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

Optimizing fp8_fp16 mixedprec gemm (#1150)

Browse Source 
* add delayed cvt

* extend fp16 gemm_splitk instances for fp8_fp16 gemm

* add f8 example

* add 128 kperblk instances for fp8

* add kpb128 instance

* added more instances into kpb128

* clean code

* clean code

* fix

* fix

* fixed

* Update example/35_splitK_gemm/splitK_gemm_xdl_fp16_fp8.cpp

Co-authored-by: Bartłomiej Kocot <barkocot@amd.com>

* Update include/ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer.hpp

Co-authored-by: Bartłomiej Kocot <barkocot@amd.com>

* Update library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_fp8_f16_mk_nk_mn_kpb128_instance.cpp

Co-authored-by: Bartłomiej Kocot <barkocot@amd.com>

---------

Co-authored-by: Jing Zhang <jizha@amd.com>
Co-authored-by: Bartłomiej Kocot <barkocot@amd.com>

[ROCm/composable_kernel commit: 602c4cc0d9]
This commit is contained in:
zjing14
2024-02-12 11:45:42 -06:00
committed by GitHub
parent cc25dc316f
commit 8063a407e7
10 changed files with 370 additions and 131 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
example/35_splitK_gemm/CMakeLists.txt
View File

@@ -10,6 +10,9 @@ foreach(gpu IN LISTS GPU_TARGETS)
add_example_executable(example_splitK_gemm_xdl_fp16 splitK_gemm_xdl_fp16.cpp)
add_example_dependencies(example_splitK_gemm_xdl example_splitK_gemm_xdl_fp16)
add_example_executable(example_splitK_gemm_xdl_fp16_fp8 splitK_gemm_xdl_fp16_fp8.cpp)
add_example_dependencies(example_splitK_gemm_xdl example_splitK_gemm_xdl_fp16_fp8)
add_example_executable(example_splitK_gemm_xdl_lds_direct_load_fp16 splitK_gemm_xdl_lds_direct_load_fp16.cpp)
add_example_dependencies(example_splitK_gemm_xdl example_splitK_gemm_xdl_lds_direct_load_fp16)

60
example/35_splitK_gemm/splitK_gemm_xdl_fp16_fp8.cpp Normal file
View File

@@ -0,0 +1,60 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2024 Advanced Micro Devices, Inc. All rights reserved.
#include <iostream>
#include <numeric>
#include <initializer_list>
#include <cstdlib>
#include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_splitk_c_shuffle.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/utility/check_err.hpp"
#include "ck/library/utility/device_memory.hpp"
#include "ck/library/utility/host_tensor.hpp"
#include "ck/library/utility/host_tensor_generator.hpp"
#include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
#include "ck/library/utility/literals.hpp"
template <ck::index_t... Is>
using S = ck::Sequence<Is...>;
using F8 = ck::f8_t;
using F16 = ck::half_t;
using F32 = float;
using Row = ck::tensor_layout::gemm::RowMajor;
using Col = ck::tensor_layout::gemm::ColumnMajor;
using PassThrough = ck::tensor_operation::element_wise::PassThrough;
using ADataType = F16;
using BDataType = F8;
using AccDataType = F32;
using CDataType = F16;
using ALayout = Row;
using BLayout = Col;
using CLayout = Row;
using AElementOp = PassThrough;
using BElementOp = PassThrough;
using CElementOp = PassThrough;
static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
using DeviceGemmInstance = ck::tensor_operation::device::DeviceGemmXdlSplitKCShuffle
// clang-format off
//######| AData| BData| CData| AccData| ALayout| BLayout| CLayout| A| B| C| GEMM| Block| MPer| NPer| KPer| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer|
//######| Type| Type| Type| Type| | | | Elementwise| Elementwise| Elementwise| Spacialization| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MXdlPerWave_MWaveMPerXdl| ScalarPerVector|
//######| | | | | | | | Operation| Operation| Operation| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NXdlPerWave_NWaveNPerXdl| _NWaveNPerXdl|
//######| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementOp, BElementOp, CElementOp, GemmDefault, 128, 16, 64, 8, 16, 16, 16, 1, 2, S<1, 8, 8, 2>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 8, 8, true, S<1, 8, 16, 1>, S<0, 2, 1, 3>, S<0, 2, 1, 3>, 3, 16, 16, true, 1, 1, S<1, 16, 1, 8>, 4, F16, ck::PipelineVersion::v1, ck::LoopScheduler::Default, ADataType, BDataType>;
// clang-format on
#include "run_splitK_gemm_example.inc"
int main(int argc, char* argv[]) { return !run_splitK_gemm_example(argc, argv); }