mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-04-20 06:49:15 +00:00
Wmma support for gemm_bias_add_reduce (#3316)
* Add tests for gemm_bias_add_reduce * Initial working implementation * Generalize implementation of reduce epilogue * Add tests for all layouts * Add instances * Fix test archs * Fix xdl bug * Remove library/profiler duplications * Fix num_byted error profiler * Fix typos * Fix copyright
This commit is contained in:
Enrico Degregori
@@ -0,0 +1,682 @@
|
||||
// Copyright (c) Advanced Micro Devices, Inc., or its affiliates.
|
||||
// SPDX-License-Identifier: MIT
|
||||
|
||||
#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_wmma_cshuffle_v3.hpp"
|
||||
#include "ck/host_utility/device_prop.hpp"
|
||||
#include "ck/host_utility/kernel_launch.hpp"
|
||||
|
||||
namespace ck {
|
||||
template <typename GridwiseGemm,
|
||||
typename ReduceTrait,
|
||||
bool HasMainKBlockLoop,
|
||||
InMemoryDataOperationEnum EGlobalMemoryDataOperation,
|
||||
index_t MinimumOccupancy = 1,
|
||||
TailNumber TailNum = TailNumber::Full>
|
||||
__global__ void
|
||||
#if CK_USE_LAUNCH_BOUNDS
|
||||
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, MinimumOccupancy)
|
||||
#endif
|
||||
kernel_gemm_bias_add_reduce_wmma_cshuffle_v3(
|
||||
typename GridwiseGemm::Argument karg,
|
||||
typename ReduceTrait::ReducePtrsGlobal_ p_reduces_grid,
|
||||
const typename ReduceTrait::ReduceInElementwiseOperations_ reduce_in_element_ops,
|
||||
const typename ReduceTrait::ReduceAccElementwiseOperations_ reduce_out_element_ops,
|
||||
const typename ReduceTrait::D0ElementwiseOperation_ d0_element_op)
|
||||
{
|
||||
#if(defined(__gfx11__) || defined(__gfx12__))
|
||||
#if defined(__gfx11__)
|
||||
// gfx11 does not support *_atomic_pk_add_f16/bf16 instructions
|
||||
using e_data_type = remove_cvref_t<remove_pointer_t<decltype(karg.p_e_grid)>>;
|
||||
if constexpr(!(EGlobalMemoryDataOperation == InMemoryDataOperationEnum::AtomicAdd &&
|
||||
(std::is_same_v<e_data_type, ck::half_t> ||
|
||||
std::is_same_v<e_data_type, ck::bhalf_t>)))
|
||||
{
|
||||
#endif
|
||||
using EpilogueType = typename GridwiseGemm::template EpilogueReduceCShuffle<ReduceTrait>;
|
||||
constexpr index_t LDS_size =
|
||||
GridwiseGemm::template GetSharedMemoryNumberOfByte<EpilogueType>();
|
||||
__shared__ char p_shared[LDS_size];
|
||||
|
||||
auto splitk_batch_offset = typename GridwiseGemm::SplitKBatchOffset(karg, blockIdx.z);
|
||||
|
||||
auto epilogue_args = EpilogueType(
|
||||
p_reduces_grid, reduce_in_element_ops, reduce_out_element_ops, karg.M, d0_element_op);
|
||||
|
||||
GridwiseGemm::template Run<HasMainKBlockLoop, EGlobalMemoryDataOperation, TailNum>(
|
||||
p_shared, splitk_batch_offset, karg, epilogue_args);
|
||||
#if defined(__gfx11__)
|
||||
}
|
||||
#endif
|
||||
#else
|
||||
ignore = karg;
|
||||
ignore = p_reduces_grid;
|
||||
ignore = reduce_in_element_ops;
|
||||
ignore = reduce_out_element_ops;
|
||||
ignore = d0_element_op;
|
||||
#endif
|
||||
}
|
||||
|
||||
} // namespace ck
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
|
||||
template <typename ALayout,
|
||||
typename BLayout,
|
||||
typename ELayout,
|
||||
typename ADataType,
|
||||
typename BDataType,
|
||||
typename EDataType,
|
||||
typename BiasDataType,
|
||||
typename D0DataType,
|
||||
typename AccDataType,
|
||||
typename CShuffleDataType,
|
||||
typename ReduceAccDataType, // Reduce
|
||||
typename ReducePtrsGlobal, // Reduce
|
||||
typename AElementwiseOperation,
|
||||
typename BElementwiseOperation,
|
||||
typename CElementwiseOperation,
|
||||
typename D0ElementwiseOperation,
|
||||
typename ReduceOperations, // Reduce
|
||||
typename ReduceInElementwiseOperations, // Reduce
|
||||
typename ReduceAccElementwiseOperations, // Reduce
|
||||
typename ReduceGlobalMemoryDataOperation, // Reduce
|
||||
GemmSpecialization GemmSpec,
|
||||
index_t BlockSize,
|
||||
index_t MPerBlock,
|
||||
index_t NPerBlock,
|
||||
index_t KPerBlock,
|
||||
index_t AK1,
|
||||
index_t BK1,
|
||||
index_t MPerWmma,
|
||||
index_t NPerWmma,
|
||||
index_t MRepeat,
|
||||
index_t NRepeat,
|
||||
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 CShuffleMRepeatPerShuffle,
|
||||
index_t CShuffleNRepeatPerShuffle,
|
||||
typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
|
||||
index_t CShuffleBlockTransferScalarPerVector,
|
||||
typename CReduceThreadClusterLengths_MPerBlock_NPerBlock, // Reduce
|
||||
index_t CReduceThreadLds2VGprCopySrcDstScalarPerVector_NPerBlock, // Reduce
|
||||
index_t CReduceThreadVgpr2GlobalCopySrcDstScalarPerVector_MPerBlock, // Reduce
|
||||
BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave,
|
||||
BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v1,
|
||||
typename ComputeTypeA = EDataType,
|
||||
typename ComputeTypeB = ComputeTypeA,
|
||||
bool PermuteA = false,
|
||||
bool PermuteB = false>
|
||||
struct DeviceGemmBiasAddReduce_Wmma_CShuffleV3
|
||||
: public DeviceGemmReduce<1, ReduceOperations::Size()>
|
||||
{
|
||||
using CDEShuffleBlockTransferScalarPerVectors = Sequence<CShuffleBlockTransferScalarPerVector,
|
||||
CShuffleBlockTransferScalarPerVector,
|
||||
CShuffleBlockTransferScalarPerVector>;
|
||||
|
||||
using GridwiseGemm = GridwiseGemm_wmma_cshuffle_v3<
|
||||
ALayout,
|
||||
BLayout,
|
||||
Tuple<ELayout, ELayout>,
|
||||
ELayout,
|
||||
Tuple<ADataType>,
|
||||
Tuple<BDataType>,
|
||||
AccDataType,
|
||||
CShuffleDataType,
|
||||
Tuple<BiasDataType, D0DataType>,
|
||||
EDataType,
|
||||
AElementwiseOperation,
|
||||
BElementwiseOperation,
|
||||
CElementwiseOperation,
|
||||
GemmSpec,
|
||||
BlockSize,
|
||||
MPerBlock,
|
||||
NPerBlock,
|
||||
KPerBlock,
|
||||
AK1,
|
||||
BK1,
|
||||
MPerWmma,
|
||||
NPerWmma,
|
||||
MRepeat,
|
||||
NRepeat,
|
||||
ABlockTransferThreadClusterLengths_AK0_M_AK1,
|
||||
ABlockTransferThreadClusterArrangeOrder,
|
||||
ABlockTransferSrcAccessOrder,
|
||||
ABlockTransferSrcVectorDim,
|
||||
ABlockTransferSrcScalarPerVector,
|
||||
ABlockTransferDstScalarPerVector_AK1,
|
||||
false,
|
||||
ABlockLdsExtraM,
|
||||
BBlockTransferThreadClusterLengths_BK0_N_BK1,
|
||||
BBlockTransferThreadClusterArrangeOrder,
|
||||
BBlockTransferSrcAccessOrder,
|
||||
BBlockTransferSrcVectorDim,
|
||||
BBlockTransferSrcScalarPerVector,
|
||||
BBlockTransferDstScalarPerVector_BK1,
|
||||
false,
|
||||
BBlockLdsExtraN,
|
||||
CShuffleMRepeatPerShuffle,
|
||||
CShuffleNRepeatPerShuffle,
|
||||
CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
|
||||
CDEShuffleBlockTransferScalarPerVectors,
|
||||
BlkGemmPipeSched,
|
||||
BlkGemmPipelineVer,
|
||||
ComputeTypeA,
|
||||
ComputeTypeB,
|
||||
PermuteA,
|
||||
PermuteB>;
|
||||
|
||||
using ReduceTrait = ReduceTrait_<ReduceAccDataType,
|
||||
ReducePtrsGlobal,
|
||||
D0ElementwiseOperation,
|
||||
ReduceOperations,
|
||||
ReduceInElementwiseOperations,
|
||||
ReduceAccElementwiseOperations,
|
||||
ReduceGlobalMemoryDataOperation,
|
||||
CReduceThreadClusterLengths_MPerBlock_NPerBlock,
|
||||
CReduceThreadLds2VGprCopySrcDstScalarPerVector_NPerBlock,
|
||||
CReduceThreadVgpr2GlobalCopySrcDstScalarPerVector_MPerBlock>;
|
||||
|
||||
// Argument
|
||||
struct Argument : public BaseArgument
|
||||
{
|
||||
Argument(const ADataType* p_a_grid,
|
||||
const BDataType* p_b_grid,
|
||||
EDataType* p_e_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_e_grid_{p_e_grid},
|
||||
p_bias_grid_{p_bias_grid},
|
||||
p_d0_grid_{p_d0_grid},
|
||||
p_reduces_grid_{p_reduces_grid},
|
||||
MRaw_{MRaw},
|
||||
NRaw_{NRaw},
|
||||
KRaw_{KRaw},
|
||||
StrideA_{StrideA},
|
||||
StrideB_{StrideB},
|
||||
StrideC_{StrideC},
|
||||
StrideC1_{StrideC1},
|
||||
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}
|
||||
{
|
||||
}
|
||||
|
||||
// private:
|
||||
const ADataType* p_a_grid_;
|
||||
const BDataType* p_b_grid_;
|
||||
EDataType* p_e_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_;
|
||||
};
|
||||
|
||||
struct Invoker : public BaseInvoker
|
||||
{
|
||||
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
|
||||
{
|
||||
typename GridwiseGemm::Argument gemm_arg{
|
||||
std::array<const void*, 1>{arg.p_a_grid_},
|
||||
std::array<const void*, 1>{arg.p_b_grid_},
|
||||
std::array<const void*, 2>{arg.p_bias_grid_, arg.p_d0_grid_},
|
||||
static_cast<EDataType*>(arg.p_e_grid_),
|
||||
arg.MRaw_,
|
||||
arg.NRaw_,
|
||||
arg.KRaw_,
|
||||
std::array<index_t, 1>{arg.StrideA_}, // StrideAs
|
||||
std::array<index_t, 1>{arg.StrideB_}, // StrideBs
|
||||
std::array<index_t, 2>{0, arg.StrideC1_}, // StrideDs
|
||||
arg.StrideC_, // StrideE
|
||||
1, // kbatch
|
||||
arg.a_element_op_,
|
||||
arg.b_element_op_,
|
||||
arg.c_element_op_};
|
||||
|
||||
if(stream_config.log_level_ > 0)
|
||||
{
|
||||
gemm_arg.Print();
|
||||
GridwiseGemm::BlockwiseGemmPipe::HotLoopInstList::Print();
|
||||
}
|
||||
|
||||
if(!GridwiseGemm::CheckValidity(gemm_arg))
|
||||
{
|
||||
throw std::runtime_error("wrong! GridwiseGemm has invalid setting");
|
||||
}
|
||||
|
||||
index_t gdx, gdy, gdz;
|
||||
std::tie(gdx, gdy, gdz) = GridwiseGemm::CalculateGridSize(arg.MRaw_, arg.NRaw_, 1);
|
||||
|
||||
float ave_time = 0;
|
||||
|
||||
index_t K_split = (arg.KRaw_ + KPerBlock - 1) / KPerBlock * KPerBlock;
|
||||
|
||||
const bool has_main_k_block_loop = GridwiseGemm::CalculateHasMainKBlockLoop(K_split);
|
||||
TailNumber TailNum = GridwiseGemm::CalculateKBlockLoopTailNum(arg.KRaw_);
|
||||
|
||||
const auto Run = [&](const auto& kernel) {
|
||||
// Note: cache flushing not supported
|
||||
|
||||
ave_time += launch_and_time_kernel(stream_config,
|
||||
kernel,
|
||||
dim3(gdx, gdy, gdz),
|
||||
dim3(BlockSize),
|
||||
0,
|
||||
gemm_arg,
|
||||
arg.p_reduces_grid_,
|
||||
arg.reduce_in_element_ops_,
|
||||
arg.reduce_out_element_ops_,
|
||||
arg.d0_element_op_);
|
||||
};
|
||||
|
||||
constexpr index_t minimum_occupancy = []() {
|
||||
if constexpr(BlkGemmPipeSched == BlockGemmPipelineScheduler::Interwave)
|
||||
{
|
||||
return 2;
|
||||
}
|
||||
else if constexpr(BlkGemmPipelineVer == BlockGemmPipelineVersion::v3)
|
||||
{
|
||||
return (MPerBlock * NPerBlock / BlockSize <= 128) ? 2 : 1;
|
||||
}
|
||||
else
|
||||
{
|
||||
return 1;
|
||||
}
|
||||
}();
|
||||
|
||||
if(has_main_k_block_loop)
|
||||
{
|
||||
// Tail number always full
|
||||
if constexpr(BlkGemmPipelineVer == BlockGemmPipelineVersion::v1 ||
|
||||
BlkGemmPipelineVer == BlockGemmPipelineVersion::v3)
|
||||
{
|
||||
if(TailNum == TailNumber::Full)
|
||||
{
|
||||
const auto kernel = kernel_gemm_bias_add_reduce_wmma_cshuffle_v3<
|
||||
GridwiseGemm,
|
||||
ReduceTrait,
|
||||
true,
|
||||
InMemoryDataOperationEnum::Set,
|
||||
minimum_occupancy>;
|
||||
Run(kernel);
|
||||
}
|
||||
else
|
||||
{
|
||||
throw std::runtime_error("wrong! Invalid pipeline setting");
|
||||
}
|
||||
}
|
||||
}
|
||||
else
|
||||
{
|
||||
if constexpr(BlkGemmPipelineVer == BlockGemmPipelineVersion::v1)
|
||||
{
|
||||
if(TailNum == TailNumber::Full)
|
||||
{
|
||||
const auto kernel = kernel_gemm_bias_add_reduce_wmma_cshuffle_v3<
|
||||
GridwiseGemm,
|
||||
ReduceTrait,
|
||||
false,
|
||||
InMemoryDataOperationEnum::Set,
|
||||
minimum_occupancy>;
|
||||
Run(kernel);
|
||||
}
|
||||
else
|
||||
{
|
||||
throw std::runtime_error("wrong! Invalid pipeline v1 setting");
|
||||
}
|
||||
}
|
||||
else if constexpr(BlkGemmPipelineVer == BlockGemmPipelineVersion::v3)
|
||||
{
|
||||
if(TailNum == TailNumber::Even)
|
||||
{
|
||||
const auto kernel = kernel_gemm_bias_add_reduce_wmma_cshuffle_v3<
|
||||
GridwiseGemm,
|
||||
ReduceTrait,
|
||||
false,
|
||||
InMemoryDataOperationEnum::Set,
|
||||
minimum_occupancy,
|
||||
TailNumber::Even>;
|
||||
Run(kernel);
|
||||
}
|
||||
else if(TailNum == TailNumber::Odd)
|
||||
{
|
||||
const auto kernel = kernel_gemm_bias_add_reduce_wmma_cshuffle_v3<
|
||||
GridwiseGemm,
|
||||
ReduceTrait,
|
||||
false,
|
||||
InMemoryDataOperationEnum::Set,
|
||||
minimum_occupancy,
|
||||
TailNumber::Odd>;
|
||||
Run(kernel);
|
||||
}
|
||||
else
|
||||
{
|
||||
throw std::runtime_error("wrong! Invalid pipeline v3 setting");
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
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::is_gfx11_supported() && !ck::is_gfx12_supported())
|
||||
{
|
||||
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
|
||||
{
|
||||
std::cout << "Device implementation supports only gfx11 and gfx12! " << __FILE__
|
||||
<< ":" << __LINE__ << ", in function: " << __func__ << std::endl;
|
||||
}
|
||||
return false;
|
||||
}
|
||||
|
||||
if constexpr(std::is_same_v<ComputeTypeA, f8_t> || std::is_same_v<ComputeTypeA, bf8_t> ||
|
||||
std::is_same_v<ComputeTypeB, f8_t> || std::is_same_v<ComputeTypeB, bf8_t>)
|
||||
{
|
||||
if(ck::is_gfx11_supported())
|
||||
{
|
||||
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
|
||||
{
|
||||
std::cout << "FP8 and BF8 not supported on gfx11! " << __FILE__ << ":"
|
||||
<< __LINE__ << ", in function: " << __func__ << std::endl;
|
||||
}
|
||||
return false;
|
||||
}
|
||||
}
|
||||
|
||||
if((arg.KRaw_ % AK1 != 0 || arg.KRaw_ % BK1 != 0) &&
|
||||
!(GemmSpec == GemmSpecialization::MKPadding ||
|
||||
GemmSpec == GemmSpecialization::NKPadding ||
|
||||
GemmSpec == GemmSpecialization::MNKPadding ||
|
||||
GemmSpec == GemmSpecialization::KPadding))
|
||||
{
|
||||
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
|
||||
{
|
||||
std::cout << "Without padding, K must be divisible by AK1 and BK1! " << __FILE__
|
||||
<< ":" << __LINE__ << ", in function: " << __func__ << std::endl;
|
||||
}
|
||||
return false;
|
||||
}
|
||||
|
||||
typename GridwiseGemm::Argument gemm_arg{
|
||||
std::array<const void*, 1>{arg.p_a_grid_},
|
||||
std::array<const void*, 1>{arg.p_b_grid_},
|
||||
std::array<const void*, 2>{arg.p_bias_grid_, arg.p_d0_grid_},
|
||||
static_cast<EDataType*>(arg.p_e_grid_),
|
||||
arg.MRaw_,
|
||||
arg.NRaw_,
|
||||
arg.KRaw_,
|
||||
std::array<index_t, 1>{arg.StrideA_}, // StrideAs
|
||||
std::array<index_t, 1>{arg.StrideB_}, // StrideBs
|
||||
std::array<index_t, 2>{0, arg.StrideC1_}, // StrideDs
|
||||
arg.StrideC_, // StrideE
|
||||
1, // kbatch
|
||||
arg.a_element_op_,
|
||||
arg.b_element_op_,
|
||||
arg.c_element_op_};
|
||||
|
||||
return GridwiseGemm::CheckValidity(gemm_arg);
|
||||
}
|
||||
|
||||
// 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<EDataType*>(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<EDataType*>(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_Wmma_CShuffleV3"
|
||||
<< "<"
|
||||
<< BlockSize << ", "
|
||||
<< MPerBlock << ", "
|
||||
<< NPerBlock << ", "
|
||||
<< KPerBlock << ", "
|
||||
<< AK1 << ", "
|
||||
<< BK1 << ", "
|
||||
<< MPerWmma << ", "
|
||||
<< NPerWmma << ", "
|
||||
<< MRepeat << ", "
|
||||
<< NRepeat << ", "
|
||||
<< ABlockTransferSrcScalarPerVector << ", "
|
||||
<< BBlockTransferSrcScalarPerVector << ", "
|
||||
<< CShuffleMRepeatPerShuffle << ", "
|
||||
<< CShuffleNRepeatPerShuffle
|
||||
<< ">";
|
||||
// clang-format on
|
||||
|
||||
return str.str();
|
||||
}
|
||||
};
|
||||
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
@@ -49,8 +49,11 @@ __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, MinimumOccupancy)
|
||||
|
||||
auto splitk_batch_offset = typename GridwiseGemm::SplitKBatchOffset(karg, blockIdx.z);
|
||||
|
||||
auto epilogue_args =
|
||||
EpilogueType(p_reduces_grid, reduce_in_element_ops, reduce_out_element_ops, karg.M);
|
||||
auto epilogue_args = EpilogueType(p_reduces_grid,
|
||||
reduce_in_element_ops,
|
||||
reduce_out_element_ops,
|
||||
karg.M,
|
||||
tensor_operation::element_wise::PassThrough{});
|
||||
|
||||
GridwiseGemm::template Run<HasMainKBlockLoop, EGlobalMemoryDataOperation, TailNum>(
|
||||
p_shared, splitk_batch_offset, karg, epilogue_args);
|
||||
@@ -188,6 +191,7 @@ struct DeviceGemmReduce_Wmma_CShuffleV3 : public DeviceGemmReduce<0, ReduceOpera
|
||||
|
||||
using ReduceTrait = ReduceTrait_<ReduceAccDataType,
|
||||
ReducePtrsGlobal,
|
||||
tensor_operation::element_wise::PassThrough,
|
||||
ReduceOperations,
|
||||
ReduceInElementwiseOperations,
|
||||
ReduceAccElementwiseOperations,
|
||||
|
||||
@@ -10,6 +10,7 @@ namespace ck {
|
||||
|
||||
template <typename ReduceAccDataType,
|
||||
typename ReducePtrsGlobal,
|
||||
typename D0ElementwiseOperation,
|
||||
typename ReduceOperations,
|
||||
typename ReduceInElementwiseOperations,
|
||||
typename ReduceAccElementwiseOperations,
|
||||
@@ -21,6 +22,7 @@ struct ReduceTrait_
|
||||
{
|
||||
using ReduceAccDataType_ = ReduceAccDataType;
|
||||
using ReducePtrsGlobal_ = ReducePtrsGlobal;
|
||||
using D0ElementwiseOperation_ = D0ElementwiseOperation;
|
||||
using ReduceOperations_ = ReduceOperations;
|
||||
using ReduceInElementwiseOperations_ = ReduceInElementwiseOperations;
|
||||
using ReduceAccElementwiseOperations_ = ReduceAccElementwiseOperations;
|
||||
@@ -148,11 +150,13 @@ struct EpilogueReduceCShuffle
|
||||
typename ReduceTrait::ReducePtrsGlobal_ p_reduces_grid_,
|
||||
const typename ReduceTrait::ReduceInElementwiseOperations_ reduce_in_element_ops_,
|
||||
const typename ReduceTrait::ReduceAccElementwiseOperations_ reduce_out_element_ops_,
|
||||
const index_t MRaw_)
|
||||
const index_t MRaw_,
|
||||
const typename ReduceTrait::D0ElementwiseOperation_ d0_element_op_)
|
||||
: p_reduces_grid(p_reduces_grid_),
|
||||
reduce_in_element_ops(reduce_in_element_ops_),
|
||||
reduce_out_element_ops(reduce_out_element_ops_),
|
||||
MRaw(MRaw_),
|
||||
d0_element_op{d0_element_op_},
|
||||
reduce_grid_desc_m{MakeReduceGridDescriptor_M(MRaw)}
|
||||
{
|
||||
}
|
||||
@@ -174,6 +178,13 @@ struct EpilogueReduceCShuffle
|
||||
const index_t& block_m_id,
|
||||
const index_t& block_n_id)
|
||||
{
|
||||
// HACK: this force m/n_block_data_idx_on_grid into SGPR
|
||||
const index_t m_block_data_idx_on_grid =
|
||||
__builtin_amdgcn_readfirstlane(block_m_id * MPerBlock);
|
||||
|
||||
const index_t n_block_data_idx_on_grid =
|
||||
__builtin_amdgcn_readfirstlane(block_n_id * NPerBlock);
|
||||
|
||||
auto reduce_grid_desc_mblock_mperblock =
|
||||
MakeReduceGridDescriptor_MBlock_MPerBlock(reduce_grid_desc_m);
|
||||
|
||||
@@ -216,29 +227,6 @@ struct EpilogueReduceCShuffle
|
||||
c_block_desc_mrepeat_mwave_msubgroup_nrepeat_nwave_nthreadpersubgroup_maccvgprs =
|
||||
GetCShuffleLDSDescriptor();
|
||||
|
||||
// tuple of reference to C/Ds tensor descriptors
|
||||
const auto c_ds_desc_refs = concat_tuple_of_reference(
|
||||
tie(c_shuffle_block_desc_mshrepeat_mpershrepeat_nshrepeat_npershrepeat),
|
||||
generate_tie([&](auto i) -> const auto& // return type should be reference
|
||||
{ return ds_grid_desc_mblock_mperblock_nblock_nperblock[i]; },
|
||||
Number<NumDTensor>{}));
|
||||
|
||||
// Thread transfer LDS to Vmem
|
||||
auto cde_shuffle_block_copy_lds_and_global =
|
||||
Base::template GetLDSToVmemEpilogueDescriptor<EGlobalMemoryDataOperation, EDataType>(
|
||||
c_ds_desc_refs,
|
||||
e_grid_desc_mblock_mperblock_nblock_nperblock,
|
||||
cde_element_op,
|
||||
block_m_id,
|
||||
block_n_id);
|
||||
|
||||
// tuple of reference to C/Ds tensor buffers
|
||||
const auto c_ds_buf_refs = concat_tuple_of_reference(
|
||||
tie(c_shuffle_block_buf),
|
||||
generate_tie([&](auto i) -> const auto& // return type should be reference
|
||||
{ return ds_grid_buf[i]; },
|
||||
Number<NumDTensor>{}));
|
||||
|
||||
// LDS c_reduce_block_desc_mperblock_nperblock
|
||||
constexpr auto c_reduce_block_desc_mperblock_nperblock = transform_tensor_descriptor(
|
||||
c_shuffle_block_desc_mshrepeat_mpershrepeat_nshrepeat_npershrepeat,
|
||||
@@ -346,6 +334,68 @@ struct EpilogueReduceCShuffle
|
||||
},
|
||||
Number<NumReduce>{});
|
||||
|
||||
// multiple Ds
|
||||
constexpr auto d_reduce_thread_desc_mblock_mperblock_nblock_nperblock =
|
||||
make_naive_tensor_descriptor_packed(
|
||||
make_tuple(I1, Number<mreduce_per_thread>{}, I1, Number<nreduce_per_thread>{}));
|
||||
|
||||
constexpr auto ds_reduce_thread_desc_mblock_mperblock_nblock_nperblock = generate_tuple(
|
||||
[&](auto) { return d_reduce_thread_desc_mblock_mperblock_nblock_nperblock; },
|
||||
Number<NumDTensor>{});
|
||||
|
||||
constexpr auto ds_thread_buf_size =
|
||||
d_reduce_thread_desc_mblock_mperblock_nblock_nperblock.GetElementSpaceSize();
|
||||
|
||||
auto c01_thread_buf =
|
||||
make_static_buffer<AddressSpaceEnum::Vgpr, typename ReduceTrait::ReduceAccDataType_>(
|
||||
Number<ds_thread_buf_size>{});
|
||||
|
||||
auto ds_thread_copy_global_to_vgpr = generate_tuple(
|
||||
[&](auto I) {
|
||||
return ThreadwiseTensorSliceTransfer_v2<
|
||||
remove_cvref_t<tuple_element_t<I.value, DsDataType>>,
|
||||
typename ReduceTrait::ReduceAccDataType_,
|
||||
decltype(ds_grid_desc_mblock_mperblock_nblock_nperblock[I]),
|
||||
remove_cvref_t<
|
||||
decltype(ds_reduce_thread_desc_mblock_mperblock_nblock_nperblock[I])>,
|
||||
Sequence<I1, mreduce_per_thread, I1, nreduce_per_thread>,
|
||||
Sequence<0, 1, 2, 3>,
|
||||
3,
|
||||
ReduceTrait::CReduceThreadLds2VGprCopySrcDstScalarPerVector_NPerBlock_,
|
||||
1,
|
||||
true>(ds_grid_desc_mblock_mperblock_nblock_nperblock[I],
|
||||
make_multi_index(
|
||||
I0,
|
||||
m_block_data_idx_on_grid + c_reduce_thread_data_idx_begin[I0],
|
||||
I0,
|
||||
n_block_data_idx_on_grid + c_reduce_thread_data_idx_begin[I1]));
|
||||
},
|
||||
Number<NumDTensor>{});
|
||||
|
||||
constexpr auto c_reduce_thread_desc_mblock_mperblock_nblock_nperblock =
|
||||
make_naive_tensor_descriptor_packed(
|
||||
make_tuple(I1, Number<mreduce_per_thread>{}, I1, Number<nreduce_per_thread>{}));
|
||||
|
||||
// Write E from Vgpr to Vmem
|
||||
auto c_reduce_thread_copy_vgpr_to_global = ThreadwiseTensorSliceTransfer_v1r3<
|
||||
typename ReduceTrait::ReduceAccDataType_,
|
||||
EDataType,
|
||||
decltype(c_reduce_thread_desc_mblock_mperblock_nblock_nperblock),
|
||||
decltype(e_grid_desc_mblock_mperblock_nblock_nperblock),
|
||||
tensor_operation::element_wise::PassThrough,
|
||||
Sequence<I1, mreduce_per_thread, I1, nreduce_per_thread>, // SliceLengths
|
||||
Sequence<0, 1, 2, 3>, // DimAccessOrder
|
||||
3, // DstVectorDim
|
||||
ReduceTrait::CReduceThreadLds2VGprCopySrcDstScalarPerVector_NPerBlock_,
|
||||
EGlobalMemoryDataOperation,
|
||||
1,
|
||||
true>{e_grid_desc_mblock_mperblock_nblock_nperblock,
|
||||
make_multi_index(I0,
|
||||
m_block_data_idx_on_grid + c_reduce_thread_data_idx_begin[I0],
|
||||
I0,
|
||||
n_block_data_idx_on_grid + c_reduce_thread_data_idx_begin[I1]),
|
||||
NumDTensor > 0 ? tensor_operation::element_wise::PassThrough{} : cde_element_op};
|
||||
|
||||
constexpr index_t num_access = sfc_c_vgpr.GetNumOfAccess();
|
||||
|
||||
static_assert(num_access == sfc_cde_global.GetNumOfAccess(), "wrong!");
|
||||
@@ -365,15 +415,6 @@ struct EpilogueReduceCShuffle
|
||||
|
||||
// make sure it's safe to read from LDS
|
||||
block_sync_lds();
|
||||
|
||||
// each block loads its C data from LDS, D from global, applies elementwise
|
||||
// operation and stores result E to global
|
||||
cde_shuffle_block_copy_lds_and_global.Run(
|
||||
c_ds_desc_refs,
|
||||
c_ds_buf_refs,
|
||||
tie(e_grid_desc_mblock_mperblock_nblock_nperblock),
|
||||
tie(e_grid_buf));
|
||||
|
||||
{
|
||||
c_reduce_thread_copy_lds_to_vgpr.Run(c_reduce_block_desc_mperblock_nperblock,
|
||||
c_shuffle_block_buf,
|
||||
@@ -381,6 +422,53 @@ struct EpilogueReduceCShuffle
|
||||
make_tuple(I0, I0),
|
||||
c_reduce_thread_buf);
|
||||
|
||||
// Note: currently multiple Ds supports only Bias + Add.
|
||||
// It needs to be generalized for other operations (currently not needed)
|
||||
if constexpr(NumDTensor > 0)
|
||||
{
|
||||
auto& d0_thread_copy_global_to_vgpr = ds_thread_copy_global_to_vgpr(I0);
|
||||
// d0 / d1 operations
|
||||
d0_thread_copy_global_to_vgpr.Run(
|
||||
ds_grid_desc_mblock_mperblock_nblock_nperblock[I0],
|
||||
ds_grid_buf[I0],
|
||||
ds_reduce_thread_desc_mblock_mperblock_nblock_nperblock[I0],
|
||||
make_tuple(I0, I0, I0, I0),
|
||||
c01_thread_buf);
|
||||
|
||||
// c = activation(c + bias)
|
||||
static_for<0, c_reduce_thread_desc_mperblock_nperblock.GetElementSize(), 1>{}(
|
||||
[&](auto i) {
|
||||
typename ReduceTrait::ReduceAccDataType_ out;
|
||||
cde_element_op(out, c_reduce_thread_buf(i) + c01_thread_buf(i));
|
||||
c_reduce_thread_buf(i) = out;
|
||||
});
|
||||
|
||||
auto& d1_thread_copy_global_to_vgpr = ds_thread_copy_global_to_vgpr(I1);
|
||||
|
||||
d1_thread_copy_global_to_vgpr.Run(
|
||||
ds_grid_desc_mblock_mperblock_nblock_nperblock[I1],
|
||||
ds_grid_buf[I1],
|
||||
ds_reduce_thread_desc_mblock_mperblock_nblock_nperblock[I1],
|
||||
make_tuple(I0, I0, I0, I0),
|
||||
c01_thread_buf);
|
||||
|
||||
// c = c + c1_function(c1)
|
||||
static_for<0, c_reduce_thread_desc_mperblock_nperblock.GetElementSize(), 1>{}(
|
||||
[&](auto i) {
|
||||
d0_element_op(c01_thread_buf(i), c01_thread_buf(i));
|
||||
c_reduce_thread_buf(i) += c01_thread_buf(i);
|
||||
});
|
||||
}
|
||||
|
||||
// Write E
|
||||
c_reduce_thread_copy_vgpr_to_global.Run(
|
||||
c_reduce_thread_desc_mblock_mperblock_nblock_nperblock,
|
||||
make_tuple(I0, I0, I0, I0),
|
||||
c_reduce_thread_buf,
|
||||
e_grid_desc_mblock_mperblock_nblock_nperblock,
|
||||
e_grid_buf);
|
||||
|
||||
// Reduction
|
||||
static_for<0, NumReduce, 1>{}([&](auto In) {
|
||||
auto& p_reduce_grid = p_reduces_grid[In];
|
||||
|
||||
@@ -448,14 +536,15 @@ struct EpilogueReduceCShuffle
|
||||
{
|
||||
constexpr auto cde_global_step = sfc_cde_global.GetForwardStep(access_id);
|
||||
// move on Ds
|
||||
static_for<0, NumDTensor, 1>{}([&](auto i) {
|
||||
cde_shuffle_block_copy_lds_and_global.MoveSrcSliceWindow(
|
||||
c_ds_desc_refs, i + I1, cde_global_step);
|
||||
static_for<0, NumDTensor, 1>{}([&](auto I) {
|
||||
auto& d_thread_copy_global_to_vgpr = ds_thread_copy_global_to_vgpr(I);
|
||||
d_thread_copy_global_to_vgpr.MoveSrcSliceWindow(
|
||||
ds_grid_desc_mblock_mperblock_nblock_nperblock[I], cde_global_step);
|
||||
});
|
||||
|
||||
// move on E
|
||||
cde_shuffle_block_copy_lds_and_global.MoveDstSliceWindow(
|
||||
tie(e_grid_desc_mblock_mperblock_nblock_nperblock), cde_global_step);
|
||||
c_reduce_thread_copy_vgpr_to_global.MoveDstSliceWindow(
|
||||
e_grid_desc_mblock_mperblock_nblock_nperblock, cde_global_step);
|
||||
}
|
||||
});
|
||||
}
|
||||
@@ -464,6 +553,7 @@ struct EpilogueReduceCShuffle
|
||||
typename ReduceTrait::ReduceInElementwiseOperations_ reduce_in_element_ops;
|
||||
typename ReduceTrait::ReduceAccElementwiseOperations_ reduce_out_element_ops;
|
||||
index_t MRaw;
|
||||
typename ReduceTrait::D0ElementwiseOperation_ d0_element_op;
|
||||
ReduceGridDesc_M reduce_grid_desc_m;
|
||||
};
|
||||
|
||||
|
||||
@@ -897,6 +897,8 @@ struct GridwiseGemmBiasAddReduce_k0mk1_k0nk1_mn_xdl_cshuffle_v1
|
||||
static_assert(num_access == sfc_c_global.GetNumOfAccess(), "wrong!");
|
||||
|
||||
static_for<0, num_access, 1>{}([&](auto access_id) {
|
||||
block_sync_lds();
|
||||
|
||||
// each thread write its data from VGPR to LDS
|
||||
c_thread_copy_vgpr_to_lds.Run(c_thread_desc_m0_n0_m1_n1_m2_m3_m4_n2,
|
||||
sfc_c_vgpr.GetIndexTupleOfNumber(access_id),
|
||||
|
||||
Reference in New Issue
Block a user