefaultBlock2CTileMap block_2_ctile_map_;
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CElementwiseOperation c_element_op_;
ReduceInElementwiseOperations reduce_in_element_ops_;
ReduceAccElementwiseOperations reduce_out_element_ops_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
#if 0
{
std::cout << "arg.a_grid_desc_ak0_m_ak1_{"
<< arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) << ", "
<< arg.a_grid_desc_ak0_m_ak1_.GetLength(I1) << ", "
<< arg.a_grid_desc_ak0_m_ak1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.b_grid_desc_bk0_n_bk1_{"
<< arg.b_grid_desc_bk0_n_bk1_.GetLength(I0) << ", "
<< arg.b_grid_desc_bk0_n_bk1_.GetLength(I1) << ", "
<< arg.b_grid_desc_bk0_n_bk1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.c_grid_desc_m_n_{ " << arg.c_grid_desc_m_n_.GetLength(I0) << ", "
<< arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
std::cout << "arg.reduce_grid_desc_m_{ " << arg.reduce_grid_desc_m_.GetLength(I0) << "}"
<< std::endl;
}
#endif
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_))
{
throw std::runtime_error("wrong! GridwiseGemm has invalid setting");
}
const index_t grid_size =
arg.block_2_ctile_map_.CalculateGridSize(arg.c_grid_desc_m_n_);
const auto K =
arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) * arg.a_grid_desc_ak0_m_ak1_.GetLength(I2);
float elapsed_time = 0.0f;
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
const auto kernel = kernel_gemm_reduce_xdl_cshuffle_v1<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
ReducePtrsGlobal,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
ReduceInElementwiseOperations,
ReduceAccElementwiseOperations,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::ReduceGridDescriptor_MBlock_MPerBlock,
typename GridwiseGemm::DefaultBlock2CTileMap,
true>;
elapsed_time =
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.p_reduces_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.reduce_in_element_ops_,
arg.reduce_out_element_ops_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.reduce_grid_desc_mblock_mperblock_,
arg.block_2_ctile_map_);
}
else
{
const auto kernel = kernel_gemm_reduce_xdl_cshuffle_v1<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
ReducePtrsGlobal,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
ReduceInElementwiseOperations,
ReduceAccElementwiseOperations,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::ReduceGridDescriptor_MBlock_MPerBlock,
typename GridwiseGemm::DefaultBlock2CTileMap,
false>;
elapsed_time =
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.p_reduces_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.reduce_in_element_ops_,
arg.reduce_out_element_ops_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.reduce_grid_desc_mblock_mperblock_,
arg.block_2_ctile_map_);
}
return elapsed_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
return GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static constexpr int NumReduce = ReduceOperations::Size();
static auto MakeArgument(const void* p_a,
const void* p_b,
const void* p_bias,
std::array<const void*, 0> p_ds,
void* p_c,
std::array<void*, NumReduce> p_reduces,
ck::index_t M,
ck::index_t N,
ck::index_t K,
ck::index_t StrideA,
ck::index_t StrideB,
ck::index_t StrideC,
std::array<ck::index_t, 0> StrideDs,
std::array<void*, 3> gemm_element_ops,
std::array<void*, 0> d_element_ops,
std::array<void*, NumReduce> reduce_in_element_op,
std::array<void*, NumReduce> reduce_out_element_op)
{
(void)p_bias;
(void)p_ds;
(void)StrideDs;
(void)d_element_ops;
ReducePtrsGlobal reduce_tuple = generate_tuple(
[&](auto I) {
auto tmp = ReducePtrsGlobal{}[I];
using T = remove_pointer_t<decltype(tmp)>;
return static_cast<T*>(p_reduces[I]);
},
Number<NumReduce>{});
ReduceInElementwiseOperations reduce_in_element_ops = generate_tuple(
[&](auto I) {
auto tmp = ReduceInElementwiseOperations{}[I];
using T = remove_pointer_t<decltype(tmp)>;
return *(static_cast<T*>(reduce_in_element_op[I]));
},
Number<NumReduce>{});
ReduceAccElementwiseOperations reduce_out_element_ops = generate_tuple(
[&](auto I) {
auto tmp = ReduceAccElementwiseOperations{}[I];
using T = remove_pointer_t<decltype(tmp)>;
return *(static_cast<T*>(reduce_out_element_op[I]));
},
Number<NumReduce>{});
AElementwiseOperation a_element_op =
*(static_cast<AElementwiseOperation*>(gemm_element_ops[0]));
BElementwiseOperation b_element_op =
*(static_cast<BElementwiseOperation*>(gemm_element_ops[1]));
CElementwiseOperation c_element_op =
*(static_cast<CElementwiseOperation*>(gemm_element_ops[2]));
return Argument{static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
static_cast<CDataType*>(p_c),
reduce_tuple,
M,
N,
K,
StrideA,
StrideB,
StrideC,
a_element_op,
b_element_op,
c_element_op,
reduce_in_element_ops,
reduce_out_element_ops};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_a,
const void* p_b,
const void* p_bias,
std::array<const void*, 0> p_ds,
void* p_c,
std::array<void*, NumReduce> p_reduces,
ck::index_t M,
ck::index_t N,
ck::index_t K,
ck::index_t StrideA,
ck::index_t StrideB,
ck::index_t StrideC,
std::array<ck::index_t, 0> StrideDs,
std::array<void*, 3> gemm_element_ops,
std::array<void*, 0> d_element_ops,
std::array<void*, NumReduce> reduce_in_element_op,
std::array<void*, NumReduce> reduce_out_element_op,
ck::index_t = 1) override
{
(void)p_bias;
(void)p_ds;
(void)StrideDs;
(void)d_element_ops;
ReducePtrsGlobal reduce_tuple = generate_tuple(
[&](auto I) {
auto tmp = ReducePtrsGlobal{}[I];
using T = remove_pointer_t<decltype(tmp)>;
return static_cast<T*>(p_reduces[I]);
},
Number<NumReduce>{});
ReduceInElementwiseOperations reduce_in_element_ops = generate_tuple(
[&](auto I) {
auto tmp = ReduceInElementwiseOperations{}[I];
using T = remove_pointer_t<decltype(tmp)>;
return *(static_cast<T*>(reduce_in_element_op[I]));
},
Number<NumReduce>{});
ReduceAccElementwiseOperations reduce_out_element_ops = generate_tuple(
[&](auto I) {
auto tmp = ReduceAccElementwiseOperations{}[I];
using T = remove_pointer_t<decltype(tmp)>;
return *(static_cast<T*>(reduce_out_element_op[I]));
},
Number<NumReduce>{});
AElementwiseOperation a_element_op =
*(static_cast<AElementwiseOperation*>(gemm_element_ops[0]));
BElementwiseOperation b_element_op =
*(static_cast<BElementwiseOperation*>(gemm_element_ops[1]));
CElementwiseOperation c_element_op =
*(static_cast<CElementwiseOperation*>(gemm_element_ops[2]));
return std::make_unique<Argument>(static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
static_cast<CDataType*>(p_c),
reduce_tuple,
M,
N,
K,
StrideA,
StrideB,
StrideC,
a_element_op,
b_element_op,
c_element_op,
reduce_in_element_ops,
reduce_out_element_ops);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceGemmReduce_Xdl_CShuffle"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< AK1 << ", "
<< BK1
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

547
include/ck/tensor_operation/gpu/device/impl/device_gemm_xdl.hpp Normal file
View File

@@ -0,0 +1,547 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_gemm.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v2r3.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <typename ADataType,
typename BDataType,
typename CDataType,
typename AccDataType,
typename ALayout,
typename BLayout,
typename CLayout,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CElementwiseOperation,
GemmSpecialization GemmSpec,
ck::index_t BlockSize,
ck::index_t MPerBlock,
ck::index_t NPerBlock,
ck::index_t K0PerBlock,
ck::index_t K1,
ck::index_t MPerXDL,
ck::index_t NPerXDL,
ck::index_t MXdlPerWave,
ck::index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_K0_M_K1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
ck::index_t ABlockTransferSrcVectorDim,
ck::index_t ABlockTransferSrcScalarPerVector,
ck::index_t ABlockTransferDstScalarPerVector_K1,
bool ABlockLdsAddExtraM,
typename BBlockTransferThreadClusterLengths_K0_N_K1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
ck::index_t BBlockTransferSrcVectorDim,
ck::index_t BBlockTransferSrcScalarPerVector,
ck::index_t BBlockTransferDstScalarPerVector_K1,
bool BBlockLdsAddExtraN,
ck::index_t CThreadTransferSrcDstVectorDim,
ck::index_t CThreadTransferDstScalarPerVector,
ck::index_t NumPrefetch = 1>
struct DeviceGemmXdl : public DeviceGemm<ALayout,
BLayout,
CLayout,
ADataType,
BDataType,
CDataType,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation>
{
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto K1Number = Number<K1>{};
static auto MakeAGridDescriptor_K0_M_K1(index_t M, index_t K, index_t StrideA)
{
assert(K % K1 == 0);
const index_t K0 = K / K1;
const auto a_grid_desc_m_k = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, ALayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, K), make_tuple(StrideA, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, ALayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, K), make_tuple(I1, StrideA));
}
}();
if constexpr(GemmSpec == GemmSpecialization::MNPadding)
{
const auto PadM = (MPerBlock - M % MPerBlock) % MPerBlock;
return transform_tensor_descriptor(
a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(K0, K1Number)),
make_right_pad_transform(M, PadM)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
else
{
return transform_tensor_descriptor(
a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(K0, K1Number)),
make_pass_through_transform(M)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
}
static auto MakeBGridDescriptor_K0_N_K1(index_t K, index_t N, index_t StrideB)
{
assert(K % K1 == 0);
const index_t K0 = K / K1;
const auto b_grid_desc_k_n = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(K, N), make_tuple(StrideB, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(K, N), make_tuple(I1, StrideB));
}
}();
if constexpr(GemmSpec == GemmSpecialization::MNPadding)
{
const auto PadN = (NPerBlock - N % NPerBlock) % NPerBlock;
return transform_tensor_descriptor(
b_grid_desc_k_n,
make_tuple(make_unmerge_transform(make_tuple(K0, K1Number)),
make_right_pad_transform(N, PadN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
else
{
return transform_tensor_descriptor(
b_grid_desc_k_n,
make_tuple(make_unmerge_transform(make_tuple(K0, K1Number)),
make_pass_through_transform(N)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
}
static auto MakeCGridDescriptor_M_N(index_t M, index_t N, index_t StrideC)
{
const auto c_grid_desc_m_n = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, N), make_tuple(StrideC, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, N), make_tuple(I1, StrideC));
}
}();
if constexpr(GemmSpec == GemmSpecialization::MNPadding)
{
const auto PadM = (MPerBlock - M % MPerBlock) % MPerBlock;
const auto PadN = (NPerBlock - N % NPerBlock) % NPerBlock;
return transform_tensor_descriptor(
c_grid_desc_m_n,
make_tuple(make_right_pad_transform(M, PadM), make_right_pad_transform(N, PadN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else
{
return transform_tensor_descriptor(
c_grid_desc_m_n,
make_tuple(make_pass_through_transform(M), make_pass_through_transform(N)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
}
using AGridDesc_K0_M_K1 = decltype(MakeAGridDescriptor_K0_M_K1(1, 1, 1));
using BGridDesc_K0_N_K1 = decltype(MakeBGridDescriptor_K0_N_K1(1, 1, 1));
using CGridDesc_M_N = decltype(MakeCGridDescriptor_M_N(1, 1, 1));
// GridwiseGemm
using GridwiseGemm = GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r3<
BlockSize,
ADataType, // TODO: distinguish A/B datatype
AccDataType,
CDataType,
InMemoryDataOperationEnum::Set,
AGridDesc_K0_M_K1,
BGridDesc_K0_N_K1,
CGridDesc_M_N,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
MPerBlock,
NPerBlock,
K0PerBlock,
MPerXDL,
NPerXDL,
K1,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_K0_M_K1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_K1,
false, // AThreadTransferSrcResetCoordinateAfterRun,
ABlockLdsAddExtraM,
BBlockTransferThreadClusterLengths_K0_N_K1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_K1,
false, // BThreadTransferSrcResetCoordinateAfterRun,
BBlockLdsAddExtraN,
Sequence<0, 2, 4, 5, 6, 1, 3, 7>, // CThreadTransferSrcDstAccessOrder,
CThreadTransferSrcDstVectorDim,
CThreadTransferDstScalarPerVector,
NumPrefetch>;
// Argument
struct Argument : public BaseArgument
{
Argument(const ADataType* p_a_grid,
const BDataType* p_b_grid,
CDataType* p_c_grid,
index_t M,
index_t N,
index_t K,
index_t StrideA,
index_t StrideB,
index_t StrideC,
index_t M01,
index_t N01,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op)
: p_a_grid_{p_a_grid},
p_b_grid_{p_b_grid},
p_c_grid_{p_c_grid},
a_grid_desc_k0_m_k1_{},
b_grid_desc_k0_n_k1_{},
c_grid_desc_m_n_{},
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_{},
block_2_ctile_map_{},
M01_{M01},
N01_{N01},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
c_element_op_{c_element_op}
{
a_grid_desc_k0_m_k1_ = DeviceGemmXdl::MakeAGridDescriptor_K0_M_K1(M, K, StrideA);
b_grid_desc_k0_n_k1_ = DeviceGemmXdl::MakeBGridDescriptor_K0_N_K1(K, N, StrideB);
c_grid_desc_m_n_ = DeviceGemmXdl::MakeCGridDescriptor_M_N(M, N, StrideC);
block_2_ctile_map_ =
GridwiseGemm::MakeDefaultBlock2CTileMap(c_grid_desc_m_n_, M01, N01);
if(GridwiseGemm::CheckValidity(a_grid_desc_k0_m_k1_,
b_grid_desc_k0_n_k1_,
c_grid_desc_m_n_,
block_2_ctile_map_))
{
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_ =
GridwiseGemm::MakeCGridDescriptor_M0_N0_M1_N1_M2_M3_M4_N2(c_grid_desc_m_n_);
}
}
// private:
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
CDataType* p_c_grid_;
AGridDesc_K0_M_K1 a_grid_desc_k0_m_k1_;
BGridDesc_K0_N_K1 b_grid_desc_k0_n_k1_;
CGridDesc_M_N c_grid_desc_m_n_;
typename GridwiseGemm::CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_;
typename GridwiseGemm::DefaultBlock2CTileMap block_2_ctile_map_;
index_t M01_;
index_t N01_;
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CElementwiseOperation c_element_op_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceGemmXdl::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
#if 0
{
std::cout << "arg.a_grid_desc_k0_m_k1_{" << arg.a_grid_desc_k0_m_k1_.GetLength(I0)
<< ", " << arg.a_grid_desc_k0_m_k1_.GetLength(I1) << ", "
<< arg.a_grid_desc_k0_m_k1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.b_grid_desc_k0_n_k1_{" << arg.b_grid_desc_k0_n_k1_.GetLength(I0)
<< ", " << arg.b_grid_desc_k0_n_k1_.GetLength(I1) << ", "
<< arg.b_grid_desc_k0_n_k1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.c_grid_desc_m_n_{ " << arg.c_grid_desc_m_n_.GetLength(I0) << ", "
<< arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
}
#endif
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_))
{
throw std::runtime_error(
"wrong! GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r3 has invalid setting");
}
const index_t grid_size =
arg.block_2_ctile_map_.CalculateGridSize(arg.c_grid_desc_m_n_);
const auto K =
arg.a_grid_desc_k0_m_k1_.GetLength(I0) * arg.a_grid_desc_k0_m_k1_.GetLength(I2);
float ave_time = 0;
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
const auto kernel = kernel_gemm_xdlops_v2r3<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceGemmXdl::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceGemmXdl::BGridDesc_K0_N_K1>,
remove_reference_t<typename GridwiseGemm::CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2>,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
remove_reference_t<typename GridwiseGemm::DefaultBlock2CTileMap>,
true>;
ave_time = launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.block_2_ctile_map_);
}
else
{
const auto kernel = kernel_gemm_xdlops_v2r3<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceGemmXdl::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceGemmXdl::BGridDesc_K0_N_K1>,
remove_reference_t<typename GridwiseGemm::CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2>,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
remove_reference_t<typename GridwiseGemm::DefaultBlock2CTileMap>,
false>;
ave_time = launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.block_2_ctile_map_);
}
return ave_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
if(ck::get_device_name() == "gfx908")
{
if constexpr(!(is_same_v<AccDataType, float> || is_same_v<AccDataType, float> ||
is_same_v<AccDataType, int32_t>))
{
return false;
}
}
else if(ck::get_device_name() == "gfx90a")
{
if constexpr(!(is_same_v<AccDataType, float> || is_same_v<AccDataType, float> ||
is_same_v<AccDataType, int32_t> || is_same_v<AccDataType, double>))
{
return false;
}
}
else
{
return false;
}
return GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const ADataType* p_a,
const BDataType* p_b,
CDataType* p_c,
index_t M,
index_t N,
index_t K,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op)
{
return Argument{p_a,
p_b,
p_c,
M,
N,
K,
StrideA,
StrideB,
StrideC,
1,
1,
a_element_op,
b_element_op,
c_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,
const void* p_b,
void* p_c,
index_t M,
index_t N,
index_t K,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op) override
{
return std::make_unique<Argument>(static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
static_cast<CDataType*>(p_c),
M,
N,
K,
StrideA,
StrideB,
StrideC,
1,
1,
a_element_op,
b_element_op,
c_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceGemmXdl"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< K0PerBlock << ", "
<< K1 << ", "
<< MPerXDL << ", "
<< NPerXDL << ", "
<< MXdlPerWave << ", "
<< NXdlPerWave
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

677
include/ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle.hpp Normal file
View File

@@ -0,0 +1,677 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_gemm.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_v1.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
// Note: inter-wave loop scheduler is rolled out to c-shuffle version first. Becuase non c-shuffle
// version currently has compiler issues with register spill which further causes validation
// failures.
template <typename ALayout,
typename BLayout,
typename CLayout,
typename ADataType,
typename BDataType,
typename CDataType,
typename GemmAccDataType,
typename CShuffleDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CElementwiseOperation,
GemmSpecialization GemmSpec,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock,
index_t KPerBlock,
index_t AK1,
index_t BK1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
bool ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
bool BBlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CShuffleBlockTransferScalarPerVector_NPerBlock,
LoopScheduler LoopSched = make_default_loop_scheduler()>
struct DeviceGemm_Xdl_CShuffle : public DeviceGemm<ALayout,
BLayout,
CLayout,
ADataType,
BDataType,
CDataType,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation>
{
using DeviceOp = DeviceGemm_Xdl_CShuffle;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static auto MakeAGridDescriptor_AK0_M_AK1(index_t MRaw, index_t KRaw, index_t StrideA)
{
const auto a_grid_desc_mraw_kraw = [&]() {
if constexpr(is_same_v<tensor_layout::gemm::RowMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(StrideA, I1));
}
else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(I1, StrideA));
}
}();
const auto M = math::integer_divide_ceil(MRaw, MPerBlock) * MPerBlock;
const auto K = math::integer_divide_ceil(KRaw, KPerBlock) * KPerBlock;
const auto MPad = M - MRaw;
const auto KPad = K - KRaw;
if constexpr(GemmSpec == GemmSpecialization::MKPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad both M and K
assert(K % AK1 == 0);
const auto AK0 = K / AK1;
const auto a_grid_desc_m_k =
transform_tensor_descriptor(a_grid_desc_mraw_kraw,
make_tuple(make_right_pad_transform(MRaw, MPad),
make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_pass_through_transform(M)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
else if constexpr(GemmSpec == GemmSpecialization::MPadding ||
GemmSpec == GemmSpecialization::MNPadding)
{
// pad M, but not K
assert(KRaw % AK1 == 0);
const auto AK0 = KRaw / AK1;
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_mraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_right_pad_transform(MRaw, MPad)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
else if constexpr(GemmSpec == GemmSpecialization::KPadding ||
GemmSpec == GemmSpecialization::NKPadding)
{
// pad K, but not M
assert(K % AK1 == 0);
const auto AK0 = K / AK1;
const auto a_grid_desc_m_k = transform_tensor_descriptor(
a_grid_desc_mraw_kraw,
make_tuple(make_pass_through_transform(MRaw), make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_pass_through_transform(MRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
else
{
// not pad M or K
assert(KRaw % AK1 == 0);
const auto AK0 = KRaw / AK1;
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_mraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_pass_through_transform(MRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
}
static auto MakeBGridDescriptor_BK0_N_BK1(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
const auto N = math::integer_divide_ceil(NRaw, NPerBlock) * NPerBlock;
const auto K = math::integer_divide_ceil(KRaw, KPerBlock) * KPerBlock;
const auto NPad = N - NRaw;
const auto KPad = K - KRaw;
if constexpr(GemmSpec == GemmSpecialization::NKPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad both N and K
assert(K % BK1 == 0);
const auto BK0 = K / BK1;
const auto b_grid_desc_n_k =
transform_tensor_descriptor(b_grid_desc_nraw_kraw,
make_tuple(make_right_pad_transform(NRaw, NPad),
make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_n_k,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_pass_through_transform(N)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
else if constexpr(GemmSpec == GemmSpecialization::NPadding ||
GemmSpec == GemmSpecialization::MNPadding)
{
// pad N, but not K
assert(KRaw % BK1 == 0);
const auto BK0 = KRaw / BK1;
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_nraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_right_pad_transform(NRaw, NPad)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
else if constexpr(GemmSpec == GemmSpecialization::KPadding ||
GemmSpec == GemmSpecialization::MKPadding)
{
// pad K, but not N
assert(K % BK1 == 0);
const auto BK0 = K / BK1;
const auto b_grid_desc_n_k = transform_tensor_descriptor(
b_grid_desc_nraw_kraw,
make_tuple(make_pass_through_transform(NRaw), make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_n_k,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_pass_through_transform(NRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
else
{
// not pad N or K
assert(KRaw % BK1 == 0);
const auto BK0 = KRaw / BK1;
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_nraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_pass_through_transform(NRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
}
static auto MakeCGridDescriptor_M_N(index_t MRaw, index_t NRaw, index_t StrideC)
{
const auto c_grid_desc_mraw_nraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(StrideC, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(I1, StrideC));
}
}();
const auto M = math::integer_divide_ceil(MRaw, MPerBlock) * MPerBlock;
const auto N = math::integer_divide_ceil(NRaw, NPerBlock) * NPerBlock;
const auto MPad = M - MRaw;
const auto NPad = N - NRaw;
if constexpr(GemmSpec == GemmSpecialization::MNPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad M and N
return transform_tensor_descriptor(c_grid_desc_mraw_nraw,
make_tuple(make_right_pad_transform(MRaw, MPad),
make_right_pad_transform(NRaw, NPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else if constexpr(GemmSpec == GemmSpecialization::MPadding ||
GemmSpec == GemmSpecialization::MKPadding)
{
// pad M, but not N
return transform_tensor_descriptor(
c_grid_desc_mraw_nraw,
make_tuple(make_right_pad_transform(MRaw, MPad), make_pass_through_transform(NRaw)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else if constexpr(GemmSpec == GemmSpecialization::NPadding ||
GemmSpec == GemmSpecialization::NKPadding)
{
// pad N, but not M
return transform_tensor_descriptor(
c_grid_desc_mraw_nraw,
make_tuple(make_pass_through_transform(MRaw), make_right_pad_transform(NRaw, NPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else
{
// not pad M or N
return c_grid_desc_mraw_nraw;
}
}
using AGridDesc_AK0_M_AK1 = decltype(MakeAGridDescriptor_AK0_M_AK1(1, 1, 1));
using BGridDesc_BK0_N_BK1 = decltype(MakeBGridDescriptor_BK0_N_BK1(1, 1, 1));
using CGridDesc_M_N = decltype(MakeCGridDescriptor_M_N(1, 1, 1));
// GridwiseGemm
using GridwiseGemm = GridwiseGemm_k0mk1_k0nk1_mn_xdl_cshuffle_v1<
ADataType, // TODO: distinguish A/B datatype
GemmAccDataType,
CShuffleDataType,
CDataType,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
InMemoryDataOperationEnum::Set,
AGridDesc_AK0_M_AK1,
BGridDesc_BK0_N_BK1,
CGridDesc_M_N,
NumGemmKPrefetchStage,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
AK1,
BK1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
false,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
false,
BBlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CShuffleBlockTransferScalarPerVector_NPerBlock,
LoopSched>;
// Argument
struct Argument : public BaseArgument
{
Argument(const ADataType* p_a_grid,
const BDataType* p_b_grid,
CDataType* p_c_grid,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op)
: p_a_grid_{p_a_grid},
p_b_grid_{p_b_grid},
p_c_grid_{p_c_grid},
a_grid_desc_ak0_m_ak1_{DeviceOp::MakeAGridDescriptor_AK0_M_AK1(MRaw, KRaw, StrideA)},
b_grid_desc_bk0_n_bk1_{DeviceOp::MakeBGridDescriptor_BK0_N_BK1(KRaw, NRaw, StrideB)},
c_grid_desc_m_n_{DeviceOp::MakeCGridDescriptor_M_N(MRaw, NRaw, StrideC)},
c_grid_desc_mblock_mperblock_nblock_nperblock_{},
block_2_ctile_map_{GridwiseGemm::MakeDefaultBlock2CTileMap(c_grid_desc_m_n_)},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
c_element_op_{c_element_op}
{
if(GridwiseGemm::CheckValidity(a_grid_desc_ak0_m_ak1_,
b_grid_desc_bk0_n_bk1_,
c_grid_desc_m_n_,
block_2_ctile_map_))
{
c_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeCGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
c_grid_desc_m_n_);
}
}
// private:
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
CDataType* p_c_grid_;
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
CGridDesc_M_N c_grid_desc_m_n_;
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
c_grid_desc_mblock_mperblock_nblock_nperblock_;
typename GridwiseGemm::DefaultBlock2CTileMap block_2_ctile_map_;
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CElementwiseOperation c_element_op_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
#if 0
{
std::cout << "arg.a_grid_desc_ak0_m_ak1_{"
<< arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) << ", "
<< arg.a_grid_desc_ak0_m_ak1_.GetLength(I1) << ", "
<< arg.a_grid_desc_ak0_m_ak1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.b_grid_desc_bk0_n_bk1_{"
<< arg.b_grid_desc_bk0_n_bk1_.GetLength(I0) << ", "
<< arg.b_grid_desc_bk0_n_bk1_.GetLength(I1) << ", "
<< arg.b_grid_desc_bk0_n_bk1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.c_grid_desc_m_n_{ " << arg.c_grid_desc_m_n_.GetLength(I0) << ", "
<< arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
}
#endif
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_))
{
throw std::runtime_error("wrong! GridwiseGemm has invalid setting");
}
const index_t grid_size =
arg.block_2_ctile_map_.CalculateGridSize(arg.c_grid_desc_m_n_);
const auto K =
arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) * arg.a_grid_desc_ak0_m_ak1_.GetLength(I2);
float ave_time = 0;
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
const auto kernel = kernel_gemm_xdl_cshuffle_v1<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::DefaultBlock2CTileMap,
true>;
ave_time =
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_ctile_map_);
}
else
{
const auto kernel = kernel_gemm_xdl_cshuffle_v1<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::DefaultBlock2CTileMap,
false>;
ave_time =
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_ctile_map_);
}
return ave_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
if(!(ck::get_device_name() == "gfx908" || ck::get_device_name() == "gfx90a"))
{
return false;
}
return GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const ADataType* p_a,
const BDataType* p_b,
CDataType* p_c,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op)
{
return Argument{p_a,
p_b,
p_c,
MRaw,
NRaw,
KRaw,
StrideA,
StrideB,
StrideC,
a_element_op,
b_element_op,
c_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,
const void* p_b,
void* p_c,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op) override
{
return std::make_unique<Argument>(static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
static_cast<CDataType*>(p_c),
MRaw,
NRaw,
KRaw,
StrideA,
StrideB,
StrideC,
a_element_op,
b_element_op,
c_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceGemm_Xdl_CShuffle"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< AK1 << ", "
<< BK1
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

773
include/ck/tensor_operation/gpu/device/impl/device_gemm_xdl_layernorm_cshuffle.hpp Normal file
View File

@@ -0,0 +1,773 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_gemm.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_layernorm_cshuffle_v1.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
// The GEMM + Layernorm implementation is a specialized kernel which allows fusing both layers
// together given the condition GEMM extents N of MNK is spanned by a single workgroup. For example,
// a kernel configured with NPerBlock = 128 allows to operate on all GEMM sizes if N <= 128
//
// Note: inter-wave loop scheduler is rolled out to c-shuffle version first. Becuase non c-shuffle
// version currently has compiler issues with register spill which further causes validation
// failures.
//
// D = Layernorm(acc_element_op(A * B + broadcast(bias)) + add) * broadcast(gamma) + broadcast(beta)
template <typename ALayout,
typename BLayout,
typename CLayout,
typename ADataType,
typename BDataType,
typename CDataType,
typename C0DataType,
typename GemmAccDataType,
typename CShuffleDataType,
typename ReduceAccDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename AccElementwiseOperation,
typename CElementwiseOperation,
GemmSpecialization GemmSpec,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock,
index_t KPerBlock,
index_t AK1,
index_t BK1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
bool ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
bool BBlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CShuffleBlockTransferScalarPerVector_NPerBlock,
typename CReduceThreadClusterLengths_MPerBlock_NPerBlock,
index_t CReduceThreadCopySrcDstScalarPerVector_NPerBlock,
LoopScheduler LoopSched = make_default_loop_scheduler()>
struct DeviceGemmLayerNorm_Xdl_CShuffle : public BaseOperator
{
using DeviceOp = DeviceGemmLayerNorm_Xdl_CShuffle;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static auto MakeAGridDescriptor_AK0_M_AK1(index_t MRaw, index_t KRaw, index_t StrideA)
{
const auto a_grid_desc_mraw_kraw = [&]() {
if constexpr(is_same_v<tensor_layout::gemm::RowMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(StrideA, I1));
}
else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(I1, StrideA));
}
}();
const auto M = math::integer_divide_ceil(MRaw, MPerBlock) * MPerBlock;
const auto K = math::integer_divide_ceil(KRaw, KPerBlock) * KPerBlock;
const auto MPad = M - MRaw;
const auto KPad = K - KRaw;
if constexpr(GemmSpec == GemmSpecialization::MKPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad both M and K
assert(K % AK1 == 0);
const auto AK0 = K / AK1;
const auto a_grid_desc_m_k =
transform_tensor_descriptor(a_grid_desc_mraw_kraw,
make_tuple(make_right_pad_transform(MRaw, MPad),
make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_pass_through_transform(M)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
else if constexpr(GemmSpec == GemmSpecialization::MPadding ||
GemmSpec == GemmSpecialization::MNPadding)
{
// pad M, but not K
assert(KRaw % AK1 == 0);
const auto AK0 = KRaw / AK1;
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_mraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_right_pad_transform(MRaw, MPad)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
else if constexpr(GemmSpec == GemmSpecialization::KPadding ||
GemmSpec == GemmSpecialization::NKPadding)
{
// pad K, but not M
assert(K % AK1 == 0);
const auto AK0 = K / AK1;
const auto a_grid_desc_m_k = transform_tensor_descriptor(
a_grid_desc_mraw_kraw,
make_tuple(make_pass_through_transform(MRaw), make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_pass_through_transform(MRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
else
{
// not pad M or K
assert(KRaw % AK1 == 0);
const auto AK0 = KRaw / AK1;
const auto a_grid_desc_ak0_m_ak1 =
transform_tensor_descriptor(a_grid_desc_mraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(AK0, AK1)),
make_pass_through_transform(MRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return a_grid_desc_ak0_m_ak1;
}
}
static auto MakeBGridDescriptor_BK0_N_BK1(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
const auto N = math::integer_divide_ceil(NRaw, NPerBlock) * NPerBlock;
const auto K = math::integer_divide_ceil(KRaw, KPerBlock) * KPerBlock;
const auto NPad = N - NRaw;
const auto KPad = K - KRaw;
if constexpr(GemmSpec == GemmSpecialization::NKPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad both N and K
assert(K % BK1 == 0);
const auto BK0 = K / BK1;
const auto b_grid_desc_n_k =
transform_tensor_descriptor(b_grid_desc_nraw_kraw,
make_tuple(make_right_pad_transform(NRaw, NPad),
make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_n_k,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_pass_through_transform(N)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
else if constexpr(GemmSpec == GemmSpecialization::NPadding ||
GemmSpec == GemmSpecialization::MNPadding)
{
// pad N, but not K
assert(KRaw % BK1 == 0);
const auto BK0 = KRaw / BK1;
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_nraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_right_pad_transform(NRaw, NPad)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
else if constexpr(GemmSpec == GemmSpecialization::KPadding ||
GemmSpec == GemmSpecialization::MKPadding)
{
// pad K, but not N
assert(K % BK1 == 0);
const auto BK0 = K / BK1;
const auto b_grid_desc_n_k = transform_tensor_descriptor(
b_grid_desc_nraw_kraw,
make_tuple(make_pass_through_transform(NRaw), make_right_pad_transform(KRaw, KPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_n_k,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_pass_through_transform(NRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
else
{
// not pad N or K
assert(KRaw % BK1 == 0);
const auto BK0 = KRaw / BK1;
const auto b_grid_desc_bk0_n_bk1 =
transform_tensor_descriptor(b_grid_desc_nraw_kraw,
make_tuple(make_unmerge_transform(make_tuple(BK0, BK1)),
make_pass_through_transform(NRaw)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
return b_grid_desc_bk0_n_bk1;
}
}
static auto MakeCGridDescriptor_M_N(index_t MRaw, index_t NRaw, index_t StrideC)
{
const auto c_grid_desc_mraw_nraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(StrideC, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(I1, StrideC));
}
}();
const auto M = math::integer_divide_ceil(MRaw, MPerBlock) * MPerBlock;
const auto N = math::integer_divide_ceil(NRaw, NPerBlock) * NPerBlock;
const auto MPad = M - MRaw;
const auto NPad = N - NRaw;
if constexpr(GemmSpec == GemmSpecialization::MNPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad M and N
return transform_tensor_descriptor(c_grid_desc_mraw_nraw,
make_tuple(make_right_pad_transform(MRaw, MPad),
make_right_pad_transform(NRaw, NPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else if constexpr(GemmSpec == GemmSpecialization::MPadding ||
GemmSpec == GemmSpecialization::MKPadding)
{
// pad M, but not N
return transform_tensor_descriptor(
c_grid_desc_mraw_nraw,
make_tuple(make_right_pad_transform(MRaw, MPad), make_pass_through_transform(NRaw)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else if constexpr(GemmSpec == GemmSpecialization::NPadding ||
GemmSpec == GemmSpecialization::NKPadding)
{
// pad N, but not M
return transform_tensor_descriptor(
c_grid_desc_mraw_nraw,
make_tuple(make_pass_through_transform(MRaw), make_right_pad_transform(NRaw, NPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else
{
// not pad M or N
return c_grid_desc_mraw_nraw;
}
}
static auto MakeGridDescriptor_N(index_t NRaw)
{
const auto grid_desc_nraw = make_naive_tensor_descriptor_packed(make_tuple(NRaw));
const auto N = math::integer_divide_ceil(NRaw, NPerBlock) * NPerBlock;
const auto NPad = N - NRaw;
if constexpr(GemmSpec == GemmSpecialization::NPadding ||
GemmSpec == GemmSpecialization::MNPadding ||
GemmSpec == GemmSpecialization::NKPadding ||
GemmSpec == GemmSpecialization::MNKPadding)
{
// pad N
return transform_tensor_descriptor(grid_desc_nraw,
make_tuple(make_right_pad_transform(NRaw, NPad)),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0>{}));
}
else
{
// not pad N
return grid_desc_nraw;
}
}
using AGridDesc_AK0_M_AK1 = decltype(MakeAGridDescriptor_AK0_M_AK1(1, 1, 1));
using BGridDesc_BK0_N_BK1 = decltype(MakeBGridDescriptor_BK0_N_BK1(1, 1, 1));
using CGridDesc_M_N = decltype(MakeCGridDescriptor_M_N(1, 1, 1));
using C0GridDesc_N = decltype(MakeGridDescriptor_N(1));
// GridwiseGemm
using GridwiseGemm = GridwiseGemmLayernorm_k0mk1_k0nk1_mn_xdl_cshuffle_v1<
ADataType, // TODO: distinguish A/B datatype
GemmAccDataType,
CShuffleDataType,
CDataType,
C0DataType,
ReduceAccDataType,
AElementwiseOperation,
BElementwiseOperation,
AccElementwiseOperation,
CElementwiseOperation,
InMemoryDataOperationEnum::Set,
AGridDesc_AK0_M_AK1,
BGridDesc_BK0_N_BK1,
CGridDesc_M_N,
C0GridDesc_N,
NumGemmKPrefetchStage,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
AK1,
BK1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
false,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
false,
BBlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CShuffleBlockTransferScalarPerVector_NPerBlock,
CReduceThreadClusterLengths_MPerBlock_NPerBlock,
CReduceThreadCopySrcDstScalarPerVector_NPerBlock,
LoopSched>;
using Block2CTileMap = typename GridwiseGemm::DefaultBlock2CTileMap;
// Argument
struct Argument : public BaseArgument
{
Argument(const ADataType* p_a_grid,
const BDataType* p_b_grid,
CDataType* p_c_grid,
const C0DataType* p_c0_grid_add,
const C0DataType* p_c0_grid_bias,
const C0DataType* p_c0_grid_gamma,
const C0DataType* p_c0_grid_beta,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
AccElementwiseOperation acc_element_op,
CElementwiseOperation c_element_op)
: p_a_grid_{p_a_grid},
p_b_grid_{p_b_grid},
p_c_grid_{p_c_grid},
p_c0_grid_bias_{p_c0_grid_bias},
p_c0_grid_add_{p_c0_grid_add},
p_c0_grid_gamma_{p_c0_grid_gamma},
p_c0_grid_beta_{p_c0_grid_beta},
a_grid_desc_ak0_m_ak1_{DeviceOp::MakeAGridDescriptor_AK0_M_AK1(MRaw, KRaw, StrideA)},
b_grid_desc_bk0_n_bk1_{DeviceOp::MakeBGridDescriptor_BK0_N_BK1(KRaw, NRaw, StrideB)},
c_grid_desc_m_n_{DeviceOp::MakeCGridDescriptor_M_N(MRaw, NRaw, StrideC)},
c0_grid_desc_n_{MakeGridDescriptor_N(NRaw)},
c_grid_desc_mblock_mperblock_nblock_nperblock_{},
c0_grid_desc_nblock_nperblock_{},
block_2_ctile_map_{Block2CTileMap(c_grid_desc_m_n_)},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
acc_element_op_{acc_element_op},
c_element_op_{c_element_op}
{
if(GridwiseGemm::CheckValidity(a_grid_desc_ak0_m_ak1_,
b_grid_desc_bk0_n_bk1_,
c_grid_desc_m_n_,
block_2_ctile_map_))
{
c_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeCGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
c_grid_desc_m_n_);
c0_grid_desc_nblock_nperblock_ =
GridwiseGemm::MakeC0GridDescriptor_NBlock_NPerBlock(c0_grid_desc_n_);
}
}
// private:
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
CDataType* p_c_grid_;
const C0DataType* p_c0_grid_bias_;
const C0DataType* p_c0_grid_add_;
const C0DataType* p_c0_grid_gamma_;
const C0DataType* p_c0_grid_beta_;
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
CGridDesc_M_N c_grid_desc_m_n_;
C0GridDesc_N c0_grid_desc_n_;
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
c_grid_desc_mblock_mperblock_nblock_nperblock_;
typename GridwiseGemm::C0GridDescriptor_NBlock_NPerBlock c0_grid_desc_nblock_nperblock_;
Block2CTileMap block_2_ctile_map_;
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
AccElementwiseOperation acc_element_op_;
CElementwiseOperation c_element_op_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
#if 0
{
std::cout << "arg.a_grid_desc_ak0_m_ak1_{"
<< arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) << ", "
<< arg.a_grid_desc_ak0_m_ak1_.GetLength(I1) << ", "
<< arg.a_grid_desc_ak0_m_ak1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.b_grid_desc_bk0_n_bk1_{"
<< arg.b_grid_desc_bk0_n_bk1_.GetLength(I0) << ", "
<< arg.b_grid_desc_bk0_n_bk1_.GetLength(I1) << ", "
<< arg.b_grid_desc_bk0_n_bk1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.c_grid_desc_m_n_{ " << arg.c_grid_desc_m_n_.GetLength(I0) << ", "
<< arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
}
#endif
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_))
{
throw std::runtime_error("wrong! GridwiseGemm has invalid setting");
}
const index_t grid_size =
arg.block_2_ctile_map_.CalculateGridSize(arg.c_grid_desc_m_n_);
const auto K =
arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) * arg.a_grid_desc_ak0_m_ak1_.GetLength(I2);
float ave_time = 0;
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
const auto kernel = kernel_gemm_layernorm_xdl_cshuffle_v1<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
C0DataType,
AElementwiseOperation,
BElementwiseOperation,
AccElementwiseOperation,
CElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::C0GridDescriptor_NBlock_NPerBlock,
Block2CTileMap,
true>;
ave_time =
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.p_c0_grid_bias_,
arg.p_c0_grid_add_,
arg.p_c0_grid_gamma_,
arg.p_c0_grid_beta_,
arg.a_element_op_,
arg.b_element_op_,
arg.acc_element_op_,
arg.c_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.c0_grid_desc_nblock_nperblock_,
arg.block_2_ctile_map_);
}
else
{
const auto kernel = kernel_gemm_layernorm_xdl_cshuffle_v1<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
C0DataType,
AElementwiseOperation,
BElementwiseOperation,
AccElementwiseOperation,
CElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::C0GridDescriptor_NBlock_NPerBlock,
Block2CTileMap,
false>;
ave_time =
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.p_c0_grid_bias_,
arg.p_c0_grid_add_,
arg.p_c0_grid_gamma_,
arg.p_c0_grid_beta_,
arg.a_element_op_,
arg.b_element_op_,
arg.acc_element_op_,
arg.c_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.c0_grid_desc_nblock_nperblock_,
arg.block_2_ctile_map_);
}
return ave_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
if(!(ck::get_device_name() == "gfx908" || ck::get_device_name() == "gfx90a"))
{
return false;
}
return GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const ADataType* p_a,
const BDataType* p_b,
CDataType* p_c,
const C0DataType* p_c0_bias,
const C0DataType* p_c0_add,
const C0DataType* p_c0_gamma,
const C0DataType* p_c0_beta,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
AccElementwiseOperation acc_element_op,
CElementwiseOperation c_element_op)
{
return Argument{p_a,
p_b,
p_c,
p_c0_bias,
p_c0_add,
p_c0_gamma,
p_c0_beta,
MRaw,
NRaw,
KRaw,
StrideA,
StrideB,
StrideC,
a_element_op,
b_element_op,
acc_element_op,
c_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,
const void* p_b,
void* p_c,
const void* p_c0_bias,
const void* p_c0_add,
const void* p_c0_gamma,
const void* p_c0_beta,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
AccElementwiseOperation acc_element_op,
CElementwiseOperation c_element_op,
index_t /* KBatch */ = 1)
{
return std::make_unique<Argument>(static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
static_cast<CDataType*>(p_c),
static_cast<const C0DataType*>(p_c0_bias),
static_cast<const C0DataType*>(p_c0_add),
static_cast<const C0DataType*>(p_c0_gamma),
static_cast<const C0DataType*>(p_c0_beta),
MRaw,
NRaw,
KRaw,
StrideA,
StrideB,
StrideC,
a_element_op,
b_element_op,
acc_element_op,
c_element_op);
}
std::unique_ptr<BaseInvoker> MakeInvokerPointer()
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceGemmLayerNorm_Xdl_CShuffle"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< AK1 << ", "
<< BK1
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

523
include/ck/tensor_operation/gpu/device/impl/device_gemm_xdl_skip_b_lds.hpp Normal file
View File

@@ -0,0 +1,523 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_gemm.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_skip_b_lds_v1.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <typename ADataType,
typename BDataType,
typename CDataType,
typename AccDataType,
typename ALayout,
typename BLayout,
typename CLayout,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CElementwiseOperation,
GemmSpecialization GemmSpec,
ck::index_t BlockSize,
ck::index_t MPerBlock,
ck::index_t NPerBlock,
ck::index_t K0PerBlock,
ck::index_t K1,
ck::index_t MPerXDL,
ck::index_t NPerXDL,
ck::index_t MXdlPerWave,
ck::index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_K0_M_K1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
ck::index_t ABlockTransferSrcVectorDim,
ck::index_t ABlockTransferSrcScalarPerVector,
ck::index_t ABlockTransferDstScalarPerVector_K1,
bool ABlockLdsAddExtraM,
ck::index_t BBlockTransferSrcScalarPerVector,
ck::index_t BBlockBufferSize,
ck::index_t CThreadTransferSrcDstVectorDim,
ck::index_t CThreadTransferDstScalarPerVector>
struct DeviceGemmXdlSkipBLds : public DeviceGemm<ALayout,
BLayout,
CLayout,
ADataType,
BDataType,
CDataType,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation>
{
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto K1Number = Number<K1>{};
static_assert(BBlockBufferSize >= 2);
static auto MakeAGridDescriptor_K0_M_K1(index_t M, index_t K, index_t StrideA)
{
assert(K % K1 == 0);
const index_t K0 = K / K1;
const auto a_grid_desc_m_k = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, ALayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, K), make_tuple(StrideA, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, ALayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, K), make_tuple(I1, StrideA));
}
}();
if constexpr(GemmSpec == GemmSpecialization::MNPadding)
{
const auto PadM = (MPerBlock - M % MPerBlock) % MPerBlock;
return transform_tensor_descriptor(
a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(K0, K1Number)),
make_right_pad_transform(M, PadM)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
else
{
return transform_tensor_descriptor(
a_grid_desc_m_k,
make_tuple(make_unmerge_transform(make_tuple(K0, K1Number)),
make_pass_through_transform(M)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
}
static auto MakeBGridDescriptor_K0_N_K1(index_t K, index_t N, index_t StrideB)
{
assert(K % K1 == 0);
const index_t K0 = K / K1;
const auto b_grid_desc_k_n = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(K, N), make_tuple(StrideB, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(K, N), make_tuple(I1, StrideB));
}
}();
if constexpr(GemmSpec == GemmSpecialization::MNPadding)
{
const auto PadN = (NPerBlock - N % NPerBlock) % NPerBlock;
return transform_tensor_descriptor(
b_grid_desc_k_n,
make_tuple(make_unmerge_transform(make_tuple(K0, K1Number)),
make_right_pad_transform(N, PadN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
else
{
return transform_tensor_descriptor(
b_grid_desc_k_n,
make_tuple(make_unmerge_transform(make_tuple(K0, K1Number)),
make_pass_through_transform(N)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
}
}
static auto MakeCGridDescriptor_M_N(index_t M, index_t N, index_t StrideC)
{
const auto c_grid_desc_m_n = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, N), make_tuple(StrideC, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, N), make_tuple(I1, StrideC));
}
}();
if constexpr(GemmSpec == GemmSpecialization::MNPadding)
{
const auto PadM = (MPerBlock - M % MPerBlock) % MPerBlock;
const auto PadN = (NPerBlock - N % NPerBlock) % NPerBlock;
return transform_tensor_descriptor(
c_grid_desc_m_n,
make_tuple(make_right_pad_transform(M, PadM), make_right_pad_transform(N, PadN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else
{
return transform_tensor_descriptor(
c_grid_desc_m_n,
make_tuple(make_pass_through_transform(M), make_pass_through_transform(N)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
}
using AGridDesc_K0_M_K1 = decltype(MakeAGridDescriptor_K0_M_K1(1, 1, 1));
using BGridDesc_K0_N_K1 = decltype(MakeBGridDescriptor_K0_N_K1(1, 1, 1));
using CGridDesc_M_N = decltype(MakeCGridDescriptor_M_N(1, 1, 1));
// GridwiseGemm
using GridwiseGemm = GridwiseGemm_k0mk1_k0nk1_mn_xdlops_skip_b_lds_v1<
BlockSize,
ADataType, // TODO: distinguish A/B datatype
AccDataType,
CDataType,
InMemoryDataOperationEnum::Set,
AGridDesc_K0_M_K1,
BGridDesc_K0_N_K1,
CGridDesc_M_N,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
MPerBlock,
NPerBlock,
K0PerBlock,
MPerXDL,
NPerXDL,
K1,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_K0_M_K1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_K1,
false, // AThreadTransferSrcResetCoordinateAfterRun,
ABlockLdsAddExtraM,
BBlockTransferSrcScalarPerVector,
false, // BThreadTransferSrcResetCoordinateAfterRun,
BBlockBufferSize,
Sequence<0, 2, 4, 5, 6, 1, 3, 7>, // CThreadTransferSrcDstAccessOrder,
CThreadTransferSrcDstVectorDim,
CThreadTransferDstScalarPerVector>;
// Argument
struct Argument : public BaseArgument
{
Argument(const ADataType* p_a_grid,
const BDataType* p_b_grid,
CDataType* p_c_grid,
index_t M,
index_t N,
index_t K,
index_t StrideA,
index_t StrideB,
index_t StrideC,
index_t M01,
index_t N01,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op)
: p_a_grid_{p_a_grid},
p_b_grid_{p_b_grid},
p_c_grid_{p_c_grid},
a_grid_desc_k0_m_k1_{},
b_grid_desc_k0_n_k1_{},
c_grid_desc_m_n_{},
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_{},
block_2_ctile_map_{},
M01_{M01},
N01_{N01},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
c_element_op_{c_element_op}
{
a_grid_desc_k0_m_k1_ =
DeviceGemmXdlSkipBLds::MakeAGridDescriptor_K0_M_K1(M, K, StrideA);
b_grid_desc_k0_n_k1_ =
DeviceGemmXdlSkipBLds::MakeBGridDescriptor_K0_N_K1(K, N, StrideB);
c_grid_desc_m_n_ = DeviceGemmXdlSkipBLds::MakeCGridDescriptor_M_N(M, N, StrideC);
if(GridwiseGemm::CheckValidity(
a_grid_desc_k0_m_k1_, b_grid_desc_k0_n_k1_, c_grid_desc_m_n_, M01_, N01_))
{
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_ =
GridwiseGemm::MakeCGridDescriptor_M0_N0_M1_N1_M2_M3_M4_N2(c_grid_desc_m_n_);
block_2_ctile_map_ =
GridwiseGemm::MakeDefaultBlock2CTileMap(c_grid_desc_m_n_, M01, N01);
b_grid_desc_k0_k1_k2_n0_n1_n2_n3_k3_ =
GridwiseGemm::MakeBGridDescriptor_K0_K1_K2_N0_N1_N2_N3_K3(b_grid_desc_k0_n_k1_);
}
}
// private:
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
CDataType* p_c_grid_;
AGridDesc_K0_M_K1 a_grid_desc_k0_m_k1_;
BGridDesc_K0_N_K1 b_grid_desc_k0_n_k1_;
CGridDesc_M_N c_grid_desc_m_n_;
typename GridwiseGemm::BGridDesc_K0_K1_K2_N0_N1_N2_N3_K3
b_grid_desc_k0_k1_k2_n0_n1_n2_n3_k3_;
typename GridwiseGemm::CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_;
typename GridwiseGemm::DefaultBlock2CTileMap block_2_ctile_map_;
index_t M01_;
index_t N01_;
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CElementwiseOperation c_element_op_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceGemmXdlSkipBLds::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
{
std::cout << "arg.a_grid_desc_k0_m_k1_{" << arg.a_grid_desc_k0_m_k1_.GetLength(I0)
<< ", " << arg.a_grid_desc_k0_m_k1_.GetLength(I1) << ", "
<< arg.a_grid_desc_k0_m_k1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.b_grid_desc_k0_n_k1_{" << arg.b_grid_desc_k0_n_k1_.GetLength(I0)
<< ", " << arg.b_grid_desc_k0_n_k1_.GetLength(I1) << ", "
<< arg.b_grid_desc_k0_n_k1_.GetLength(I2) << "}" << std::endl;
std::cout << "arg.c_grid_desc_m_n_{ " << arg.c_grid_desc_m_n_.GetLength(I0) << ", "
<< arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
}
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.M01_,
arg.N01_))
{
throw std::runtime_error(
"wrong! GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r3 has invalid setting");
}
const index_t grid_size = GridwiseGemm::CalculateGridSize(arg.c_grid_desc_m_n_);
const auto K0 = arg.a_grid_desc_k0_m_k1_.GetLength(I0);
const bool has_main_k0_block_loop = GridwiseGemm::CalculateHasMainK0BlockLoop(K0);
float ave_time = 0;
if(has_main_k0_block_loop)
{
const auto kernel = kernel_gemm_xdlops_skip_b_lds_v1<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceGemmXdlSkipBLds::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceGemmXdlSkipBLds::BGridDesc_K0_N_K1>,
remove_reference_t<typename GridwiseGemm::BGridDesc_K0_K1_K2_N0_N1_N2_N3_K3>,
remove_reference_t<typename GridwiseGemm::CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2>,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
remove_reference_t<typename GridwiseGemm::DefaultBlock2CTileMap>,
true>;
ave_time = launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_k1_k2_n0_n1_n2_n3_k3_,
arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.block_2_ctile_map_);
}
else
{
const auto kernel = kernel_gemm_xdlops_skip_b_lds_v1<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceGemmXdlSkipBLds::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceGemmXdlSkipBLds::BGridDesc_K0_N_K1>,
remove_reference_t<typename GridwiseGemm::BGridDesc_K0_K1_K2_N0_N1_N2_N3_K3>,
remove_reference_t<typename GridwiseGemm::CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2>,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
remove_reference_t<typename GridwiseGemm::DefaultBlock2CTileMap>,
false>;
ave_time = launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_k1_k2_n0_n1_n2_n3_k3_,
arg.c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.block_2_ctile_map_);
}
return ave_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
return GridwiseGemm::CheckValidity(arg.a_grid_desc_k0_m_k1_,
arg.b_grid_desc_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.M01_,
arg.N01_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const ADataType* p_a,
const BDataType* p_b,
CDataType* p_c,
index_t M,
index_t N,
index_t K,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op)
{
return Argument{p_a,
p_b,
p_c,
M,
N,
K,
StrideA,
StrideB,
StrideC,
1,
1,
a_element_op,
b_element_op,
c_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,
const void* p_b,
void* p_c,
index_t M,
index_t N,
index_t K,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op) override
{
return std::make_unique<Argument>(static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
static_cast<CDataType*>(p_c),
M,
N,
K,
StrideA,
StrideB,
StrideC,
1,
1,
a_element_op,
b_element_op,
c_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceGemmXdlSkipBLds"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< K0PerBlock << ", "
<< K1 << ", "
<< MPerXDL << ", "
<< NPerXDL << ", "
<< MXdlPerWave << ", "
<< NXdlPerWave
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

650
include/ck/tensor_operation/gpu/device/impl/device_gemm_xdl_splitk_c_shuffle.hpp Normal file
View File

@@ -0,0 +1,650 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_gemm_splitk.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v2r4r2.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <typename ADataType,
typename BDataType,
typename CDataType,
typename AccDataType,
typename ALayout,
typename BLayout,
typename CLayout,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CElementwiseOperation,
GemmSpecialization GemmSpec,
ck::index_t BlockSize,
ck::index_t MPerBlock,
ck::index_t NPerBlock,
ck::index_t K0PerBlock,
ck::index_t K1,
ck::index_t MPerXDL,
ck::index_t NPerXDL,
ck::index_t MXdlPerWave,
ck::index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_K0_M_K1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
ck::index_t ABlockTransferSrcVectorDim,
ck::index_t ABlockTransferSrcScalarPerVector,
ck::index_t ABlockTransferDstScalarPerVector_K1,
bool ABlockLdsAddExtraM,
typename BBlockTransferThreadClusterLengths_K0_N_K1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
ck::index_t BBlockTransferSrcVectorDim,
ck::index_t BBlockTransferSrcScalarPerVector,
ck::index_t BBlockTransferDstScalarPerVector_K1,
bool BBlockLdsAddExtraN,
index_t CShuffleMRepeatPerShuffle,
index_t CShuffleNRepeatPerShuffle,
typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CBlockTransferScalarPerVector_NWaveNPerXDL>
struct DeviceGemmXdlSplitKCShuffle : public DeviceGemmSplitK<ALayout,
BLayout,
CLayout,
ADataType,
BDataType,
CDataType,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation>
{
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr auto K1Number = Number<K1>{};
static auto
MakeAGridDescriptor_KBatch_K0_M_K1(index_t M, index_t K, index_t StrideA, int KBatch, int KPad)
{
assert(KPad % (K1 * KBatch) == 0);
const index_t K0 = KPad / (K1 * KBatch);
const auto a_grid_desc_m_k = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, ALayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, K), make_tuple(StrideA, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, ALayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, K), make_tuple(I1, StrideA));
}
}();
const auto a_grid_desc_m_kpad = transform_tensor_descriptor(
a_grid_desc_m_k,
make_tuple(make_pass_through_transform(M), make_right_pad_transform(K, KPad - K)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
if constexpr(GemmSpec == GemmSpecialization::MNPadding)
{
const auto PadM = (MPerBlock - M % MPerBlock) % MPerBlock;
return transform_tensor_descriptor(
a_grid_desc_m_kpad,
make_tuple(make_unmerge_transform(make_tuple(KBatch, K0, K1Number)),
make_right_pad_transform(M, PadM)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 1, 3>{}, Sequence<2>{}));
}
else
{
return transform_tensor_descriptor(
a_grid_desc_m_kpad,
make_tuple(make_unmerge_transform(make_tuple(KBatch, K0, K1Number)),
make_pass_through_transform(M)),
make_tuple(Sequence<1>{}, Sequence<0>{}),
make_tuple(Sequence<0, 1, 3>{}, Sequence<2>{}));
}
}
static auto
MakeBGridDescriptor_KBatch_K0_N_K1(index_t K, index_t N, index_t StrideB, int KBatch, int KPad)
{
assert(KPad % (K1 * KBatch) == 0);
const index_t K0 = KPad / (K1 * KBatch);
const auto b_grid_desc_k_n = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(K, N), make_tuple(StrideB, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(K, N), make_tuple(I1, StrideB));
}
}();
const auto b_grid_desc_kpad_n = transform_tensor_descriptor(
b_grid_desc_k_n,
make_tuple(make_right_pad_transform(K, KPad - K), make_pass_through_transform(N)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
if constexpr(GemmSpec == GemmSpecialization::MNPadding)
{
const auto PadN = (NPerBlock - N % NPerBlock) % NPerBlock;
return transform_tensor_descriptor(
b_grid_desc_kpad_n,
make_tuple(make_unmerge_transform(make_tuple(KBatch, K0, K1Number)),
make_right_pad_transform(N, PadN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 1, 3>{}, Sequence<2>{}));
}
else
{
return transform_tensor_descriptor(
b_grid_desc_kpad_n,
make_tuple(make_unmerge_transform(make_tuple(KBatch, K0, K1Number)),
make_pass_through_transform(N)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 1, 3>{}, Sequence<2>{}));
}
}
static auto MakeCGridDescriptor_M_N(index_t M, index_t N, index_t StrideC)
{
const auto c_grid_desc_m_n = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, N), make_tuple(StrideC, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, CLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(M, N), make_tuple(I1, StrideC));
}
}();
if constexpr(GemmSpec == GemmSpecialization::MNPadding)
{
const auto PadM = (MPerBlock - M % MPerBlock) % MPerBlock;
const auto PadN = (NPerBlock - N % NPerBlock) % NPerBlock;
return transform_tensor_descriptor(
c_grid_desc_m_n,
make_tuple(make_right_pad_transform(M, PadM), make_right_pad_transform(N, PadN)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
else
{
return transform_tensor_descriptor(
c_grid_desc_m_n,
make_tuple(make_pass_through_transform(M), make_pass_through_transform(N)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
}
static auto GetKPad(index_t K, index_t KBatch)
{
const index_t K0 = math::integer_divide_ceil(K, K1 * K0PerBlock * KBatch) * K0PerBlock;
const index_t KPad = KBatch * K0 * K1;
return KPad;
}
using AGridDesc_K0_M_K1 = decltype(MakeAGridDescriptor_KBatch_K0_M_K1(1, 1, 1, 1, 1));
using BGridDesc_K0_N_K1 = decltype(MakeBGridDescriptor_KBatch_K0_N_K1(1, 1, 1, 1, 1));
using CGridDesc_M_N = decltype(MakeCGridDescriptor_M_N(1, 1, 1));
// GridwiseGemm
using GridwiseGemm = GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_v2r4r2<
BlockSize,
ADataType, // TODO: distinguish A/B datatype
AccDataType,
CDataType,
InMemoryDataOperationEnum::Set,
AGridDesc_K0_M_K1,
BGridDesc_K0_N_K1,
CGridDesc_M_N,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
MPerBlock,
NPerBlock,
K0PerBlock,
MPerXDL,
NPerXDL,
K1,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_K0_M_K1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_K1,
false, // AThreadTransferSrcResetCoordinateAfterRun,
ABlockLdsAddExtraM,
BBlockTransferThreadClusterLengths_K0_N_K1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_K1,
false, // BThreadTransferSrcResetCoordinateAfterRun,
BBlockLdsAddExtraN,
CShuffleMRepeatPerShuffle,
CShuffleNRepeatPerShuffle,
CBlockTransferScalarPerVector_NWaveNPerXDL,
CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock>;
// GridwiseGemm
using GridwiseGemmAtomicAdd = GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_v2r4r2<
BlockSize,
ADataType, // TODO: distinguish A/B datatype
AccDataType,
CDataType,
InMemoryDataOperationEnum::AtomicAdd,
AGridDesc_K0_M_K1,
BGridDesc_K0_N_K1,
CGridDesc_M_N,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
MPerBlock,
NPerBlock,
K0PerBlock,
MPerXDL,
NPerXDL,
K1,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_K0_M_K1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_K1,
false, // AThreadTransferSrcResetCoordinateAfterRun,
ABlockLdsAddExtraM,
BBlockTransferThreadClusterLengths_K0_N_K1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_K1,
false, // BThreadTransferSrcResetCoordinateAfterRun,
BBlockLdsAddExtraN,
CShuffleMRepeatPerShuffle,
CShuffleNRepeatPerShuffle,
CBlockTransferScalarPerVector_NWaveNPerXDL,
CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock>;
using CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock =
decltype(GridwiseGemm::MakeCGridDesc_MBlock_MPerBlock_NBlock_NPerBlock(CGridDesc_M_N{}));
using Block2CTileMap = typename GridwiseGemm::CBlockClusterAdaptor;
// Argument
struct Argument : public BaseArgument
{
Argument(const ADataType* p_a_grid,
const BDataType* p_b_grid,
CDataType* p_c_grid,
index_t M,
index_t N,
index_t K,
index_t StrideA,
index_t StrideB,
index_t StrideC,
index_t M01,
index_t N01,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op,
index_t k_batch)
: p_a_grid_{p_a_grid},
p_b_grid_{p_b_grid},
p_c_grid_{p_c_grid},
a_grid_desc_kbatch_k0_m_k1_{},
b_grid_desc_kbatch_k0_n_k1_{},
c_grid_desc_m_n_{},
c_grid_desc_mblock_mperblock_nblock_nperblock_{},
block_2_ctile_map_{},
M01_{M01},
N01_{N01},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
c_element_op_{c_element_op},
k_batch_{k_batch}
{
int KPad = DeviceGemmXdlSplitKCShuffle::GetKPad(K, k_batch_);
a_grid_desc_kbatch_k0_m_k1_ =
DeviceGemmXdlSplitKCShuffle::MakeAGridDescriptor_KBatch_K0_M_K1(
M, K, StrideA, k_batch_, KPad);
b_grid_desc_kbatch_k0_n_k1_ =
DeviceGemmXdlSplitKCShuffle::MakeBGridDescriptor_KBatch_K0_N_K1(
K, N, StrideB, k_batch_, KPad);
c_grid_desc_m_n_ = DeviceGemmXdlSplitKCShuffle::MakeCGridDescriptor_M_N(M, N, StrideC);
block_2_ctile_map_ =
GridwiseGemm::MakeCBlockClusterAdaptor(c_grid_desc_m_n_, M01, N01, k_batch_);
if(GridwiseGemm::CheckValidity(a_grid_desc_kbatch_k0_m_k1_,
b_grid_desc_kbatch_k0_n_k1_,
c_grid_desc_m_n_,
block_2_ctile_map_))
{
c_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeCGridDesc_MBlock_MPerBlock_NBlock_NPerBlock(c_grid_desc_m_n_);
}
}
// private:
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
CDataType* p_c_grid_;
AGridDesc_K0_M_K1 a_grid_desc_kbatch_k0_m_k1_;
BGridDesc_K0_N_K1 b_grid_desc_kbatch_k0_n_k1_;
CGridDesc_M_N c_grid_desc_m_n_;
CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock c_grid_desc_mblock_mperblock_nblock_nperblock_;
Block2CTileMap block_2_ctile_map_;
index_t M01_;
index_t N01_;
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CElementwiseOperation c_element_op_;
index_t k_batch_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceGemmXdlSplitKCShuffle::Argument;
void ShowInfo(const Argument& arg)
{
std::cout << "arg.a_grid_desc_kbatch_k0_m_k1_{"
<< arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I0) << ", "
<< arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I1) << ", "
<< arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I2) << ", "
<< arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I3) << "}" << std::endl;
std::cout << "arg.b_grid_desc_kbatch_k0_n_k1_{"
<< arg.b_grid_desc_kbatch_k0_n_k1_.GetLength(I0) << ", "
<< arg.b_grid_desc_kbatch_k0_n_k1_.GetLength(I1) << ", "
<< arg.b_grid_desc_kbatch_k0_n_k1_.GetLength(I2) << ", "
<< arg.b_grid_desc_kbatch_k0_n_k1_.GetLength(I3) << "}" << std::endl;
std::cout << "arg.c_grid_desc_m_n_{ " << arg.c_grid_desc_m_n_.GetLength(I0) << ", "
<< arg.c_grid_desc_m_n_.GetLength(I1) << "}" << std::endl;
}
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
ShowInfo(arg);
const auto kbatch = arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I0);
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_kbatch_k0_m_k1_,
arg.b_grid_desc_kbatch_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_))
{
throw std::runtime_error(
"wrong! GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_v2r4r2 has invalid setting");
}
const index_t grid_size =
arg.block_2_ctile_map_.CalculateGridSize(arg.c_grid_desc_m_n_);
const auto K0 = arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I1);
const bool has_main_k0_block_loop = GridwiseGemm::CalculateHasMainK0BlockLoop(K0);
float ave_time = 0;
const auto Run = [&](const auto& kernel) {
hipGetErrorString(hipMemset(
arg.p_c_grid_,
0,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_.GetElementSpaceSize() *
sizeof(CDataType)));
ave_time =
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_grid_desc_kbatch_k0_m_k1_,
arg.b_grid_desc_kbatch_k0_n_k1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.block_2_ctile_map_);
};
if(has_main_k0_block_loop)
{
if(kbatch == 1)
{
const auto kernel = kernel_gemm_xdlops_v2r4r2<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceGemmXdlSplitKCShuffle::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceGemmXdlSplitKCShuffle::BGridDesc_K0_N_K1>,
remove_reference_t<DeviceGemmXdlSplitKCShuffle::
CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
remove_reference_t<DeviceGemmXdlSplitKCShuffle::Block2CTileMap>,
true>;
Run(kernel);
}
else
{
const auto kernel = kernel_gemm_xdlops_v2r4r2<
GridwiseGemmAtomicAdd,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceGemmXdlSplitKCShuffle::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceGemmXdlSplitKCShuffle::BGridDesc_K0_N_K1>,
remove_reference_t<DeviceGemmXdlSplitKCShuffle::
CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
remove_reference_t<DeviceGemmXdlSplitKCShuffle::Block2CTileMap>,
true>;
Run(kernel);
}
}
else
{
if(kbatch == 1)
{
const auto kernel = kernel_gemm_xdlops_v2r4r2<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceGemmXdlSplitKCShuffle::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceGemmXdlSplitKCShuffle::BGridDesc_K0_N_K1>,
remove_reference_t<DeviceGemmXdlSplitKCShuffle::
CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
remove_reference_t<DeviceGemmXdlSplitKCShuffle::Block2CTileMap>,
false>;
Run(kernel);
}
else
{
const auto kernel = kernel_gemm_xdlops_v2r4r2<
GridwiseGemmAtomicAdd,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceGemmXdlSplitKCShuffle::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceGemmXdlSplitKCShuffle::BGridDesc_K0_N_K1>,
remove_reference_t<DeviceGemmXdlSplitKCShuffle::
CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
remove_reference_t<DeviceGemmXdlSplitKCShuffle::Block2CTileMap>,
false>;
Run(kernel);
}
}
return ave_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
return GridwiseGemm::CheckValidity(arg.a_grid_desc_kbatch_k0_m_k1_,
arg.b_grid_desc_kbatch_k0_n_k1_,
arg.c_grid_desc_m_n_,
arg.block_2_ctile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const ADataType* p_a,
const BDataType* p_b,
CDataType* p_c,
index_t M,
index_t N,
index_t K,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op,
index_t KBatch)
{
return Argument{p_a,
p_b,
p_c,
M,
N,
K,
StrideA,
StrideB,
StrideC,
1,
1,
a_element_op,
b_element_op,
c_element_op,
KBatch};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,
const void* p_b,
void* p_c,
index_t M,
index_t N,
index_t K,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op,
ck::index_t KBatch = 1) override
{
return std::make_unique<Argument>(static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
static_cast<CDataType*>(p_c),
M,
N,
K,
StrideA,
StrideB,
StrideC,
1,
1,
a_element_op,
b_element_op,
c_element_op,
KBatch);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceGemmXdlSplitKCShuffle"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< K0PerBlock
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

907
include/ck/tensor_operation/gpu/device/impl/device_grouped_contraction_multiple_d_xdl_cshuffle.hpp Normal file
View File

@@ -0,0 +1,907 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_grouped_contraction_multiple_d.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/tensor_specialization.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
template <typename GridwiseGemm,
typename ContractionMultiDKernelArg,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_grouped_contraction_multiple_d_xdl_cshuffle(
const void CK_CONSTANT_ADDRESS_SPACE* contraction_args,
const index_t group_count,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const CDEElementwiseOperation cde_element_op)
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__))
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
const index_t block_id = get_block_1d_id();
const auto contraction_arg_ptr = reinterpret_cast<const ContractionMultiDKernelArg*>(
cast_pointer_to_generic_address_space(contraction_args));
index_t left = 0;
index_t right = group_count;
index_t group_id = index_t((left + right) / 2);
while((!(block_id >= contraction_arg_ptr[group_id].block_start_ &&
block_id < contraction_arg_ptr[group_id].block_end_)) &&
left <= right)
{
if(block_id < contraction_arg_ptr[group_id].block_start_)
{
right = group_id;
}
else
{
left = group_id;
}
group_id = index_t((left + right) / 2);
}
GridwiseGemm::template Run<HasMainKBlockLoop>(
contraction_arg_ptr[group_id].p_a_grid_,
contraction_arg_ptr[group_id].p_b_grid_,
contraction_arg_ptr[group_id].p_ds_grid_,
contraction_arg_ptr[group_id].p_e_grid_,
p_shared,
a_element_op,
b_element_op,
cde_element_op,
contraction_arg_ptr[group_id].a_grid_desc_ak0_m_ak1_,
contraction_arg_ptr[group_id].b_grid_desc_bk0_n_bk1_,
contraction_arg_ptr[group_id].ds_grid_desc_mblock_mperblock_nblock_nperblock_,
contraction_arg_ptr[group_id].e_grid_desc_mblock_mperblock_nblock_nperblock_,
contraction_arg_ptr[group_id].block_2_etile_map_);
#else
ignore = contraction_args;
ignore = group_count;
ignore = a_element_op;
ignore = b_element_op;
ignore = cde_element_op;
#endif
}
} // namespace ck
namespace ck {
namespace tensor_operation {
namespace device {
// Tensor Contraction:
// input : A
// input : B
// input : D0, D1, ...
// output : E
// C = a_op(A) * b_op(B)
// E = cde_op(C, D0, D1, ...)
// Assume:
// A[M0, M1, M2, ..., K0, K1, K2, ...]
// B[N0, N1, N2, ..., K0, K1, K2, ...]
// D[M0, M1, M2, ..., N0, N1, N2, ...]
// E[M0, M1, M2, ..., N0, N1, N2, ...]
template <index_t NumDimM,
index_t NumDimN,
index_t NumDimK,
typename ADataType,
typename BDataType,
typename AccDataType,
typename CShuffleDataType,
typename DsDataType,
typename EDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
GemmSpecialization GemmSpec,
TensorSpecialization ASpec,
TensorSpecialization BSpec,
TensorSpecialization DESpec,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock,
index_t KPerBlock,
index_t AK1,
index_t BK1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
bool ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
bool BBlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CDEBlockTransferScalarPerVector_NPerBlock,
LoopScheduler LoopSched = make_default_loop_scheduler()>
struct DeviceGroupedContractionMultipleD_Xdl_CShuffle
: public DeviceGroupedContractionMultipleD<NumDimM,
NumDimN,
NumDimK,
ADataType,
BDataType,
DsDataType,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation>
{
using DeviceOp = DeviceGroupedContractionMultipleD_Xdl_CShuffle;
static constexpr index_t NumDTensor = DsDataType::Size();
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr auto matrix_padder =
MatrixPadder<GemmSpec, index_t, index_t, index_t>{MPerBlock, NPerBlock, KPerBlock};
// Assume: A[M0, M1, M2, ..., K0, K1, K2, ...]
static auto MakeAGridDescriptor_M_K(const std::vector<index_t>& a_ms_ks_lengths_vec,
const std::vector<index_t>& a_ms_ks_strides_vec)
{
assert(a_ms_ks_lengths_vec.size() == NumDimM + NumDimK &&
a_ms_ks_strides_vec.size() == NumDimM + NumDimK);
const auto to_tuple = [&](auto& vec, auto num) {
return generate_tuple([&](auto i) { return vec[i]; }, num);
};
const auto a_ms_ks_lengths = to_tuple(a_ms_ks_lengths_vec, Number<NumDimM + NumDimK>{});
const auto a_ms_ks_strides = to_tuple(a_ms_ks_strides_vec, Number<NumDimM + NumDimK>{});
// dimension Ids for M0, M1, ...
constexpr auto mDimIds = typename arithmetic_sequence_gen<0, NumDimM, 1>::type{};
// dimension Ids for K0, K1, ...
constexpr auto kDimIds =
typename arithmetic_sequence_gen<NumDimM, NumDimM + NumDimK, 1>::type{};
// lengths for M0, M1, ...
const auto mLengths = get_container_subset(a_ms_ks_lengths, mDimIds);
// lengths for K0, K1, ...
const auto kLengths = get_container_subset(a_ms_ks_lengths, kDimIds);
if constexpr(ASpec == TensorSpecialization::Packed)
{
auto M = container_reduce(mLengths, math::multiplies{}, Number<1>{});
auto K = container_reduce(kLengths, math::multiplies{}, Number<1>{});
const auto a_grid_desc_mraw_kraw = make_naive_tensor_descriptor(
make_tuple(M, K),
make_tuple(a_ms_ks_strides[Number<NumDimM - 1>{}],
a_ms_ks_strides[Number<NumDimM + NumDimK - 1>{}]));
return matrix_padder.PadADescriptor_M_K(a_grid_desc_mraw_kraw);
}
else
{
// naive tensor A[M0, M1, M2, ..., K0, K1, K2...]
const auto a_grid_desc_ms_ks =
make_naive_tensor_descriptor(a_ms_ks_lengths, a_ms_ks_strides);
// transformed tensor A[MRaw = M0 * M1 * M2 * ... , KRaw = K0 * K1 * K2 * ...]
const auto a_grid_desc_mraw_kraw = transform_tensor_descriptor(
a_grid_desc_ms_ks,
make_tuple(make_merge_transform(mLengths), make_merge_transform(kLengths)),
make_tuple(mDimIds, kDimIds),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return matrix_padder.PadADescriptor_M_K(a_grid_desc_mraw_kraw);
}
}
// Assume: B[N0, N1, N2, ..., K0, K1, K2, ...]
static auto MakeBGridDescriptor_N_K(const std::vector<index_t>& b_ns_ks_lengths_vec,
const std::vector<index_t>& b_ns_ks_strides_vec)
{
assert(b_ns_ks_lengths_vec.size() == NumDimN + NumDimK &&
b_ns_ks_strides_vec.size() == NumDimN + NumDimK);
const auto to_tuple = [&](auto& vec, auto num) {
return generate_tuple([&](auto i) { return vec[i]; }, num);
};
const auto b_ns_ks_lengths = to_tuple(b_ns_ks_lengths_vec, Number<NumDimN + NumDimK>{});
const auto b_ns_ks_strides = to_tuple(b_ns_ks_strides_vec, Number<NumDimN + NumDimK>{});
// dimension Ids for N0, N1, ...
constexpr auto nDimIds = typename arithmetic_sequence_gen<0, NumDimN, 1>::type{};
// dimension Ids for K0, K1, ...
constexpr auto kDimIds =
typename arithmetic_sequence_gen<NumDimN, NumDimN + NumDimK, 1>::type{};
// lengths for K0, K1, ...
const auto kLengths = get_container_subset(b_ns_ks_lengths, kDimIds);
// lengths for N0, N1, ...
const auto nLengths = get_container_subset(b_ns_ks_lengths, nDimIds);
if constexpr(BSpec == TensorSpecialization::Packed)
{
auto N = container_reduce(nLengths, math::multiplies{}, Number<1>{});
auto K = container_reduce(kLengths, math::multiplies{}, Number<1>{});
const auto b_grid_desc_nraw_kraw = make_naive_tensor_descriptor(
make_tuple(N, K),
make_tuple(b_ns_ks_strides[Number<NumDimN - 1>{}],
b_ns_ks_strides[Number<NumDimN + NumDimK - 1>{}]));
return matrix_padder.PadBDescriptor_N_K(b_grid_desc_nraw_kraw);
}
else
{
// naive tensor B[N0, N1, N2, ..., K0, K1, K2, ...]
const auto b_grid_desc_ns_ks =
make_naive_tensor_descriptor(b_ns_ks_lengths, b_ns_ks_strides);
// transformed tensor B[NRaw = N0 * N1 * N2 * ..., KRaw = K0 * K1 * K2 * ...]
const auto b_grid_desc_nraw_kraw = transform_tensor_descriptor(
b_grid_desc_ns_ks,
make_tuple(make_merge_transform(nLengths), make_merge_transform(kLengths)),
make_tuple(nDimIds, kDimIds),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return matrix_padder.PadBDescriptor_N_K(b_grid_desc_nraw_kraw);
}
}
// assume E[M0, M1, M2, ..., N0, N1, N2...]
static auto MakeEGridDescriptor_M_N(const std::vector<index_t>& e_ms_ns_lengths_vec,
const std::vector<index_t>& e_ms_ns_strides_vec)
{
assert(e_ms_ns_lengths_vec.size() == NumDimM + NumDimN &&
e_ms_ns_strides_vec.size() == NumDimM + NumDimN);
const auto to_tuple = [&](auto& vec, auto num) {
return generate_tuple([&](auto i) { return vec[i]; }, num);
};
const auto e_ms_ns_lengths = to_tuple(e_ms_ns_lengths_vec, Number<NumDimM + NumDimN>{});
const auto e_ms_ns_strides = to_tuple(e_ms_ns_strides_vec, Number<NumDimM + NumDimN>{});
// dimension Ids for M0, M1, ...
constexpr auto mDimIds = typename arithmetic_sequence_gen<0, NumDimM, 1>::type{};
// dimension Ids for N0, N1, ...
constexpr auto nDimIds =
typename arithmetic_sequence_gen<NumDimM, NumDimM + NumDimN, 1>::type{};
// lengths for M0, M1, ...
const auto mLengths = get_container_subset(e_ms_ns_lengths, mDimIds);
// lengths for K0, K1, ...
const auto nLengths = get_container_subset(e_ms_ns_lengths, nDimIds);
if constexpr(DESpec == TensorSpecialization::Packed)
{
auto M = container_reduce(mLengths, math::multiplies{}, Number<1>{});
auto N = container_reduce(nLengths, math::multiplies{}, Number<1>{});
const auto e_grid_desc_mraw_nraw = make_naive_tensor_descriptor(
make_tuple(M, N),
make_tuple(e_ms_ns_strides[Number<NumDimM - 1>{}],
e_ms_ns_strides[Number<NumDimM + NumDimN - 1>{}]));
return matrix_padder.PadCDescriptor_M_N(e_grid_desc_mraw_nraw);
}
else
{
// naive tensor E[M0, M1, M2, ..., N0, N1, N2...]
const auto e_grid_desc_ms_ns =
make_naive_tensor_descriptor(e_ms_ns_lengths, e_ms_ns_strides);
// transformed tensor E[MRaw = M0 * M1 * M2 * ... , NRaw = N0 * N1 * N2 * ...]
const auto e_grid_desc_mraw_nraw = transform_tensor_descriptor(
e_grid_desc_ms_ns,
make_tuple(make_merge_transform(mLengths), make_merge_transform(nLengths)),
make_tuple(mDimIds, nDimIds),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return matrix_padder.PadCDescriptor_M_N(e_grid_desc_mraw_nraw);
}
}
static auto MakeDsGridDescriptor_M_N(
const std::array<std::vector<index_t>, NumDTensor>& ds_ms_ns_lengths_vec,
const std::array<std::vector<index_t>, NumDTensor>& ds_ms_ns_strides_vec)
{
return generate_tuple(
[&](auto i) {
return DeviceOp::MakeEGridDescriptor_M_N(ds_ms_ns_lengths_vec[i],
ds_ms_ns_strides_vec[i]);
},
Number<NumDTensor>{});
}
using AGridDesc_M_K = decltype(MakeAGridDescriptor_M_K({}, {}));
using BGridDesc_N_K = decltype(MakeBGridDescriptor_N_K({}, {}));
using DsGridDesc_M_N = remove_cvref_t<decltype(MakeDsGridDescriptor_M_N({{}}, {{}}))>;
using EGridDesc_M_N = decltype(MakeEGridDescriptor_M_N({}, {}));
// GridwiseGemm
using GridwiseGemm = GridwiseGemmMultipleD_xdl_cshuffle<
ADataType, // TODO: distinguish A/B datatype
AccDataType,
CShuffleDataType,
DsDataType,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
InMemoryDataOperationEnum::Set,
NumGemmKPrefetchStage,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
AK1,
BK1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
false,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
false,
BBlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CDEBlockTransferScalarPerVector_NPerBlock,
LoopSched>;
// desc for blockwise copy
using AGridDesc_AK0_M_AK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(AGridDesc_M_K{}))>;
using BGridDesc_BK0_N_BK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(BGridDesc_N_K{}))>;
using DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock = remove_cvref_t<decltype(
GridwiseGemm::MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(DsGridDesc_M_N{}))>;
using EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock = remove_cvref_t<decltype(
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(EGridDesc_M_N{}))>;
struct GroupedContractionBlock2ETileMap
{
// block-to-e-tile map
using Block2ETileMap =
remove_cvref_t<decltype(GridwiseGemm::MakeDefaultBlock2ETileMap(EGridDesc_M_N{}))>;
GroupedContractionBlock2ETileMap(const EGridDesc_M_N& e_grid_desc_m_n,
ck::index_t BlockStart)
{
default_block_2_etile_map_ = GridwiseGemm::MakeDefaultBlock2ETileMap(e_grid_desc_m_n);
block_start_ = BlockStart;
}
template <typename TopIdx>
__host__ __device__ constexpr auto CalculateBottomIndex(const TopIdx& idx_top) const
{
return default_block_2_etile_map_.CalculateBottomIndex(
make_multi_index(idx_top[I0] - block_start_));
}
// it's actually E-Tile
template <typename CTileIdx, typename CTileDim>
__host__ __device__ bool ValidCTileIndex(const CTileIdx& c_tile_idx,
const CTileDim& c_tile_dim) const
{
return default_block_2_etile_map_.ValidCTileIndex(c_tile_idx, c_tile_dim);
}
__host__ bool CheckValidity(const EGridDesc_M_N& e_grid_desc_m_n) const
{
return default_block_2_etile_map_.CheckValidity(e_grid_desc_m_n);
}
Block2ETileMap default_block_2_etile_map_;
ck::index_t block_start_;
};
struct ContractionMultiDKernelArg
{
// pointers
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
typename GridwiseGemm::DsGridPointer p_ds_grid_;
EDataType* p_e_grid_;
// tensor descriptors for block/thread-wise copy
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock
ds_grid_desc_mblock_mperblock_nblock_nperblock_;
EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock e_grid_desc_mblock_mperblock_nblock_nperblock_;
// lock-to-e-tile map
GroupedContractionBlock2ETileMap block_2_etile_map_;
ck::index_t block_start_, block_end_;
};
struct ContractionMultiDDeviceArg
{
// tensor descriptors for problem definiton
AGridDesc_M_K a_grid_desc_m_k_;
BGridDesc_N_K b_grid_desc_n_k_;
DsGridDesc_M_N ds_grid_desc_m_n_;
EGridDesc_M_N e_grid_desc_m_n_;
// Strides for the last M/N/K dimensions of A/B/Ds/E
// for sanity check of vector load/store
index_t a_mz_stride_;
index_t a_kz_stride_;
index_t b_nz_stride_;
index_t b_kz_stride_;
std::array<index_t, NumDTensor> ds_nz_stride_;
// index_t e_mz_stride_;
index_t e_nz_stride_;
};
// Argument
struct Argument : public BaseArgument
{
Argument(std::vector<const void*> p_a_vec,
std::vector<const void*> p_b_vec,
std::vector<std::array<const void*, NumDTensor>> p_ds_vec,
std::vector<void*> p_e_vec,
std::vector<ContractionDesc<NumDTensor>> contraction_descs,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op)
: a_element_op_{a_element_op},
b_element_op_{b_element_op},
cde_element_op_{cde_element_op}
{
group_count_ = contraction_descs.size();
if(!(group_count_ == p_a_vec.size() && group_count_ == p_b_vec.size() &&
group_count_ == p_e_vec.size()))
{
throw std::runtime_error("wrong! group_count_ != a/b/e_vec.size");
}
contraction_multi_d_kernel_args_.reserve(group_count_);
grid_size_ = 0;
for(std::size_t i = 0; i < group_count_; i++)
{
const auto p_a_grid = static_cast<const ADataType*>(p_a_vec[i]);
const auto p_b_grid = static_cast<const BDataType*>(p_b_vec[i]);
const auto p_e_grid = static_cast<EDataType*>(p_e_vec[i]);
const auto a_grid_desc_m_k = DeviceOp::MakeAGridDescriptor_M_K(
contraction_descs[i].a_ms_ks_lengths, contraction_descs[i].a_ms_ks_strides);
const auto b_grid_desc_n_k = DeviceOp::MakeBGridDescriptor_N_K(
contraction_descs[i].b_ns_ks_lengths, contraction_descs[i].b_ns_ks_strides);
DsGridDesc_M_N ds_grid_desc_m_n;
typename GridwiseGemm::DsGridPointer p_ds_grid;
// populate pointer, batch stride, desc for Ds
static_for<0, NumDTensor, 1>{}([&](auto j) {
using DDataType = remove_cvref_t<tuple_element_t<j.value, DsDataType>>;
// D pointer
p_ds_grid(j) = static_cast<const DDataType*>(p_ds_vec[i][j]);
// D desc
ds_grid_desc_m_n(j) =
DeviceOp::MakeEGridDescriptor_M_N(contraction_descs[i].ds_ms_ns_lengths[j],
contraction_descs[i].ds_ms_ns_strides[j]);
});
const auto e_grid_desc_m_n = DeviceOp::MakeEGridDescriptor_M_N(
contraction_descs[i].e_ms_ns_lengths, contraction_descs[i].e_ms_ns_strides);
const auto a_grid_desc_ak0_m_ak1 =
GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(a_grid_desc_m_k);
const auto b_grid_desc_bk0_n_bk1 =
GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(b_grid_desc_n_k);
const auto ds_grid_desc_mblock_mperblock_nblock_nperblock =
GridwiseGemm::MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
ds_grid_desc_m_n);
const auto e_grid_desc_mblock_mperblock_nblock_nperblock =
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
e_grid_desc_m_n);
const index_t grid_size_grp =
GridwiseGemm::MakeDefaultBlock2ETileMap(e_grid_desc_m_n)
.CalculateGridSize(e_grid_desc_m_n);
const index_t BlockStart = grid_size_;
const index_t BlockEnd = grid_size_ + grid_size_grp;
grid_size_ += grid_size_grp;
const auto block_2_etile_map =
GroupedContractionBlock2ETileMap(e_grid_desc_m_n, BlockStart);
// for sanity check of vector memory access
const index_t a_mz_stride = contraction_descs[i].a_ms_ks_strides[NumDimM - 1];
const index_t a_kz_stride =
contraction_descs[i].a_ms_ks_strides[NumDimM + NumDimK - 1];
const index_t b_nz_stride = contraction_descs[i].b_ns_ks_strides[NumDimN - 1];
const index_t b_kz_stride =
contraction_descs[i].b_ns_ks_strides[NumDimN + NumDimK - 1];
std::array<index_t, NumDTensor> ds_nz_stride;
for(index_t j = 0; j < NumDTensor; ++j)
{
ds_nz_stride[j] =
contraction_descs[i].ds_ms_ns_strides[j][NumDimM + NumDimN - 1];
}
const index_t e_nz_stride =
contraction_descs[i].e_ms_ns_strides[NumDimM + NumDimN - 1];
if(GridwiseGemm::CheckValidity(a_grid_desc_m_k,
b_grid_desc_n_k,
ds_grid_desc_m_n,
e_grid_desc_m_n,
block_2_etile_map))
{
contraction_multi_d_kernel_args_.push_back(
{p_a_grid,
p_b_grid,
p_ds_grid,
p_e_grid,
a_grid_desc_ak0_m_ak1,
b_grid_desc_bk0_n_bk1,
ds_grid_desc_mblock_mperblock_nblock_nperblock,
e_grid_desc_mblock_mperblock_nblock_nperblock,
block_2_etile_map,
BlockStart,
BlockEnd});
contraction_multi_d_device_args_.push_back({a_grid_desc_m_k,
b_grid_desc_n_k,
ds_grid_desc_m_n,
e_grid_desc_m_n,
a_mz_stride,
a_kz_stride,
b_nz_stride,
b_kz_stride,
ds_nz_stride,
e_nz_stride});
}
}
}
std::vector<ContractionMultiDKernelArg> contraction_multi_d_kernel_args_;
std::vector<ContractionMultiDDeviceArg> contraction_multi_d_device_args_;
std::size_t group_count_;
index_t grid_size_;
// element-wise op
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CDEElementwiseOperation cde_element_op_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
bool has_main_k_block_loop = true;
for(std::size_t i = 0; i < arg.group_count_; i++)
{
const auto K =
arg.contraction_multi_d_kernel_args_[i].a_grid_desc_ak0_m_ak1_.GetLength(I0) *
arg.contraction_multi_d_kernel_args_[i].a_grid_desc_ak0_m_ak1_.GetLength(I2);
if(GridwiseGemm::CalculateHasMainKBlockLoop(K) != has_main_k_block_loop)
{
throw std::runtime_error("wrong! not all gemm has_main_k_block_loop");
}
}
hipGetErrorString(hipMemcpy(arg.p_workspace_,
arg.contraction_multi_d_kernel_args_.data(),
arg.contraction_multi_d_kernel_args_.size() *
sizeof(ContractionMultiDKernelArg),
hipMemcpyHostToDevice));
float ave_time = 0;
auto launch_kernel = [&](auto has_main_k_block_loop_) {
const auto kernel =
kernel_grouped_contraction_multiple_d_xdl_cshuffle<GridwiseGemm,
ContractionMultiDKernelArg,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
has_main_k_block_loop_>;
return launch_and_time_kernel(
stream_config,
kernel,
dim3(arg.grid_size_),
dim3(BlockSize),
0,
cast_pointer_to_constant_address_space(arg.p_workspace_),
arg.group_count_,
arg.a_element_op_,
arg.b_element_op_,
arg.cde_element_op_);
};
if(has_main_k_block_loop)
{
ave_time = launch_kernel(integral_constant<bool, true>{});
}
else
{
ave_time = launch_kernel(integral_constant<bool, false>{});
}
return ave_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static bool IsSupportedArgument(const Argument& arg)
{
if(!(ck::get_device_name() == "gfx908" || ck::get_device_name() == "gfx90a"))
{
return false;
}
for(std::size_t i = 0; i < arg.group_count_; i++)
{
const auto a_grid_desc_m_k_ = arg.contraction_multi_d_device_args_[i].a_grid_desc_m_k_;
const auto b_grid_desc_n_k_ = arg.contraction_multi_d_device_args_[i].b_grid_desc_n_k_;
const auto ds_grid_desc_m_n_ =
arg.contraction_multi_d_device_args_[i].ds_grid_desc_m_n_;
const auto e_grid_desc_m_n_ = arg.contraction_multi_d_device_args_[i].e_grid_desc_m_n_;
const auto a_grid_desc_ak0_m_ak1_ =
arg.contraction_multi_d_kernel_args_[i].a_grid_desc_ak0_m_ak1_;
const auto b_grid_desc_bk0_n_bk1_ =
arg.contraction_multi_d_kernel_args_[i].b_grid_desc_bk0_n_bk1_;
const auto ds_grid_desc_mblock_mperblock_nblock_nperblock_ =
arg.contraction_multi_d_kernel_args_[i]
.ds_grid_desc_mblock_mperblock_nblock_nperblock_;
const auto e_grid_desc_mblock_mperblock_nblock_nperblock_ =
arg.contraction_multi_d_kernel_args_[i]
.e_grid_desc_mblock_mperblock_nblock_nperblock_;
const auto block_2_etile_map_ =
arg.contraction_multi_d_kernel_args_[i].block_2_etile_map_;
const auto a_mz_stride_ = arg.contraction_multi_d_device_args_[i].a_mz_stride_;
const auto a_kz_stride_ = arg.contraction_multi_d_device_args_[i].a_kz_stride_;
const auto b_nz_stride_ = arg.contraction_multi_d_device_args_[i].b_nz_stride_;
const auto b_kz_stride_ = arg.contraction_multi_d_device_args_[i].b_kz_stride_;
const auto ds_nz_stride_ = arg.contraction_multi_d_device_args_[i].ds_nz_stride_;
const auto e_nz_stride_ = arg.contraction_multi_d_device_args_[i].e_nz_stride_;
if(!GridwiseGemm::CheckValidity(a_grid_desc_m_k_,
b_grid_desc_n_k_,
ds_grid_desc_m_n_,
e_grid_desc_m_n_,
block_2_etile_map_))
{
return false;
}
// check vector access
static_assert((ABlockTransferSrcVectorDim == 1 || ABlockTransferSrcVectorDim == 2) &&
(BBlockTransferSrcVectorDim == 1 || BBlockTransferSrcVectorDim == 2),
"wrong!");
// vector memory access of A: could be on M or AK1 dimension
if constexpr(ABlockTransferSrcVectorDim == 1)
{
if(!(a_mz_stride_ == 1 &&
a_grid_desc_ak0_m_ak1_.GetLength(I1) % ABlockTransferSrcScalarPerVector == 0))
{
return false;
}
}
else
{
if(!(a_kz_stride_ == 1 &&
a_grid_desc_ak0_m_ak1_.GetLength(I2) % ABlockTransferSrcScalarPerVector == 0))
{
return false;
}
}
// vector memory access of B: could be on N or BK1 dimension
if constexpr(BBlockTransferSrcVectorDim == 1)
{
if(!(b_nz_stride_ == 1 &&
b_grid_desc_bk0_n_bk1_.GetLength(I1) % BBlockTransferSrcScalarPerVector == 0))
{
return false;
}
}
else
{
if(!(b_kz_stride_ == 1 &&
b_grid_desc_bk0_n_bk1_.GetLength(I2) % BBlockTransferSrcScalarPerVector == 0))
{
return false;
}
}
// vector memory access of Ds: always on NPerBlock dimension
bool valid_d_access = true;
static_for<0, NumDTensor, 1>{}([&](auto j) {
if(!(ds_nz_stride_[j] == 1 &&
ds_grid_desc_mblock_mperblock_nblock_nperblock_[j].GetLength(I3) %
CDEBlockTransferScalarPerVector_NPerBlock ==
0))
{
valid_d_access = false;
}
});
if(valid_d_access == false)
{
return false;
}
// vector memory access of E: always on NPerBlock dimension
if(!(e_nz_stride_ == 1 && e_grid_desc_mblock_mperblock_nblock_nperblock_.GetLength(I3) %
CDEBlockTransferScalarPerVector_NPerBlock ==
0))
{
return false;
}
}
return true;
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(std::vector<const void*> p_a_vec,
std::vector<const void*> p_b_vec,
std::vector<std::array<const void*, NumDTensor>> p_ds_vec,
std::vector<void*> p_e_vec,
std::vector<ContractionDesc<NumDTensor>> contraction_descs,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op)
{
return Argument{p_a_vec,
p_b_vec,
p_ds_vec,
p_e_vec,
contraction_descs,
a_element_op,
b_element_op,
cde_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument>
MakeArgumentPointer(std::vector<const void*> p_a_vec,
std::vector<const void*> p_b_vec,
std::vector<std::array<const void*, NumDTensor>> p_ds_vec,
std::vector<void*> p_e_vec,
std::vector<ContractionDesc<NumDTensor>> contraction_descs,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op) override
{
return std::make_unique<Argument>(p_a_vec,
p_b_vec,
p_ds_vec,
p_e_vec,
contraction_descs,
a_element_op,
b_element_op,
cde_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceGroupedContractionMultipleD_Xdl_CShuffle"
<< "<"
<< NumDimM << ", "
<< NumDimN << ", "
<< NumDimK << ", "
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< AK1 << ", "
<< BK1 << ", "
<< ABlockTransferSrcVectorDim << ", "
<< BBlockTransferSrcVectorDim
<< ">";
// clang-format on
return str.str();
}
size_t GetWorkSpaceSize(const BaseArgument* p_arg) const override
{
return dynamic_cast<const Argument*>(p_arg)->group_count_ *
sizeof(ContractionMultiDKernelArg);
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

77
include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_multiple_d_multiple_r.hpp Normal file
View File

@@ -0,0 +1,77 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <vector>
#include "ck/tensor_operation/gpu/device/device_base.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
// Grouped Convolution Forward:
// input : input image A[G, N, C, Hi, Wi],
// input : weight B[G, K, C, Y, X],
// input : D0[G, N, K, Ho, Wo], D1[G, N, K, Ho, Wo], ...
// output : output image E[G, N, K, Ho, Wo]
// output : R0[G, N, Ho, Wo], R1[G, N, Ho, Wo], ...
// C = a_op(A) * b_op(B)
// E = cde_op(C, D0, D1, ...)
// Q0 = reduce0(q_op0(E)), Q1 = reduce1(q_op0(E)), ...
// R0 = r_op0(Q0), R1 = r_op1(Q1), ...
// Assume:
// D0, D1, ... and E have the same layout
template <index_t NDimSpatial,
typename ALayout,
typename BLayout,
typename DELayout,
typename RLayout,
typename ADataType,
typename BDataType,
typename DsDataType,
typename EDataType,
typename RsDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
typename QsElementwiseOperation,
typename RsElementwiseOperation>
struct DeviceGroupedConvFwdMultipleDMultipleR : public BaseOperator
{
static constexpr index_t NumDTensor = DsDataType::Size();
static constexpr index_t NumRTensor = RsDataType::Size();
virtual std::unique_ptr<BaseArgument> MakeArgumentPointer(
const void* p_a,
const void* p_b,
const std::array<const void*, NumDTensor>& p_ds,
void* p_e,
std::array<void*, NumRTensor> p_rs,
const std::array<index_t, NDimSpatial + 3>& a_g_n_c_wis_lengths,
const std::array<index_t, NDimSpatial + 3>& a_g_n_c_wis_strides,
const std::array<index_t, NDimSpatial + 3>& b_g_k_c_xs_lengths,
const std::array<index_t, NDimSpatial + 3>& b_g_k_c_xs_strides,
const std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor>& ds_g_n_k_wos_lengths,
const std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor>& ds_g_n_k_wos_strides,
const std::array<index_t, NDimSpatial + 3>& e_g_n_k_wos_lengths,
const std::array<index_t, NDimSpatial + 3>& e_g_n_k_wos_strides,
const std::array<index_t, NDimSpatial + 2>& r_g_n_wos_lengths,
const std::array<index_t, NDimSpatial + 2>& r_g_n_wos_strides,
const std::array<index_t, NDimSpatial>& conv_filter_strides,
const std::array<index_t, NDimSpatial>& conv_filter_dilations,
const std::array<index_t, NDimSpatial>& input_left_pads,
const std::array<index_t, NDimSpatial>& input_right_pads,
const AElementwiseOperation& a_element_op,
const BElementwiseOperation& b_element_op,
const CDEElementwiseOperation& cde_element_op,
const QsElementwiseOperation& qs_element_op,
const RsElementwiseOperation& rs_element_op) = 0;
virtual std::unique_ptr<BaseInvoker> MakeInvokerPointer() = 0;
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

1106
include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_multiple_d_multiple_r_xdl_cshuffle.hpp Normal file
View File

File diff suppressed because it is too large Load Diff

952
include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_multiple_d_xdl_cshuffle.hpp Normal file
View File

@@ -0,0 +1,952 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <functional>
#include <iostream>
#include <iterator>
#include <numeric>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/convolution_forward_specialization.hpp"
#include "ck/tensor_operation/operator_transform/transform_conv_fwd_to_gemm.hpp"
#include "ck/tensor_operation/gpu/device/device_grouped_conv_fwd_multiple_d.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
#include "ck/host_utility/io.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace {
template <index_t NumDTensor>
struct ComputePtrOffsetOfStridedBatch
{
ComputePtrOffsetOfStridedBatch() = default;
ComputePtrOffsetOfStridedBatch(index_t BatchStrideA,
index_t BatchStrideB,
Array<ck::index_t, NumDTensor> BatchStrideDs,
index_t BatchStrideE)
: BatchStrideA_(BatchStrideA),
BatchStrideB_(BatchStrideB),
BatchStrideDs_(BatchStrideDs),
BatchStrideE_(BatchStrideE)
{
}
__host__ __device__ constexpr long_index_t GetAPtrOffset(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideA_);
}
__host__ __device__ constexpr long_index_t GetBPtrOffset(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideB_);
}
__host__ __device__ constexpr auto GetDsPtrOffset(index_t g_idx) const
{
Array<long_index_t, NumDTensor> ds_offset;
static_for<0, NumDTensor, 1>{}(
[&](auto i) { ds_offset(i) = g_idx * static_cast<long_index_t>(BatchStrideDs_[i]); });
return ds_offset;
}
__host__ __device__ constexpr long_index_t GetEPtrOffset(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideE_);
}
index_t BatchStrideA_;
index_t BatchStrideB_;
Array<ck::index_t, NumDTensor> BatchStrideDs_;
index_t BatchStrideE_;
};
/*
* \brief Wrapper function of GridwiseGemm::Run to realize BatchedGEMM.
*
* \tparam ComputePtrOffsetOfBatch Class that computes the base pointer offsets of A, B, C matrix
* given the batch. For example, ComputePtrOffsetOfStridedBatch() computes the offsets of evenly
* strided batched, but we can easily extend to other layouts. The returned offset can be either \p
* index_t or \p long_index_t. If it returns \p long_index_t, we are not subject to the 2GB
* limitations.
*
* \tparam Block2ETileMap Block2ETileMap::CalculateBottomIndex() takes in id of a workgroup and
* returns the 2D index of the tile that it computes. \see
* GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r3::Run().
*
* \note Using \p ComputePtrOffsetOfBatch gives us the flexibility that 2 workgroups can compute 2
* tiles from different matrices. Keep in mind that these 2 matrices can share the same grid
* descriptor (like in BatchedGEMM), or use their own grid descriptors (in GroupedGemm). \link
* impl/device_conv3d_fwd_xdl_ndhwc_kzyxc_ndhwk.hpp kernel_gemm_xdlops_v2r3_for_conv3d \endlink for
* \link DeviceConv3d \endlink uses the same concept, but currently does NOT encapsulate the
* computing of pointer offset into \p ComputePtrOffsetOfStridedBatch.
*
* \note \p Block2ETileMap allows customized mapping between a workgroup and the C-tile it computes.
* Together with \p ComputePtrOffsetOfBatch, we can reuse GridwiseGemm (and GridwiseGemm fusion ) to
* realize BatchedGemm and GroupedGemm (and the corresponding GEMM fusion).
*
*/
template <typename GridwiseGemm,
typename ABDataType,
typename DsPointer,
typename EDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
typename AGridDesc_AK0_M_AK1,
typename BGridDesc_BK0_N_BK1,
typename DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
typename Block2ETileMap,
typename ComputePtrOffsetOfBatch,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_grouped_conv_fwd_multiple_d_xdl_cshuffle(
const ABDataType* __restrict__ p_a_grid,
const ABDataType* __restrict__ p_b_grid,
DsPointer p_ds_grid,
EDataType* __restrict__ p_e_grid,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const CDEElementwiseOperation cde_element_op,
const index_t batch_count,
const AGridDesc_AK0_M_AK1 a_grid_desc_k0_m_k1,
const BGridDesc_BK0_N_BK1 b_grid_desc_k0_n_k1,
const DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
ds_grid_desc_mblock_mperblock_nblock_nperblock,
const EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock
e_grid_desc_mblock_mperblock_nblock_nperblock_,
const Block2ETileMap block_2_ctile_map,
const ComputePtrOffsetOfBatch compute_ptr_offset_of_batch)
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__))
// offset base pointer for each work-group
const index_t num_blocks_per_batch =
__builtin_amdgcn_readfirstlane(get_grid_size() / batch_count);
const index_t g_idx = __builtin_amdgcn_readfirstlane(get_block_1d_id() / num_blocks_per_batch);
const long_index_t a_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_ptr_offset_of_batch.GetAPtrOffset(g_idx)));
const long_index_t b_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_ptr_offset_of_batch.GetBPtrOffset(g_idx)));
const long_index_t e_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_ptr_offset_of_batch.GetEPtrOffset(g_idx)));
const auto ds_batch_offset = compute_ptr_offset_of_batch.GetDsPtrOffset(g_idx);
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
DsPointer p_ds_grid_grp;
static constexpr index_t NumDTensor =
DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock::Size();
static_for<0, NumDTensor, 1>{}(
[&](auto i) { p_ds_grid_grp(i) = p_ds_grid[i] + ds_batch_offset[i]; });
GridwiseGemm::template Run<HasMainKBlockLoop>(p_a_grid + a_batch_offset,
p_b_grid + b_batch_offset,
p_ds_grid_grp,
p_e_grid + e_batch_offset,
p_shared,
a_element_op,
b_element_op,
cde_element_op,
a_grid_desc_k0_m_k1,
b_grid_desc_k0_n_k1,
ds_grid_desc_mblock_mperblock_nblock_nperblock,
e_grid_desc_mblock_mperblock_nblock_nperblock_,
block_2_ctile_map);
#else
ignore = p_a_grid;
ignore = p_b_grid;
ignore = p_ds_grid;
ignore = p_e_grid;
ignore = batch_count;
ignore = a_grid_desc_k0_m_k1;
ignore = b_grid_desc_k0_n_k1;
ignore = ds_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = e_grid_desc_mblock_mperblock_nblock_nperblock_;
ignore = a_element_op;
ignore = b_element_op;
ignore = cde_element_op;
ignore = compute_ptr_offset_of_batch;
ignore = block_2_ctile_map;
#endif
}
} // namespace
//
// @brief Device Convolution operation.
//
// Supports:
// @li Forward convolution with up to 3 spatial dimentions
// @li Input tensor in GNWC data format
// @li Weight tensor in GKXC data format
// @li Output tensor in GNWK data format
//
// 1D:
// out[N, Wo, K] = in[N, Wi, C] * wei[K, X, C]
// 2D:
// out[N, Ho, Wo, K] = in[N, Hi, Wi, C] * wei[K, Y, X, C]
// 3D:
// out[N, Do, Ho, Wo, K] = in[N, Di, Hi, Wi, C] * wei[K, Z, Y, X, C]
//
template <index_t NDimSpatial,
typename ALayout,
typename BLayout,
typename DsLayout,
typename ELayout,
typename ADataType,
typename BDataType,
typename AccDataType,
typename CShuffleDataType,
typename DsDataType,
typename EDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
ConvolutionForwardSpecialization ConvForwardSpecialization,
GemmSpecialization GemmSpec,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock,
index_t KPerBlock,
index_t AK1,
index_t BK1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
index_t ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
index_t BBlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CDEBlockTransferScalarPerVector_NPerBlock,
LoopScheduler LoopSched = make_default_loop_scheduler()>
struct DeviceGroupedConvFwdMultipleD_Xdl_CShuffle
: public DeviceGroupedConvFwdMultipleD<NDimSpatial,
ALayout,
BLayout,
DsLayout,
ELayout,
ADataType,
BDataType,
DsDataType,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation>
{
using DeviceOp = DeviceGroupedConvFwdMultipleD_Xdl_CShuffle;
static constexpr index_t NumDTensor = DsDataType::Size();
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr auto conv_to_gemm_transformer =
TransformConvFwdToGemm<NDimSpatial, ConvForwardSpecialization>{};
static constexpr auto matrix_padder =
MatrixPadder<GemmSpec, index_t, index_t, index_t>{MPerBlock, NPerBlock, KPerBlock};
template <typename ALay>
static auto
MakeAGridDescriptor_M_K(const std::array<index_t, NDimSpatial + 3>& a_g_n_c_wis_lengths,
const std::array<index_t, NDimSpatial + 3>& a_g_n_c_wis_strides,
const std::array<index_t, NDimSpatial + 3>& b_g_k_c_xs_lengths,
const std::array<index_t, NDimSpatial + 3>& b_g_k_c_xs_strides,
const std::array<index_t, NDimSpatial + 3>& e_g_n_k_wos_lengths,
const std::array<index_t, NDimSpatial + 3>& e_g_n_k_wos_strides,
const std::array<index_t, NDimSpatial>& conv_filter_strides,
const std::array<index_t, NDimSpatial>& conv_filter_dilations,
const std::array<index_t, NDimSpatial>& input_left_pads,
const std::array<index_t, NDimSpatial>& input_right_pads)
{
const auto in_gemmmraw_gemmkraw_desc =
conv_to_gemm_transformer.template MakeADescriptor_M_K<ALay>(a_g_n_c_wis_lengths,
a_g_n_c_wis_strides,
b_g_k_c_xs_lengths,
b_g_k_c_xs_strides,
e_g_n_k_wos_lengths,
e_g_n_k_wos_strides,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads);
const auto in_gemmm_gemmk_desc =
matrix_padder.PadADescriptor_M_K(in_gemmmraw_gemmkraw_desc);
return in_gemmm_gemmk_desc;
}
template <typename BLay>
static auto
MakeBGridDescriptor_N_K(const std::array<index_t, NDimSpatial + 3>& b_g_k_c_xs_lengths,
const std::array<index_t, NDimSpatial + 3>& b_g_k_c_xs_strides)
{
const auto wei_gemmnraw_gemmkraw_desc =
conv_to_gemm_transformer.template MakeBDescriptor_N_K<BLay>(b_g_k_c_xs_lengths,
b_g_k_c_xs_strides);
const auto wei_gemmn_gemmk_desc =
matrix_padder.PadBDescriptor_N_K(wei_gemmnraw_gemmkraw_desc);
return wei_gemmn_gemmk_desc;
}
template <typename ELay>
static auto
MakeEGridDescriptor_M_N(const std::array<index_t, NDimSpatial + 3>& e_g_n_k_wos_lengths,
const std::array<index_t, NDimSpatial + 3>& e_g_n_k_wos_strides)
{
const auto out_gemmmraw_gemmnraw_desc =
conv_to_gemm_transformer.template MakeCDescriptor_M_N<ELay>(e_g_n_k_wos_lengths,
e_g_n_k_wos_strides);
const auto out_gemmm_gemmn_desc =
matrix_padder.PadCDescriptor_M_N(out_gemmmraw_gemmnraw_desc);
return out_gemmm_gemmn_desc;
}
static auto MakeDsGridDescriptor_M_N(
const std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor>& ds_g_n_k_wos_lengths,
const std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor>& ds_g_n_k_wos_strides)
{
return generate_tuple(
[&](auto i) {
using DLayout = remove_cvref_t<tuple_element_t<i.value, DsLayout>>;
return DeviceOp::MakeEGridDescriptor_M_N<DLayout>(ds_g_n_k_wos_lengths[i],
ds_g_n_k_wos_strides[i]);
},
Number<NumDTensor>{});
}
// desc for problem definition
using AGridDesc_M_K = remove_cvref_t<decltype(
MakeAGridDescriptor_M_K<ALayout>({}, {}, {}, {}, {}, {}, {}, {}, {}, {}))>;
using BGridDesc_N_K = remove_cvref_t<decltype(MakeBGridDescriptor_N_K<BLayout>({}, {}))>;
using DsGridDesc_M_N = remove_cvref_t<decltype(MakeDsGridDescriptor_M_N({}, {}))>;
using EGridDesc_M_N = remove_cvref_t<decltype(MakeEGridDescriptor_M_N<ELayout>({}, {}))>;
// GridwiseGemm
using GridwiseGemm = GridwiseGemmMultipleD_xdl_cshuffle<
ADataType, // TODO: distinguish A/B datatype
AccDataType,
CShuffleDataType,
DsDataType,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
InMemoryDataOperationEnum::Set,
NumGemmKPrefetchStage,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
AK1,
BK1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
false,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
false,
BBlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CDEBlockTransferScalarPerVector_NPerBlock,
LoopSched>;
// desc for blockwise copy
using AGridDesc_AK0_M_AK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(AGridDesc_M_K{}))>;
using BGridDesc_BK0_N_BK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(BGridDesc_N_K{}))>;
using DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock = remove_cvref_t<decltype(
GridwiseGemm::MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(DsGridDesc_M_N{}))>;
using EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock = remove_cvref_t<decltype(
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(EGridDesc_M_N{}))>;
// block-to-e-tile map
using Block2ETileMap =
remove_cvref_t<decltype(GridwiseGemm::MakeDefaultBlock2ETileMap(EGridDesc_M_N{}))>;
// Argument
struct Argument : public BaseArgument
{
Argument(const void* p_a,
const void* p_b,
const std::array<const void*, NumDTensor>& p_ds,
void* p_e,
const std::array<index_t, NDimSpatial + 3>& a_g_n_c_wis_lengths,
const std::array<index_t, NDimSpatial + 3>& a_g_n_c_wis_strides,
const std::array<index_t, NDimSpatial + 3>& b_g_k_c_xs_lengths,
const std::array<index_t, NDimSpatial + 3>& b_g_k_c_xs_strides,
const std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor>&
ds_g_n_k_wos_lengths,
const std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor>&
ds_g_n_k_wos_strides,
const std::array<index_t, NDimSpatial + 3>& e_g_n_k_wos_lengths,
const std::array<index_t, NDimSpatial + 3>& e_g_n_k_wos_strides,
const std::array<index_t, NDimSpatial>& conv_filter_strides,
const std::array<index_t, NDimSpatial>& conv_filter_dilations,
const std::array<index_t, NDimSpatial>& input_left_pads,
const std::array<index_t, NDimSpatial>& input_right_pads,
const AElementwiseOperation& a_element_op,
const BElementwiseOperation& b_element_op,
const CDEElementwiseOperation& cde_element_op)
: p_a_grid_{static_cast<const ADataType*>(p_a)},
p_b_grid_{static_cast<const BDataType*>(p_b)},
p_ds_grid_{},
p_e_grid_{static_cast<EDataType*>(p_e)},
num_group_{a_g_n_c_wis_lengths[0]},
a_grid_desc_m_k_{DeviceOp::MakeAGridDescriptor_M_K<ALayout>(a_g_n_c_wis_lengths,
a_g_n_c_wis_strides,
b_g_k_c_xs_lengths,
b_g_k_c_xs_strides,
e_g_n_k_wos_lengths,
e_g_n_k_wos_strides,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads)},
b_grid_desc_n_k_{DeviceOp::MakeBGridDescriptor_N_K<BLayout>(b_g_k_c_xs_lengths,
b_g_k_c_xs_strides)},
ds_grid_desc_m_n_{},
e_grid_desc_m_n_{DeviceOp::MakeEGridDescriptor_M_N<ELayout>(e_g_n_k_wos_lengths,
e_g_n_k_wos_strides)},
a_grid_desc_ak0_m_ak1_{
GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(a_grid_desc_m_k_)},
b_grid_desc_bk0_n_bk1_{
GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(b_grid_desc_n_k_)},
ds_grid_desc_mblock_mperblock_nblock_nperblock_{},
e_grid_desc_mblock_mperblock_nblock_nperblock_{},
block_2_etile_map_{GridwiseGemm::MakeDefaultBlock2ETileMap(e_grid_desc_m_n_)},
compute_ptr_offset_of_batch_{},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
cde_element_op_{cde_element_op},
a_g_n_c_wis_lengths_{a_g_n_c_wis_lengths},
a_g_n_c_wis_strides_{a_g_n_c_wis_strides},
b_g_k_c_xs_lengths_{b_g_k_c_xs_lengths},
b_g_k_c_xs_strides_{b_g_k_c_xs_strides},
ds_g_n_k_wos_lengths_{ds_g_n_k_wos_lengths},
ds_g_n_k_wos_strides_{ds_g_n_k_wos_strides},
e_g_n_k_wos_lengths_{e_g_n_k_wos_lengths},
e_g_n_k_wos_strides_{e_g_n_k_wos_strides},
conv_filter_strides_{conv_filter_strides},
conv_filter_dilations_{conv_filter_dilations},
input_left_pads_{input_left_pads},
input_right_pads_{input_right_pads}
{
// A/B/E Batch Stride
compute_ptr_offset_of_batch_.BatchStrideA_ = a_g_n_c_wis_strides[0];
compute_ptr_offset_of_batch_.BatchStrideB_ = b_g_k_c_xs_strides[0];
compute_ptr_offset_of_batch_.BatchStrideE_ = e_g_n_k_wos_strides[0];
// populate pointer, batch stride, desc for Ds
static_for<0, NumDTensor, 1>{}([&](auto i) {
using DLayout = remove_cvref_t<tuple_element_t<i.value, DsLayout>>;
using DDataType = remove_cvref_t<tuple_element_t<i.value, DsDataType>>;
// D pointer
p_ds_grid_(i) = static_cast<const DDataType*>(p_ds[i]);
// D batch stride
compute_ptr_offset_of_batch_.BatchStrideDs_(i) = ds_g_n_k_wos_strides[i][0];
// D desc
ds_grid_desc_m_n_(i) = DeviceOp::MakeEGridDescriptor_M_N<DLayout>(
ds_g_n_k_wos_lengths[i], ds_g_n_k_wos_strides[i]);
});
// populate desc for Ds/E
if(GridwiseGemm::CheckValidity(a_grid_desc_m_k_,
b_grid_desc_n_k_,
ds_grid_desc_m_n_,
e_grid_desc_m_n_,
block_2_etile_map_))
{
e_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
e_grid_desc_m_n_);
ds_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
ds_grid_desc_m_n_);
}
}
void Print() const
{
std::cout << "A[M, K]: " << a_grid_desc_m_k_ << std::endl;
std::cout << "B[N, K]: " << b_grid_desc_n_k_ << std::endl;
static_for<0, NumDTensor, 1>{}(
[&](auto i) { std::cout << "Ds[M, N]: " << ds_grid_desc_m_n_[i] << std::endl; });
std::cout << "E[M, N]: " << e_grid_desc_m_n_ << std::endl;
}
// private:
// pointers
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
typename GridwiseGemm::DsGridPointer p_ds_grid_;
EDataType* p_e_grid_;
// tensor descriptors for problem definiton
index_t num_group_;
AGridDesc_M_K a_grid_desc_m_k_;
BGridDesc_N_K b_grid_desc_n_k_;
DsGridDesc_M_N ds_grid_desc_m_n_;
EGridDesc_M_N e_grid_desc_m_n_;
// tensor descriptors for block/thread-wise copy
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock
ds_grid_desc_mblock_mperblock_nblock_nperblock_;
EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock e_grid_desc_mblock_mperblock_nblock_nperblock_;
// block-to-e-tile map
Block2ETileMap block_2_etile_map_;
// for computing batch offset
ComputePtrOffsetOfStridedBatch<NumDTensor> compute_ptr_offset_of_batch_;
// element-wise op
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CDEElementwiseOperation cde_element_op_;
// for checking IsSupportedArgument()
std::array<index_t, NDimSpatial + 3> a_g_n_c_wis_lengths_;
std::array<index_t, NDimSpatial + 3> a_g_n_c_wis_strides_;
std::array<index_t, NDimSpatial + 3> b_g_k_c_xs_lengths_;
std::array<index_t, NDimSpatial + 3> b_g_k_c_xs_strides_;
std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor> ds_g_n_k_wos_lengths_;
std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor> ds_g_n_k_wos_strides_;
std::array<index_t, NDimSpatial + 3> e_g_n_k_wos_lengths_;
std::array<index_t, NDimSpatial + 3> e_g_n_k_wos_strides_;
std::array<index_t, NDimSpatial> conv_filter_strides_;
std::array<index_t, NDimSpatial> conv_filter_dilations_;
std::array<index_t, NDimSpatial> input_left_pads_;
std::array<index_t, NDimSpatial> input_right_pads_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(stream_config.log_level_ > 0)
{
arg.Print();
}
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_m_k_,
arg.b_grid_desc_n_k_,
arg.ds_grid_desc_m_n_,
arg.e_grid_desc_m_n_,
arg.block_2_etile_map_))
{
throw std::runtime_error(
"wrong! GridwiseGemmMultipleD_xdl_cshuffle has invalid setting");
}
const index_t grid_size =
arg.block_2_etile_map_.CalculateGridSize(arg.e_grid_desc_m_n_) * arg.num_group_;
const auto K =
arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) * arg.a_grid_desc_ak0_m_ak1_.GetLength(I2);
auto launch_kernel = [&](auto has_main_k_block_loop) {
constexpr bool has_main_loop = has_main_k_block_loop.value;
const auto kernel = kernel_grouped_conv_fwd_multiple_d_xdl_cshuffle<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
typename GridwiseGemm::DsGridPointer,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
DeviceOp::DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
DeviceOp::EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
Block2ETileMap,
ComputePtrOffsetOfStridedBatch<NumDTensor>,
has_main_loop>;
return launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_ds_grid_,
arg.p_e_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.cde_element_op_,
arg.a_g_n_c_wis_lengths_[0], // Group count
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.ds_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.e_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_etile_map_,
arg.compute_ptr_offset_of_batch_);
};
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
return launch_kernel(integral_constant<bool, true>{});
}
else
{
return launch_kernel(integral_constant<bool, false>{});
}
}
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static bool IsSupportedArgument(const Argument& arg)
{
namespace ctc = tensor_layout::convolution;
// check device
if(get_device_name() == "gfx908")
{
if constexpr(!(is_same_v<AccDataType, float> || is_same_v<AccDataType, float> ||
is_same_v<AccDataType, int32_t>))
{
return false;
}
}
else if(get_device_name() == "gfx90a")
{
if constexpr(!(is_same_v<AccDataType, float> || is_same_v<AccDataType, float> ||
is_same_v<AccDataType, int32_t> || is_same_v<AccDataType, double>))
{
return false;
}
}
else
{
return false;
}
// check ConvolutionForwardSpecialization
if constexpr(ConvForwardSpecialization ==
ConvolutionForwardSpecialization::Filter1x1Stride1Pad0)
{
// check if it's 1x1, stride=1 conv
for(index_t i = 0; i < NDimSpatial; ++i)
{
const index_t X = arg.b_g_k_c_xs_lengths_[i + 2];
const index_t ConvStride = arg.conv_filter_strides_[i];
const index_t LeftPad = arg.input_left_pads_[i];
const index_t RightPad = arg.input_right_pads_[i];
if(!(X == 1 && ConvStride == 1 && LeftPad == 0 && RightPad == 0))
{
return false;
}
}
}
else if constexpr(ConvForwardSpecialization ==
ConvolutionForwardSpecialization::Filter1x1Pad0)
{
// check if it's 1x1 conv
for(index_t i = 0; i < NDimSpatial; ++i)
{
const index_t X = arg.b_g_k_c_xs_lengths_[i + 2];
const index_t LeftPad = arg.input_left_pads_[i];
const index_t RightPad = arg.input_right_pads_[i];
if(!(X == 1 && LeftPad == 0 && RightPad == 0))
{
return false;
}
}
}
// check vector access of A
// FIXME: layout
if constexpr(is_same_v<ALayout, ctc::G_NW_C> || is_same_v<ALayout, ctc::G_NHW_C> ||
is_same_v<ALayout, ctc::G_NDHW_C> || is_same_v<ALayout, ctc::GNWC> ||
is_same_v<ALayout, ctc::GNHWC> || is_same_v<ALayout, ctc::GNDHWC> ||
is_same_v<ALayout, ctc::NWGC> || is_same_v<ALayout, ctc::NHWGC> ||
is_same_v<ALayout, ctc::NDHWGC>)
{
const index_t C = arg.a_g_n_c_wis_lengths_[2];
if(!(ABlockTransferSrcVectorDim == 2 && C % ABlockTransferSrcScalarPerVector == 0))
{
return false;
}
}
else
{
return false;
}
// check vector access of B
// FIXME: layout
if constexpr(is_same_v<BLayout, ctc::G_K_X_C> || is_same_v<BLayout, ctc::G_K_YX_C> ||
is_same_v<BLayout, ctc::G_K_ZYX_C> || is_same_v<BLayout, ctc::GKXC> ||
is_same_v<BLayout, ctc::GKYXC> || is_same_v<BLayout, ctc::GKZYXC> ||
is_same_v<BLayout, ctc::KXGC> || is_same_v<BLayout, ctc::KYXGC> ||
is_same_v<BLayout, ctc::KZYXGC>)
{
const index_t C = arg.b_g_k_c_xs_lengths_[2];
if(!(BBlockTransferSrcVectorDim == 2 && C % BBlockTransferSrcScalarPerVector == 0))
{
return false;
}
}
else
{
return false;
}
// check vector access of Ds
bool valid = true;
static_for<0, NumDTensor, 1>{}([&](auto i) {
using DLayout = remove_cvref_t<tuple_element_t<i.value, DsLayout>>;
// FIXME: layout
if constexpr(is_same_v<DLayout, ctc::G_NW_K> || is_same_v<DLayout, ctc::G_NHW_K> ||
is_same_v<DLayout, ctc::G_NDHW_K> || is_same_v<DLayout, ctc::GNWK> ||
is_same_v<DLayout, ctc::GNHWK> || is_same_v<DLayout, ctc::GNDHWK> ||
is_same_v<DLayout, ctc::NWGK> || is_same_v<DLayout, ctc::NHWGK> ||
is_same_v<DLayout, ctc::NDHWGK> || is_same_v<DLayout, ctc::GK> ||
is_same_v<DLayout, ctc::G_K>)
{
const index_t K = arg.ds_g_n_k_wos_lengths_[i][2];
if(!(K % CDEBlockTransferScalarPerVector_NPerBlock == 0))
{
valid = false;
}
}
else
{
valid = false;
}
});
if(!valid)
{
return false;
}
// check vector access of E
if constexpr(is_same_v<ELayout, ctc::G_NW_K> || is_same_v<ELayout, ctc::G_NHW_K> ||
is_same_v<ELayout, ctc::G_NDHW_K> || is_same_v<ELayout, ctc::GNWK> ||
is_same_v<ELayout, ctc::GNHWK> || is_same_v<ELayout, ctc::GNDHWK> ||
is_same_v<ELayout, ctc::NWGK> || is_same_v<ELayout, ctc::NHWGK> ||
is_same_v<ELayout, ctc::NDHWGK>)
{
const index_t K = arg.e_g_n_k_wos_lengths_[2];
if(!(K % CDEBlockTransferScalarPerVector_NPerBlock == 0))
{
return false;
}
}
else
{
return false;
}
// check Gridwise GEMM
return GridwiseGemm::CheckValidity(arg.a_grid_desc_m_k_,
arg.b_grid_desc_n_k_,
arg.ds_grid_desc_m_n_,
arg.e_grid_desc_m_n_,
arg.block_2_etile_map_);
}
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(
const void* p_a,
const void* p_b,
const std::array<const void*, NumDTensor>& p_ds,
void* p_e,
const std::array<index_t, NDimSpatial + 3>& a_g_n_c_wis_lengths,
const std::array<index_t, NDimSpatial + 3>& a_g_n_c_wis_strides,
const std::array<index_t, NDimSpatial + 3>& b_g_k_c_xs_lengths,
const std::array<index_t, NDimSpatial + 3>& b_g_k_c_xs_strides,
const std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor>& ds_g_n_k_wos_lengths,
const std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor>& ds_g_n_k_wos_strides,
const std::array<index_t, NDimSpatial + 3>& e_g_n_k_wos_lengths,
const std::array<index_t, NDimSpatial + 3>& e_g_n_k_wos_strides,
const std::array<index_t, NDimSpatial>& conv_filter_strides,
const std::array<index_t, NDimSpatial>& conv_filter_dilations,
const std::array<index_t, NDimSpatial>& input_left_pads,
const std::array<index_t, NDimSpatial>& input_right_pads,
const AElementwiseOperation& a_element_op,
const BElementwiseOperation& b_element_op,
const CDEElementwiseOperation& cde_element_op)
{
return Argument{p_a,
p_b,
p_ds,
p_e,
a_g_n_c_wis_lengths,
a_g_n_c_wis_strides,
b_g_k_c_xs_lengths,
b_g_k_c_xs_strides,
ds_g_n_k_wos_lengths,
ds_g_n_k_wos_strides,
e_g_n_k_wos_lengths,
e_g_n_k_wos_strides,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
a_element_op,
b_element_op,
cde_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
std::unique_ptr<BaseArgument> MakeArgumentPointer(
const void* p_a,
const void* p_b,
const std::array<const void*, NumDTensor>& p_ds,
void* p_e,
const std::array<index_t, NDimSpatial + 3>& a_g_n_c_wis_lengths,
const std::array<index_t, NDimSpatial + 3>& a_g_n_c_wis_strides,
const std::array<index_t, NDimSpatial + 3>& b_g_k_c_xs_lengths,
const std::array<index_t, NDimSpatial + 3>& b_g_k_c_xs_strides,
const std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor>& ds_g_n_k_wos_lengths,
const std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor>& ds_g_n_k_wos_strides,
const std::array<index_t, NDimSpatial + 3>& e_g_n_k_wos_lengths,
const std::array<index_t, NDimSpatial + 3>& e_g_n_k_wos_strides,
const std::array<index_t, NDimSpatial>& conv_filter_strides,
const std::array<index_t, NDimSpatial>& conv_filter_dilations,
const std::array<index_t, NDimSpatial>& input_left_pads,
const std::array<index_t, NDimSpatial>& input_right_pads,
const AElementwiseOperation& a_element_op,
const BElementwiseOperation& b_element_op,
const CDEElementwiseOperation& cde_element_op) override
{
return std::make_unique<Argument>(p_a,
p_b,
p_ds,
p_e,
a_g_n_c_wis_lengths,
a_g_n_c_wis_strides,
b_g_k_c_xs_lengths,
b_g_k_c_xs_strides,
ds_g_n_k_wos_lengths,
ds_g_n_k_wos_strides,
e_g_n_k_wos_lengths,
e_g_n_k_wos_strides,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
a_element_op,
b_element_op,
cde_element_op);
}
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceGroupedConvFwdMultipleD_Xdl_CShuffle"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< getConvForwardSpecializationString(ConvForwardSpecialization)
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

666
include/ck/tensor_operation/gpu/device/impl/device_grouped_gemm_xdl.hpp Normal file
View File

@@ -0,0 +1,666 @@
#pragma once
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_grouped_gemm.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <typename GridwiseGemm,
typename GemmDesc,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_grouped_gemm_xdl(const void CK_CONSTANT_ADDRESS_SPACE* gemm_descs_const,
const index_t group_count,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const CDEElementwiseOperation c_element_op)
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__))
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
const index_t block_id = get_block_1d_id();
const auto gemm_desc_ptr =
reinterpret_cast<const GemmDesc*>(cast_pointer_to_generic_address_space(gemm_descs_const));
index_t left = 0;
index_t right = group_count;
index_t group_id = index_t((left + right) / 2);
while((!(block_id >= gemm_desc_ptr[group_id].BlockStart_ &&
block_id < gemm_desc_ptr[group_id].BlockEnd_)) &&
left <= right)
{
if(block_id < gemm_desc_ptr[group_id].BlockStart_)
{
right = group_id;
}
else
{
left = group_id;
}
group_id = index_t((left + right) / 2);
}
GridwiseGemm::template Run<HasMainKBlockLoop>(
gemm_desc_ptr[group_id].a_ptr_,
gemm_desc_ptr[group_id].b_ptr_,
gemm_desc_ptr[group_id].ds_ptr_,
gemm_desc_ptr[group_id].e_ptr_,
p_shared,
a_element_op,
b_element_op,
c_element_op,
gemm_desc_ptr[group_id].a_grid_desc_ak0_m_ak1_,
gemm_desc_ptr[group_id].b_grid_desc_bk0_n_bk1_,
gemm_desc_ptr[group_id].ds_grid_desc_mblock_mperblock_nblock_nperblock_,
gemm_desc_ptr[group_id].e_grid_desc_mblock_mperblock_nblock_nperblock_,
gemm_desc_ptr[group_id].block_2_etile_map_);
#else
ignore = gemm_descs_const;
ignore = group_count;
ignore = a_element_op;
ignore = b_element_op;
ignore = c_element_op;
#endif
}
template <typename ALayout,
typename BLayout,
typename DsLayout,
typename ELayout,
typename ADataType,
typename BDataType,
typename AccDataType,
typename CShuffleDataType,
typename DsDataType,
typename EDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
GemmSpecialization GemmSpec,
ck::index_t NumPrefetch,
ck::index_t BlockSize,
ck::index_t MPerBlock,
ck::index_t NPerBlock,
ck::index_t KPerBlock,
ck::index_t AK1,
ck::index_t BK1,
ck::index_t MPerXDL,
ck::index_t NPerXDL,
ck::index_t MXdlPerWave,
ck::index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_K0_M_K1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
ck::index_t ABlockTransferSrcVectorDim,
ck::index_t ABlockTransferSrcScalarPerVector,
ck::index_t ABlockTransferDstScalarPerVector_K1,
bool ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_K0_N_K1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
ck::index_t BBlockTransferSrcVectorDim,
ck::index_t BBlockTransferSrcScalarPerVector,
ck::index_t BBlockTransferDstScalarPerVector_K1,
bool BBlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CDEBlockTransferScalarPerVector_NPerBlock,
LoopScheduler LoopSched = make_default_loop_scheduler()>
struct DeviceGroupedGemm_Xdl : public DeviceGroupedGemm<ALayout,
BLayout,
DsLayout,
ELayout,
ADataType,
BDataType,
DsDataType,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation>
{
using DeviceOp = DeviceGroupedGemm_Xdl;
static constexpr index_t NumDTensor = DsDataType::Size();
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto matrix_padder =
MatrixPadder<GemmSpec, index_t, index_t, index_t>{MPerBlock, NPerBlock, KPerBlock};
static auto MakeAGridDescriptor_M_K(index_t MRaw, index_t KRaw, index_t StrideA)
{
const auto a_grid_desc_mraw_kraw = [&]() {
if constexpr(is_same_v<tensor_layout::gemm::RowMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(StrideA, I1));
}
else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, ALayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(I1, StrideA));
}
}();
return matrix_padder.PadADescriptor_M_K(a_grid_desc_mraw_kraw);
}
static auto MakeBGridDescriptor_N_K(index_t KRaw, index_t NRaw, index_t StrideB)
{
const auto b_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideB));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideB, I1));
}
}();
return matrix_padder.PadBDescriptor_N_K(b_grid_desc_nraw_kraw);
}
template <typename ELay>
static auto MakeEGridDescriptor_M_N(index_t MRaw, index_t NRaw, index_t StrideE)
{
const auto e_grid_desc_mraw_nraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, ELay>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(StrideE, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, ELay>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(I1, StrideE));
}
}();
return matrix_padder.PadCDescriptor_M_N(e_grid_desc_mraw_nraw);
}
static auto MakeDsGridDescriptor_M_N(const std::array<index_t, NumDTensor>& MRaws,
const std::array<index_t, NumDTensor>& NRaws,
const std::array<index_t, NumDTensor>& DsStride)
{
return generate_tuple(
[&](auto i) {
using DLayout = remove_cvref_t<tuple_element_t<i.value, DsLayout>>;
return DeviceOp::MakeEGridDescriptor_M_N<DLayout>(MRaws[i], NRaws[i], DsStride[i]);
},
Number<NumDTensor>{});
}
using AGridDesc_M_K = decltype(MakeAGridDescriptor_M_K(1, 1, 1));
using BGridDesc_N_K = decltype(MakeBGridDescriptor_N_K(1, 1, 1));
using DsGridDesc_M_N = remove_cvref_t<decltype(MakeDsGridDescriptor_M_N({}, {}, {}))>;
using EGridDesc_M_N = decltype(MakeEGridDescriptor_M_N<ELayout>(1, 1, 1));
// GridwiseGemm
using GridwiseGemm = GridwiseGemmMultipleD_xdl_cshuffle<
ADataType, // TODO: distinguish A/B datatype
AccDataType,
CShuffleDataType,
DsDataType,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
InMemoryDataOperationEnum::Set,
NumPrefetch, // NumGemmKPrefetchStage
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
AK1,
BK1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_K0_M_K1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_K1,
false, // AThreadTransferSrcResetCoordinateAfterRun,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_K0_N_K1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_K1,
false, // BThreadTransferSrcResetCoordinateAfterRun,
BBlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CDEBlockTransferScalarPerVector_NPerBlock,
LoopSched>;
using AGridDesc_AK0_M_AK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(AGridDesc_M_K{}))>;
using BGridDesc_BK0_N_BK1 = remove_cvref_t<decltype(
GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(BGridDesc_N_K{}))>;
using DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock = remove_cvref_t<decltype(
GridwiseGemm::MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(DsGridDesc_M_N{}))>;
using EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock = remove_cvref_t<decltype(
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(EGridDesc_M_N{}))>;
struct GroupedGemmBlock2ETileMap
{
using Block2ETileMap =
remove_cvref_t<decltype(GridwiseGemm::MakeDefaultBlock2ETileMap(EGridDesc_M_N{}))>;
GroupedGemmBlock2ETileMap()
{
block_2_etile_map_ = GridwiseGemm::MakeDefaultBlock2ETileMap(EGridDesc_M_N{});
BlockStart_ = -1;
}
GroupedGemmBlock2ETileMap(const EGridDesc_M_N& e_grid_desc_m_n, ck::index_t BlockStart)
{
block_2_etile_map_ = GridwiseGemm::MakeDefaultBlock2ETileMap(e_grid_desc_m_n);
BlockStart_ = BlockStart;
}
template <typename TopIdx>
__host__ __device__ constexpr auto CalculateBottomIndex(const TopIdx& idx_top) const
{
return block_2_etile_map_.CalculateBottomIndex(
make_multi_index(idx_top[I0] - BlockStart_));
}
// it's actually E-Tile
template <typename CTileIdx, typename CTileDim>
__host__ __device__ bool ValidCTileIndex(const CTileIdx& c_tile_idx,
const CTileDim& c_tile_dim) const
{
return block_2_etile_map_.ValidCTileIndex(c_tile_idx, c_tile_dim);
}
__host__ bool CheckValidity(const EGridDesc_M_N& e_grid_desc_m_n) const
{
return block_2_etile_map_.CheckValidity(e_grid_desc_m_n);
}
Block2ETileMap block_2_etile_map_;
ck::index_t BlockStart_;
};
struct GemmBiasTransKernelArg
{
// pointers
const ADataType* a_ptr_;
const BDataType* b_ptr_;
typename GridwiseGemm::DsGridPointer ds_ptr_;
EDataType* e_ptr_;
// tensor descriptors for problem definiton
AGridDesc_M_K a_grid_desc_m_k_;
BGridDesc_N_K b_grid_desc_n_k_;
DsGridDesc_M_N ds_grid_desc_m_n_;
EGridDesc_M_N e_grid_desc_m_n_;
// tensor descriptors for block/thread-wise copy
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock
ds_grid_desc_mblock_mperblock_nblock_nperblock_;
EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock e_grid_desc_mblock_mperblock_nblock_nperblock_;
// block-to-e-tile map
GroupedGemmBlock2ETileMap block_2_etile_map_;
ck::index_t BlockStart_, BlockEnd_;
};
// Argument
struct Argument : public BaseArgument
{
Argument(std::vector<const void*>& p_As,
std::vector<const void*>& p_Bs,
std::vector<std::array<const void*, NumDTensor>>& p_Ds,
std::vector<void*>& p_Es,
std::vector<GemmDesc>& gemm_descs,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation c_element_op)
: a_element_op_{a_element_op}, b_element_op_{b_element_op}, c_element_op_{c_element_op}
{
grid_size_ = 0;
group_count_ = ck::type_convert<ck::index_t>(gemm_descs.size());
if(!(group_count_ == ck::type_convert<ck::index_t>(p_As.size()) &&
group_count_ == ck::type_convert<ck::index_t>(p_Bs.size()) &&
group_count_ == ck::type_convert<ck::index_t>(p_Es.size())))
{
throw std::runtime_error("wrong! group_count_ != p_As/b/c.size");
}
gemm_desc_kernel_arg_.reserve(group_count_);
for(std::size_t i = 0; i < gemm_descs.size(); i++)
{
const index_t M = gemm_descs[i].M_;
const index_t N = gemm_descs[i].N_;
const index_t K = gemm_descs[i].K_;
const index_t StrideA = gemm_descs[i].stride_A_;
const index_t StrideB = gemm_descs[i].stride_B_;
const index_t StrideC = gemm_descs[i].stride_C_;
// pointer
typename GridwiseGemm::DsGridPointer p_ds_grid{};
static_for<0, NumDTensor, 1>{}([&](auto j) {
using DDataType = remove_cvref_t<tuple_element_t<j.value, DsDataType>>;
p_ds_grid(j) = static_cast<const DDataType*>(p_Ds[i][j]);
});
// tensor descriptors for problem definiton
const auto a_grid_desc_m_k = DeviceOp::MakeAGridDescriptor_M_K(M, K, StrideA);
const auto b_grid_desc_n_k = DeviceOp::MakeBGridDescriptor_N_K(K, N, StrideB);
DsGridDesc_M_N ds_grid_desc_m_n;
static_for<0, NumDTensor, 1>{}([&](auto j) {
using DLayout = remove_cvref_t<tuple_element_t<j.value, DsLayout>>;
ds_grid_desc_m_n(j) = DeviceOp::MakeEGridDescriptor_M_N<DLayout>(
M, N, gemm_descs[i].stride_Ds_[j]);
});
const auto e_grid_desc_m_n =
DeviceOp::MakeEGridDescriptor_M_N<ELayout>(M, N, StrideC);
// tensor descriptors for block/thread-wise copy
const auto a_grid_desc_ak0_m_ak1 =
GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(a_grid_desc_m_k);
const auto b_grid_desc_bk0_n_bk1 =
GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(b_grid_desc_n_k);
const index_t grid_size_grp =
GroupedGemmBlock2ETileMap(e_grid_desc_m_n, 0)
.block_2_etile_map_.CalculateGridSize(e_grid_desc_m_n);
const index_t BlockStart = grid_size_;
const index_t BlockEnd = grid_size_ + grid_size_grp;
grid_size_ += grid_size_grp;
// block-to-e-tile map
const auto block_2_etile_map =
GroupedGemmBlock2ETileMap(e_grid_desc_m_n, BlockStart);
if(GridwiseGemm::CheckValidity(a_grid_desc_m_k,
b_grid_desc_n_k,
ds_grid_desc_m_n,
e_grid_desc_m_n,
block_2_etile_map))
{
// tensor descriptors for block/thread-wise copy
DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock
ds_grid_desc_mblock_mperblock_nblock_nperblock;
static_for<0, NumDTensor, 1>{}([&](auto j) {
ds_grid_desc_mblock_mperblock_nblock_nperblock(j) =
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
ds_grid_desc_m_n[j]);
});
const auto e_grid_desc_mblock_mperblock_nblock_nperblock =
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
e_grid_desc_m_n);
gemm_desc_kernel_arg_.push_back(
GemmBiasTransKernelArg{static_cast<const ADataType*>(p_As[i]),
static_cast<const BDataType*>(p_Bs[i]),
p_ds_grid,
static_cast<EDataType*>(p_Es[i]),
a_grid_desc_m_k,
b_grid_desc_n_k,
ds_grid_desc_m_n,
e_grid_desc_m_n,
a_grid_desc_ak0_m_ak1,
b_grid_desc_bk0_n_bk1,
ds_grid_desc_mblock_mperblock_nblock_nperblock,
e_grid_desc_mblock_mperblock_nblock_nperblock,
block_2_etile_map,
BlockStart,
BlockEnd});
}
}
}
// private:
index_t group_count_;
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CDEElementwiseOperation c_element_op_;
std::vector<GemmBiasTransKernelArg> gemm_desc_kernel_arg_;
index_t grid_size_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
bool has_main_k_block_loop = true;
for(std::size_t i = 0; i < arg.gemm_desc_kernel_arg_.size(); i++)
{
std::cout << "group: " << i << " arg.a_grid_desc_ak0_m_ak1_{"
<< arg.gemm_desc_kernel_arg_[i].a_grid_desc_ak0_m_ak1_.GetLength(I0)
<< ", "
<< arg.gemm_desc_kernel_arg_[i].a_grid_desc_ak0_m_ak1_.GetLength(I1)
<< ", "
<< arg.gemm_desc_kernel_arg_[i].a_grid_desc_ak0_m_ak1_.GetLength(I2)
<< "}";
std::cout << ", arg.b_grid_desc_bk0_n_bk1_{"
<< arg.gemm_desc_kernel_arg_[i].b_grid_desc_bk0_n_bk1_.GetLength(I0)
<< ", "
<< arg.gemm_desc_kernel_arg_[i].b_grid_desc_bk0_n_bk1_.GetLength(I1)
<< ", "
<< arg.gemm_desc_kernel_arg_[i].b_grid_desc_bk0_n_bk1_.GetLength(I2)
<< "}";
std::cout << ", arg.e_grid_desc_m_n_{ "
<< arg.gemm_desc_kernel_arg_[i].e_grid_desc_m_n_.GetLength(I0) << ", "
<< arg.gemm_desc_kernel_arg_[i].e_grid_desc_m_n_.GetLength(I1) << "}"
<< std::endl;
if(!GridwiseGemm::CheckValidity(arg.gemm_desc_kernel_arg_[i].a_grid_desc_m_k_,
arg.gemm_desc_kernel_arg_[i].b_grid_desc_n_k_,
arg.gemm_desc_kernel_arg_[i].ds_grid_desc_m_n_,
arg.gemm_desc_kernel_arg_[i].e_grid_desc_m_n_,
arg.gemm_desc_kernel_arg_[i].block_2_etile_map_))
{
throw std::runtime_error(
"wrong! GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r3 has invalid setting");
}
const auto K = arg.gemm_desc_kernel_arg_[i].a_grid_desc_ak0_m_ak1_.GetLength(I0) *
arg.gemm_desc_kernel_arg_[i].a_grid_desc_ak0_m_ak1_.GetLength(I2);
if(GridwiseGemm::CalculateHasMainKBlockLoop(K) != has_main_k_block_loop)
{
throw std::runtime_error("wrong! not all gemm has_main_k_block_loop");
}
}
hipGetErrorString(
hipMemcpy(arg.p_workspace_,
arg.gemm_desc_kernel_arg_.data(),
arg.gemm_desc_kernel_arg_.size() * sizeof(GemmBiasTransKernelArg),
hipMemcpyHostToDevice));
float ave_time = 0;
auto launch_kernel = [&](auto has_main_k_block_loop_) {
const auto kernel = kernel_grouped_gemm_xdl<GridwiseGemm,
GemmBiasTransKernelArg,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
has_main_k_block_loop_>;
return launch_and_time_kernel(
stream_config,
kernel,
dim3(arg.grid_size_),
dim3(BlockSize),
0,
cast_pointer_to_constant_address_space(arg.p_workspace_),
arg.gemm_desc_kernel_arg_.size(),
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_);
};
if(has_main_k_block_loop)
{
ave_time = launch_kernel(integral_constant<bool, true>{});
}
else
{
ave_time = launch_kernel(integral_constant<bool, false>{});
}
return ave_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static bool IsSupportedArgument(const Argument& arg)
{
if(ck::type_convert<ck::index_t>(arg.gemm_desc_kernel_arg_.size()) != arg.group_count_)
{
return false;
}
return true;
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(std::vector<const void*>& p_As,
std::vector<const void*>& p_Bs,
std::vector<std::array<const void*, NumDTensor>>& p_Ds,
std::vector<void*>& p_Es,
std::vector<GemmDesc> gemm_descs,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation c_element_op)
{
return Argument{
p_As, p_Bs, p_Ds, p_Es, gemm_descs, a_element_op, b_element_op, c_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument>
MakeArgumentPointer(std::vector<const void*>& p_As,
std::vector<const void*>& p_Bs,
std::vector<std::array<const void*, NumDTensor>>& p_Ds,
std::vector<void*>& p_Es,
std::vector<GemmDesc>& gemm_descs,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation c_element_op) override
{
return std::make_unique<Argument>(
p_As, p_Bs, p_Ds, p_Es, gemm_descs, a_element_op, b_element_op, c_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceGroupedGemm_Xdl"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< AK1 << ", "
<< BK1 << ", "
<< MPerXDL << ", "
<< NPerXDL << ", "
<< MXdlPerWave << ", "
<< NXdlPerWave
<< ">";
// clang-format on
return str.str();
}
size_t GetWorkSpaceSize(const BaseArgument* p_arg) const override
{
return dynamic_cast<const Argument*>(p_arg)->group_count_ * sizeof(GemmBiasTransKernelArg);
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

595
include/ck/tensor_operation/gpu/device/impl/device_multiple_reduce_multiblock.hpp Normal file
View File

@@ -0,0 +1,595 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/sequence.hpp"
#include "ck/utility/reduction_operator.hpp"
#include "ck/tensor_operation/gpu/device/device_base.hpp"
#include "ck/tensor_operation/gpu/device/device_multiple_reduce.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_reduce_common.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_2d_multiple_reduction_multiblock.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_set_multiple_buffer_value.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <index_t NumReduction,
typename InDataType,
typename AccDataType,
typename OutDataTypeTuple,
index_t Rank,
index_t NumReduceDim,
typename ReduceOperation,
typename InElementwiseOperationTuple,
typename AccElementwiseOperationTuple,
InMemoryDataOperationEnum OutMemoryDataOperation,
bool PropagateNan,
index_t BlockSize,
index_t MThreadClusterSize,
index_t KThreadClusterSize,
index_t MThreadSliceSize,
index_t KThreadSliceSize,
index_t InSrcVectorDim,
index_t InSrcVectorSize,
typename OutDstVectorSizeSeq>
struct DeviceMultipleReduceMultiBlock : public DeviceMultipleReduce<Rank,
NumReduceDim,
NumReduction,
InElementwiseOperationTuple,
AccElementwiseOperationTuple>
{
static_assert(Rank <= 6, "Bigger Rank size is not supported!");
static_assert(BlockSize == MThreadClusterSize * KThreadClusterSize,
"Invalid thread cluster size assignments!");
static_assert((InSrcVectorDim == 0 && MThreadSliceSize % InSrcVectorSize == 0) ||
(InSrcVectorDim == 1 && KThreadSliceSize % InSrcVectorSize == 0),
"Invalid thread slice sizes and/or vector sizes configuration, please check!");
static_assert(NumReduction == OutDataTypeTuple::Size() &&
NumReduction == InElementwiseOperationTuple::Size() &&
NumReduction == AccElementwiseOperationTuple::Size() &&
NumReduction == OutDstVectorSizeSeq::Size(),
"All tuple should have the same size as the number of Reductions!");
static_assert(sequence_all_of(OutDstVectorSizeSeq{},
[](auto vectorSize) {
return (MThreadSliceSize % vectorSize == 0);
}),
"The OutDstVectorSize should completely divide the MThreadSliceSize!");
static constexpr bool CheckDataTypeTuple()
{
bool flag = true;
static_for<0, NumReduction, 1>{}([&](auto I) {
using OutDataType = remove_cvref_t<decltype(OutDataTypeTuple{}[I])>;
flag =
flag && ck::reduce::InMemoryDataOperatonSupportedOnDataType<OutMemoryDataOperation,
OutDataType>::value;
});
return flag;
};
static_assert(CheckDataTypeTuple(),
"The OutDataType must support the specified OutMemoryDataOperation!");
static constexpr index_t NumInvariantDim = Rank - NumReduceDim;
static constexpr index_t NumInputDim = Rank;
static constexpr index_t NumOutputDim = (NumInvariantDim == 0) ? 1 : NumInvariantDim;
static constexpr bool reduceAllDim = (NumInvariantDim == 0);
// So far, only AtomicAdd is considered, other Atomic Operation like AtomicMax can be added
// later
static constexpr bool use_multiblock =
(OutMemoryDataOperation == InMemoryDataOperationEnum::AtomicAdd);
static_assert(
ReduceOperation::IsCompatibleInMemoryDataOperation(OutMemoryDataOperation),
"The reduction accumulation operation must be compatible with the OutMemoryDataOperation!");
static constexpr index_t M_BlockTileSize = MThreadClusterSize * MThreadSliceSize;
static constexpr index_t K_BlockTileSize = KThreadClusterSize * KThreadSliceSize;
static auto GenerateOutDataTypePointerTuple()
{
return generate_tuple(
[&](auto I) {
using DataType = remove_cvref_t<decltype(OutDataTypeTuple{}[I])>;
return static_cast<DataType*>(nullptr);
},
Number<NumReduction>{});
};
using OutDataTypePointerTuple = decltype(GenerateOutDataTypePointerTuple());
static auto MakeSrc2dDescriptor(const std::array<index_t, NumInputDim>& inLengths,
const std::array<index_t, NumInputDim>& inStrides,
int blkGroupSize,
int numBlockTileIteration)
{
const auto tupleSrcLengths =
generate_tuple([&](auto I) { return inLengths[I]; }, Number<NumInputDim>{});
const auto tupleSrcStrides =
generate_tuple([&](auto I) { return inStrides[I]; }, Number<NumInputDim>{});
const auto inDesc = make_naive_tensor_descriptor(tupleSrcLengths, tupleSrcStrides);
const auto in_grid_desc_m_k = [&]() {
if constexpr(reduceAllDim)
{
const auto one_dim_inDesc = transform_tensor_descriptor(
inDesc,
make_tuple(make_merge_transform(tupleSrcLengths)),
make_tuple(typename arithmetic_sequence_gen<0, NumInputDim, 1>::type{}),
make_tuple(Sequence<0>{}));
return transform_tensor_descriptor(one_dim_inDesc,
make_tuple(make_unmerge_transform(make_tuple(
1, one_dim_inDesc.GetLength(Number<0>{})))),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0, 1>{}));
}
else
{
using InvariantDims = typename arithmetic_sequence_gen<0, NumInvariantDim, 1>::type;
using ReduceDims = typename arithmetic_sequence_gen<NumInvariantDim, Rank, 1>::type;
const auto reduceDimLengths = generate_tuple(
[&](auto I) { return inLengths[NumInvariantDim + I]; }, Number<NumReduceDim>{});
const auto invariantDimLengths =
generate_tuple([&](auto I) { return inLengths[I]; }, Number<NumInvariantDim>{});
return transform_tensor_descriptor(
inDesc,
make_tuple(make_merge_transform(invariantDimLengths),
make_merge_transform(reduceDimLengths)),
make_tuple(InvariantDims{}, ReduceDims{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
}();
const auto invariantLength = in_grid_desc_m_k.GetLength(Number<0>{});
const auto reduceLength = in_grid_desc_m_k.GetLength(Number<1>{});
const int reduceSizePerBlock = K_BlockTileSize * numBlockTileIteration;
const auto inPad_M =
math::integer_least_multiple(invariantLength, M_BlockTileSize) - invariantLength;
const auto inPad_K = reduceSizePerBlock * blkGroupSize - reduceLength;
auto in_grid_desc_m_k_padded = transform_tensor_descriptor(
in_grid_desc_m_k,
make_tuple(make_right_pad_transform(invariantLength, inPad_M),
make_right_pad_transform(reduceLength, inPad_K)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return (in_grid_desc_m_k_padded);
};
static auto MakeDst1dDescriptor(const std::array<index_t, NumOutputDim>& outLengths,
const std::array<index_t, NumOutputDim>& outStrides)
{
const auto tupleDstLengths =
generate_tuple([&](auto I) { return outLengths[I]; }, Number<NumOutputDim>{});
const auto tupleDstStrides =
generate_tuple([&](auto I) { return outStrides[I]; }, Number<NumOutputDim>{});
auto outDesc = make_naive_tensor_descriptor(tupleDstLengths, tupleDstStrides);
auto out_grid_desc_m = transform_tensor_descriptor(
outDesc,
make_tuple(make_merge_transform(tupleDstLengths)),
make_tuple(typename arithmetic_sequence_gen<0, NumOutputDim, 1>::type{}),
make_tuple(Sequence<0>{}));
const auto invariantLength = out_grid_desc_m.GetLength(Number<0>{});
const auto outPad =
math::integer_least_multiple(invariantLength, M_BlockTileSize) - invariantLength;
auto out_grid_desc_m_padded = transform_tensor_descriptor(
out_grid_desc_m,
make_tuple(make_right_pad_transform(invariantLength, outPad)),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0>{}));
return (out_grid_desc_m_padded);
};
static auto GenerateOutGrid1dDescTuple()
{
return generate_tuple(
[&](auto I) {
(void)I;
return MakeDst1dDescriptor(std::array<index_t, NumOutputDim>{},
std::array<index_t, NumOutputDim>{});
},
Number<NumReduction>{});
};
using InGridDesc_M_K = decltype(MakeSrc2dDescriptor(
std::array<index_t, NumInputDim>{}, std::array<index_t, NumInputDim>{}, 1, 1));
using OutGridDesc_M_Tuple = decltype(GenerateOutGrid1dDescTuple());
static auto MakeDst1dDescriptorForBufferSet(const std::array<index_t, NumOutputDim>& outLengths,
const std::array<index_t, NumOutputDim>& outStrides)
{
const auto tupleDstLengths =
generate_tuple([&](auto I) { return outLengths[I]; }, Number<NumOutputDim>{});
const auto tupleDstStrides =
generate_tuple([&](auto I) { return outStrides[I]; }, Number<NumOutputDim>{});
auto outDesc = make_naive_tensor_descriptor(tupleDstLengths, tupleDstStrides);
auto out_grid_desc_m = transform_tensor_descriptor(
outDesc,
make_tuple(make_merge_transform(tupleDstLengths)),
make_tuple(typename arithmetic_sequence_gen<0, NumOutputDim, 1>::type{}),
make_tuple(Sequence<0>{}));
const auto length = out_grid_desc_m.GetLength(Number<0>{});
const auto pad = math::integer_least_multiple(length, BlockSize) - length;
auto out_grid_desc_m_padded =
transform_tensor_descriptor(out_grid_desc_m,
make_tuple(make_right_pad_transform(length, pad)),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0>{}));
return (out_grid_desc_m_padded);
};
static auto GenerateOutGrid1dDescTuple_2()
{
return generate_tuple(
[&](auto I) {
(void)I;
return MakeDst1dDescriptorForBufferSet(std::array<index_t, NumOutputDim>{},
std::array<index_t, NumOutputDim>{});
},
Number<NumReduction>{});
};
using OutGridDesc_M_Tuple_2 = decltype(GenerateOutGrid1dDescTuple_2());
struct Argument : public BaseArgument
{
Argument(const std::array<index_t, NumInputDim>& inLengths,
const std::array<index_t, NumInputDim>& inStrides,
const std::array<index_t, NumOutputDim>& outLengths,
const std::array<std::array<index_t, NumOutputDim>, NumReduction>& outStridesArray,
const std::array<int, NumReduceDim>& reduceDims,
const std::array<const void*, NumReduction>& alphas,
const std::array<const void*, NumReduction>& betas,
const void* in_dev,
const std::array<void*, NumReduction>& out_dev_buffers,
const InElementwiseOperationTuple in_elementwise_op_tuple,
const AccElementwiseOperationTuple acc_elementwise_op_tuple)
: outLengths_{outLengths},
outStridesArray_{outStridesArray},
in_elementwise_op_tuple_{in_elementwise_op_tuple},
acc_elementwise_op_tuple_{acc_elementwise_op_tuple}
{
inLengths_ = shuffle_tensor_dimensions<Rank, NumReduceDim>(inLengths, reduceDims);
inStrides_ = shuffle_tensor_dimensions<Rank, NumReduceDim>(inStrides, reduceDims);
for(size_t i = 0; i < NumReduction; i++)
{
alpha_values_(i) = *static_cast<const AccDataType*>(alphas[i]);
beta_values_(i) = *static_cast<const AccDataType*>(betas[i]);
};
in_dev_ = static_cast<const InDataType*>(in_dev);
out_dev_buffers_ = generate_tuple(
[&](auto iR) {
using OutDataTypePointer =
remove_cvref_t<decltype(OutDataTypePointerTuple{}[iR])>;
using OutDataType = remove_cvref_t<remove_pointer_t<OutDataTypePointer>>;
return static_cast<OutDataType*>(out_dev_buffers[iR]);
},
Number<NumReduction>{});
std::tie(invariant_total_length, reduce_total_length) =
get_2d_lengths<Rank, NumReduceDim>(inLengths_);
if constexpr(use_multiblock)
{
int iterations = 1;
while(true)
{
int testBlkGroupSize =
(reduce_total_length + (K_BlockTileSize * iterations) - 1) /
(K_BlockTileSize * iterations);
// we want the blkGroupSize be not more than 128
if(testBlkGroupSize <= 128)
break;
iterations++;
};
blkGroupSize = (reduce_total_length + (K_BlockTileSize * iterations) - 1) /
(K_BlockTileSize * iterations);
numBlockTileIteration = iterations;
}
else
{
blkGroupSize = 1;
numBlockTileIteration =
(reduce_total_length + K_BlockTileSize - 1) / K_BlockTileSize;
};
in_grid_desc_m_k =
MakeSrc2dDescriptor(inLengths_, inStrides_, blkGroupSize, numBlockTileIteration);
out_grid_desc_m_tuple = generate_tuple(
[&](auto I) { return MakeDst1dDescriptor(outLengths, outStridesArray[I]); },
Number<NumReduction>{});
out_grid_desc_m_tuple_2 = generate_tuple(
[&](auto I) {
return MakeDst1dDescriptorForBufferSet(outLengths, outStridesArray[I]);
},
Number<NumReduction>{});
gridSize = math::integer_least_multiple(invariant_total_length, M_BlockTileSize) /
M_BlockTileSize * blkGroupSize;
gridSize_pre =
math::integer_least_multiple(invariant_total_length, BlockSize) / BlockSize;
}
std::array<index_t, NumInputDim> inLengths_;
std::array<index_t, NumInputDim> inStrides_;
std::array<index_t, NumOutputDim> outLengths_;
std::array<std::array<index_t, NumOutputDim>, NumReduction> outStridesArray_;
Array<AccDataType, NumReduction> alpha_values_;
Array<AccDataType, NumReduction> beta_values_;
const InDataType* in_dev_;
OutDataTypePointerTuple out_dev_buffers_;
InGridDesc_M_K in_grid_desc_m_k;
OutGridDesc_M_Tuple out_grid_desc_m_tuple;
OutGridDesc_M_Tuple_2 out_grid_desc_m_tuple_2;
InElementwiseOperationTuple in_elementwise_op_tuple_;
AccElementwiseOperationTuple acc_elementwise_op_tuple_;
long_index_t invariant_total_length;
long_index_t reduce_total_length;
int blkGroupSize;
int numBlockTileIteration;
size_t gridSize;
size_t gridSize_pre;
};
struct Invoker : public BaseInvoker
{
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
using GridwiseMultipleReduce =
GridwiseMultipleReduction_mk_to_m_multiblock<NumReduction,
InDataType,
OutDataTypePointerTuple,
AccDataType,
InGridDesc_M_K,
OutGridDesc_M_Tuple,
ReduceOperation,
InElementwiseOperationTuple,
AccElementwiseOperationTuple,
OutMemoryDataOperation,
PropagateNan,
BlockSize,
MThreadClusterSize,
KThreadClusterSize,
MThreadSliceSize,
KThreadSliceSize,
InSrcVectorDim,
InSrcVectorSize,
OutDstVectorSizeSeq>;
const auto kernel_main =
kernel_multiple_reduce_multiblock<GridwiseMultipleReduce,
NumReduction,
InDataType,
OutDataTypePointerTuple,
AccDataType,
InGridDesc_M_K,
OutGridDesc_M_Tuple,
InElementwiseOperationTuple,
AccElementwiseOperationTuple>;
float avg_time = 0;
if constexpr(use_multiblock)
{
auto identity_values = generate_tuple(
[&](auto iR) {
using OutDataType = remove_cvref_t<decltype(OutDataTypeTuple{}[iR])>;
return ck::reduce::GetIdentityValueForInMemoryDataOperation<OutDataType>(
OutMemoryDataOperation);
},
Number<NumReduction>{});
const auto kernel_pre = kernel_multiple_buffer_set_value<OutGridDesc_M_Tuple_2,
NumReduction,
BlockSize,
OutDataTypePointerTuple,
OutDataTypeTuple>;
avg_time += launch_and_time_kernel(stream_config,
kernel_pre,
dim3(arg.gridSize_pre),
dim3(BlockSize),
0,
arg.out_grid_desc_m_tuple_2,
arg.out_dev_buffers_,
identity_values);
};
avg_time += launch_and_time_kernel(stream_config,
kernel_main,
dim3(arg.gridSize),
dim3(BlockSize),
0,
arg.in_grid_desc_m_k,
arg.out_grid_desc_m_tuple,
arg.in_elementwise_op_tuple_,
arg.acc_elementwise_op_tuple_,
arg.blkGroupSize,
arg.numBlockTileIteration,
arg.alpha_values_,
arg.in_dev_,
arg.beta_values_,
arg.out_dev_buffers_);
return (avg_time);
};
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
};
};
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
const Argument* pArg = dynamic_cast<const Argument*>(p_arg);
if constexpr(use_multiblock)
{
for(size_t i = 0; i < pArg->beta_values_.Size(); i++)
if(pArg->beta_values_[i] != 0.0f)
return (false);
};
if constexpr(InSrcVectorDim == 0)
{
if constexpr(NumInvariantDim == 0)
{
return (false);
}
else
{
if(pArg->inStrides_[NumInvariantDim - 1] != 1 && InSrcVectorSize != 1)
return (false);
if(pArg->inLengths_[NumInvariantDim - 1] % InSrcVectorSize != 0)
return (false);
};
}
else
{
if(pArg->inStrides_[Rank - 1] != 1 && InSrcVectorSize != 1)
return (false);
if(pArg->inLengths_[Rank - 1] % InSrcVectorSize != 0)
return (false);
};
// To improve
bool valid = true;
static_for<0, NumReduction, 1>{}([&](auto I) {
if(pArg->outStridesArray_[I.value][NumOutputDim - 1] != 1 &&
OutDstVectorSizeSeq::At(I) != 1)
valid = false;
if(pArg->outLengths_[NumOutputDim - 1] % OutDstVectorSizeSeq::At(I) != 0)
valid = false;
});
if(!valid)
return (false);
if constexpr(use_multiblock)
{
// blkGroupSize of 1 should be handled by Blockwise path using
// InMemoryDataOperationEnum::Set
if(pArg->blkGroupSize == 1)
return (false);
// This is very strong restriction, but needed to avoid some failure
if(pArg->outLengths_[NumOutputDim - 1] % M_BlockTileSize != 0)
return (false);
}
else
{
// cases with very small reduce_total_length should be handled by ThreadWise kernel
if(pArg->reduce_total_length / KThreadSliceSize < 2)
return (false);
};
return (true);
};
std::unique_ptr<BaseArgument> MakeArgumentPointer(
const std::array<index_t, NumInputDim> inLengths,
const std::array<index_t, NumInputDim> inStrides,
const std::array<index_t, NumOutputDim> outLengths,
const std::array<std::array<index_t, NumOutputDim>, NumReduction> outStridesArray,
const std::array<int, NumReduceDim> reduceDims,
const std::array<const void*, NumReduction> alphas,
const std::array<const void*, NumReduction> betas,
const void* in_dev,
const std::array<void*, NumReduction> out_dev_buffers,
const InElementwiseOperationTuple in_elementwise_op_tuple,
const AccElementwiseOperationTuple acc_elementwise_op_tuple) override
{
return std::make_unique<Argument>(inLengths,
inStrides,
outLengths,
outStridesArray,
reduceDims,
alphas,
betas,
in_dev,
out_dev_buffers,
in_elementwise_op_tuple,
acc_elementwise_op_tuple);
};
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>();
};
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << (OutMemoryDataOperation == InMemoryDataOperationEnum::Set? "DeviceMultipleReduceBlockWise<" : "DeviceMultipleReduceMultiBlock<") << BlockSize << ",";
str << "M_C" << MThreadClusterSize << "_S" << MThreadSliceSize << ",";
str << "K_C" << KThreadClusterSize << "_S" << KThreadSliceSize << ",";
str << "InSrcVectorDim_" << InSrcVectorDim << "_InSrcVectorSize_" << InSrcVectorSize << ",";
str << "OutDstVectorSize";
static_for<0, OutDstVectorSizeSeq::Size(), 1>{}([&](auto I) {str << "_" << OutDstVectorSizeSeq::At(I); });
str << ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

422
include/ck/tensor_operation/gpu/device/impl/device_multiple_reduce_threadwise.hpp Normal file
View File

@@ -0,0 +1,422 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/sequence.hpp"
#include "ck/utility/reduction_operator.hpp"
#include "ck/tensor_operation/gpu/device/device_base.hpp"
#include "ck/tensor_operation/gpu/device/device_multiple_reduce.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_reduce_common.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_2d_multiple_reduction_threadwise.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <index_t NumReduction,
typename InDataType,
typename AccDataType,
typename OutDataTypeTuple,
index_t Rank,
index_t NumReduceDim,
typename ReduceOperation,
typename InElementwiseOperationTuple,
typename AccElementwiseOperationTuple,
bool PropagateNan,
index_t BlockSize,
index_t MThreadSliceSize,
index_t KThreadSliceSize,
index_t InSrcVectorDim,
index_t InSrcVectorSize,
typename OutDstVectorSizeSeq>
struct DeviceMultipleReduceThreadWise : public DeviceMultipleReduce<Rank,
NumReduceDim,
NumReduction,
InElementwiseOperationTuple,
AccElementwiseOperationTuple>
{
static_assert(Rank <= 6, "Bigger Rank size is not supported!");
static_assert((InSrcVectorDim == 0 && MThreadSliceSize % InSrcVectorSize == 0) ||
(InSrcVectorDim == 1 && KThreadSliceSize % InSrcVectorSize == 0),
"Invalid thread slice sizes and/or vector sizes configuration, please check!");
static_assert(NumReduction == OutDataTypeTuple::Size() &&
NumReduction == InElementwiseOperationTuple::Size() &&
NumReduction == AccElementwiseOperationTuple::Size() &&
NumReduction == OutDstVectorSizeSeq::Size(),
"All tuple should have the same size as the number of Reductions!");
static_assert(sequence_all_of(OutDstVectorSizeSeq{},
[](auto vectorSize) {
return (MThreadSliceSize % vectorSize == 0);
}),
"The OutDstVectorSize should completely divide the MThreadSliceSize!");
static constexpr index_t NumInvariantDim = Rank - NumReduceDim;
static constexpr index_t NumInputDim = Rank;
static constexpr index_t NumOutputDim = (NumInvariantDim == 0) ? 1 : NumInvariantDim;
static constexpr bool reduceAllDim = (NumInvariantDim == 0);
static constexpr index_t M_BlockTileSize = BlockSize * MThreadSliceSize;
static constexpr index_t K_BlockTileSize = 1 * KThreadSliceSize;
static auto GenerateOutDataTypePointerTuple()
{
return generate_tuple(
[&](auto I) {
using DataType = remove_cvref_t<decltype(OutDataTypeTuple{}[I])>;
return static_cast<DataType*>(nullptr);
},
Number<NumReduction>{});
};
using OutDataTypePointerTuple = decltype(GenerateOutDataTypePointerTuple());
static auto MakeSrc2dDescriptor(const std::array<index_t, NumInputDim>& inLengths,
const std::array<index_t, NumInputDim>& inStrides)
{
const auto tupleSrcLengths =
generate_tuple([&](auto I) { return inLengths[I]; }, Number<NumInputDim>{});
const auto tupleSrcStrides =
generate_tuple([&](auto I) { return inStrides[I]; }, Number<NumInputDim>{});
const auto inDesc = make_naive_tensor_descriptor(tupleSrcLengths, tupleSrcStrides);
const auto in_grid_desc_m_k = [&]() {
if constexpr(reduceAllDim)
{
const auto one_dim_inDesc = transform_tensor_descriptor(
inDesc,
make_tuple(make_merge_transform(tupleSrcLengths)),
make_tuple(typename arithmetic_sequence_gen<0, NumInputDim, 1>::type{}),
make_tuple(Sequence<0>{}));
return transform_tensor_descriptor(one_dim_inDesc,
make_tuple(make_unmerge_transform(make_tuple(
1, one_dim_inDesc.GetLength(Number<0>{})))),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0, 1>{}));
}
else
{
using InvariantDims = typename arithmetic_sequence_gen<0, NumInvariantDim, 1>::type;
using ReduceDims = typename arithmetic_sequence_gen<NumInvariantDim, Rank, 1>::type;
const auto reduceDimLengths = generate_tuple(
[&](auto I) { return inLengths[NumInvariantDim + I]; }, Number<NumReduceDim>{});
const auto invariantDimLengths =
generate_tuple([&](auto I) { return inLengths[I]; }, Number<NumInvariantDim>{});
return transform_tensor_descriptor(
inDesc,
make_tuple(make_merge_transform(invariantDimLengths),
make_merge_transform(reduceDimLengths)),
make_tuple(InvariantDims{}, ReduceDims{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
}();
const auto invariantLength = in_grid_desc_m_k.GetLength(Number<0>{});
const auto reduceLength = in_grid_desc_m_k.GetLength(Number<1>{});
const auto inPad_M =
math::integer_least_multiple(invariantLength, M_BlockTileSize) - invariantLength;
const auto inPad_K =
math::integer_least_multiple(reduceLength, K_BlockTileSize) - reduceLength;
auto in_grid_desc_m_k_padded = transform_tensor_descriptor(
in_grid_desc_m_k,
make_tuple(make_right_pad_transform(invariantLength, inPad_M),
make_right_pad_transform(reduceLength, inPad_K)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return (in_grid_desc_m_k_padded);
};
static auto MakeDst1dDescriptor(const std::array<index_t, NumOutputDim>& outLengths,
const std::array<index_t, NumOutputDim>& outStrides)
{
const auto tupleDstLengths =
generate_tuple([&](auto I) { return outLengths[I]; }, Number<NumOutputDim>{});
const auto tupleDstStrides =
generate_tuple([&](auto I) { return outStrides[I]; }, Number<NumOutputDim>{});
auto outDesc = make_naive_tensor_descriptor(tupleDstLengths, tupleDstStrides);
auto out_grid_desc_m = transform_tensor_descriptor(
outDesc,
make_tuple(make_merge_transform(tupleDstLengths)),
make_tuple(typename arithmetic_sequence_gen<0, NumOutputDim, 1>::type{}),
make_tuple(Sequence<0>{}));
const auto invariantLength = out_grid_desc_m.GetLength(Number<0>{});
const auto outPad =
math::integer_least_multiple(invariantLength, M_BlockTileSize) - invariantLength;
auto out_grid_desc_m_padded = transform_tensor_descriptor(
out_grid_desc_m,
make_tuple(make_right_pad_transform(invariantLength, outPad)),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0>{}));
return (out_grid_desc_m_padded);
};
static auto GenerateOutGrid1dDescTuple()
{
return generate_tuple(
[&](auto I) {
(void)I;
return MakeDst1dDescriptor(std::array<index_t, NumOutputDim>{},
std::array<index_t, NumOutputDim>{});
},
Number<NumReduction>{});
};
using InGridDesc_M_K = decltype(MakeSrc2dDescriptor(std::array<index_t, NumInputDim>{},
std::array<index_t, NumInputDim>{}));
using OutGridDesc_M_Tuple = decltype(GenerateOutGrid1dDescTuple());
struct Argument : public BaseArgument
{
Argument(const std::array<index_t, NumInputDim>& inLengths,
const std::array<index_t, NumInputDim>& inStrides,
const std::array<index_t, NumOutputDim>& outLengths,
const std::array<std::array<index_t, NumOutputDim>, NumReduction>& outStridesArray,
const std::array<int, NumReduceDim>& reduceDims,
const std::array<const void*, NumReduction>& alphas,
const std::array<const void*, NumReduction>& betas,
const void* in_dev,
const std::array<void*, NumReduction>& out_dev_buffers,
const InElementwiseOperationTuple in_elementwise_op_tuple,
const AccElementwiseOperationTuple acc_elementwise_op_tuple)
: outLengths_{outLengths},
outStridesArray_{outStridesArray},
in_elementwise_op_tuple_{in_elementwise_op_tuple},
acc_elementwise_op_tuple_{acc_elementwise_op_tuple}
{
inLengths_ = shuffle_tensor_dimensions<Rank, NumReduceDim>(inLengths, reduceDims);
inStrides_ = shuffle_tensor_dimensions<Rank, NumReduceDim>(inStrides, reduceDims);
for(size_t i = 0; i < NumReduction; i++)
{
alpha_values_(i) = *static_cast<const AccDataType*>(alphas[i]);
beta_values_(i) = *static_cast<const AccDataType*>(betas[i]);
};
in_dev_ = static_cast<const InDataType*>(in_dev);
out_dev_buffers_ = generate_tuple(
[&](auto iR) {
using OutDataTypePointer =
remove_cvref_t<decltype(OutDataTypePointerTuple{}[iR])>;
using OutDataType = remove_cvref_t<remove_pointer_t<OutDataTypePointer>>;
return static_cast<OutDataType*>(out_dev_buffers[iR]);
},
Number<NumReduction>{});
std::tie(invariant_total_length, reduce_total_length) =
get_2d_lengths<Rank, NumReduceDim>(inLengths_);
in_grid_desc_m_k = MakeSrc2dDescriptor(inLengths_, inStrides_);
out_grid_desc_m_tuple = generate_tuple(
[&](auto I) { return MakeDst1dDescriptor(outLengths, outStridesArray[I]); },
Number<NumReduction>{});
gridSize = math::integer_least_multiple(invariant_total_length, M_BlockTileSize) /
M_BlockTileSize;
}
std::array<index_t, NumInputDim> inLengths_;
std::array<index_t, NumInputDim> inStrides_;
std::array<index_t, NumOutputDim> outLengths_;
std::array<std::array<index_t, NumOutputDim>, NumReduction> outStridesArray_;
Array<AccDataType, NumReduction> alpha_values_;
Array<AccDataType, NumReduction> beta_values_;
const InDataType* in_dev_;
OutDataTypePointerTuple out_dev_buffers_;
InGridDesc_M_K in_grid_desc_m_k;
OutGridDesc_M_Tuple out_grid_desc_m_tuple;
InElementwiseOperationTuple in_elementwise_op_tuple_;
AccElementwiseOperationTuple acc_elementwise_op_tuple_;
long_index_t invariant_total_length;
long_index_t reduce_total_length;
size_t gridSize;
};
struct Invoker : public BaseInvoker
{
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
using GridwiseMultipleReduce =
GridwiseMultipleReduction_mk_to_m_threadwise<NumReduction,
InDataType,
OutDataTypePointerTuple,
AccDataType,
InGridDesc_M_K,
OutGridDesc_M_Tuple,
ReduceOperation,
InElementwiseOperationTuple,
AccElementwiseOperationTuple,
InMemoryDataOperationEnum::Set,
PropagateNan,
BlockSize,
MThreadSliceSize,
KThreadSliceSize,
InSrcVectorDim,
InSrcVectorSize,
OutDstVectorSizeSeq>;
const auto kernel_main =
kernel_multiple_reduce_threadwise<GridwiseMultipleReduce,
NumReduction,
InDataType,
OutDataTypePointerTuple,
AccDataType,
InGridDesc_M_K,
OutGridDesc_M_Tuple,
InElementwiseOperationTuple,
AccElementwiseOperationTuple>;
float avg_time = 0;
avg_time += launch_and_time_kernel(stream_config,
kernel_main,
dim3(arg.gridSize),
dim3(BlockSize),
0,
arg.in_grid_desc_m_k,
arg.out_grid_desc_m_tuple,
arg.in_elementwise_op_tuple_,
arg.acc_elementwise_op_tuple_,
arg.alpha_values_,
arg.in_dev_,
arg.beta_values_,
arg.out_dev_buffers_);
return (avg_time);
};
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
};
};
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
const Argument* pArg = dynamic_cast<const Argument*>(p_arg);
if constexpr(InSrcVectorDim == 0)
{
if constexpr(NumInvariantDim == 0)
{
return (false);
}
else
{
if(pArg->inStrides_[NumInvariantDim - 1] != 1 && InSrcVectorSize != 1)
return (false);
if(pArg->inLengths_[NumInvariantDim - 1] % InSrcVectorSize != 0)
return (false);
};
}
else
{
if(pArg->inStrides_[Rank - 1] != 1 && InSrcVectorSize != 1)
return (false);
if(pArg->inLengths_[Rank - 1] % InSrcVectorSize != 0)
return (false);
};
// To improve
bool valid = true;
static_for<0, NumReduction, 1>{}([&](auto I) {
if(pArg->outStridesArray_[I.value][NumOutputDim - 1] != 1 &&
OutDstVectorSizeSeq::At(I) != 1)
valid = false;
if(pArg->outLengths_[NumOutputDim - 1] % OutDstVectorSizeSeq::At(I) != 0)
valid = false;
});
if(!valid)
return (false);
return (true);
};
std::unique_ptr<BaseArgument> MakeArgumentPointer(
const std::array<index_t, NumInputDim> inLengths,
const std::array<index_t, NumInputDim> inStrides,
const std::array<index_t, NumOutputDim> outLengths,
const std::array<std::array<index_t, NumOutputDim>, NumReduction> outStridesArray,
const std::array<int, NumReduceDim> reduceDims,
const std::array<const void*, NumReduction> alphas,
const std::array<const void*, NumReduction> betas,
const void* in_dev,
const std::array<void*, NumReduction> out_dev_buffers,
const InElementwiseOperationTuple in_elementwise_op_tuple,
const AccElementwiseOperationTuple acc_elementwise_op_tuple) override
{
return std::make_unique<Argument>(inLengths,
inStrides,
outLengths,
outStridesArray,
reduceDims,
alphas,
betas,
in_dev,
out_dev_buffers,
in_elementwise_op_tuple,
acc_elementwise_op_tuple);
};
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>();
};
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceMultipleReduceThreadwise<" << BlockSize << ",";
str << "M_C" << BlockSize << "_S" << MThreadSliceSize << ",";
str << "K_C" << 1 << "_S" << KThreadSliceSize << ",";
str << "InSrcVectorDim_" << InSrcVectorDim << "_InSrcVectorSize_" << InSrcVectorSize << ",";
str << "OutDstVectorSize";
static_for<0, OutDstVectorSizeSeq::Size(), 1>{}([&](auto I) {str << "_" << OutDstVectorSizeSeq::At(I); });
str << ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

468
include/ck/tensor_operation/gpu/device/impl/device_normalization_impl.hpp Normal file
View File

@@ -0,0 +1,468 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/reduction_operator.hpp"
#include "ck/tensor_operation/gpu/device/device_normalization.hpp"
#include "ck/tensor_operation/gpu/device/device_reduce.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_reduce_common.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_layernorm_welford_variance.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_set_buffer_value.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
template <typename GridwiseReduction,
typename XDataType,
typename GammaDataType,
typename BetaDataType,
typename YDataType,
typename AccDataType,
typename AccElementwiseOperation,
typename GridDesc_M_K>
__global__ void kernel_layernorm(const GridDesc_M_K x_grid_desc_m_k,
const GridDesc_M_K gamma_grid_desc_m_k,
const GridDesc_M_K beta_grid_desc_m_k,
const GridDesc_M_K y_grid_desc_m_k,
index_t num_k_block_tile_iteration,
AccDataType epsilon,
const XDataType* const __restrict__ p_x_global,
const GammaDataType* const __restrict__ p_gamma_global,
const BetaDataType* const __restrict__ p_beta_global,
YDataType* const __restrict__ p_y_global,
const AccElementwiseOperation acc_elementwise_op)
{
GridwiseReduction::Run(x_grid_desc_m_k,
gamma_grid_desc_m_k,
beta_grid_desc_m_k,
y_grid_desc_m_k,
num_k_block_tile_iteration,
epsilon,
p_x_global,
p_gamma_global,
p_beta_global,
p_y_global,
acc_elementwise_op);
};
} // namespace ck
namespace ck {
namespace tensor_operation {
namespace device {
// Y = LayerNorm(X, Beta, Gamma)
template <typename XDataType,
typename GammaDataType,
typename BetaDataType,
typename AccDataType,
typename YDataType,
typename AccElementwiseOperation,
index_t Rank,
index_t NumReduceDim,
index_t BlockSize,
index_t MThreadClusterSize,
index_t KThreadClusterSize,
index_t MThreadSliceSize,
index_t KThreadSliceSize,
index_t XYSrcVectorDim,
index_t XSrcVectorSize,
index_t GammaSrcVectorDim,
index_t GammaSrcVectorSize,
index_t BetaSrcVectorDim,
index_t BetaSrcVectorSize,
index_t YDstVectorSize>
struct DeviceNormalizationImpl : public DeviceNormalization<XDataType,
GammaDataType,
BetaDataType,
AccDataType,
YDataType,
AccElementwiseOperation,
Rank,
NumReduceDim>
{
static_assert(
((GammaSrcVectorDim == 0 && MThreadSliceSize % GammaSrcVectorSize == 0) ||
(GammaSrcVectorDim == 1 && KThreadSliceSize % GammaSrcVectorSize == 0)),
"Invalid thread slice sizes and/or gamma vector sizes configuration, please check!");
static_assert(
((BetaSrcVectorDim == 0 && MThreadSliceSize % BetaSrcVectorSize == 0) ||
(BetaSrcVectorDim == 1 && KThreadSliceSize % BetaSrcVectorSize == 0)),
"Invalid thread slice sizes and/or beta vector sizes configuration, please check!");
using PassThrough = tensor_operation::element_wise::PassThrough;
static constexpr index_t M_BlockTileSize = MThreadClusterSize * MThreadSliceSize;
static constexpr index_t K_BlockTileSize = KThreadClusterSize * KThreadSliceSize;
static auto MakeSrc2dDescriptor(const std::vector<index_t>& inLengths,
const std::vector<index_t>& inStrides,
int blkGroupSize,
int numBlockTileIteration)
{
constexpr index_t NumInvariantDim = Rank - NumReduceDim;
static constexpr index_t numSrcDim = Rank;
static constexpr bool reduceAllDim = (NumInvariantDim == 0);
const auto tupleSrcLengths = make_tuple_from_array(inLengths, Number<numSrcDim>{});
const auto tupleSrcStrides = make_tuple_from_array(inStrides, Number<numSrcDim>{});
const auto inDesc = make_naive_tensor_descriptor(tupleSrcLengths, tupleSrcStrides);
const auto in_grid_desc_m_k = [&]() {
if constexpr(reduceAllDim)
{
const auto one_dim_inDesc = transform_tensor_descriptor(
inDesc,
make_tuple(make_merge_transform(tupleSrcLengths)),
make_tuple(typename arithmetic_sequence_gen<0, numSrcDim, 1>::type{}),
make_tuple(Sequence<0>{}));
return transform_tensor_descriptor(one_dim_inDesc,
make_tuple(make_unmerge_transform(make_tuple(
1, one_dim_inDesc.GetLength(Number<0>{})))),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0, 1>{}));
}
else
{
using InvariantDims = typename arithmetic_sequence_gen<0, NumInvariantDim, 1>::type;
using ReduceDims = typename arithmetic_sequence_gen<NumInvariantDim, Rank, 1>::type;
const auto reduceDimLengths =
make_tuple_from_array_and_index_seq(inLengths, ReduceDims{});
const auto invariantDimLengths =
make_tuple_from_array_and_index_seq(inLengths, InvariantDims{});
return transform_tensor_descriptor(
inDesc,
make_tuple(make_merge_transform(invariantDimLengths),
make_merge_transform(reduceDimLengths)),
make_tuple(InvariantDims{}, ReduceDims{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
}();
const auto invariantLength = in_grid_desc_m_k.GetLength(Number<0>{});
const auto reduceLength = in_grid_desc_m_k.GetLength(Number<1>{});
const int reduceSizePerBlock = K_BlockTileSize * numBlockTileIteration;
const auto inPad_M =
math::integer_least_multiple(invariantLength, M_BlockTileSize) - invariantLength;
const auto inPad_K = reduceSizePerBlock * blkGroupSize - reduceLength;
auto in_grid_desc_m_k_padded = transform_tensor_descriptor(
in_grid_desc_m_k,
make_tuple(make_right_pad_transform(invariantLength, inPad_M),
make_right_pad_transform(reduceLength, inPad_K)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return (in_grid_desc_m_k_padded);
};
using GridDesc_M_K = decltype(MakeSrc2dDescriptor({1}, {1}, 1, 1));
using GridwiseReduceLayernormGeneric =
GridwiseLayernormWelfordVariance_mk_to_mk<XDataType,
GammaDataType,
BetaDataType,
YDataType,
AccDataType,
AccElementwiseOperation,
GridDesc_M_K,
BlockSize,
MThreadClusterSize,
KThreadClusterSize,
MThreadSliceSize,
KThreadSliceSize,
XYSrcVectorDim,
XSrcVectorSize,
GammaSrcVectorDim,
GammaSrcVectorSize,
BetaSrcVectorDim,
BetaSrcVectorSize,
XYSrcVectorDim,
YDstVectorSize,
false>;
using GridwiseReduceLayernormSweepOnce =
GridwiseLayernormWelfordVariance_mk_to_mk<XDataType,
GammaDataType,
BetaDataType,
YDataType,
AccDataType,
AccElementwiseOperation,
GridDesc_M_K,
BlockSize,
MThreadClusterSize,
KThreadClusterSize,
MThreadSliceSize,
KThreadSliceSize,
XYSrcVectorDim,
XSrcVectorSize,
GammaSrcVectorDim,
GammaSrcVectorSize,
BetaSrcVectorDim,
BetaSrcVectorSize,
XYSrcVectorDim,
YDstVectorSize,
true>;
struct Argument : public BaseArgument
{
Argument(const std::vector<index_t> lengths,
const std::vector<index_t> xStrides,
const std::vector<index_t> gammaStrides,
const std::vector<index_t> betaStrides,
const std::vector<index_t> yStrides,
const std::vector<index_t> reduceDims,
AccElementwiseOperation acc_elementwise_op,
AccDataType epsilon,
const XDataType* p_x,
const GammaDataType* p_gamma,
const BetaDataType* p_beta,
YDataType* p_y)
: epsilon_(epsilon),
p_x_(p_x),
p_gamma_(p_gamma),
p_beta_(p_beta),
p_y_(p_y),
acc_elementwise_op_(acc_elementwise_op)
{
Lengths_ = shuffle_tensor_dimensions<Rank, NumReduceDim>(lengths, reduceDims);
xStrides_ = shuffle_tensor_dimensions<Rank, NumReduceDim>(xStrides, reduceDims);
yStrides_ = shuffle_tensor_dimensions<Rank, NumReduceDim>(yStrides, reduceDims);
gammaStrides_ = shuffle_tensor_dimensions<Rank, NumReduceDim>(gammaStrides, reduceDims);
betaStrides_ = shuffle_tensor_dimensions<Rank, NumReduceDim>(betaStrides, reduceDims);
long_index_t invariant_total_length;
long_index_t reduce_total_length;
std::tie(invariant_total_length, reduce_total_length) =
get_2d_lengths<Rank, NumReduceDim>(Lengths_);
blkGroupSize_ = 1;
numBlockTileIteration_ = (reduce_total_length + K_BlockTileSize - 1) / K_BlockTileSize;
gridSize_ = math::integer_least_multiple(invariant_total_length, M_BlockTileSize) /
M_BlockTileSize * blkGroupSize_;
x_grid_desc_m_k_ =
MakeSrc2dDescriptor(Lengths_, xStrides_, blkGroupSize_, numBlockTileIteration_);
gamma_grid_desc_m_k_ =
MakeSrc2dDescriptor(Lengths_, gammaStrides_, blkGroupSize_, numBlockTileIteration_);
beta_grid_desc_m_k_ =
MakeSrc2dDescriptor(Lengths_, betaStrides_, blkGroupSize_, numBlockTileIteration_);
y_grid_desc_m_k_ =
MakeSrc2dDescriptor(Lengths_, yStrides_, blkGroupSize_, numBlockTileIteration_);
isSweeponce_ =
x_grid_desc_m_k_.GetLength(Number<1>{}) <= KThreadClusterSize * KThreadSliceSize;
}
AccDataType epsilon_;
const XDataType* p_x_;
const GammaDataType* p_gamma_;
const BetaDataType* p_beta_;
YDataType* p_y_;
std::vector<index_t> Lengths_;
std::vector<index_t> xStrides_;
std::vector<index_t> gammaStrides_;
std::vector<index_t> betaStrides_;
std::vector<index_t> yStrides_;
AccElementwiseOperation acc_elementwise_op_;
int blkGroupSize_;
int numBlockTileIteration_;
size_t gridSize_;
GridDesc_M_K x_grid_desc_m_k_;
GridDesc_M_K gamma_grid_desc_m_k_;
GridDesc_M_K beta_grid_desc_m_k_;
GridDesc_M_K y_grid_desc_m_k_;
bool isSweeponce_;
};
struct Invoker : public BaseInvoker
{
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
const auto kernel_main = arg.isSweeponce_
? kernel_layernorm<GridwiseReduceLayernormSweepOnce,
XDataType,
GammaDataType,
BetaDataType,
YDataType,
AccDataType,
AccElementwiseOperation,
GridDesc_M_K>
: kernel_layernorm<GridwiseReduceLayernormGeneric,
XDataType,
GammaDataType,
BetaDataType,
YDataType,
AccDataType,
AccElementwiseOperation,
GridDesc_M_K>;
float avg_time = 0;
avg_time += launch_and_time_kernel(stream_config,
kernel_main,
dim3(arg.gridSize_),
dim3(BlockSize),
0,
arg.x_grid_desc_m_k_,
arg.gamma_grid_desc_m_k_,
arg.beta_grid_desc_m_k_,
arg.y_grid_desc_m_k_,
arg.numBlockTileIteration_,
arg.epsilon_,
arg.p_x_,
arg.p_gamma_,
arg.p_beta_,
arg.p_y_,
arg.acc_elementwise_op_);
return (avg_time);
};
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
};
};
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
const Argument* p_arg_ = dynamic_cast<const Argument*>(p_arg);
constexpr index_t NumInvariantDim = Rank - NumReduceDim;
if constexpr(XYSrcVectorDim == 0)
{
if constexpr(NumInvariantDim == 0)
{
return false;
}
else
{
if(p_arg_->xStrides_[NumInvariantDim - 1] != 1)
return false;
if(p_arg_->invariant_lowest_length % XSrcVectorSize != 0)
return false;
};
}
else
{
if(p_arg_->xStrides_[Rank - 1] != 1)
return false;
if(p_arg_->Lengths_[Rank - 1] % XSrcVectorSize != 0)
return false;
};
if(p_arg_->Lengths_[Rank - 1] % YDstVectorSize != 0)
{
return false;
}
// if fastest dim is not reduced
if constexpr(GammaSrcVectorDim == 0)
{
if(p_arg_->gammaStrides_[NumInvariantDim - 1] != 1)
return (false);
if(p_arg_->Lengths_[Rank - 1] % GammaSrcVectorSize != 0)
return (false);
}
else // if fastest dim is reduced
{
if(p_arg_->gammaStrides_[Rank - 1] != 1)
return (false);
if(p_arg_->Lengths_[Rank - 1] % GammaSrcVectorSize != 0)
return (false);
}
// if fastest dim is not reduced
if constexpr(BetaSrcVectorDim == 0)
{
if(p_arg_->betaStrides_[NumInvariantDim - 1] != 1)
return (false);
if(p_arg_->invariant_lowest_length % BetaSrcVectorSize != 0)
return (false);
}
else // if fastest dim is reduced
{
if(p_arg_->betaStrides_[Rank - 1] != 1)
return (false);
if(p_arg_->Lengths_[Rank - 1] % BetaSrcVectorSize != 0)
return (false);
}
return true;
};
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const std::vector<index_t> lengths,
const std::vector<index_t> xStrides,
const std::vector<index_t> gammaStrides,
const std::vector<index_t> betaStrides,
const std::vector<index_t> yStrides,
const std::vector<index_t> reduceDims,
AccDataType epsilon,
const void* p_x,
const void* p_gamma,
const void* p_beta,
void* p_y,
AccElementwiseOperation acc_elementwise_op) override
{
return std::make_unique<Argument>(lengths,
xStrides,
gammaStrides,
betaStrides,
yStrides,
reduceDims,
acc_elementwise_op,
epsilon,
static_cast<const XDataType*>(p_x),
static_cast<const GammaDataType*>(p_gamma),
static_cast<const BetaDataType*>(p_beta),
static_cast<YDataType*>(p_y));
};
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>();
};
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceNormalizationImpl<" << BlockSize << ",";
str << "M_C" << MThreadClusterSize << "_S" << MThreadSliceSize << ",";
str << "K_C" << KThreadClusterSize << "_S" << KThreadSliceSize << ",";
str << "XYSrcVectorDim_" << XYSrcVectorDim << ",";
str << "VectorSize_X" << XSrcVectorSize << "_Gamma" << GammaSrcVectorSize << "_Beta" << BetaSrcVectorSize << "_Y" << YDstVectorSize << ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

327
include/ck/tensor_operation/gpu/device/impl/device_pool2d_fwd_nhwc_nhwc.hpp Normal file
View File

@@ -0,0 +1,327 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/reduction_operator_mapping.hpp"
#include "ck/tensor_operation/gpu/device/device_pool2d_fwd.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_2d_reduction_threadwise.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <typename InDataType,
typename OutDataType,
typename AccDataType,
ck::ReduceTensorOp ReduceOpId,
bool OuputIndex,
ck::index_t BlockSize,
ck::index_t ReduceMThreadClusterSize,
ck::index_t ReduceKThreadClusterSize,
ck::index_t ReduceMThreadSliceSize,
ck::index_t ReduceKThreadSliceSize,
ck::index_t InSrcOutDstVectorSize>
struct DevicePool2dFwd_Input_N_Hi_Wi_C_Output_N_Ho_Wo_C : public DevicePool2dFwd<ReduceOpId>
{
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr auto I4 = Number<4>{};
static constexpr auto I5 = Number<5>{};
using IndexDataType = int32_t;
using ReduceOperation = typename reduce_binary_operator<ReduceOpId>::opType;
using InElementwiseOperation =
typename reduce_unary_operator<ReduceOpId, true, true>::InElementwiseOperation;
using AccElementwiseOperation =
typename reduce_unary_operator<ReduceOpId, true, true>::AccElementwiseOperation;
static constexpr index_t InSrcOutDstVectorDim =
0; // for NHWC, the dim C is the vector Dim for both input and output in memory, which is
// not reduced.
static constexpr ck::index_t ReduceM_BlockTileSize =
ReduceMThreadClusterSize * ReduceMThreadSliceSize;
static constexpr ck::index_t ReduceK_BlockTileSize =
ReduceKThreadClusterSize * ReduceKThreadSliceSize;
static auto MakeABGridDescriptor_A_M_K_B_M(ck::index_t N,
ck::index_t C,
std::array<ck::index_t, 2> input_spatial_lengths,
std::array<ck::index_t, 2> window_spatial_lengths,
std::array<ck::index_t, 2> output_spatial_lengths,
std::array<ck::index_t, 2> window_strides,
std::array<ck::index_t, 2> input_left_pads,
std::array<ck::index_t, 2> input_right_pads)
{
const index_t Hi = input_spatial_lengths[0];
const index_t Wi = input_spatial_lengths[1];
const index_t Ho = output_spatial_lengths[0];
const index_t Wo = output_spatial_lengths[1];
const index_t Y = window_spatial_lengths[0];
const index_t X = window_spatial_lengths[1];
const index_t ConvStrideH = window_strides[0];
const index_t ConvStrideW = window_strides[1];
const index_t InLeftPadH = input_left_pads[0];
const index_t InLeftPadW = input_left_pads[1];
const index_t InRightPadH = input_right_pads[0];
const index_t InRightPadW = input_right_pads[1];
const index_t ReduceMRaw = N * Ho * Wo * C;
const index_t ReduceMPad =
math::integer_least_multiple(ReduceMRaw, ReduceM_BlockTileSize) - ReduceMRaw;
const index_t ReduceKRaw = Y * X;
const index_t ReduceKPad =
math::integer_least_multiple(ReduceKRaw, ReduceK_BlockTileSize) - ReduceKRaw;
// A[ReduceM, ReduceK]
const auto in_grid_desc_n_hi_wi_c =
make_naive_tensor_descriptor_packed(make_tuple(N, Hi, Wi, C));
const auto in_grid_desc_n_hip_wip_c = transform_tensor_descriptor(
in_grid_desc_n_hi_wi_c,
make_tuple(make_pass_through_transform(N),
make_pad_transform(Hi, InLeftPadH, InRightPadH),
make_pad_transform(Wi, InLeftPadW, InRightPadW),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
const auto in_grid_desc_n_y_ho_x_wo_c = transform_tensor_descriptor(
in_grid_desc_n_hip_wip_c,
make_tuple(make_pass_through_transform(N),
make_embed_transform(make_tuple(Y, Ho), make_tuple(I1, ConvStrideH)),
make_embed_transform(make_tuple(X, Wo), make_tuple(I1, ConvStrideW)),
make_pass_through_transform(C)),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1, 2>{}, Sequence<3, 4>{}, Sequence<5>{}));
const auto in_grid_desc_reducemraw_reducekraw =
transform_tensor_descriptor(in_grid_desc_n_y_ho_x_wo_c,
make_tuple(make_merge_transform(make_tuple(N, Ho, Wo, C)),
make_merge_transform(make_tuple(Y, X))),
make_tuple(Sequence<0, 2, 4, 5>{}, Sequence<1, 3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto in_grid_desc_reducem_reducek = transform_tensor_descriptor(
in_grid_desc_reducemraw_reducekraw,
make_tuple(make_right_pad_transform(ReduceMRaw, ReduceMPad),
make_right_pad_transform(ReduceKRaw, ReduceKPad)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
// B[ReduceM]
const auto out_grid_desc_reducemraw =
make_naive_tensor_descriptor_packed(make_tuple(N * Ho * Wo * C));
const auto out_grid_desc_reducem = transform_tensor_descriptor(
out_grid_desc_reducemraw,
make_tuple(make_right_pad_transform(ReduceMRaw, ReduceMPad)),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0>{}));
return make_tuple(in_grid_desc_reducem_reducek, out_grid_desc_reducem);
}
using ABGridDescs = decltype(
MakeABGridDescriptor_A_M_K_B_M(1, 1, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}));
using AGridDesc_M_K = remove_cvref_t<decltype(ABGridDescs{}[I0])>;
using BGridDesc_M = remove_cvref_t<decltype(ABGridDescs{}[I1])>;
// TODO
struct Argument : public BaseArgument
{
Argument(const InDataType* p_in_dev,
OutDataType* p_out_dev,
int* p_out_indices_dev,
ck::index_t N,
ck::index_t C,
std::array<ck::index_t, 2>& input_spatial_lengths,
std::array<ck::index_t, 2>& window_spatial_lengths,
std::array<ck::index_t, 2>& output_spatial_lengths,
std::array<ck::index_t, 2>& window_strides,
std::array<ck::index_t, 2>& input_left_pads,
std::array<ck::index_t, 2>& input_right_pads)
: p_in_dev_{p_in_dev},
p_out_dev_{p_out_dev},
p_out_indices_dev_{p_out_indices_dev},
a_grid_desc_m_k_{},
b_grid_desc_m_{}
{
const auto descs = MakeABGridDescriptor_A_M_K_B_M(N,
C,
input_spatial_lengths,
window_spatial_lengths,
output_spatial_lengths,
window_strides,
input_left_pads,
input_right_pads);
a_grid_desc_m_k_ = descs[I0];
b_grid_desc_m_ = descs[I1];
invariant_lowest_length_ = C;
reduce_lowest_length_ = window_spatial_lengths[1];
int32_t reduceLength = window_spatial_lengths[0] * window_spatial_lengths[1];
std::tie(in_element_op_, acc_element_op_) =
reduce_unary_operator<ReduceOpId, true, true>::GetElementwiseOperator(reduceLength);
}
const InDataType* p_in_dev_;
OutDataType* p_out_dev_;
int* p_out_indices_dev_;
AGridDesc_M_K a_grid_desc_m_k_;
BGridDesc_M b_grid_desc_m_;
InElementwiseOperation in_element_op_;
AccElementwiseOperation acc_element_op_;
// for checking vector load/store
ck::index_t invariant_lowest_length_;
ck::index_t reduce_lowest_length_;
};
struct Invoker : public BaseInvoker
{
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
using gridwise_reduce =
GridwiseReduction_mk_to_m_threadwise<InDataType,
OutDataType,
AccDataType,
IndexDataType,
AGridDesc_M_K,
BGridDesc_M,
ReduceOperation,
InElementwiseOperation,
AccElementwiseOperation,
InMemoryDataOperationEnum::Set,
false, // propagate_nan
BlockSize,
ReduceMThreadSliceSize,
ReduceKThreadSliceSize,
InSrcOutDstVectorDim,
InSrcOutDstVectorSize,
InSrcOutDstVectorSize>;
const auto kernel = kernel_reduce_threadwise<gridwise_reduce,
OuputIndex,
false, // don't have index input
InDataType,
OutDataType,
AccDataType,
IndexDataType,
AGridDesc_M_K,
BGridDesc_M,
InElementwiseOperation,
AccElementwiseOperation>;
ck::index_t ReduceM = arg.a_grid_desc_m_k_.GetLength(I0);
const index_t grid_size = (ReduceM / ReduceM_BlockTileSize);
return launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.a_grid_desc_m_k_,
arg.b_grid_desc_m_,
arg.in_element_op_,
arg.acc_element_op_,
float(1),
arg.p_in_dev_,
nullptr,
float(0),
arg.p_out_dev_,
arg.p_out_indices_dev_);
}
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
const Argument* pArg = dynamic_cast<const Argument*>(p_arg);
if(pArg->invariant_lowest_length_ % InSrcOutDstVectorSize != 0)
{
return (false);
}
return (true);
}
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_in_dev,
void* p_out_dev,
void* p_out_indices_dev,
ck::index_t N,
ck::index_t C,
std::array<ck::index_t, 2> input_spatial_lengths,
std::array<ck::index_t, 2> window_spatial_lengths,
std::array<ck::index_t, 2> output_spatial_lengths,
std::array<ck::index_t, 2> window_strides,
std::array<ck::index_t, 2> input_left_pads,
std::array<ck::index_t, 2> input_right_pads) override
{
return std::make_unique<Argument>(static_cast<const InDataType*>(p_in_dev),
static_cast<OutDataType*>(p_out_dev),
static_cast<int*>(p_out_indices_dev),
N,
C,
input_spatial_lengths,
window_spatial_lengths,
output_spatial_lengths,
window_strides,
input_left_pads,
input_right_pads);
}
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DevicePool2dFwd_Input_N_Hi_Wi_C_Output_N_Ho_Wo_C<" << BlockSize << ",";
str << "M_C" << ReduceMThreadClusterSize << "_S" << ReduceMThreadSliceSize << ",";
str << "K_C" << ReduceKThreadClusterSize << "_S" << ReduceKThreadSliceSize << ",";
str <<"InSrcOutDstVectorSize_" << InSrcOutDstVectorSize << ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

142
include/ck/tensor_operation/gpu/device/impl/device_reduce_common.hpp Normal file
View File

@@ -0,0 +1,142 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <vector>
#include <cassert>
#include "ck/utility/common_header.hpp"
#include "ck/utility/reduction_enums.hpp"
#include "ck/utility/reduction_operator.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
// here, inLengths[] is already shuffled so that lengths of invariant dims are included before those
// of reduce dims
template <index_t Rank, int NumReduceDim>
std::pair<long_index_t, long_index_t> get_2d_lengths(const std::vector<index_t>& inLengths)
{
static_assert(Rank <= 6, "bigger Rank size not supported!");
long_index_t invariant_total_length = 1;
long_index_t reduce_total_length = 1;
constexpr int NumInvariantDim = Rank - NumReduceDim;
for(int i = NumInvariantDim; i < Rank; i++)
reduce_total_length *= inLengths[i];
for(int i = 0; i < NumInvariantDim; i++)
invariant_total_length *= inLengths[i];
return std::make_pair(invariant_total_length, reduce_total_length);
};
template <index_t Rank, int NumReduceDim>
std::pair<long_index_t, long_index_t> get_2d_lengths(const std::array<index_t, Rank>& inLengths)
{
static_assert(Rank <= 6, "bigger Rank size not supported!");
long_index_t invariant_total_length = 1;
long_index_t reduce_total_length = 1;
constexpr int NumInvariantDim = Rank - NumReduceDim;
for(int i = NumInvariantDim; i < Rank; i++)
reduce_total_length *= inLengths[i];
for(int i = 0; i < NumInvariantDim; i++)
invariant_total_length *= inLengths[i];
return std::make_pair(invariant_total_length, reduce_total_length);
};
// helper functions using variadic template arguments
template <index_t... Ns>
auto make_tuple_from_array_and_index_seq(const std::vector<index_t>& lengths, Sequence<Ns...>)
{
return make_tuple(static_cast<index_t>(lengths[Ns])...);
};
template <index_t arraySize>
auto make_tuple_from_array(const std::vector<index_t>& lengths, Number<arraySize>)
{
static_assert(arraySize >= 1 && arraySize <= 6, "The tensor should have 1 to 6 dimensions");
constexpr auto index_seq = typename arithmetic_sequence_gen<0, arraySize, 1>::type{};
return make_tuple_from_array_and_index_seq(lengths, index_seq);
};
template <index_t Rank, index_t NumReduceDim>
std::vector<index_t> shuffle_tensor_dimensions(const std::vector<index_t>& origLengthsStrides,
const std::vector<int>& reduceDims)
{
std::vector<index_t> newLengthsStrides;
assert(Rank == origLengthsStrides.size() && NumReduceDim == reduceDims.size());
int reduceFlag = 0;
// flag the bits for the reduceDims
for(int i = 0; i < NumReduceDim; i++)
{
reduceFlag |= 1 << reduceDims[i];
};
// collect invariant dimensions
for(int i = 0; i < Rank; i++)
if((reduceFlag & (1 << i)) == 0)
{
newLengthsStrides.push_back(origLengthsStrides[i]);
};
// collect reduce dimensions
for(int i = 0; i < Rank; i++)
if((reduceFlag & (1 << i)) > 0)
{
newLengthsStrides.push_back(origLengthsStrides[i]);
};
return newLengthsStrides;
};
template <index_t Rank, index_t NumReduceDim>
std::array<index_t, Rank>
shuffle_tensor_dimensions(const std::array<index_t, Rank>& origLengthsStrides,
const std::array<int, NumReduceDim>& reduceDims)
{
std::array<index_t, Rank> newLengthsStrides;
int reduceFlag = 0;
// flag the bits for the reduceDims
for(int i = 0; i < NumReduceDim; i++)
{
reduceFlag |= 1 << reduceDims[i];
};
// collect invariant dimensions
int pos = 0;
for(int i = 0; i < Rank; i++)
if((reduceFlag & (1 << i)) == 0)
{
newLengthsStrides[pos++] = origLengthsStrides[i];
};
// collect reduce dimensions
for(int i = 0; i < Rank; i++)
if((reduceFlag & (1 << i)) > 0)
{
newLengthsStrides[pos++] = origLengthsStrides[i];
};
return newLengthsStrides;
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

513
include/ck/tensor_operation/gpu/device/impl/device_reduce_multiblock.hpp Normal file
View File

@@ -0,0 +1,513 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/utility/reduction_operator.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/device_reduce.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_reduce_common.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_2d_reduction_multiblock.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_set_buffer_value.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <typename InDataType,
typename AccDataType,
typename OutDataType,
index_t Rank,
index_t NumReduceDim,
typename ReduceOperation,
typename InElementwiseOperation,
typename AccElementwiseOperation,
InMemoryDataOperationEnum OutMemoryDataOperation,
bool PropagateNan,
bool OutputIndex,
bool HaveIndexInputIfOutputIndex,
index_t BlockSize,
index_t MThreadClusterSize,
index_t KThreadClusterSize,
index_t MThreadSliceSize,
index_t KThreadSliceSize,
index_t InSrcVectorDim,
index_t InSrcVectorSize,
index_t OutDstVectorSize>
struct DeviceReduceMultiBlock : public DeviceReduce<InElementwiseOperation, AccElementwiseOperation>
{
static_assert(Rank <= 6, "Bigger Rank size is not supported!");
static_assert(BlockSize == MThreadClusterSize * KThreadClusterSize,
"Invalid thread cluster size assignments!");
static_assert(((InSrcVectorDim == 0 && MThreadSliceSize % InSrcVectorSize == 0) ||
(InSrcVectorDim == 1 && KThreadSliceSize % InSrcVectorSize == 0)) &&
(MThreadSliceSize % OutDstVectorSize == 0),
"Invalid thread slice sizes and/or vector sizes configuration, please check!");
using IndexDataType = int32_t;
static constexpr bool HaveIndexInput = OutputIndex && HaveIndexInputIfOutputIndex;
static constexpr index_t NumInvariantDim = Rank - NumReduceDim;
static constexpr index_t numSrcDim = Rank;
static constexpr index_t numDstDim = (NumInvariantDim == 0) ? 1 : NumInvariantDim;
static constexpr bool reduceAllDim = (NumInvariantDim == 0);
// So far, only AtomicAdd is considered, other Atomic Operation like AtomicMax can be added
// later
static constexpr bool use_multiblock =
(OutMemoryDataOperation == InMemoryDataOperationEnum::AtomicAdd);
static_assert(ck::reduce::InMemoryDataOperatonSupportedOnDataType<OutMemoryDataOperation,
OutDataType>::value,
"The OutDataType must support the specified OutMemoryDataOperation!");
static_assert(!use_multiblock || (use_multiblock && !OutputIndex),
"MultiBlock reduction can only be used when outputing index is not required");
static_assert(
ReduceOperation::IsCompatibleInMemoryDataOperation(OutMemoryDataOperation),
"The reduction accumulation operation must be compatible with the OutMemoryDataOperation!");
static constexpr index_t M_BlockTileSize = MThreadClusterSize * MThreadSliceSize;
static constexpr index_t K_BlockTileSize = KThreadClusterSize * KThreadSliceSize;
static auto MakeSrc2dDescriptor(const std::vector<index_t>& inLengths,
const std::vector<index_t>& inStrides,
int blkGroupSize,
int numBlockTileIteration)
{
const auto tupleSrcLengths = make_tuple_from_array(inLengths, Number<numSrcDim>{});
const auto tupleSrcStrides = make_tuple_from_array(inStrides, Number<numSrcDim>{});
const auto inDesc = make_naive_tensor_descriptor(tupleSrcLengths, tupleSrcStrides);
const auto in_grid_desc_m_k = [&]() {
if constexpr(reduceAllDim)
{
const auto one_dim_inDesc = transform_tensor_descriptor(
inDesc,
make_tuple(make_merge_transform(tupleSrcLengths)),
make_tuple(typename arithmetic_sequence_gen<0, numSrcDim, 1>::type{}),
make_tuple(Sequence<0>{}));
return transform_tensor_descriptor(one_dim_inDesc,
make_tuple(make_unmerge_transform(make_tuple(
1, one_dim_inDesc.GetLength(Number<0>{})))),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0, 1>{}));
}
else
{
using InvariantDims = typename arithmetic_sequence_gen<0, NumInvariantDim, 1>::type;
using ReduceDims = typename arithmetic_sequence_gen<NumInvariantDim, Rank, 1>::type;
const auto reduceDimLengths =
make_tuple_from_array_and_index_seq(inLengths, ReduceDims{});
const auto invariantDimLengths =
make_tuple_from_array_and_index_seq(inLengths, InvariantDims{});
return transform_tensor_descriptor(
inDesc,
make_tuple(make_merge_transform(invariantDimLengths),
make_merge_transform(reduceDimLengths)),
make_tuple(InvariantDims{}, ReduceDims{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
}();
const auto invariantLength = in_grid_desc_m_k.GetLength(Number<0>{});
const auto reduceLength = in_grid_desc_m_k.GetLength(Number<1>{});
const int reduceSizePerBlock = K_BlockTileSize * numBlockTileIteration;
const auto inPad_M =
math::integer_least_multiple(invariantLength, M_BlockTileSize) - invariantLength;
const auto inPad_K = reduceSizePerBlock * blkGroupSize - reduceLength;
auto in_grid_desc_m_k_padded = transform_tensor_descriptor(
in_grid_desc_m_k,
make_tuple(make_right_pad_transform(invariantLength, inPad_M),
make_right_pad_transform(reduceLength, inPad_K)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return (in_grid_desc_m_k_padded);
};
static auto MakeDst1dDescriptor(const std::vector<index_t>& outLengths,
const std::vector<index_t>& outStrides)
{
const auto tupleDstLengths = make_tuple_from_array(outLengths, Number<numDstDim>{});
const auto tupleDstStrides = make_tuple_from_array(outStrides, Number<numDstDim>{});
auto outDesc = make_naive_tensor_descriptor(tupleDstLengths, tupleDstStrides);
auto out_grid_desc_m = transform_tensor_descriptor(
outDesc,
make_tuple(make_merge_transform(tupleDstLengths)),
make_tuple(typename arithmetic_sequence_gen<0, numDstDim, 1>::type{}),
make_tuple(Sequence<0>{}));
const auto invariantLength = out_grid_desc_m.GetLength(Number<0>{});
const auto outPad =
math::integer_least_multiple(invariantLength, M_BlockTileSize) - invariantLength;
auto out_grid_desc_m_padded = transform_tensor_descriptor(
out_grid_desc_m,
make_tuple(make_right_pad_transform(invariantLength, outPad)),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0>{}));
return (out_grid_desc_m_padded);
};
static auto MakeDst1dDescriptorForBufferSet(const std::vector<index_t>& outLengths,
const std::vector<index_t>& outStrides)
{
const auto tupleDstLengths = make_tuple_from_array(outLengths, Number<numDstDim>{});
const auto tupleDstStrides = make_tuple_from_array(outStrides, Number<numDstDim>{});
auto outDesc = make_naive_tensor_descriptor(tupleDstLengths, tupleDstStrides);
auto out_grid_desc_m = transform_tensor_descriptor(
outDesc,
make_tuple(make_merge_transform(tupleDstLengths)),
make_tuple(typename arithmetic_sequence_gen<0, numDstDim, 1>::type{}),
make_tuple(Sequence<0>{}));
const auto length = out_grid_desc_m.GetLength(Number<0>{});
const auto pad = math::integer_least_multiple(length, BlockSize) - length;
auto out_grid_desc_m_padded =
transform_tensor_descriptor(out_grid_desc_m,
make_tuple(make_right_pad_transform(length, pad)),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0>{}));
return (out_grid_desc_m_padded);
};
struct Argument : public BaseArgument
{
Argument(const std::vector<index_t> inLengths,
const std::vector<index_t> inStrides,
const std::vector<index_t> outLengths,
const std::vector<index_t> outStrides,
const std::vector<int> reduceDims,
float alpha,
float beta,
const InDataType* in_dev,
const IndexDataType* in_index_dev,
OutDataType* out_dev,
IndexDataType* out_index_dev,
const InElementwiseOperation in_elementwise_op,
const AccElementwiseOperation acc_elementwise_op)
: outLengths_{outLengths},
outStrides_{outStrides},
in_dev_{in_dev},
in_index_dev_{in_index_dev},
out_dev_{out_dev},
out_index_dev_{out_index_dev},
in_elementwise_op_{in_elementwise_op},
acc_elementwise_op_{acc_elementwise_op}
{
inLengths_ = shuffle_tensor_dimensions<Rank, NumReduceDim>(inLengths, reduceDims);
inStrides_ = shuffle_tensor_dimensions<Rank, NumReduceDim>(inStrides, reduceDims);
alpha_ = type_convert<AccDataType>(alpha);
beta_ = type_convert<AccDataType>(beta);
std::tie(invariant_total_length, reduce_total_length) =
get_2d_lengths<Rank, NumReduceDim>(inLengths_);
if constexpr(NumInvariantDim == 0)
invariant_lowest_length = 1;
else
invariant_lowest_length = inLengths_[NumInvariantDim - 1];
reduce_lowest_length = inLengths_[Rank - 1];
if constexpr(use_multiblock)
{
int iterations = 1;
while(true)
{
int testBlkGroupSize =
(reduce_total_length + (K_BlockTileSize * iterations) - 1) /
(K_BlockTileSize * iterations);
// we want the blkGroupSize be not more than 128
if(testBlkGroupSize <= 128)
break;
iterations++;
};
blkGroupSize = (reduce_total_length + (K_BlockTileSize * iterations) - 1) /
(K_BlockTileSize * iterations);
numBlockTileIteration = iterations;
}
else
{
blkGroupSize = 1;
numBlockTileIteration =
(reduce_total_length + K_BlockTileSize - 1) / K_BlockTileSize;
};
gridSize = math::integer_least_multiple(invariant_total_length, M_BlockTileSize) /
M_BlockTileSize * blkGroupSize;
gridSize_pre =
math::integer_least_multiple(invariant_total_length, BlockSize) / BlockSize;
}
std::vector<index_t> inLengths_;
std::vector<index_t> inStrides_;
std::vector<index_t> outLengths_;
std::vector<index_t> outStrides_;
AccDataType alpha_;
AccDataType beta_;
const InDataType* in_dev_;
const IndexDataType* in_index_dev_;
OutDataType* out_dev_;
IndexDataType* out_index_dev_;
InElementwiseOperation in_elementwise_op_;
AccElementwiseOperation acc_elementwise_op_;
index_t invariant_lowest_length;
index_t reduce_lowest_length;
long_index_t invariant_total_length;
long_index_t reduce_total_length;
int blkGroupSize;
int numBlockTileIteration;
size_t gridSize;
size_t gridSize_pre;
};
struct Invoker : public BaseInvoker
{
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
const auto in_grid_desc_m_k = DeviceReduceMultiBlock::MakeSrc2dDescriptor(
arg.inLengths_, arg.inStrides_, arg.blkGroupSize, arg.numBlockTileIteration);
const auto out_grid_desc_m =
DeviceReduceMultiBlock::MakeDst1dDescriptor(arg.outLengths_, arg.outStrides_);
const auto out_grid_desc_m_2 = DeviceReduceMultiBlock::MakeDst1dDescriptorForBufferSet(
arg.outLengths_, arg.outStrides_);
using InGridDesc_M_K = decltype(in_grid_desc_m_k);
using OutGridDesc_M = decltype(out_grid_desc_m);
using OutGridDesc_M_2 = decltype(out_grid_desc_m_2);
using GridwiseReduce = GridwiseReduction_mk_to_m_multiblock<InDataType,
OutDataType,
AccDataType,
IndexDataType,
InGridDesc_M_K,
OutGridDesc_M,
ReduceOperation,
InElementwiseOperation,
AccElementwiseOperation,
OutMemoryDataOperation,
PropagateNan,
BlockSize,
MThreadClusterSize,
KThreadClusterSize,
MThreadSliceSize,
KThreadSliceSize,
InSrcVectorDim,
InSrcVectorSize,
OutDstVectorSize>;
const auto kernel_main = kernel_reduce_multiblock<GridwiseReduce,
OutputIndex,
HaveIndexInput,
InDataType,
OutDataType,
AccDataType,
int32_t,
InGridDesc_M_K,
OutGridDesc_M,
InElementwiseOperation,
AccElementwiseOperation>;
float avg_time = 0;
if constexpr(use_multiblock)
{
const auto identityVal =
ck::reduce::GetIdentityValueForInMemoryDataOperation<OutDataType>(
OutMemoryDataOperation);
const auto kernel_pre =
kernel_buffer_set_value<BlockSize, OutDataType, OutGridDesc_M_2>;
avg_time += launch_and_time_kernel(stream_config,
kernel_pre,
dim3(arg.gridSize_pre),
dim3(BlockSize),
0,
out_grid_desc_m_2,
arg.out_dev_,
identityVal);
};
avg_time += launch_and_time_kernel(stream_config,
kernel_main,
dim3(arg.gridSize),
dim3(BlockSize),
0,
in_grid_desc_m_k,
out_grid_desc_m,
arg.in_elementwise_op_,
arg.acc_elementwise_op_,
arg.blkGroupSize,
arg.numBlockTileIteration,
arg.alpha_,
arg.in_dev_,
arg.in_index_dev_,
arg.beta_,
arg.out_dev_,
arg.out_index_dev_);
return (avg_time);
};
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
};
};
static bool IsSupportedArgument(const Argument* pArg)
{
if constexpr(use_multiblock)
{
if(static_cast<float>(pArg->beta_) != 0.0f)
return (false);
};
if constexpr(InSrcVectorDim == 0)
{
if constexpr(NumInvariantDim == 0)
{
return (false);
}
else
{
if(pArg->inStrides_[NumInvariantDim - 1] != 1)
return (false);
if(pArg->invariant_lowest_length % InSrcVectorSize != 0)
return (false);
};
}
else
{
if(pArg->inStrides_[Rank - 1] != 1)
return (false);
if(pArg->reduce_lowest_length % InSrcVectorSize != 0)
return (false);
};
// To improve
if(pArg->invariant_lowest_length % OutDstVectorSize != 0)
return (false);
if constexpr(use_multiblock)
{
// blkGroupSize of 1 should be handled by Blockwise path using
// InMemoryDataOperationEnum::Set
if(pArg->blkGroupSize == 1)
return (false);
// This is very strong restriction, but needed to avoid some failure
if(pArg->invariant_lowest_length % M_BlockTileSize != 0)
return (false);
}
else
{
// cases with very small reduce_total_length should be handled by ThreadWise kernel
// if(pArg->reduce_total_length / KThreadSliceSize < 2)
// return (false);
};
return (true);
}
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(dynamic_cast<const Argument*>(p_arg));
};
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const std::vector<index_t> inLengths,
const std::vector<index_t> inStrides,
const std::vector<index_t> outLengths,
const std::vector<index_t> outStrides,
const std::vector<int> reduceDims,
float alpha,
float beta,
const void* in_dev,
const void* in_index_dev,
void* out_dev,
void* out_index_dev,
const InElementwiseOperation in_elementwise_op,
const AccElementwiseOperation acc_elementwise_op) override
{
return std::make_unique<Argument>(inLengths,
inStrides,
outLengths,
outStrides,
reduceDims,
alpha,
beta,
static_cast<const InDataType*>(in_dev),
static_cast<const IndexDataType*>(in_index_dev),
static_cast<OutDataType*>(out_dev),
static_cast<IndexDataType*>(out_index_dev),
in_elementwise_op,
acc_elementwise_op);
};
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>();
};
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << (OutMemoryDataOperation == InMemoryDataOperationEnum::Set? "DeviceReduceBlockWise<" : "DeviceReduceMultiBlock<") << BlockSize << ",";
str << "M_C" << MThreadClusterSize << "_S" << MThreadSliceSize << ",";
str << "K_C" << KThreadClusterSize << "_S" << KThreadSliceSize << ",";
str << "InSrcVectorDim_" << InSrcVectorDim << "_InSrcVectorSize_" << InSrcVectorSize << "_OutDstVectorSize_" << OutDstVectorSize << ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

376
include/ck/tensor_operation/gpu/device/impl/device_reduce_threadwise.hpp Normal file
View File

@@ -0,0 +1,376 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
#include "ck/tensor_operation/gpu/device/device_reduce.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_reduce_common.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_2d_reduction_multiblock.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_2d_reduction_threadwise.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <typename InDataType,
typename AccDataType,
typename OutDataType,
index_t Rank,
index_t NumReduceDim,
typename ReduceOperation,
typename InElementwiseOperation,
typename AccElementwiseOperation,
bool PropagateNan,
bool OutputIndex,
bool HaveIndexInputIfOutputIndex,
index_t BlockSize,
index_t MThreadSliceSize,
index_t KThreadSliceSize,
index_t InSrcVectorDim,
index_t InSrcVectorSize,
index_t OutDstVectorSize>
struct DeviceReduceThreadWise : public DeviceReduce<InElementwiseOperation, AccElementwiseOperation>
{
static_assert(Rank <= 6, "Bigger Rank size is not supported!");
static_assert(((InSrcVectorDim == 0 && MThreadSliceSize % InSrcVectorSize == 0) ||
(InSrcVectorDim == 1 && KThreadSliceSize % InSrcVectorSize == 0)) &&
(MThreadSliceSize % OutDstVectorSize == 0),
"Invalid thread slice sizes and/or vector sizes configuration, please check!");
using IndexDataType = int32_t;
static constexpr bool HaveIndexInput = OutputIndex && HaveIndexInputIfOutputIndex;
static constexpr index_t NumInvariantDim = Rank - NumReduceDim;
static constexpr index_t numSrcDim = Rank;
static constexpr index_t numDstDim = (NumInvariantDim == 0) ? 1 : NumInvariantDim;
static constexpr bool reduceAllDim = (NumInvariantDim == 0);
static constexpr index_t M_BlockTileSize = BlockSize * MThreadSliceSize;
static constexpr index_t K_BlockTileSize = 1 * KThreadSliceSize;
static auto MakeSrc2dDescriptor(const std::vector<index_t>& inLengths,
const std::vector<index_t>& inStrides)
{
const auto tupleSrcLengths = make_tuple_from_array(inLengths, Number<numSrcDim>{});
const auto tupleSrcStrides = make_tuple_from_array(inStrides, Number<numSrcDim>{});
const auto inDesc = make_naive_tensor_descriptor(tupleSrcLengths, tupleSrcStrides);
const auto in_grid_desc_m_k = [&]() {
if constexpr(reduceAllDim)
{
const auto one_dim_inDesc = transform_tensor_descriptor(
inDesc,
make_tuple(make_merge_transform(tupleSrcLengths)),
make_tuple(typename arithmetic_sequence_gen<0, numSrcDim, 1>::type{}),
make_tuple(Sequence<0>{}));
return transform_tensor_descriptor(one_dim_inDesc,
make_tuple(make_unmerge_transform(make_tuple(
1, one_dim_inDesc.GetLength(Number<0>{})))),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0, 1>{}));
}
else
{
using InvariantDims = typename arithmetic_sequence_gen<0, NumInvariantDim, 1>::type;
using ReduceDims = typename arithmetic_sequence_gen<NumInvariantDim, Rank, 1>::type;
const auto reduceDimLengths =
make_tuple_from_array_and_index_seq(inLengths, ReduceDims{});
const auto invariantDimLengths =
make_tuple_from_array_and_index_seq(inLengths, InvariantDims{});
return transform_tensor_descriptor(
inDesc,
make_tuple(make_merge_transform(invariantDimLengths),
make_merge_transform(reduceDimLengths)),
make_tuple(InvariantDims{}, ReduceDims{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
}
}();
const auto invariantLength = in_grid_desc_m_k.GetLength(Number<0>{});
const auto reduceLength = in_grid_desc_m_k.GetLength(Number<1>{});
const auto inPad_M =
math::integer_least_multiple(invariantLength, M_BlockTileSize) - invariantLength;
const auto inPad_K =
math::integer_least_multiple(reduceLength, K_BlockTileSize) - reduceLength;
auto in_grid_desc_m_k_padded = transform_tensor_descriptor(
in_grid_desc_m_k,
make_tuple(make_right_pad_transform(invariantLength, inPad_M),
make_right_pad_transform(reduceLength, inPad_K)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
return (in_grid_desc_m_k_padded);
};
static auto MakeDst1dDescriptor(const std::vector<index_t>& outLengths,
const std::vector<index_t>& outStrides)
{
const auto tupleDstLengths = make_tuple_from_array(outLengths, Number<numDstDim>{});
const auto tupleDstStrides = make_tuple_from_array(outStrides, Number<numDstDim>{});
auto outDesc = make_naive_tensor_descriptor(tupleDstLengths, tupleDstStrides);
auto out_grid_desc_m = transform_tensor_descriptor(
outDesc,
make_tuple(make_merge_transform(tupleDstLengths)),
make_tuple(typename arithmetic_sequence_gen<0, numDstDim, 1>::type{}),
make_tuple(Sequence<0>{}));
const auto invariantLength = out_grid_desc_m.GetLength(Number<0>{});
const auto outPad =
math::integer_least_multiple(invariantLength, M_BlockTileSize) - invariantLength;
auto out_grid_desc_m_padded = transform_tensor_descriptor(
out_grid_desc_m,
make_tuple(make_right_pad_transform(invariantLength, outPad)),
make_tuple(Sequence<0>{}),
make_tuple(Sequence<0>{}));
return (out_grid_desc_m_padded);
};
struct Argument : public BaseArgument
{
Argument(const std::vector<index_t> inLengths,
const std::vector<index_t> inStrides,
const std::vector<index_t> outLengths,
const std::vector<index_t> outStrides,
const std::vector<int> reduceDims,
float alpha,
float beta,
const InDataType* in_dev,
OutDataType* out_dev,
IndexDataType* out_index_dev,
const InElementwiseOperation in_elementwise_op,
const AccElementwiseOperation acc_elementwise_op)
: outLengths_{outLengths},
outStrides_{outStrides},
in_dev_{in_dev},
out_dev_{out_dev},
out_index_dev_{out_index_dev},
in_elementwise_op_{in_elementwise_op},
acc_elementwise_op_{acc_elementwise_op}
{
inLengths_ = shuffle_tensor_dimensions<Rank, NumReduceDim>(inLengths, reduceDims);
inStrides_ = shuffle_tensor_dimensions<Rank, NumReduceDim>(inStrides, reduceDims);
alpha_ = type_convert<AccDataType>(alpha);
beta_ = type_convert<AccDataType>(beta);
std::tie(invariant_total_length, reduce_total_length) =
get_2d_lengths<Rank, NumReduceDim>(inLengths_);
if constexpr(NumInvariantDim == 0)
invariant_lowest_length = 1;
else
invariant_lowest_length = inLengths_[NumInvariantDim - 1];
reduce_lowest_length = inLengths_[Rank - 1];
numBlockTileIteration = (reduce_total_length + K_BlockTileSize - 1) / K_BlockTileSize;
gridSize = math::integer_least_multiple(invariant_total_length, M_BlockTileSize) /
M_BlockTileSize;
}
std::vector<index_t> inLengths_;
std::vector<index_t> inStrides_;
std::vector<index_t> outLengths_;
std::vector<index_t> outStrides_;
AccDataType alpha_;
AccDataType beta_;
const InDataType* in_dev_;
OutDataType* out_dev_;
IndexDataType* out_index_dev_;
InElementwiseOperation in_elementwise_op_;
AccElementwiseOperation acc_elementwise_op_;
index_t invariant_lowest_length;
index_t reduce_lowest_length;
long_index_t invariant_total_length;
long_index_t reduce_total_length;
int numBlockTileIteration;
size_t gridSize;
};
struct Invoker : public BaseInvoker
{
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
const auto in_grid_desc_m_k =
DeviceReduceThreadWise::MakeSrc2dDescriptor(arg.inLengths_, arg.inStrides_);
const auto out_grid_desc_m =
DeviceReduceThreadWise::MakeDst1dDescriptor(arg.outLengths_, arg.outStrides_);
using InGridDesc_M_K = decltype(in_grid_desc_m_k);
using OutGridDesc_M = decltype(out_grid_desc_m);
float avg_time = 0;
using GridwiseReduce =
GridwiseReduction_mk_to_m_threadwise<InDataType,
OutDataType,
AccDataType,
IndexDataType,
InGridDesc_M_K,
OutGridDesc_M,
ReduceOperation,
InElementwiseOperation,
AccElementwiseOperation,
InMemoryDataOperationEnum::Set,
PropagateNan,
BlockSize,
MThreadSliceSize,
KThreadSliceSize,
InSrcVectorDim,
InSrcVectorSize,
OutDstVectorSize>;
const auto kernel = kernel_reduce_threadwise<GridwiseReduce,
OutputIndex,
HaveIndexInput,
InDataType,
OutDataType,
AccDataType,
IndexDataType,
InGridDesc_M_K,
OutGridDesc_M,
InElementwiseOperation,
AccElementwiseOperation>;
avg_time = launch_and_time_kernel(stream_config,
kernel,
dim3(arg.gridSize),
dim3(BlockSize),
0,
in_grid_desc_m_k,
out_grid_desc_m,
arg.in_elementwise_op_,
arg.acc_elementwise_op_,
arg.alpha_,
arg.in_dev_,
nullptr,
arg.beta_,
arg.out_dev_,
arg.out_index_dev_);
return (avg_time);
};
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
};
};
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
const Argument* pArg = dynamic_cast<const Argument*>(p_arg);
if constexpr(InSrcVectorDim == 0)
{
if constexpr(NumInvariantDim == 0)
{
return (false);
}
else
{
if(pArg->inStrides_[NumInvariantDim - 1] != 1)
return (false);
if(pArg->invariant_lowest_length % InSrcVectorSize != 0)
return (false);
};
}
else
{
if(pArg->inStrides_[Rank - 1] != 1)
return (false);
if(pArg->reduce_lowest_length % InSrcVectorSize != 0)
return (false);
};
// To improve
if(pArg->invariant_lowest_length % OutDstVectorSize != 0)
return (false);
// cases with big reduce_total_length should be handled by Blockwise kernel
if(pArg->reduce_total_length / KThreadSliceSize >= 32)
return (false);
return (true);
};
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const std::vector<index_t> inLengths,
const std::vector<index_t> inStrides,
const std::vector<index_t> outLengths,
const std::vector<index_t> outStrides,
const std::vector<int> reduceDims,
float alpha,
float beta,
const void* in_dev,
const void* in_index_dev,
void* out_dev,
void* out_index_dev,
const InElementwiseOperation in_elementwise_op,
const AccElementwiseOperation acc_elementwise_op) override
{
(void)in_index_dev;
return std::make_unique<Argument>(inLengths,
inStrides,
outLengths,
outStrides,
reduceDims,
alpha,
beta,
static_cast<const InDataType*>(in_dev),
static_cast<OutDataType*>(out_dev),
static_cast<IndexDataType*>(out_index_dev),
in_elementwise_op,
acc_elementwise_op);
};
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>();
};
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceReduceThreadWise<" << BlockSize << ",";
str << "M_C" << BlockSize << "_S" << MThreadSliceSize << ",";
str << "K_C" << 1 << "_S" << KThreadSliceSize << ",";
str << "InSrcVectorDim_" << InSrcVectorDim << "_InSrcVectorSize_" << InSrcVectorSize << "_OutDstVectorSize_" << OutDstVectorSize << ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck

4
include/ck/tensor_operation/gpu/device/impl/device_softmax_impl.hpp
View File

@@ -10,8 +10,8 @@
#include "ck/tensor_operation/gpu/device/device_base.hpp"
#include "ck/tensor_operation/gpu/device/device_reduce.hpp"
#include "ck/tensor_operation/gpu/device/device_softmax.hpp"
#include "ck/tensor_operation/gpu/device/device_reduce_multiblock.hpp"
#include "ck/tensor_operation/gpu/device/device_reduce_common.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_reduce_common.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_reduce_multiblock.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_softmax.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_set_buffer_value.hpp"
#include "ck/host_utility/device_prop.hpp"

210
include/ck/tensor_operation/gpu/device/impl/device_sparse_embedding3_forward_layernorm.hpp Normal file
View File

@@ -0,0 +1,210 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
#include "ck/tensor_operation/gpu/device/device_base.hpp"
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_sparse_embedding3_forward_layernorm.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <typename EmbType,
typename IndexType,
typename GammaDataType,
typename BetaDataType,
typename AccDataType,
typename OutType,
ck::index_t BlockSize,
ck::index_t DimClusterSize,
ck::index_t RowClusterSize,
ck::index_t DimPerBlock,
ck::index_t RowPerBlock,
ck::index_t DimThreadSize,
ck::index_t RowVectorSize>
struct DeviceSparseEmbedding3ForwardLayernorm : public BaseOperator
{
static auto MakeOutputDescriptor(const index_t index_length, const index_t rows)
{
return make_naive_tensor_descriptor_packed(make_tuple(index_length, rows));
}
struct Argument : public BaseArgument
{
Argument(OutType* p_out,
const EmbType* p_emb_a,
const EmbType* p_emb_b,
const EmbType* p_emb_c,
const IndexType* p_index_a,
const IndexType* p_index_b,
const IndexType* p_index_c,
const GammaDataType* p_gamma,
const BetaDataType* p_beta,
const ck::index_t NumRows,
const ck::index_t EmbeddingDim,
const ck::index_t IndexLength,
const AccDataType epsilon)
: p_out_(p_out),
p_emb_a_(p_emb_a),
p_emb_b_(p_emb_b),
p_emb_c_(p_emb_c),
p_index_a_(p_index_a),
p_index_b_(p_index_b),
p_index_c_(p_index_c),
p_gamma_(p_gamma),
p_beta_(p_beta),
NumRows_(NumRows),
EmbeddingDim_(EmbeddingDim),
IndexLength_(IndexLength),
epsilon_(epsilon)
{
grid_size_ = (IndexLength + DimClusterSize - 1) / DimClusterSize;
}
OutType* p_out_;
const EmbType* p_emb_a_;
const EmbType* p_emb_b_;
const EmbType* p_emb_c_;
const IndexType* p_index_a_;
const IndexType* p_index_b_;
const IndexType* p_index_c_;
const GammaDataType* p_gamma_;
const BetaDataType* p_beta_;
ck::index_t NumRows_;
ck::index_t EmbeddingDim_;
ck::index_t IndexLength_;
AccDataType epsilon_;
size_t grid_size_;
};
virtual std::unique_ptr<BaseArgument> MakeArgumentPointer(void* p_out,
const void* p_emb_a,
const void* p_emb_b,
const void* p_emb_c,
const void* p_index_a,
const void* p_index_b,
const void* p_index_c,
const void* p_gamma,
const void* p_beta,
ck::index_t NumRows,
ck::index_t EmbeddingDim,
ck::index_t IndexLength,
const AccDataType epsilon)
{
return std::make_unique<Argument>(reinterpret_cast<OutType*>(p_out),
reinterpret_cast<const EmbType*>(p_emb_a),
reinterpret_cast<const EmbType*>(p_emb_b),
reinterpret_cast<const EmbType*>(p_emb_c),
reinterpret_cast<const IndexType*>(p_index_a),
reinterpret_cast<const IndexType*>(p_index_b),
reinterpret_cast<const IndexType*>(p_index_c),
reinterpret_cast<const GammaDataType*>(p_gamma),
reinterpret_cast<const BetaDataType*>(p_beta),
NumRows,
EmbeddingDim,
IndexLength,
epsilon);
}
using GridwiseSparseEmbedding =
GridwiseSparseEmbedding3ForwardLayernorm<EmbType,
IndexType,
GammaDataType,
BetaDataType,
AccDataType,
OutType,
decltype(MakeOutputDescriptor(1, 1)),
BlockSize,
DimClusterSize,
RowClusterSize,
DimPerBlock,
RowPerBlock,
DimThreadSize,
RowVectorSize>;
struct Invoker : public BaseInvoker
{
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
auto out_desc = MakeOutputDescriptor(arg.IndexLength_, arg.EmbeddingDim_);
const auto kernel_main =
kernel_sparse_embedding3_forward_layernorm<GridwiseSparseEmbedding,
EmbType,
IndexType,
GammaDataType,
BetaDataType,
AccDataType,
OutType,
decltype(out_desc)>;
float avg_time = 0;
avg_time += launch_and_time_kernel(stream_config,
kernel_main,
dim3(arg.grid_size_),
dim3(BlockSize),
0,
arg.p_out_,
arg.p_emb_a_,
arg.p_emb_b_,
arg.p_emb_c_,
arg.p_index_a_,
arg.p_index_b_,
arg.p_index_c_,
arg.p_gamma_,
arg.p_beta_,
out_desc,
arg.epsilon_);
return (avg_time);
}
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
};
};
static bool IsSupportedArgument(const Argument* p_arg)
{
return (RowPerBlock == p_arg->EmbeddingDim_) && (p_arg->NumRows_ % DimPerBlock == 0);
}
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(dynamic_cast<const Argument*>(p_arg));
}
virtual std::unique_ptr<BaseInvoker> MakeInvokerPointer()
{
return std::make_unique<Invoker>();
}
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceSparseEmbedding3ForwardLayernorm_"<< BlockSize << "_" <<
DimClusterSize << "x" << RowClusterSize << "_" <<
DimPerBlock << "x" << RowPerBlock << "_" <<
DimThreadSize << "x" << RowVectorSize;
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck