mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-07-08 16:17:16 +00:00
Code update of template parameters modified.
This commit is contained in:
apoorva
@@ -1,5 +1,5 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#pragma once
|
||||
|
||||
@@ -56,7 +56,7 @@ void add_device_gemm_add_wmma_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_instances(
|
||||
PassThrough,
|
||||
Add>>>&);
|
||||
|
||||
void add_device_gemm_add_wmma_c_shuffle_bf16_f16_bf16_bf16_mk_kn_mn_mn_instances(
|
||||
void add_device_gemm_add_wmma_c_shuffle_bf16_bf16_bf16_bf16_mk_kn_mn_mn_instances(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleD<Row,
|
||||
Row,
|
||||
Row_Tuple,
|
||||
|
||||
@@ -3,7 +3,7 @@ add_instance_library(device_gemm_add_instance
|
||||
device_gemm_add_xdl_c_shuffle_f16_i8_f16_f16_mk_kn_mn_mn_instance.cpp
|
||||
device_gemm_add_xdl_c_shuffle_bf16_i8_bf16_bf16_mk_kn_mn_mn_instance.cpp
|
||||
device_gemm_add_wmma_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_instance.cpp
|
||||
device_gemm_add_wmma_c_shuffle_bf16_f16_bf16_bf16_mk_kn_mn_mn_instance.cpp
|
||||
device_gemm_add_wmma_c_shuffle_bf16_bf16_bf16_bf16_mk_kn_mn_mn_instance.cpp
|
||||
device_gemm_add_wmma_c_shuffle_v3_f16_f16_f16_f16_mk_kn_mn_mn_instance.cpp
|
||||
device_gemm_add_wmma_c_shuffle_v3_bf16_f16_bf16_bf16_mk_kn_mn_mn_instance.cpp
|
||||
device_gemm_add_wmma_c_shuffle_v3_bf16_bf16_bf16_bf16_mk_kn_mn_mn_instance.cpp
|
||||
)
|
||||
|
||||
@@ -20,7 +20,7 @@ static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecial
|
||||
// e = elementwise((a * b), d0, d1)
|
||||
// outout: e[m, n]
|
||||
// input: a[m, k], b[k, n], d0[m, n], d1[m, n]
|
||||
using device_gemm_add_wmma_c_shuffle_bf16_f16_bf16_bf16_mk_kn_mn_mn_generic_instances = std::tuple<
|
||||
using device_gemm_add_wmma_c_shuffle_bf16_bf16_bf16_bf16_mk_kn_mn_mn_generic_instances = std::tuple<
|
||||
// clang-format off
|
||||
// M/N/K padding
|
||||
//################################| A| B| Ds| E| AData| BData| AccData| CShuffle| DsData| EData| A| B| CDE| GEMM| Prefetch| Block| MPer| NPer| K0Per| K1| MPer| NPer| MRepeat| NRepeat| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer|
|
||||
@@ -32,7 +32,7 @@ using device_gemm_add_wmma_c_shuffle_bf16_f16_bf16_bf16_mk_kn_mn_mn_generic_inst
|
||||
// clang-format on
|
||||
>;
|
||||
|
||||
using add_device_gemm_add_wmma_c_shuffle_bf16_f16_bf16_bf16_mk_kn_mn_mn_instances = std::tuple<
|
||||
using add_device_gemm_add_wmma_c_shuffle_bf16_bf16_bf16_bf16_mk_kn_mn_mn_instances = std::tuple<
|
||||
// clang-format off
|
||||
// M/N/K padding
|
||||
//################################| A| B| Ds| E| AData| BData| AccData| CShuffle| DsData| EData| A| B| CDE| GEMM| Prefetch| Block| MPer| NPer| K0Per| K1| MPer| NPer| MRepeat| NRepeat| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer|
|
||||
@@ -46,7 +46,7 @@ using add_device_gemm_add_wmma_c_shuffle_bf16_f16_bf16_bf16_mk_kn_mn_mn_instance
|
||||
// clang-format on
|
||||
>;
|
||||
|
||||
void device_gemm_add_wmma_c_shuffle_bf16_f16_bf16_bf16_mk_kn_mn_mn_instances(
|
||||
void device_gemm_add_wmma_c_shuffle_bf16_bf16_bf16_bf16_mk_kn_mn_mn_instances(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleD<Row,
|
||||
Row,
|
||||
Row_Tuple,
|
||||
@@ -61,9 +61,9 @@ void device_gemm_add_wmma_c_shuffle_bf16_f16_bf16_bf16_mk_kn_mn_mn_instances(
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_add_wmma_c_shuffle_bf16_f16_bf16_bf16_mk_kn_mn_mn_generic_instances{});
|
||||
device_gemm_add_wmma_c_shuffle_bf16_bf16_bf16_bf16_mk_kn_mn_mn_generic_instances{});
|
||||
add_device_operation_instances(
|
||||
instances, add_device_gemm_add_wmma_c_shuffle_bf16_f16_bf16_bf16_mk_kn_mn_mn_instances{});
|
||||
instances, add_device_gemm_add_wmma_c_shuffle_bf16_bf16_bf16_bf16_mk_kn_mn_mn_instances{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
@@ -1,5 +1,5 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
|
||||
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
|
||||
#include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
|
||||
@@ -28,7 +28,7 @@ using device_gemm_add_wmma_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_generic_instanc
|
||||
//################################| | | | | | | | | | | Operation| Operation| Operation| | | | | | | | | | | | Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerXdl| _NWaveNPerXdl|
|
||||
//################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
|
||||
// TODO: these template variables need to be adjusted
|
||||
DeviceGemmMultipleD_Wmma_CShuffle< Row, Row, Row_Tuple, Row, F16, F16, F32, F32,F16_Tuple, F16, PassThrough, PassThrough, Add, GemmMNKPadding, 1, 256, 64, 128, 32, 8, 16, 16, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 1, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 1, 1, 1, S<1, 32, 1, 8>, 1>
|
||||
DeviceGemmMultipleD_Wmma_CShuffle< Row, Row, Row_Tuple, Row, F16, F16, F32, F32,F16_Tuple, F16, PassThrough, PassThrough, Add, GemmMNKPadding, 1, 256, 128, 128, 64, 8, 16, 16, 4, 2, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 2, S<1, 32, 1, 8>, 8>
|
||||
// clang-format on
|
||||
>;
|
||||
|
||||
@@ -40,9 +40,9 @@ using device_gemm_add_wmma_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_instances = std
|
||||
//################################| | | | | | | | | | | Operation| Operation| Operation| | | | | | | | | | | | Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerXdl| _NWaveNPerXdl|
|
||||
//################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
|
||||
// TODO: these template variables need to be adjusted
|
||||
DeviceGemmMultipleD_Wmma_CShuffle< Row, Row, Row_Tuple, Row, F16, F16, F32, F32,F16_Tuple, F16, PassThrough, PassThrough, Add, GemmMNKPadding, 1, 256, 16, 128, 32, 8, 16, 16, 1, 2, S<4, 16, 4>, S<1, 0, 2>, S<1, 0, 2>, 2, 2, 2, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 1, 1, 1, S<1, 16, 1, 4>, 1, LoopScheduler::Default, PipelineVersion::v1>,
|
||||
DeviceGemmMultipleD_Wmma_CShuffle< Row, Row, Row_Tuple, Row, F16, F16, F32, F32,F16_Tuple, F16, PassThrough, PassThrough, Add, GemmMNKPadding, 1, 64, 16, 16, 32, 8, 16, 16, 1, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 1, 1, 1, 1, S<1, 16, 1, 4>, 1, LoopScheduler::Default, PipelineVersion::v1>,
|
||||
DeviceGemmMultipleD_Wmma_CShuffle< Row, Row, Row_Tuple, Row, F16, F16, F32, F32,F16_Tuple, F16, PassThrough, PassThrough, Add, GemmMNKPadding, 1, 64, 16, 16, 64, 8, 16, 16, 1, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 1, 1, 1, 1, S<1, 16, 1, 4>, 1, LoopScheduler::Default, PipelineVersion::v1>
|
||||
DeviceGemmMultipleD_Wmma_CShuffle< Row, Row, Row_Tuple, Row, F16, F16, F32, F32,F16_Tuple, F16, PassThrough, PassThrough, Add, GemmMNKPadding, 1, 256, 128, 128, 64, 8, 16, 16, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 2, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
|
||||
DeviceGemmMultipleD_Wmma_CShuffle< Row, Row, Row_Tuple, Row, F16, F16, F32, F32,F16_Tuple, F16, PassThrough, PassThrough, Add, GemmMNKPadding, 1, 128, 64, 64, 64, 8, 16, 16, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 2, S<1, 32, 1, 4>, 8, LoopScheduler::Default, PipelineVersion::v1>,
|
||||
DeviceGemmMultipleD_Wmma_CShuffle< Row, Row, Row_Tuple, Row, F16, F16, F32, F32,F16_Tuple, F16, PassThrough, PassThrough, Add, GemmMNKPadding, 1, 128, 64, 64, 64, 8, 16, 16, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 2, S<1, 32, 1, 4>, 8, LoopScheduler::Default, PipelineVersion::v1>
|
||||
// clang-format on
|
||||
>;
|
||||
|
||||
|
||||
@@ -1,55 +1,44 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2018-2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include <cstdlib>
|
||||
|
||||
#include "ck/ck.hpp"
|
||||
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
|
||||
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
|
||||
#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_wmma_cshuffle_v3.hpp"
|
||||
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
|
||||
#include "ck/utility/blkgemmpipe_scheduler.hpp"
|
||||
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
|
||||
#include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
|
||||
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
|
||||
#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_wmma_cshuffle_v3.hpp"
|
||||
#include "ck/utility/sequence.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
using BF16 = ck::bhalf_t;
|
||||
using F32 = float;
|
||||
using BF16_Tuple = ck::Tuple<BF16, BF16>;
|
||||
|
||||
using Row = ck::tensor_layout::gemm::RowMajor;
|
||||
using Col = ck::tensor_layout::gemm::ColumnMajor;
|
||||
using Row_Tuple = ck::Tuple<Row, Row>;
|
||||
|
||||
template <ck::index_t... Is>
|
||||
using S = ck::Sequence<Is...>;
|
||||
|
||||
using PassThrough = ck::tensor_operation::element_wise::PassThrough;
|
||||
using AddMultiply = ck::tensor_operation::element_wise::AddMultiply;
|
||||
|
||||
static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
|
||||
static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
|
||||
|
||||
static constexpr auto Intrawave = BlockGemmPipelineScheduler::Intrawave;
|
||||
static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
|
||||
static constexpr auto Intrawave = BlockGemmPipelineScheduler::Intrawave;
|
||||
|
||||
using device_gemm_add_wmma_c_shuffle_v3_bf16_bf16_bf16_bf16_bf16_mk_kn_mn_mn_mn_instances =
|
||||
std::tuple<
|
||||
// clang-format off
|
||||
//##################################| ALayout| BLayout| DsLayout| ELayout| AData| BData| DsData| EData| AccData| CShuffle| A| B| CDE| GemmSpec| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MRepeat| NRepeat| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CShuffleBlockTransfer| CDEShuffleBlockTransfer| BlkGemm| BlkGemm|
|
||||
//##################################| | | | | Type| Type| Type| Type| Type| DataType| Elementwise| Elementwise| Elementwise| | Size| Block| Block| Block| | | Wmma| Wmma| | | ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| ExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| ExtraN| MRepeat| NRepeat| ClusterLengths| ScalarPerVectors| PipeSched| PipelineVer|
|
||||
//##################################| | | | | | | | | | | Operation| Operation| Operation| | | | | | | | | | | | Lengths_AK0_M_AK1| ArrangeOrder| | | PerVector| PerVector_AK1| | Lengths_BK0_N_BK1| ArrangeOrder| | | PerVector| PerVector_BK1| | PerShuffle| PerShuffle| _MBlock_MPerBlock| | | |
|
||||
//##################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | _NBlock_NPerBlock| | | |
|
||||
DeviceGemmMultipleD_Wmma_CShuffleV3< Row, Row, Row_Tuple, Row, BF16, BF16, BF16_Tuple, BF16, F32, BF16, PassThrough, PassThrough, AddMultiply, GemmDefault, 256, 128, 128, 64, 8, 8, 16, 16, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, true, 1, 1, S<1, 32, 1, 8>, S<8, 8, 8>, Intrawave, BlockGemmPipelineVersion::v1>,
|
||||
DeviceGemmMultipleD_Wmma_CShuffleV3< Row, Row, Row_Tuple, Row, BF16, BF16, BF16_Tuple, BF16, F32, BF16, PassThrough, PassThrough, AddMultiply, GemmMNKPadding, 256, 128, 128, 64, 8, 8, 16, 16, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, true, 1, 1, S<1, 32, 1, 8>, S<8, 8, 8>, Interwave, BlockGemmPipelineVersion::v1>
|
||||
// clang-format on
|
||||
>;
|
||||
static constexpr auto V1 = BlockGemmPipelineVersion::v1;
|
||||
static constexpr auto V3 = BlockGemmPipelineVersion::v3;
|
||||
|
||||
void add_device_gemm_add_wmma_c_shuffle_bf16_bf16_bf16_bf16_bf16_mk_kn_mn_mn_mn_instances(
|
||||
template <GemmSpecialization GemmSpec>
|
||||
using device_gemm_add_wmma_c_shuffle_bf16_bf16_bf16_bf16_mk_kn_mn_mn_instances = std::tuple<
|
||||
// clang-format off
|
||||
//##################################| ALayout| BLayout| DsLayout| ELayout| AData| BData| DsData| EData| AccData| CShuffle| A| B| CDE| GemmSpec| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MRepeat| NRepeat| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CShuffleBlockTransfer| CDEShuffleBlockTransfer| BlkGemm| BlkGemm|
|
||||
//##################################| | | | | Type| Type| Type| Type| Type| DataType| Elementwise| Elementwise| Elementwise| | Size| Block| Block| Block| | | Wmma| Wmma| | | ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| ExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| ExtraN| MRepeat| NRepeat| ClusterLengths| ScalarPerVectors| PipeSched| PipelineVer|
|
||||
//##################################| | | | | | | | | | | Operation| Operation| Operation| | | | | | | | | | | | Lengths_AK0_M_AK1| ArrangeOrder| | | PerVector| PerVector_AK1| | Lengths_BK0_N_BK1| ArrangeOrder| | | PerVector| PerVector_BK1| | PerShuffle| PerShuffle| _MBlock_MPerBlock| | | |
|
||||
//##################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | _NBlock_NPerBlock| | | |
|
||||
DeviceGemmMultipleD_Wmma_CShuffleV3< Row, Row, Row_Tuple, Row, BF16, BF16, BF16_Tuple, BF16, F32, F32, PassThrough, PassThrough, Add, GemmSpec, 256, 128, 128, 32, 8, 8, 16, 16, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 0, 1, 1, S<1, 32, 1, 8>, S<8, 8, 8>, Interwave, V1>,
|
||||
DeviceGemmMultipleD_Wmma_CShuffleV3< Row, Row, Row_Tuple, Row, BF16, BF16, BF16_Tuple, BF16, F32, F32, PassThrough, PassThrough, Add, GemmSpec, 128, 128, 64, 64, 8, 8, 16, 16, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 0, 1, 1, S<1, 32, 1, 4>, S<8, 8, 8>, Intrawave, V1>,
|
||||
DeviceGemmMultipleD_Wmma_CShuffleV3< Row, Row, Row_Tuple, Row, BF16, BF16, BF16_Tuple, BF16, F32, F32, PassThrough, PassThrough, Add, GemmSpec, 256, 128, 128, 32, 8, 8, 16, 16, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 1, 1, 1, S<1, 32, 1, 8>, S<8, 8, 8>, Intrawave, V3>,
|
||||
DeviceGemmMultipleD_Wmma_CShuffleV3< Row, Row, Row_Tuple, Row, BF16, BF16, BF16_Tuple, BF16, F32, F32, PassThrough, PassThrough, Add, GemmSpec, 128, 64, 64, 32, 8, 8, 16, 16, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 4, 1, 1, 1, S<1, 32, 1, 2>, S<8, 8, 8>, Intrawave, V3>
|
||||
// clang-format on
|
||||
>;
|
||||
|
||||
void add_device_gemm_add_wmma_c_shuffle_bf16_bf16_bf16_bf16_mk_kn_mn_mn_instances(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitK<Row,
|
||||
Row,
|
||||
Row_Tuple,
|
||||
@@ -60,14 +49,17 @@ void add_device_gemm_add_wmma_c_shuffle_bf16_bf16_bf16_bf16_bf16_mk_kn_mn_mn_mn_
|
||||
BF16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
AddMultiply>>>& instances)
|
||||
Add>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_add_wmma_c_shuffle_v3_bf16_bf16_bf16_bf16_bf16_mk_kn_mn_mn_mn_instances{});
|
||||
device_gemm_add_wmma_c_shuffle_bf16_bf16_bf16_bf16_mk_kn_mn_mn_instances<GemmDefault>{});
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_add_wmma_c_shuffle_bf16_bf16_bf16_bf16_mk_kn_mn_mn_instances<GemmMNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
} // namespace ck
|
||||
|
||||
@@ -1,54 +1,44 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
// Copyright (c) 2018-2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include <cstdlib>
|
||||
|
||||
#include "ck/ck.hpp"
|
||||
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
|
||||
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
|
||||
#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_wmma_cshuffle_v3.hpp"
|
||||
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
|
||||
#include "ck/utility/blkgemmpipe_scheduler.hpp"
|
||||
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
|
||||
|
||||
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
|
||||
#include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
|
||||
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
|
||||
#include "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_wmma_cshuffle_v3.hpp"
|
||||
#include "ck/utility/sequence.hpp"
|
||||
|
||||
namespace ck {
|
||||
namespace tensor_operation {
|
||||
namespace device {
|
||||
namespace instance {
|
||||
|
||||
using F16 = ck::half_t;
|
||||
using F32 = float;
|
||||
using F16_Tuple = ck::Tuple<F16, F16>;
|
||||
|
||||
using Row = ck::tensor_layout::gemm::RowMajor;
|
||||
using Col = ck::tensor_layout::gemm::ColumnMajor;
|
||||
using Row_Tuple = ck::Tuple<Row, Row>;
|
||||
|
||||
template <ck::index_t... Is>
|
||||
using S = ck::Sequence<Is...>;
|
||||
|
||||
using PassThrough = ck::tensor_operation::element_wise::PassThrough;
|
||||
using AddMultiply = ck::tensor_operation::element_wise::AddMultiply;
|
||||
|
||||
static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
|
||||
static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
|
||||
|
||||
static constexpr auto Intrawave = BlockGemmPipelineScheduler::Intrawave;
|
||||
static constexpr auto Interwave = BlockGemmPipelineScheduler::Interwave;
|
||||
static constexpr auto Intrawave = BlockGemmPipelineScheduler::Intrawave;
|
||||
|
||||
using device_gemm_add_wmma_c_shuffle_v3_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_instances = std::tuple<
|
||||
static constexpr auto V1 = BlockGemmPipelineVersion::v1;
|
||||
static constexpr auto V3 = BlockGemmPipelineVersion::v3;
|
||||
|
||||
template <GemmSpecialization GemmSpec>
|
||||
using device_gemm_add_wmma_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_instances = std::tuple<
|
||||
// clang-format off
|
||||
//##################################| ALayout| BLayout| DsLayout| ELayout| AData| BData| DsData| EData| AccData| CShuffle| A| B| CDE| GemmSpec| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MRepeat| NRepeat| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CShuffleBlockTransfer| CDEShuffleBlockTransfer| BlkGemm| BlkGemm|
|
||||
//##################################| | | | | Type| Type| Type| Type| Type| DataType| Elementwise| Elementwise| Elementwise| | Size| Block| Block| Block| | | Wmma| Wmma| | | ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| ExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| ExtraN| MRepeat| NRepeat| ClusterLengths| ScalarPerVectors| PipeSched| PipelineVer|
|
||||
//##################################| | | | | | | | | | | Operation| Operation| Operation| | | | | | | | | | | | Lengths_AK0_M_AK1| ArrangeOrder| | | PerVector| PerVector_AK1| | Lengths_BK0_N_BK1| ArrangeOrder| | | PerVector| PerVector_BK1| | PerShuffle| PerShuffle| _MBlock_MPerBlock| | | |
|
||||
//##################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | _NBlock_NPerBlock| | | |
|
||||
DeviceGemmMultipleD_Wmma_CShuffleV3< Row, Row, Row_Tuple, Row, F16, F16, F16_Tuple, F16, F32, F16, PassThrough, PassThrough, AddMultiply, GemmDefault, 256, 128, 128, 64, 8, 8, 16, 16, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, true, 1, 1, S<1, 32, 1, 8>, S<8, 8, 8>, Intrawave, BlockGemmPipelineVersion::v1>,
|
||||
DeviceGemmMultipleD_Wmma_CShuffleV3< Row, Row, Row_Tuple, Row, F16, F16, F16_Tuple, F16, F32, F16, PassThrough, PassThrough, AddMultiply, GemmMNKPadding, 256, 128, 128, 64, 8, 8, 16, 16, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, true, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, true, 1, 1, S<1, 32, 1, 8>, S<8, 8, 8>, Interwave, BlockGemmPipelineVersion::v1>
|
||||
//##################################| ALayout| BLayout| DsLayout| ELayout| AData| BData| DsData| EData| AccData| CShuffle| A| B| CDE| GemmSpec| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MRepeat| NRepeat| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CShuffleBlockTransfer| CDEShuffleBlockTransfer| BlkGemm| BlkGemm|
|
||||
//##################################| | | | | Type| Type| Type| Type| Type| DataType| Elementwise| Elementwise| Elementwise| | Size| Block| Block| Block| | | Wmma| Wmma| | | ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| ExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| ExtraN| MRepeat| NRepeat| ClusterLengths| ScalarPerVectors| PipeSched| PipelineVer|
|
||||
//##################################| | | | | | | | | | | Operation| Operation| Operation| | | | | | | | | | | | Lengths_AK0_M_AK1| ArrangeOrder| | | PerVector| PerVector_AK1| | Lengths_BK0_N_BK1| ArrangeOrder| | | PerVector| PerVector_BK1| | PerShuffle| PerShuffle| _MBlock_MPerBlock| | | |
|
||||
//##################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | _NBlock_NPerBlock| | | |
|
||||
DeviceGemmMultipleD_Wmma_CShuffleV3< Row, Row, Row_Tuple, Row, F16, F16, F16_Tuple, F16, F32, F32, PassThrough, PassThrough, Add, GemmSpec, 256, 128, 128, 32, 8, 8, 16, 16, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 0, 1, 1, S<1, 32, 1, 8>, S<8, 8, 8>, Interwave, V1>,
|
||||
DeviceGemmMultipleD_Wmma_CShuffleV3< Row, Row, Row_Tuple, Row, F16, F16, F16_Tuple, F16, F32, F32, PassThrough, PassThrough, Add, GemmSpec, 128, 128, 64, 64, 8, 8, 16, 16, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 0, 1, 1, S<1, 32, 1, 4>, S<8, 8, 8>, Intrawave, V1>,
|
||||
DeviceGemmMultipleD_Wmma_CShuffleV3< Row, Row, Row_Tuple, Row, F16, F16, F16_Tuple, F16, F32, F32, PassThrough, PassThrough, Add, GemmSpec, 256, 128, 128, 32, 8, 8, 16, 16, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 1, 1, 1, S<1, 32, 1, 8>, S<8, 8, 8>, Intrawave, V3>,
|
||||
DeviceGemmMultipleD_Wmma_CShuffleV3< Row, Row, Row_Tuple, Row, F16, F16, F16_Tuple, F16, F32, F32, PassThrough, PassThrough, Add, GemmSpec, 128, 64, 64, 32, 8, 8, 16, 16, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 0, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 4, 1, 1, 1, S<1, 32, 1, 2>, S<8, 8, 8>, Intrawave, V3>
|
||||
// clang-format on
|
||||
>;
|
||||
|
||||
void add_device_gemm_add_wmma_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_instances(
|
||||
void add_device_gemm_add_wmma_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_instances(
|
||||
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitK<Row,
|
||||
Row,
|
||||
Row_Tuple,
|
||||
@@ -59,14 +49,17 @@ void add_device_gemm_add_wmma_c_shuffle_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_insta
|
||||
F16,
|
||||
PassThrough,
|
||||
PassThrough,
|
||||
AddMultiply>>>& instances)
|
||||
Add>>>& instances)
|
||||
{
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_add_wmma_c_shuffle_v3_f16_f16_f16_f16_f16_mk_kn_mn_mn_mn_instances{});
|
||||
device_gemm_add_wmma_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_instances<GemmDefault>{});
|
||||
add_device_operation_instances(
|
||||
instances,
|
||||
device_gemm_add_wmma_c_shuffle_f16_f16_f16_f16_mk_kn_mn_mn_instances<GemmMNKPadding>{});
|
||||
}
|
||||
|
||||
} // namespace instance
|
||||
} // namespace device
|
||||
} // namespace tensor_operation
|
||||
} // namespace ck
|
||||
} // namespace ck
|
||||
|
||||
Reference in New Issue
Block a user