ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-07-17 09:08:35 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'develop' into amd-develop

Browse Source
This commit is contained in:
Jun Liu
2024-02-29 12:58:44 -08:00
parent b30d416c97 a776978cbe
commit 6d9a07d799
195 changed files with 6614 additions and 1422 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
Dockerfile
View File

@@ -44,7 +44,6 @@ ENV PATH=$PATH:${SCCACHE_INSTALL_LOCATION}
RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-unauthenticated \
build-essential \
cmake \
ccache \
git \
hip-rocclr \
iputils-ping \
@@ -74,6 +73,10 @@ RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-
apt-get clean && \
rm -rf /var/lib/apt/lists/*
#Install latest ccache
RUN git clone https://github.com/ccache/ccache.git && \
cd ccache && mkdir build && cd build && cmake .. && make install
#Install ninja build tracing tools
RUN wget -qO /usr/local/bin/ninja.gz https://github.com/ninja-build/ninja/releases/latest/download/ninja-linux.zip
RUN gunzip /usr/local/bin/ninja.gz
@@ -111,7 +114,7 @@ ENV LANG=C.UTF-8
RUN groupadd -f render
# Install the new rocm-cmake version
RUN git clone -b master https://github.com/RadeonOpenCompute/rocm-cmake.git && \
RUN git clone -b master https://github.com/ROCm/rocm-cmake.git && \
cd rocm-cmake && mkdir build && cd build && \
cmake .. && cmake --build . && cmake --build . --target install
@@ -123,7 +126,7 @@ RUN sh -c "echo compiler version = '$compiler_version'"
RUN sh -c "echo compiler commit = '$compiler_commit'"
RUN if ( [ "$compiler_version" = "amd-staging" ] || [ "$compiler_version" = "amd-mainline-open" ] ) && [ "$compiler_commit" = "" ]; then \
git clone -b "$compiler_version" https://github.com/RadeonOpenCompute/llvm-project.git && \
git clone -b "$compiler_version" https://github.com/ROCm/llvm-project.git && \
cd llvm-project && mkdir build && cd build && \
cmake -DCMAKE_INSTALL_PREFIX=/opt/rocm/llvm -DCMAKE_BUILD_TYPE=Release -DLLVM_ENABLE_ASSERTIONS=1 -DLLVM_TARGETS_TO_BUILD="AMDGPU;X86" -DLLVM_ENABLE_PROJECTS="clang;lld" -DLLVM_ENABLE_RUNTIMES="compiler-rt" ../llvm && \
make -j 8 ; \
@@ -131,7 +134,7 @@ RUN if ( [ "$compiler_version" = "amd-staging" ] || [ "$compiler_version" = "amd
fi
RUN if ( [ "$compiler_version" = "amd-staging" ] || [ "$compiler_version" = "amd-mainline-open" ] ) && [ "$compiler_commit" != "" ]; then \
git clone -b "$compiler_version" https://github.com/RadeonOpenCompute/llvm-project.git && \
git clone -b "$compiler_version" https://github.com/ROCm/llvm-project.git && \
cd llvm-project && git checkout "$compiler_commit" && echo "checking out commit $compiler_commit" && mkdir build && cd build && \
cmake -DCMAKE_INSTALL_PREFIX=/opt/rocm/llvm -DCMAKE_BUILD_TYPE=Release -DLLVM_ENABLE_ASSERTIONS=1 -DLLVM_TARGETS_TO_BUILD="AMDGPU;X86" -DLLVM_ENABLE_PROJECTS="clang;lld" -DLLVM_ENABLE_RUNTIMES="compiler-rt" ../llvm && \
make -j 8 ; \

8
Jenkinsfile vendored
View File

@@ -134,7 +134,9 @@ def buildDocker(install_prefix){
//force building the new docker if that parameter is true
echo "Building image: ${image_name}"
retimage = docker.build("${image_name}", dockerArgs + ' .')
retimage.push()
withDockerRegistry([ credentialsId: "docker_test_cred", url: "" ]) {
retimage.push()
}
sh 'docker images -q -f dangling=true | xargs --no-run-if-empty docker rmi'
}
else{
@@ -146,7 +148,9 @@ def buildDocker(install_prefix){
catch(Exception ex){
echo "Unable to locate image: ${image_name}. Building image now"
retimage = docker.build("${image_name}", dockerArgs + ' .')
retimage.push()
withDockerRegistry([ credentialsId: "docker_test_cred", url: "" ]) {
retimage.push()
}
}
}

4
client_example/01_gemm/gemm.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>
@@ -83,7 +83,7 @@ int main(int argc, char* argv[])
[](std::size_t nRow, std::size_t nCol, std::size_t stride, auto layout) {
using Layout = decltype(layout);
if constexpr(std::is_same<Layout, ck::tensor_layout::gemm::RowMajor>::value)
if constexpr(std::is_same<Layout, Row>::value)
{
return (nRow - 1) * stride + nCol;
}

4
client_example/02_gemm_add_add_fastgelu/gemm_add_add_fastgelu.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>
@@ -92,7 +92,7 @@ int main(int argc, char* argv[])
[](std::size_t nRow, std::size_t nCol, std::size_t stride, auto layout) {
using Layout = decltype(layout);
if constexpr(std::is_same<Layout, ck::tensor_layout::gemm::RowMajor>::value)
if constexpr(std::is_same<Layout, Row>::value)
{
return (nRow - 1) * stride + nCol;
}

4
client_example/02_gemm_add_add_fastgelu/gemm_add_add_fastgelu_generic.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>
@@ -93,7 +93,7 @@ int main(int argc, char* argv[])
[](std::size_t nRow, std::size_t nCol, std::size_t stride, auto layout) {
using Layout = decltype(layout);
if constexpr(std::is_same<Layout, ck::tensor_layout::gemm::RowMajor>::value)
if constexpr(std::is_same<Layout, Row>::value)
{
return (nRow - 1) * stride + nCol;
}

4
client_example/02_gemm_add_add_fastgelu/gemm_add_fastgelu.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>
@@ -88,7 +88,7 @@ int main(int argc, char* argv[])
[](std::size_t nRow, std::size_t nCol, std::size_t stride, auto layout) {
using Layout = decltype(layout);
if constexpr(std::is_same<Layout, ck::tensor_layout::gemm::RowMajor>::value)
if constexpr(std::is_same<Layout, Row>::value)
{
return (nRow - 1) * stride + nCol;
}

4
client_example/02_gemm_add_add_fastgelu/gemm_add_fastgelu_generic.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>
@@ -89,7 +89,7 @@ int main(int argc, char* argv[])
[](std::size_t nRow, std::size_t nCol, std::size_t stride, auto layout) {
using Layout = decltype(layout);
if constexpr(std::is_same<Layout, ck::tensor_layout::gemm::RowMajor>::value)
if constexpr(std::is_same<Layout, Row>::value)
{
return (nRow - 1) * stride + nCol;
}

4
client_example/02_gemm_add_add_fastgelu/gemm_fastgelu.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>
@@ -84,7 +84,7 @@ int main(int argc, char* argv[])
[](std::size_t nRow, std::size_t nCol, std::size_t stride, auto layout) {
using Layout = decltype(layout);
if constexpr(std::is_same<Layout, ck::tensor_layout::gemm::RowMajor>::value)
if constexpr(std::is_same<Layout, Row>::value)
{
return (nRow - 1) * stride + nCol;
}

4
client_example/02_gemm_add_add_fastgelu/gemm_fastgelu_generic.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>
@@ -85,7 +85,7 @@ int main(int argc, char* argv[])
[](std::size_t nRow, std::size_t nCol, std::size_t stride, auto layout) {
using Layout = decltype(layout);
if constexpr(std::is_same<Layout, ck::tensor_layout::gemm::RowMajor>::value)
if constexpr(std::is_same<Layout, Row>::value)
{
return (nRow - 1) * stride + nCol;
}

6
client_example/03_gemm_layernorm/gemm_add_add_layernorm_naive.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>
@@ -17,6 +17,8 @@
using F16 = ck::half_t;
using F32 = float;
using Row = ck::tensor_layout::gemm::RowMajor;
using ADataType = F16;
using BDataType = F16;
using BiasDataType = F32;
@@ -191,7 +193,7 @@ int main()
[](std::size_t nRow, std::size_t nCol, std::size_t stride, auto layout) {
using Layout = decltype(layout);
if constexpr(std::is_same<Layout, ck::tensor_layout::gemm::RowMajor>::value)
if constexpr(std::is_same<Layout, Row>::value)
{
return (nRow - 1) * stride + nCol;
}

4
client_example/03_gemm_layernorm/gemm_add_relu_add_layernorm_welford.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <iostream>
@@ -78,7 +78,7 @@ int main(int argc, char* argv[])
[](std::size_t nRow, std::size_t nCol, std::size_t stride, auto layout) {
using Layout = decltype(layout);
if constexpr(std::is_same<Layout, ck::tensor_layout::gemm::RowMajor>::value)
if constexpr(std::is_same<Layout, Row>::value)
{
return (nRow - 1) * stride + nCol;
}

2
client_example/04_contraction/contraction_bilinear_fp32.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <numeric>

2
client_example/04_contraction/contraction_bilinear_fp64.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <numeric>

2
client_example/04_contraction/contraction_g1m2n3k1_add_xdl_fp16.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <numeric>

2
client_example/04_contraction/contraction_scale_fp32.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <numeric>

2
client_example/04_contraction/contraction_scale_fp64.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <numeric>

2
client_example/05_layernorm/layernorm2d_bwd_data.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>

2
client_example/05_layernorm/layernorm2d_bwd_gamma_beta.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>

2
client_example/05_layernorm/layernorm2d_fwd.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>

2
client_example/05_layernorm/layernorm4d_fwd.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>

2
client_example/06_softmax/softmax4d.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <functional>
#include <numeric>

2
client_example/07_grouped_convnd_fwd/grouped_conv1d_fwd.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib>
#include <iomanip>

2
client_example/07_grouped_convnd_fwd/grouped_conv2d_fwd.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib>
#include <iomanip>

2
client_example/08_fused_attention/fused_attention.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iostream>
#include <vector>

2
client_example/08_fused_attention/fused_attention_bias.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iostream>
#include <vector>

2
client_example/09_quantization/conv2d_fwd_bias_relu_perchannel_quantization.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <iostream>

2
client_example/09_quantization/conv2d_fwd_bias_relu_perlayer_quantization.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <iostream>

2
client_example/09_quantization/conv2d_fwd_bias_tanh_perchannel_quantization.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <iostream>

2
client_example/09_quantization/conv2d_fwd_bias_tanh_perlayer_quantization.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <iostream>

2
client_example/09_quantization/conv2d_fwd_perchannel_quantization.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <iostream>

2
client_example/09_quantization/conv2d_fwd_perlayer_quantization.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <iostream>

2
client_example/09_quantization/gemm_quantization.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <iostream>

2
client_example/10_grouped_convnd_bwd_data/grouped_conv2d_bwd_data.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib>
#include <iomanip>

2
client_example/10_grouped_convnd_bwd_data/grouped_conv3d_bwd_data.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib>
#include <iomanip>

2
client_example/10_grouped_convnd_bwd_data/grouped_conv3d_bwd_data_input_fp16_comp_bf8f8.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib>
#include <iomanip>

2
client_example/11_grouped_conv_bwd_weight/grouped_conv1d_bwd_weight_fp16.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp"

2
client_example/11_grouped_conv_bwd_weight/grouped_conv2d_bwd_weight_fp16.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp"

2
client_example/11_grouped_conv_bwd_weight/grouped_conv3d_bwd_weight_fp16.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp"

2
client_example/11_grouped_conv_bwd_weight/grouped_conv3d_bwd_weight_fp16_comp_bf8_fp8.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp"

2
client_example/11_grouped_conv_bwd_weight/grouped_conv3d_bwd_weight_fp32.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp"

2
client_example/12_elementwise_normalization/elementwise_layernorm2d.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>

2
client_example/13_batchnorm/batchnorm_bwd_nhwc.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <functional>
#include <numeric>

2
client_example/13_batchnorm/batchnorm_fwd_nhwc.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <functional>
#include <numeric>

2
client_example/13_batchnorm/batchnorm_infer_nhwc.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <functional>
#include <numeric>

2
client_example/14_instance_id/batchnorm_fwd_instance_id.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <functional>
#include <numeric>

2
client_example/15_convnd_bwd_data/conv3d_bwd_data_fp16.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp"

2
client_example/15_convnd_bwd_data/conv3d_bwd_data_fp32.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp"

2
client_example/15_gemm_add_multiply/gemm_add_multiply.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>

2
client_example/15_reduce/reduce_nhwc_c.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <functional>
#include <numeric>

2
client_example/16_convnd_fwd/conv3d_fwd_fp16.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp"

2
client_example/16_convnd_fwd/conv3d_fwd_fp16_comp_fp8.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp"

2
client_example/16_convnd_fwd/conv3d_fwd_fp32.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include "common.hpp"

2
client_example/17_grouped_gemm_fastgelu/grouped_gemm_fastgelu.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <iostream>

2
client_example/18_groupnorm/groupnorm_bwd_data.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>

2
client_example/18_groupnorm/groupnorm_bwd_gamma_beta.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>

2
client_example/18_groupnorm/groupnorm_swish_fwd.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>

2
client_example/19_pool/avg_pool3d_bwd.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>

2
client_example/19_pool/avg_pool3d_fwd.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>

2
client_example/19_pool/max_pool2d_bwd.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>

2
client_example/19_pool/max_pool2d_fwd.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>

4
client_example/20_splitk_gemm/splitK_gemm_fp16_f8.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>
@@ -88,7 +88,7 @@ int main(int argc, char* argv[])
[](std::size_t nRow, std::size_t nCol, std::size_t stride, auto layout) {
using Layout = decltype(layout);
if constexpr(std::is_same<Layout, ck::tensor_layout::gemm::RowMajor>::value)
if constexpr(std::is_same<Layout, Row>::value)
{
return (nRow - 1) * stride + nCol;
}

4
client_example/21_grouped_gemm_bias/grouped_gemm_fixed_nk_bias_fp16.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <iostream>
@@ -79,7 +79,7 @@ int main()
[](std::size_t nRow, std::size_t nCol, std::size_t stride, auto layout) {
using Layout = decltype(layout);
if constexpr(std::is_same<Layout, ck::tensor_layout::gemm::RowMajor>::value)
if constexpr(std::is_same<Layout, Row>::value)
{
return (nRow - 1) * stride + nCol;
}

3
client_example/22_grouped_gemm/CMakeLists.txt
View File

@@ -6,3 +6,6 @@ target_link_libraries(client_grouped_gemm_fixed_nk_fp8 PRIVATE composable_kernel
add_executable(client_grouped_gemm_fixed_nk_i8 grouped_gemm_fixed_nk_i8.cpp)
target_link_libraries(client_grouped_gemm_fixed_nk_i8 PRIVATE composable_kernel::device_gemm_operations)
add_executable(client_grouped_gemm_fixed_nk_bf16 grouped_gemm_fixed_nk_bf16.cpp)
target_link_libraries(client_grouped_gemm_fixed_nk_bf16 PRIVATE composable_kernel::device_gemm_operations)

237
client_example/22_grouped_gemm/grouped_gemm_fixed_nk_bf16.cpp Normal file
View File

@@ -0,0 +1,237 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <iostream>
#include <vector>
#include <random>
#include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_grouped_gemm_fixed_nk.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_gemm_fixed_nk.hpp"
using I8 = int8_t;
using BF16 = ck::bhalf_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 = BF16;
using BDataType = I8;
using DsDataType = ck::Tuple<>;
using EDataType = BF16;
using ALayout = Row;
using BLayout = Row;
using DsLayout = ck::Tuple<>;
using ELayout = Row;
using AElementOp = PassThrough;
using BElementOp = PassThrough;
using CDEElementOp = PassThrough;
struct SimpleDeviceMem
{
SimpleDeviceMem() = delete;
SimpleDeviceMem(std::size_t mem_size) : p_mem_{}
{
(void)hipMalloc(static_cast<void**>(&p_mem_), mem_size);
}
void* GetDeviceBuffer() { return p_mem_; }
~SimpleDeviceMem() { (void)hipFree(p_mem_); }
void* p_mem_;
};
int main()
{
std::vector<int> Ms, Ns, Ks, StrideAs, StrideBs, StrideEs;
int sum_of_m = 0;
const int group_count = 16;
for(int i = 0; i < group_count; ++i)
{
Ms.push_back(256 + 256 * i);
Ns.push_back(128 + 128 * i);
Ks.push_back(128 + 64 * i);
StrideAs.push_back(std::is_same<Row, ALayout>::value ? Ks[i] : Ms[i]);
StrideBs.push_back(std::is_same<Row, BLayout>::value ? Ns[i] : Ks[i]);
StrideEs.push_back(std::is_same<Row, ELayout>::value ? Ns[i] : Ms[i]);
sum_of_m += Ms[i];
}
auto f_matrix_space_size =
[](std::size_t nRow, std::size_t nCol, std::size_t stride, auto layout) {
using Layout = decltype(layout);
if constexpr(std::is_same<Layout, Row>::value)
{
return (nRow - 1) * stride + nCol;
}
else
{
return (nCol - 1) * stride + nRow;
}
};
std::vector<SimpleDeviceMem> a_dev_bufs, b_dev_bufs, e_dev_bufs;
a_dev_bufs.reserve(group_count);
b_dev_bufs.reserve(group_count);
e_dev_bufs.reserve(group_count);
std::vector<void*> p_e;
p_e.reserve(group_count);
std::vector<ck::tensor_operation::device::GemmDesc> gemm_descs;
gemm_descs.reserve(group_count);
std::vector<ck::tensor_operation::device::GroupedGemmKernelArgument<1>>
grouped_gemm_kernel_args_;
grouped_gemm_kernel_args_.reserve(group_count);
for(int i = 0; i < group_count; ++i)
{
a_dev_bufs.emplace_back(sizeof(ADataType) *
f_matrix_space_size(Ms[i], Ks[i], StrideAs[i], ALayout{}));
b_dev_bufs.emplace_back(sizeof(BDataType) *
f_matrix_space_size(Ks[i], Ns[i], StrideBs[i], BLayout{}));
e_dev_bufs.emplace_back(sizeof(EDataType) *
f_matrix_space_size(Ms[i], Ns[i], StrideEs[i], ELayout{}));
gemm_descs.push_back({sum_of_m, Ns[i], Ks[i], 1, StrideBs[i], 1, {0}});
p_e.push_back(e_dev_bufs[i].GetDeviceBuffer());
grouped_gemm_kernel_args_.push_back({a_dev_bufs[i].GetDeviceBuffer(),
b_dev_bufs[i].GetDeviceBuffer(),
{},
e_dev_bufs[i].GetDeviceBuffer(),
Ms[i],
Ns[i],
Ks[i],
StrideAs[i],
StrideBs[i],
{},
StrideEs[i]});
}
using DeviceOp = ck::tensor_operation::device::DeviceGroupedGemmFixedNK<ALayout,
BLayout,
DsLayout,
ELayout,
ADataType,
BDataType,
DsDataType,
EDataType,
AElementOp,
BElementOp,
CDEElementOp>;
// get device op instances
const auto op_ptrs = ck::tensor_operation::device::instance::DeviceOperationInstanceFactory<
DeviceOp>::GetInstances();
std::cout << "found " << op_ptrs.size() << " instances" << std::endl;
const auto a_element_op = AElementOp{};
const auto b_element_op = BElementOp{};
const auto cde_element_op = CDEElementOp{};
std::string best_op_name;
bool found = false;
int best_op_id = -1;
float best_ave_time = 0;
float best_tflops = 0;
float best_gb_per_sec = 0;
// profile device operation instances
std::cout << "Run all instances and do timing" << std::endl;
std::vector<const void*> p_a = {}, p_b = {};
std::vector<std::array<const void*, 0>> p_ds = {};
for(int i = 0; i < op_ptrs.size(); ++i)
{
auto& op_ptr = op_ptrs[i];
auto argument_ptr = op_ptr->MakeArgumentPointer(
p_a, p_b, p_ds, p_e, gemm_descs, a_element_op, b_element_op, cde_element_op);
auto invoker_ptr = op_ptr->MakeInvokerPointer();
SimpleDeviceMem grouped_gemm_kernel_args_dev(
op_ptr->GetDeviceKernelArgSize(argument_ptr.get()));
SimpleDeviceMem grouped_gemm_workspace_dev(op_ptr->GetWorkSpaceSize(argument_ptr.get()));
std::string op_name = op_ptr->GetTypeString();
hipGetErrorString(hipMemcpy(grouped_gemm_kernel_args_dev.GetDeviceBuffer(),
grouped_gemm_kernel_args_.data(),
op_ptr->GetDeviceKernelArgSize(argument_ptr.get()),
hipMemcpyHostToDevice));
op_ptr->SetWorkSpacePointer(argument_ptr.get(),
grouped_gemm_workspace_dev.GetDeviceBuffer());
op_ptr->SetDeviceKernelArgs(argument_ptr.get(),
grouped_gemm_kernel_args_dev.GetDeviceBuffer());
op_ptr->SetKBatch(argument_ptr.get(), 1);
if(op_ptr->IsSupportedArgument(argument_ptr.get()))
{
float ave_time = invoker_ptr->Run(argument_ptr.get(), StreamConfig{nullptr, true});
std::size_t flop = 0, num_btype = 0;
for(std::size_t j = 0; j < gemm_descs.size(); ++j)
{
flop += std::size_t(2) * Ms[j] * Ns[j] * Ks[j];
num_btype += sizeof(ADataType) * Ms[j] * Ks[j] + sizeof(BDataType) * Ks[j] * Ns[j] +
sizeof(EDataType) * Ms[j] * Ns[j];
}
float tflops = static_cast<float>(flop) / 1.E9 / ave_time;
float gb_per_sec = num_btype / 1.E6 / ave_time;
std::cout << "Perf: " << std::setw(10) << ave_time << " ms, " << tflops << " TFlops, "
<< gb_per_sec << " GB/s, " << op_name << std::endl;
if(tflops > best_tflops)
{
found = true;
best_op_id = i;
best_op_name = op_name;
best_tflops = tflops;
best_ave_time = ave_time;
best_gb_per_sec = gb_per_sec;
}
}
else
{
std::cout << op_name << " does not support this problem" << std::endl;
}
}
std::cout << "Best Perf: " << best_ave_time << " ms, " << best_tflops << " TFlops, "
<< best_gb_per_sec << " GB/s, " << best_op_name << std::endl;
return 0;
}

4
client_example/22_grouped_gemm/grouped_gemm_fixed_nk_fp16.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <iostream>
@@ -76,7 +76,7 @@ int main()
[](std::size_t nRow, std::size_t nCol, std::size_t stride, auto layout) {
using Layout = decltype(layout);
if constexpr(std::is_same<Layout, ck::tensor_layout::gemm::RowMajor>::value)
if constexpr(std::is_same<Layout, Row>::value)
{
return (nRow - 1) * stride + nCol;
}

4
client_example/22_grouped_gemm/grouped_gemm_fixed_nk_fp8.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <iostream>
@@ -77,7 +77,7 @@ int main()
[](std::size_t nRow, std::size_t nCol, std::size_t stride, auto layout) {
using Layout = decltype(layout);
if constexpr(std::is_same<Layout, ck::tensor_layout::gemm::RowMajor>::value)
if constexpr(std::is_same<Layout, Row>::value)
{
return (nRow - 1) * stride + nCol;
}

4
client_example/22_grouped_gemm/grouped_gemm_fixed_nk_i8.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <iostream>
@@ -77,7 +77,7 @@ int main()
[](std::size_t nRow, std::size_t nCol, std::size_t stride, auto layout) {
using Layout = decltype(layout);
if constexpr(std::is_same<Layout, ck::tensor_layout::gemm::RowMajor>::value)
if constexpr(std::is_same<Layout, Row>::value)
{
return (nRow - 1) * stride + nCol;
}

2
client_example/22_im2col_col2im/image_to_column.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib>
#include <iomanip>

2
client_example/23_elementwise_transpose/elementwise_transpose_3d.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <iomanip>
#include <vector>

11
client_example/23_grouped_convnd_fwd_scaleadd_scaleadd_relu/CMakeLists.txt
View File

@@ -1,11 +0,0 @@
add_executable(client_grouped_convnd_fwd_scaleadd_scaleadd_relu_fp32 grouped_conv_fwd_scaleadd_scaleadd_relu_fp32.cpp)
target_link_libraries(client_grouped_convnd_fwd_scaleadd_scaleadd_relu_fp32 PRIVATE composable_kernel::device_conv_operations)
add_executable(client_grouped_convnd_fwd_scaleadd_scaleadd_relu_fp16 grouped_conv_fwd_scaleadd_scaleadd_relu_fp16.cpp)
target_link_libraries(client_grouped_convnd_fwd_scaleadd_scaleadd_relu_fp16 PRIVATE composable_kernel::device_conv_operations)
add_executable(client_grouped_convnd_fwd_scaleadd_scaleadd_relu_bf16 grouped_conv_fwd_scaleadd_scaleadd_relu_bf16.cpp)
target_link_libraries(client_grouped_convnd_fwd_scaleadd_scaleadd_relu_bf16 PRIVATE composable_kernel::device_conv_operations)
add_executable(client_grouped_convnd_fwd_scaleadd_scaleadd_relu_int8 grouped_conv_fwd_scaleadd_scaleadd_relu_int8.cpp)
target_link_libraries(client_grouped_convnd_fwd_scaleadd_scaleadd_relu_int8 PRIVATE composable_kernel::device_conv_operations)

40
client_example/24_grouped_conv_activation/CMakeLists.txt Normal file
View File

@@ -0,0 +1,40 @@
# Fwd scaleadd scaleadd relu
add_executable(client_grouped_convnd_fwd_scaleadd_scaleadd_relu_fp32
grouped_convnd_fwd_scaleadd_scaleadd_relu/grouped_conv_fwd_scaleadd_scaleadd_relu_fp32.cpp)
target_link_libraries(client_grouped_convnd_fwd_scaleadd_scaleadd_relu_fp32 PRIVATE composable_kernel::device_conv_operations)
add_executable(client_grouped_convnd_fwd_scaleadd_scaleadd_relu_fp16
grouped_convnd_fwd_scaleadd_scaleadd_relu/grouped_conv_fwd_scaleadd_scaleadd_relu_fp16.cpp)
target_link_libraries(client_grouped_convnd_fwd_scaleadd_scaleadd_relu_fp16 PRIVATE composable_kernel::device_conv_operations)
add_executable(client_grouped_convnd_fwd_scaleadd_scaleadd_relu_bf16
grouped_convnd_fwd_scaleadd_scaleadd_relu/grouped_conv_fwd_scaleadd_scaleadd_relu_bf16.cpp)
target_link_libraries(client_grouped_convnd_fwd_scaleadd_scaleadd_relu_bf16 PRIVATE composable_kernel::device_conv_operations)
add_executable(client_grouped_convnd_fwd_scaleadd_scaleadd_relu_int8
grouped_convnd_fwd_scaleadd_scaleadd_relu/grouped_conv_fwd_scaleadd_scaleadd_relu_int8.cpp)
target_link_libraries(client_grouped_convnd_fwd_scaleadd_scaleadd_relu_int8 PRIVATE composable_kernel::device_conv_operations)
# Fwd scaleadd AB
add_executable(client_grouped_convnd_fwd_scaleadd_ab_fp32
grouped_convnd_fwd_scaleadd_ab/grouped_conv_fwd_scaleadd_ab_fp32.cpp)
target_link_libraries(client_grouped_convnd_fwd_scaleadd_ab_fp32 PRIVATE composable_kernel::device_conv_operations)
add_executable(client_grouped_convnd_fwd_scaleadd_ab_fp16
grouped_convnd_fwd_scaleadd_ab/grouped_conv_fwd_scaleadd_ab_fp16.cpp)
target_link_libraries(client_grouped_convnd_fwd_scaleadd_ab_fp16 PRIVATE composable_kernel::device_conv_operations)
add_executable(client_grouped_convnd_fwd_scaleadd_ab_bf16
grouped_convnd_fwd_scaleadd_ab/grouped_conv_fwd_scaleadd_ab_bf16.cpp)
target_link_libraries(client_grouped_convnd_fwd_scaleadd_ab_bf16 PRIVATE composable_kernel::device_conv_operations)
add_executable(client_grouped_convnd_fwd_scaleadd_ab_int8
grouped_convnd_fwd_scaleadd_ab/grouped_conv_fwd_scaleadd_ab_int8.cpp)
target_link_libraries(client_grouped_convnd_fwd_scaleadd_ab_int8 PRIVATE composable_kernel::device_conv_operations)
# Fwd bilinear
add_executable(client_grouped_convnd_fwd_bilinear_residual_fp16
grouped_convnd_fwd_bilinear/grouped_conv_fwd_bilinear_residual_fp16.cpp)
target_link_libraries(client_grouped_convnd_fwd_bilinear_residual_fp16 PRIVATE composable_kernel::device_conv_operations)
# Bwd data bilinear
add_executable(client_grouped_convnd_bwd_data_bilinear_residual_fp16
grouped_convnd_bwd_data_bilinear/grouped_conv_bwd_data_bilinear_residual_fp16.cpp)
target_link_libraries(client_grouped_convnd_bwd_data_bilinear_residual_fp16 PRIVATE composable_kernel::device_conv_operations)

217
client_example/24_grouped_conv_activation/grouped_convnd_bwd_data_bilinear/grouped_conv_bwd_data_bilinear_residual_fp16.cpp Normal file
View File

@@ -0,0 +1,217 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
#include <tuple>
#include <cstdlib>
#include <iomanip>
#include <iostream>
#include <iterator>
#include <numeric>
#include <vector>
#include "ck/utility/data_type.hpp"
#include "ck/utility/tuple.hpp"
#include "ck/ck.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_data_bilinear.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
using InDataType = ck::half_t;
using WeiDataType = ck::half_t;
using OutDataType = ck::half_t;
// Use std tuple instead of ck tuple to avoid clang
// implicit instantiation of undefined template error.
using DDataTypes = std::tuple<ck::half_t>;
using InLayout = ck::tensor_layout::convolution::NDHWGC;
using WeiLayout = ck::tensor_layout::convolution::GKZYXC;
using OutLayout = ck::tensor_layout::convolution::NDHWGK;
using PassThrough = ck::tensor_operation::element_wise::PassThrough;
using Bilinear = ck::tensor_operation::element_wise::Bilinear;
static constexpr ck::index_t NumDimSpatial = 3;
static constexpr ck::index_t G = 32;
static constexpr ck::index_t N = 64; // batch size
static constexpr ck::index_t K = 64; // output channel
static constexpr ck::index_t C = 32; // input channel (per group)
static constexpr ck::index_t Z = 3; // filter D
static constexpr ck::index_t Y = 3; // filter H
static constexpr ck::index_t X = 3; // filter W
static constexpr ck::index_t Di = 14; // input D
static constexpr ck::index_t Hi = 14; // input H
static constexpr ck::index_t Wi = 14; // input W
static constexpr ck::index_t Do = 14; // output D
static constexpr ck::index_t Ho = 14; // output H
static constexpr ck::index_t Wo = 14; // output W
struct SimpleDeviceMem
{
SimpleDeviceMem() = delete;
SimpleDeviceMem(std::size_t mem_size) : p_mem_{}
{
(void)hipMalloc(static_cast<void**>(&p_mem_), mem_size);
}
void* GetDeviceBuffer() { return p_mem_; }
~SimpleDeviceMem() { (void)hipFree(p_mem_); }
void* p_mem_;
};
int execute_conv_bwd_data_bilinear()
{
std::array<ck::index_t, NumDimSpatial + 3> in_lengths{G, N, C, Di, Hi, Wi};
std::array<ck::index_t, NumDimSpatial + 3> in_strides{
C, Di * Hi * Wi * G * C, 1, Hi * Wi * G * C, Wi * G * C, G * C};
std::array<ck::index_t, NumDimSpatial + 3> wei_lengths{G, K, C, Z, Y, X};
std::array<ck::index_t, NumDimSpatial + 3> wei_strides{
K * Z * Y * X * C, Z * Y * X * C, 1, Y * X * C, X * C, C};
std::array<ck::index_t, NumDimSpatial + 3> out_lengths{G, N, K, Do, Ho, Wo};
std::array<ck::index_t, NumDimSpatial + 3> out_strides{
K, Do * Ho * Wo * G * K, 1, Ho * Wo * G * K, Wo * G * K, G * K};
std::array<ck::index_t, NumDimSpatial> filter_strides{1, 1, 1};
std::array<ck::index_t, NumDimSpatial> filter_dilations{1, 1, 1};
std::array<ck::index_t, NumDimSpatial> input_left_pads{1, 1, 1};
std::array<ck::index_t, NumDimSpatial> input_right_pads{1, 1, 1};
SimpleDeviceMem in(sizeof(InDataType) * G * N * Di * Hi * Wi * C);
SimpleDeviceMem wei(sizeof(WeiDataType) * G * K * Z * Y * X * C);
SimpleDeviceMem out(sizeof(OutDataType) * G * N * Do * Ho * Wo * K);
using DeviceOp =
ck::tensor_operation::device::DeviceGroupedConvBwdDataMultipleD<NumDimSpatial,
OutLayout,
WeiLayout,
ck::Tuple<InLayout>,
InLayout,
OutDataType,
WeiDataType,
ck::Tuple<InDataType>,
InDataType,
PassThrough,
PassThrough,
Bilinear>;
// get device op instances
const auto op_ptrs = ck::tensor_operation::device::instance::DeviceOperationInstanceFactory<
DeviceOp>::GetInstances();
std::cout << "found " << op_ptrs.size() << " instances" << std::endl;
std::string best_op_name;
int best_op_id = -1;
float best_avg_time = std::numeric_limits<float>::max();
float best_gb_per_sec = 0;
float best_tflops = 0;
// profile device operation instances
std::cout << "Run all instances and do timing" << std::endl;
for(int i = 0; i < op_ptrs.size(); ++i)
{
auto& op_ptr = op_ptrs[i];
auto argument_ptr = op_ptr->MakeArgumentPointer(out.GetDeviceBuffer(),
wei.GetDeviceBuffer(),
{in.GetDeviceBuffer()},
in.GetDeviceBuffer(),
out_lengths,
out_strides,
wei_lengths,
wei_strides,
{in_lengths},
{in_strides},
in_lengths,
in_strides,
filter_strides,
filter_dilations,
input_left_pads,
input_right_pads,
PassThrough{},
PassThrough{},
Bilinear{2.f, 2.f});
auto invoker_ptr = op_ptr->MakeInvokerPointer();
std::string op_name = op_ptr->GetTypeString();
if(op_ptr->IsSupportedArgument(argument_ptr.get()))
{
float avg_time = invoker_ptr->Run(argument_ptr.get(), StreamConfig{nullptr, true});
std::size_t flop = std::size_t(2) * G * N * K * C * Do * Ho * Wo * Y * X +
3 * G * N * Di * Hi * Wi * C;
std::size_t num_bytes = 2 * sizeof(InDataType) * G * N * Di * Hi * Wi * C +
sizeof(WeiDataType) * G * K * Z * Y * X * C +
sizeof(OutDataType) * G * N * Do * Ho * Wo * K;
float tflops = static_cast<float>(flop) / 1.E9 / avg_time;
float gb_per_sec = num_bytes / 1.E6 / avg_time;
std::cout << "Perf: " << std::setw(10) << avg_time << " ms, " << tflops << " TFlops, "
<< gb_per_sec << " GB/s, " << op_name << std::endl;
if(tflops > best_tflops)
{
best_op_id = i;
best_op_name = op_name;
best_avg_time = avg_time;
best_gb_per_sec = gb_per_sec;
best_tflops = tflops;
}
}
else
{
std::cerr << op_name << " does not support this problem" << std::endl;
}
}
if(best_op_id < 0)
{
std::cerr << "no suitable instance" << std::endl;
return EXIT_FAILURE;
}
std::cout << "Best Perf: " << std::setw(10) << best_avg_time << " ms, " << best_tflops
<< " TFlops, " << best_gb_per_sec << " GB/s, " << best_op_name << std::endl;
// run the best intance
{
auto& op_ptr = op_ptrs[best_op_id];
std::cout << "Run the best instance without timing: " << op_ptr->GetTypeString()
<< std::endl;
auto argument_ptr = op_ptr->MakeArgumentPointer(out.GetDeviceBuffer(),
wei.GetDeviceBuffer(),
{in.GetDeviceBuffer()},
in.GetDeviceBuffer(),
out_lengths,
out_strides,
wei_lengths,
wei_strides,
{in_lengths},
{in_strides},
in_lengths,
in_strides,
filter_strides,
filter_dilations,
input_left_pads,
input_right_pads,
PassThrough{},
PassThrough{},
Bilinear{2.f, 2.f});
auto invoker_ptr = op_ptr->MakeInvokerPointer();
if(op_ptr->IsSupportedArgument(argument_ptr.get()))
{
invoker_ptr->Run(argument_ptr.get(), StreamConfig{nullptr, false});
}
std::cout << "Done" << std::endl;
}
return 0;
}
int main() { return execute_conv_bwd_data_bilinear(); }

221
client_example/24_grouped_conv_activation/grouped_convnd_fwd_bilinear/grouped_conv_fwd_bilinear_residual_fp16.cpp Normal file
View File

@@ -0,0 +1,221 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
#include <tuple>
#include <cstdlib>
#include <iomanip>
#include <iostream>
#include <iterator>
#include <numeric>
#include <vector>
#include "ck/utility/data_type.hpp"
#include "ck/utility/tuple.hpp"
#include "ck/ck.hpp"
#include "ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_bilinear.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
using InDataType = ck::half_t;
using WeiDataType = ck::half_t;
using OutDataType = ck::half_t;
// Use std tuple instead of ck tuple to avoid clang
// implicit instantiation of undefined template error.
using DDataTypes = std::tuple<ck::half_t>;
using InLayout = ck::tensor_layout::convolution::NDHWGC;
using WeiLayout = ck::tensor_layout::convolution::GKZYXC;
using OutLayout = ck::tensor_layout::convolution::NDHWGK;
using PassThrough = ck::tensor_operation::element_wise::PassThrough;
using Bilinear = ck::tensor_operation::element_wise::Bilinear;
static constexpr ck::index_t NumDimSpatial = 3;
static constexpr ck::index_t G = 32;
static constexpr ck::index_t N = 64; // batch size
static constexpr ck::index_t K = 64; // output channel
static constexpr ck::index_t C = 32; // input channel (per group)
static constexpr ck::index_t Z = 3; // filter D
static constexpr ck::index_t Y = 3; // filter H
static constexpr ck::index_t X = 3; // filter W
static constexpr ck::index_t Di = 14; // input D
static constexpr ck::index_t Hi = 14; // input H
static constexpr ck::index_t Wi = 14; // input W
static constexpr ck::index_t Do = 14; // output D
static constexpr ck::index_t Ho = 14; // output H
static constexpr ck::index_t Wo = 14; // output W
struct SimpleDeviceMem
{
SimpleDeviceMem() = delete;
SimpleDeviceMem(std::size_t mem_size) : p_mem_{}
{
(void)hipMalloc(static_cast<void**>(&p_mem_), mem_size);
}
void* GetDeviceBuffer() { return p_mem_; }
~SimpleDeviceMem() { (void)hipFree(p_mem_); }
void* p_mem_;
};
int execute_conv_fwd_bilinear()
{
// We have NHWGC/GKYXC/NHWGK (x, weight, y) in memory space.
// However, CK's API only accepts lengths and strides with order of GNCDHW/GKCZYX/GNKDHW.
// Hence, we need to adjust the order of strides.
std::array<ck::index_t, 6> in_lengths{G, N, C, Di, Hi, Wi};
std::array<ck::index_t, 6> in_strides{
C, Di * Hi * Wi * G * C, 1, Hi * Wi * G * C, Wi * G * C, G * C};
std::array<ck::index_t, 6> wei_lengths{G, K, C, Z, Y, X};
std::array<ck::index_t, 6> wei_strides{
K * Z * Y * X * C, Z * Y * X * C, 1, Y * X * C, X * C, C};
std::array<ck::index_t, 6> out_lengths{G, N, K, Do, Ho, Wo};
std::array<ck::index_t, 6> out_strides{
K, Do * Ho * Wo * G * K, 1, Ho * Wo * G * K, Wo * G * K, G * K};
// Logical broadcast bias (we have to pass bias lengths in the same format as output - GNKDHW)
std::array<ck::index_t, 6> bias_lengths{G, 1, K, 1, 1, 1};
std::array<ck::index_t, 6> bias_strides{K, 0, 1, 0, 0, 0};
std::array<ck::index_t, NumDimSpatial> filter_strides{1, 1, 1};
std::array<ck::index_t, NumDimSpatial> filter_dilations{1, 1, 1};
std::array<ck::index_t, NumDimSpatial> input_left_pads{1, 1, 1};
std::array<ck::index_t, NumDimSpatial> input_right_pads{1, 1, 1};
SimpleDeviceMem in(sizeof(InDataType) * N * Di * Hi * Wi * G * C);
SimpleDeviceMem wei(sizeof(WeiDataType) * G * K * Z * Y * X * C);
SimpleDeviceMem out(sizeof(OutDataType) * N * Do * Ho * Wo * G * K);
using DeviceOp =
ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD<NumDimSpatial,
InLayout,
WeiLayout,
ck::Tuple<OutLayout>,
OutLayout,
InDataType,
WeiDataType,
ck::Tuple<OutDataType>,
OutDataType,
PassThrough,
PassThrough,
Bilinear>;
// get device op instances
const auto op_ptrs = ck::tensor_operation::device::instance::DeviceOperationInstanceFactory<
DeviceOp>::GetInstances();
std::cout << "found " << op_ptrs.size() << " instances" << std::endl;
std::string best_op_name;
int best_op_id = -1;
float best_avg_time = std::numeric_limits<float>::max();
float best_gb_per_sec = 0;
float best_tflops = 0;
// profile device operation instances
std::cout << "Run all instances and do timing" << std::endl;
for(int i = 0; i < op_ptrs.size(); ++i)
{
auto& op_ptr = op_ptrs[i];
auto argument_ptr = op_ptr->MakeArgumentPointer(in.GetDeviceBuffer(),
wei.GetDeviceBuffer(),
{out.GetDeviceBuffer()},
out.GetDeviceBuffer(),
in_lengths,
in_strides,
wei_lengths,
wei_strides,
{out_lengths},
{out_strides},
out_lengths,
out_strides,
filter_strides,
filter_dilations,
input_left_pads,
input_right_pads,
PassThrough{},
PassThrough{},
Bilinear{2.f, 2.f});
auto invoker_ptr = op_ptr->MakeInvokerPointer();
std::string op_name = op_ptr->GetTypeString();
if(op_ptr->IsSupportedArgument(argument_ptr.get()))
{
float avg_time = invoker_ptr->Run(argument_ptr.get(), StreamConfig{nullptr, true});
std::size_t flop =
std::size_t(2) * G * N * K * C * Ho * Wo * Y * X + 3 * N * Ho * Wo * G * K;
std::size_t num_bytes = sizeof(InDataType) * N * Hi * Wi * G * C +
sizeof(WeiDataType) * G * K * Y * X * C +
sizeof(OutDataType) * 2 * N * Ho * Wo * G * K;
float tflops = static_cast<float>(flop) / 1.E9 / avg_time;
float gb_per_sec = num_bytes / 1.E6 / avg_time;
std::cout << "Perf: " << std::setw(10) << avg_time << " ms, " << tflops << " TFlops, "
<< gb_per_sec << " GB/s, " << op_name << std::endl;
if(tflops > best_tflops)
{
best_op_id = i;
best_op_name = op_name;
best_avg_time = avg_time;
best_gb_per_sec = gb_per_sec;
best_tflops = tflops;
}
}
else
{
std::cerr << op_name << " does not support this problem" << std::endl;
}
}
if(best_op_id < 0)
{
std::cerr << "no suitable instance" << std::endl;
return EXIT_FAILURE;
}
std::cout << "Best Perf: " << std::setw(10) << best_avg_time << " ms, " << best_tflops
<< " TFlops, " << best_gb_per_sec << " GB/s, " << best_op_name << std::endl;
// run the best intance
{
auto& op_ptr = op_ptrs[best_op_id];
std::cout << "Run the best instance without timing: " << op_ptr->GetTypeString()
<< std::endl;
auto argument_ptr = op_ptr->MakeArgumentPointer(in.GetDeviceBuffer(),
wei.GetDeviceBuffer(),
{out.GetDeviceBuffer()},
out.GetDeviceBuffer(),
in_lengths,
in_strides,
wei_lengths,
wei_strides,
{out_lengths},
{out_strides},
out_lengths,
out_strides,
filter_strides,
filter_dilations,
input_left_pads,
input_right_pads,
PassThrough{},
PassThrough{},
Bilinear{2.f, 2.f});
auto invoker_ptr = op_ptr->MakeInvokerPointer();
if(op_ptr->IsSupportedArgument(argument_ptr.get()))
{
invoker_ptr->Run(argument_ptr.get(), StreamConfig{nullptr, false});
}
std::cout << "Done" << std::endl;
}
return 0;
}
int main() { return execute_conv_fwd_bilinear(); }

2
client_example/24_grouped_convnd_fwd_scaleadd_ab/grouped_conv_fwd_scaleadd_ab.inc → client_example/24_grouped_conv_activation/grouped_convnd_fwd_scaleadd_ab/grouped_conv_fwd_scaleadd_ab.inc
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2023-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib>
#include <iomanip>

2
client_example/24_grouped_convnd_fwd_scaleadd_ab/grouped_conv_fwd_scaleadd_ab_bf16.cpp → client_example/24_grouped_conv_activation/grouped_convnd_fwd_scaleadd_ab/grouped_conv_fwd_scaleadd_ab_bf16.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2023-2024, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/utility/data_type.hpp"
#include "ck/utility/tuple.hpp"

2
client_example/24_grouped_convnd_fwd_scaleadd_ab/grouped_conv_fwd_scaleadd_ab_fp16.cpp → client_example/24_grouped_conv_activation/grouped_convnd_fwd_scaleadd_ab/grouped_conv_fwd_scaleadd_ab_fp16.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2023-2024, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/utility/data_type.hpp"
#include "ck/utility/tuple.hpp"

2
client_example/24_grouped_convnd_fwd_scaleadd_ab/grouped_conv_fwd_scaleadd_ab_fp32.cpp → client_example/24_grouped_conv_activation/grouped_convnd_fwd_scaleadd_ab/grouped_conv_fwd_scaleadd_ab_fp32.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2023-2024, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/utility/data_type.hpp"
#include "ck/utility/tuple.hpp"

2
client_example/24_grouped_convnd_fwd_scaleadd_ab/grouped_conv_fwd_scaleadd_ab_int8.cpp → client_example/24_grouped_conv_activation/grouped_convnd_fwd_scaleadd_ab/grouped_conv_fwd_scaleadd_ab_int8.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2023-2024, Advanced Micro Devices, Inc. All rights reserved.
#include "ck/utility/data_type.hpp"
#include "ck/utility/tuple.hpp"

2
client_example/23_grouped_convnd_fwd_scaleadd_scaleadd_relu/grouped_conv_fwd_scaleadd_scaleadd_relu.inc → client_example/24_grouped_conv_activation/grouped_convnd_fwd_scaleadd_scaleadd_relu/grouped_conv_fwd_scaleadd_scaleadd_relu.inc
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2023-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib>
#include <iomanip>

2
client_example/23_grouped_convnd_fwd_scaleadd_scaleadd_relu/grouped_conv_fwd_scaleadd_scaleadd_relu_bf16.cpp → client_example/24_grouped_conv_activation/grouped_convnd_fwd_scaleadd_scaleadd_relu/grouped_conv_fwd_scaleadd_scaleadd_relu_bf16.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2023-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <tuple>

2
client_example/23_grouped_convnd_fwd_scaleadd_scaleadd_relu/grouped_conv_fwd_scaleadd_scaleadd_relu_fp16.cpp → client_example/24_grouped_conv_activation/grouped_convnd_fwd_scaleadd_scaleadd_relu/grouped_conv_fwd_scaleadd_scaleadd_relu_fp16.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2023-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <tuple>

2
client_example/23_grouped_convnd_fwd_scaleadd_scaleadd_relu/grouped_conv_fwd_scaleadd_scaleadd_relu_fp32.cpp → client_example/24_grouped_conv_activation/grouped_convnd_fwd_scaleadd_scaleadd_relu/grouped_conv_fwd_scaleadd_scaleadd_relu_fp32.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2023-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <tuple>

2
client_example/23_grouped_convnd_fwd_scaleadd_scaleadd_relu/grouped_conv_fwd_scaleadd_scaleadd_relu_int8.cpp → client_example/24_grouped_conv_activation/grouped_convnd_fwd_scaleadd_scaleadd_relu/grouped_conv_fwd_scaleadd_scaleadd_relu_int8.cpp
View File

@@ -1,5 +1,5 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2023-2024, Advanced Micro Devices, Inc. All rights reserved.
#include <tuple>

11
client_example/24_grouped_convnd_fwd_scaleadd_ab/CMakeLists.txt
View File

@@ -1,11 +0,0 @@
add_executable(client_grouped_convnd_fwd_scaleadd_ab_fp32 grouped_conv_fwd_scaleadd_ab_fp32.cpp)
target_link_libraries(client_grouped_convnd_fwd_scaleadd_ab_fp32 PRIVATE composable_kernel::device_conv_operations)
add_executable(client_grouped_convnd_fwd_scaleadd_ab_fp16 grouped_conv_fwd_scaleadd_ab_fp16.cpp)
target_link_libraries(client_grouped_convnd_fwd_scaleadd_ab_fp16 PRIVATE composable_kernel::device_conv_operations)
add_executable(client_grouped_convnd_fwd_scaleadd_ab_bf16 grouped_conv_fwd_scaleadd_ab_bf16.cpp)
target_link_libraries(client_grouped_convnd_fwd_scaleadd_ab_bf16 PRIVATE composable_kernel::device_conv_operations)
add_executable(client_grouped_convnd_fwd_scaleadd_ab_int8 grouped_conv_fwd_scaleadd_ab_int8.cpp)
target_link_libraries(client_grouped_convnd_fwd_scaleadd_ab_int8 PRIVATE composable_kernel::device_conv_operations)

8
client_example/25_wrapper/CMakeLists.txt
View File

@@ -2,3 +2,11 @@ add_executable(client_tensor_transform_using_wrapper tensor_transform_using_wrap
target_link_libraries(client_tensor_transform_using_wrapper PRIVATE composable_kernel::device_other_operations)
add_executable(client_wrapper_img2col wrapper_img2col.cpp)
target_link_libraries(client_wrapper_img2col PRIVATE composable_kernel::device_other_operations)
if(GPU_TARGETS MATCHES "gfx908" OR GPU_TARGETS MATCHES "gfx90a" OR
GPU_TARGETS MATCHES "gfx940" OR GPU_TARGETS MATCHES "gfx941" OR
GPU_TARGETS MATCHES "gfx942")
add_executable(client_wrapper_basic_gemm wrapper_basic_gemm.cpp)
target_link_libraries(client_wrapper_basic_gemm PRIVATE composable_kernel::device_other_operations)
add_executable(client_wrapper_optimized_gemm wrapper_optimized_gemm.cpp)
target_link_libraries(client_wrapper_optimized_gemm PRIVATE composable_kernel::device_other_operations)
endif()

177
client_example/25_wrapper/README.md Normal file
View File

@@ -0,0 +1,177 @@
# Composable Kernel wrapper GEMM tutorial
This tutorial demonstrates how to implement matrix multiplication using Composable Kernel (CK)
wrapper. We present the base version of GEMM without most of the available optimizations; however,
it's worth noting that CK has kernels with different optimizations.
To implement these optimizations, you can use the CK wrapper or directly use available instances in
CK. You can also refer to the
[optimized GEMM example](https://github.com/ROCm/composable_kernel/blob/develop/client_example/25_wrapper/wrapper_optimized_gemm.cpp),
that uses CK wrapper based on the
[`gridwise_gemm_xdlops_v2r3`](https://github.com/ROCm/composable_kernel/blob/develop/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v2r3.hpp) implementation.
The kernel definition should look similar to:
```cpp
template <typename DataType,
typename GemmTraits,
ck::index_t scalar_per_vector,
typename BlockShape,
typename ThreadLayout>
__global__ void __CK_WRAPPER_LAUNCH_BOUNDS__ DeviceGemm(const void* p_a,
const void* p_b,
void* p_c,
const ck::index_t M,
const ck::index_t N,
const ck::index_t K,
const BlockShape tile_shape,
const ThreadLayout thread_layout)
```
We pass pointers to global memory and matrix dimensions via arguments. Additionally, we pass
selected lengths of processed data through each block (`tile_shape`) and thread layout
(`thread_layout`). For compilation time parameters, we define the data type,
[traits for the GEMM operation](https://github.com/ROCm/composable_kernel/blob/develop/include/ck/wrapper/traits/blockwise_gemm_xdl_traits.hpp)
and scalar per vector value during copy.
Step 1: Create layouts for global and LDS memory.
```cpp
// Specify layouts for global memory.
const auto a_global_layout =
ck::wrapper::make_layout(ck::make_tuple(M, K), ck::make_tuple(K, 1));
const auto b_global_layout =
ck::wrapper::make_layout(ck::make_tuple(N, K), ck::make_tuple(K, 1));
const auto c_global_layout =
ck::wrapper::make_layout(ck::make_tuple(M, N), ck::make_tuple(N, 1));
// Specify layouts for tiles.
constexpr auto a_tile_layout = ck::wrapper::make_layout(
ck::make_tuple(MPerBlock, KPerBlock), ck::make_tuple(KPerBlock, ck::Number<1>{}));
constexpr auto b_tile_layout = ck::wrapper::make_layout(
ck::make_tuple(NPerBlock, KPerBlock), ck::make_tuple(KPerBlock, ck::Number<1>{}));
constexpr auto c_tile_layout = ck::wrapper::make_layout(
ck::make_tuple(MPerBlock, NPerBlock), ck::make_tuple(NPerBlock, ck::Number<1>{}));
// Apply padding for global memory.
auto a_global_layout_padded = ck::wrapper::pad(a_global_layout, shape(a_tile_layout));
auto b_global_layout_padded = ck::wrapper::pad(b_global_layout, shape(b_tile_layout));
auto c_global_layout_padded = ck::wrapper::pad(c_global_layout, shape(c_tile_layout));
```
We pad layouts for global tensors in case M, N, and K are not divisible by `MPerBlock`, `NPerBlock`, or
`KPerBlock`.
Step 2: Create tensors for global and LDS memory.
```cpp
// Make tensors for global memory.
auto a_global_tensor = ck::wrapper::make_tensor<ck::wrapper::MemoryTypeEnum::Global>(
static_cast<const DataType*>(p_a), a_global_layout_padded);
auto b_global_tensor = ck::wrapper::make_tensor<ck::wrapper::MemoryTypeEnum::Global>(
static_cast<const DataType*>(p_b), b_global_layout_padded);
auto c_global_tensor = ck::wrapper::make_tensor<ck::wrapper::MemoryTypeEnum::Global>(
static_cast<DataType*>(p_c), c_global_layout_padded);
// Allocate LDS memory.
__shared__ DataType lds_a[ck::wrapper::size(a_tile_layout)];
__shared__ DataType lds_b[ck::wrapper::size(b_tile_layout)];
// Make tensors for lds memory.
auto a_lds_tensor = ck::wrapper::make_tensor<ck::wrapper::MemoryTypeEnum::Lds>(
static_cast<DataType*>(lds_a), a_tile_layout);
auto b_lds_tensor = ck::wrapper::make_tensor<ck::wrapper::MemoryTypeEnum::Lds>(
static_cast<DataType*>(lds_b), b_tile_layout);
```
We must specify parameters for copy and convert block indexes to tuple:
```cpp
// Specify block index as tuple.
const auto block_idxs = ck::make_tuple(static_cast<ck::index_t>(blockIdx.x),
static_cast<ck::index_t>(blockIdx.y),
ck::wrapper::slice());
// Specify access parameters for copy.
using DimAccessOrder = ck::Tuple<ck::Number<0>, ck::Number<1>>;
constexpr ck::index_t vector_dim = 1;
```
We create a local tile (per block) and local partitions (per thread) for the global memory (`C`). We also
define and clear an output register (`c_vgpr_reg`) for the accumulation.
```cpp
auto c_global_local_tile = ck::wrapper::make_local_tile(
c_global_tensor,
tile_shape,
block_idxs,
make_tuple(ck::Number<1>{}, ck::Number<1>{}, ck::wrapper::slice(KPerBlock)));
auto c_global_local_partition =
ck::wrapper::make_blockwise_gemm_xdl_c_local_partition<DataType,
decltype(a_tile_layout),
decltype(b_tile_layout),
ck::wrapper::size(thread_layout),
GemmTraits>(c_global_local_tile);
// Create C vgpr to accumulate results.
auto c_vgpr_reg = ck::wrapper::make_blockwise_gemm_xdl_c_vgpr<DataType,
decltype(a_tile_layout),
decltype(b_tile_layout),
ck::wrapper::size(thread_layout),
GemmTraits>();
// Clear C vgpr.
ck::wrapper::clear(c_vgpr_reg);
```
We use two specific functions for `blockwise_gemm`: `make_blockwise_gemm_xdl_c_local_partition` and
`make_blockwise_gemm_xdl_c_vgpr`. This helps to choose the appropriate partition for the `C` output
and define tensors with specific layouts for `blockwise_gemm`. In the following step, we use only
generic functions for the CK wrapper.
Step 3: Create the compute loop.
```cpp
const ck::index_t num_loop = ck::math::integer_divide_ceil(K, KPerBlock);
ck::index_t i = 0;
do
{
// Get KPerBlock slice.
const auto k_slice = ck::wrapper::slice(i * KPerBlock, (i + 1) * KPerBlock);
auto a_global_tensor_k_slice = a_global_tensor(ck::wrapper::slice(), k_slice);
auto b_global_tensor_k_slice = b_global_tensor(ck::wrapper::slice(), k_slice);
// Create local tiles for A and B.
auto a_global_local_tile = ck::wrapper::make_local_tile(
a_global_tensor_k_slice,
tile_shape,
block_idxs,
make_tuple(ck::Number<1>{}, ck::wrapper::slice(N), ck::Number<1>{}));
auto b_global_local_tile = ck::wrapper::make_local_tile(
b_global_tensor_k_slice,
tile_shape,
block_idxs,
make_tuple(ck::wrapper::slice(M), ck::Number<1>{}, ck::Number<1>{}));
// Copy from global to LDS.
ck::wrapper::blockwise_copy<DimAccessOrder, vector_dim, scalar_per_vector>(
a_global_local_tile, a_lds_tensor, thread_layout);
ck::wrapper::blockwise_copy<DimAccessOrder, vector_dim, scalar_per_vector>(
b_global_local_tile, b_lds_tensor, thread_layout);
// Synchronize lds.
ck::block_sync_lds();
// Execute blockwise GEMM.
ck::wrapper::blockwise_gemm_xdl<DataType, ck::wrapper::size(thread_layout), GemmTraits>(
a_lds_tensor, b_lds_tensor, c_vgpr_reg);
++i;
} while(i < num_loop);
```
Loop iterate over `K / KPerBlock`. Each time a local tile is created for A and B tensors (tensor per block),
data is copied from global memory to LDS. The `blockwise_gemm` function performs the GEMM
operation on `a_lds_tensor` and `b_lds_tensor`, and stores results in `c_vgpr_reg`.
The end result from `c_vgpr_reg` is stored in the `C` local partition (tensor per thread):
```cpp
ck::wrapper::copy(c_vgpr_reg, c_global_local_partition);
```
If you want to dive deep into the details, you can find the entire example
[here](https://github.com/ROCm/composable_kernel/blob/develop/client_example/25_wrapper/wrapper_basic_gemm.cpp).

215
client_example/25_wrapper/wrapper_basic_gemm.cpp Normal file
View File

@@ -0,0 +1,215 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
#include <numeric>
#include <cstdlib>
#include <iostream>
#include <initializer_list>
#include <vector>
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/utility/host_tensor.hpp"
#include "ck/host_utility/kernel_launch.hpp"
#include "ck/library/utility/device_memory.hpp"
#include "ck/library/utility/check_err.hpp"
#include "ck/utility/common_header.hpp"
#include "ck/library/utility/fill.hpp"
#include "ck/wrapper/layout.hpp"
#include "ck/wrapper/tensor.hpp"
#include "ck/wrapper/operations/copy.hpp"
#include "ck/wrapper/operations/gemm.hpp"
#include "ck/wrapper/utils/kernel_utils.hpp"
struct SimpleDeviceMem
{
SimpleDeviceMem() = delete;
SimpleDeviceMem(std::size_t mem_size) : p_mem_{}
{
(void)hipMalloc(static_cast<void**>(&p_mem_), mem_size);
}
void* GetDeviceBuffer() { return p_mem_; }
~SimpleDeviceMem() { (void)hipFree(p_mem_); }
void* p_mem_;
};
template <typename DataType,
typename GemmTraits,
ck::index_t scalar_per_vector,
typename BlockShape,
typename ThreadLayout>
__global__ void __CK_WRAPPER_LAUNCH_BOUNDS__ DeviceGemm(const void* p_a,
const void* p_b,
void* p_c,
const ck::index_t M,
const ck::index_t N,
const ck::index_t K,
const BlockShape tile_shape,
const ThreadLayout thread_layout)
{
constexpr auto MPerBlock = ck::wrapper::size<0>(tile_shape);
constexpr auto NPerBlock = ck::wrapper::size<1>(tile_shape);
constexpr auto KPerBlock = ck::wrapper::size<2>(tile_shape);
// Specify layouts for global memory.
const auto a_global_layout =
ck::wrapper::make_layout(ck::make_tuple(M, K), ck::make_tuple(K, 1));
const auto b_global_layout =
ck::wrapper::make_layout(ck::make_tuple(N, K), ck::make_tuple(K, 1));
const auto c_global_layout =
ck::wrapper::make_layout(ck::make_tuple(M, N), ck::make_tuple(N, 1));
// Specify layouts for tiles.
constexpr auto a_tile_layout = ck::wrapper::make_layout(
ck::make_tuple(MPerBlock, KPerBlock), ck::make_tuple(KPerBlock, ck::Number<1>{}));
constexpr auto b_tile_layout = ck::wrapper::make_layout(
ck::make_tuple(NPerBlock, KPerBlock), ck::make_tuple(KPerBlock, ck::Number<1>{}));
constexpr auto c_tile_layout = ck::wrapper::make_layout(
ck::make_tuple(MPerBlock, NPerBlock), ck::make_tuple(NPerBlock, ck::Number<1>{}));
// Apply padding for global memory.
auto a_global_layout_padded = ck::wrapper::pad(a_global_layout, shape(a_tile_layout));
auto b_global_layout_padded = ck::wrapper::pad(b_global_layout, shape(b_tile_layout));
auto c_global_layout_padded = ck::wrapper::pad(c_global_layout, shape(c_tile_layout));
// Make tensors for global memory.
auto a_global_tensor = ck::wrapper::make_tensor<ck::wrapper::MemoryTypeEnum::Global>(
static_cast<const DataType*>(p_a), a_global_layout_padded);
auto b_global_tensor = ck::wrapper::make_tensor<ck::wrapper::MemoryTypeEnum::Global>(
static_cast<const DataType*>(p_b), b_global_layout_padded);
auto c_global_tensor = ck::wrapper::make_tensor<ck::wrapper::MemoryTypeEnum::Global>(
static_cast<DataType*>(p_c), c_global_layout_padded);
// Allocate lds memory.
__shared__ DataType lds_a[ck::wrapper::size(a_tile_layout)];
__shared__ DataType lds_b[ck::wrapper::size(b_tile_layout)];
// Make tensors for lds memory.
auto a_lds_tensor = ck::wrapper::make_tensor<ck::wrapper::MemoryTypeEnum::Lds>(
static_cast<DataType*>(lds_a), a_tile_layout);
auto b_lds_tensor = ck::wrapper::make_tensor<ck::wrapper::MemoryTypeEnum::Lds>(
static_cast<DataType*>(lds_b), b_tile_layout);
// Specify block index as tuple.
const auto block_idxs = ck::make_tuple(static_cast<ck::index_t>(blockIdx.x),
static_cast<ck::index_t>(blockIdx.y),
ck::wrapper::slice());
// Specify access parameters for copy.
using DimAccessOrder = ck::Tuple<ck::Number<0>, ck::Number<1>>;
constexpr ck::index_t vector_dim = 1;
// Create tile and partition for C. Use specific function for blockwise_gemm to assign the
// appropriate partitions.
auto c_global_local_tile = ck::wrapper::make_local_tile(
c_global_tensor,
tile_shape,
block_idxs,
make_tuple(ck::Number<1>{}, ck::Number<1>{}, ck::wrapper::slice(KPerBlock)));
auto c_global_local_partition =
ck::wrapper::make_blockwise_gemm_xdl_c_local_partition<DataType,
decltype(a_tile_layout),
decltype(b_tile_layout),
ck::wrapper::size(thread_layout),
GemmTraits>(c_global_local_tile);
// Create C vgpr to accumulate results.
auto c_vgpr_reg = ck::wrapper::make_blockwise_gemm_xdl_c_vgpr<DataType,
decltype(a_tile_layout),
decltype(b_tile_layout),
ck::wrapper::size(thread_layout),
GemmTraits>();
// Clear C vgpr.
ck::wrapper::clear(c_vgpr_reg);
// Iterate over K with KPerBlock step.
const ck::index_t num_loop = ck::math::integer_divide_ceil(K, KPerBlock);
ck::index_t i = 0;
do
{
// Get KPerBlock slice.
const auto k_slice = ck::wrapper::slice(i * KPerBlock, (i + 1) * KPerBlock);
auto a_global_tensor_k_slice = a_global_tensor(ck::wrapper::slice(), k_slice);
auto b_global_tensor_k_slice = b_global_tensor(ck::wrapper::slice(), k_slice);
// Create local tiles for A and B.
auto a_global_local_tile = ck::wrapper::make_local_tile(
a_global_tensor_k_slice,
tile_shape,
block_idxs,
make_tuple(ck::Number<1>{}, ck::wrapper::slice(N), ck::Number<1>{}));
auto b_global_local_tile = ck::wrapper::make_local_tile(
b_global_tensor_k_slice,
tile_shape,
block_idxs,
make_tuple(ck::wrapper::slice(M), ck::Number<1>{}, ck::Number<1>{}));
// Copy from global to lds.
ck::wrapper::blockwise_copy<DimAccessOrder, vector_dim, scalar_per_vector>(
a_global_local_tile, a_lds_tensor, thread_layout);
ck::wrapper::blockwise_copy<DimAccessOrder, vector_dim, scalar_per_vector>(
b_global_local_tile, b_lds_tensor, thread_layout);
// Synchronize lds.
ck::block_sync_lds();
// Execute blockwise gemm.
ck::wrapper::blockwise_gemm_xdl<DataType, ck::wrapper::size(thread_layout), GemmTraits>(
a_lds_tensor, b_lds_tensor, c_vgpr_reg);
++i;
} while(i < num_loop);
// Copy vgpr results to C global memory.
ck::wrapper::copy(c_vgpr_reg, c_global_local_partition);
}
template <typename DataType,
typename GemmTraits,
ck::index_t scalar_per_vector,
typename BlockShape,
typename ThreadLayout>
void PerformGemm(const ck::index_t M,
const ck::index_t N,
const ck::index_t K,
const BlockShape& tile_shape,
const ThreadLayout& thread_layout)
{
// Global memory buffers
SimpleDeviceMem a_mem(M * K * sizeof(DataType));
SimpleDeviceMem b_mem(K * N * sizeof(DataType));
SimpleDeviceMem c_mem(M * N * sizeof(DataType));
const ck::index_t grid_size_x =
ck::math::integer_divide_ceil(M, ck::wrapper::size<0>(tile_shape));
const ck::index_t grid_size_y =
ck::math::integer_divide_ceil(N, ck::wrapper::size<1>(tile_shape));
const auto kernel =
DeviceGemm<DataType, GemmTraits, scalar_per_vector, BlockShape, ThreadLayout>;
const float avg_time = launch_and_time_kernel(StreamConfig{nullptr, true},
kernel,
dim3(grid_size_x, grid_size_y, 1),
dim3(ck::wrapper::size(thread_layout)),
0,
a_mem.GetDeviceBuffer(),
b_mem.GetDeviceBuffer(),
c_mem.GetDeviceBuffer(),
M,
N,
K,
tile_shape,
thread_layout);
std::size_t flop = std::size_t(2) * M * N * K;
std::size_t num_btype =
sizeof(DataType) * M * K + sizeof(DataType) * K * N + sizeof(DataType) * M * N;
float tflops = static_cast<float>(flop) / 1.E9 / avg_time;
float gb_per_sec = num_btype / 1.E6 / avg_time;
std::cout << "Perf: " << std::setw(10) << avg_time << " ms, " << tflops << " TFlops, "
<< gb_per_sec << " GB/s, " << std::endl;
}
int main(int argc, char* argv[])
{
using DataType = ck::half_t;
const auto thread_layout =
ck::wrapper::make_layout(ck::make_tuple(ck::Number<64>{}, ck::Number<4>{}),
ck::make_tuple(ck::Number<4>{}, ck::Number<1>{}));
const auto tile_shape = ck::make_tuple(ck::Number<256>{}, ck::Number<128>{}, ck::Number<32>{});
PerformGemm<DataType, ck::wrapper::BlockwisGemmXdlTraits_32x32Xdl_4x2XdlPerWave_8K1, 8>(
3840, 4096, 4096, tile_shape, thread_layout);
return 0;
}

42
client_example/25_wrapper/wrapper_img2col.cpp
View File

@@ -15,6 +15,7 @@
#include "ck/wrapper/layout.hpp"
#include "ck/wrapper/tensor.hpp"
#include "ck/wrapper/operations/copy.hpp"
#include "ck/wrapper/utils/kernel_utils.hpp"
static constexpr ck::index_t NumDimSpatial = 3;
using DataType = float;
@@ -36,21 +37,20 @@ struct SimpleDeviceMem
void* p_mem_;
};
// Test copy from Global to Global through LDS and VGPR
template <typename InputTensor,
typename OutputTensor,
typename BlockShape,
typename ThreadLayoutShape>
__global__ void DeviceImageToColumnPad0(InputTensor input_tensor,
OutputTensor output_tensor,
const BlockShape tile_shape,
const ThreadLayoutShape thread_layout)
template <typename InputTensor, typename OutputTensor, typename BlockShape, typename ThreadLayout>
__global__ void __CK_WRAPPER_LAUNCH_BOUNDS__
DeviceImageToColumnPad0(InputTensor input_tensor,
OutputTensor output_tensor,
const BlockShape tile_shape,
const ThreadLayout thread_layout)
{
const ck::index_t block_idx = static_cast<ck::index_t>(blockIdx.x);
// grid layout (dim1, dim0)
const auto block_idxs =
ck::make_tuple(static_cast<ck::index_t>(blockIdx.y), static_cast<ck::index_t>(blockIdx.x));
// Get local tiles for global memory
auto input_local_tile = ck::wrapper::make_local_tile(input_tensor, tile_shape, block_idx);
auto output_local_tile = ck::wrapper::make_local_tile(output_tensor, tile_shape, block_idx);
auto input_local_tile = ck::wrapper::make_local_tile(input_tensor, tile_shape, block_idxs);
auto output_local_tile = ck::wrapper::make_local_tile(output_tensor, tile_shape, block_idxs);
// Get partition per thread
const auto input_local_partition =
@@ -112,9 +112,11 @@ void PerformImageToColumnPad0(const ck::index_t G,
SimpleDeviceMem out_buf(ck::wrapper::size(out_layout) * sizeof(DataType));
// User can choose appropriate number of threads and sizes per block
const auto thread_layout = ck::make_tuple(ck::Number<8>{}, ck::Number<16>{});
const auto thread_layout =
ck::wrapper::make_layout(ck::make_tuple(ck::Number<8>{}, ck::Number<16>{}),
ck::make_tuple(ck::Number<16>{}, ck::Number<1>{}));
// This example doesn't support padding, user should select tile sizes
// which divides the shape completely
// which are divisible by the shape.
const auto tile_shape = ck::make_tuple(ck::Number<32>{}, ck::Number<64>{});
// Create buffers for global memory
@@ -123,10 +125,11 @@ void PerformImageToColumnPad0(const ck::index_t G,
auto output_tensor_global = ck::wrapper::make_tensor<ck::wrapper::MemoryTypeEnum::Global>(
static_cast<DataType*>(out_buf.GetDeviceBuffer()), out_layout);
const ck::index_t grid_size = ck::math::integer_divide_ceil(ck::wrapper::size<0>(in_layout),
ck::wrapper::size<0>(tile_shape)) *
ck::math::integer_divide_ceil(ck::wrapper::size<1>(in_layout),
ck::wrapper::size<1>(tile_shape));
// grid layout (dim1, dim0)
const ck::index_t grid_size_x = ck::math::integer_divide_ceil(ck::wrapper::size<1>(in_layout),
ck::wrapper::size<1>(tile_shape));
const ck::index_t grid_size_y = ck::math::integer_divide_ceil(ck::wrapper::size<0>(in_layout),
ck::wrapper::size<0>(tile_shape));
const auto kernel = DeviceImageToColumnPad0<decltype(input_tensor_global),
decltype(output_tensor_global),
@@ -134,7 +137,7 @@ void PerformImageToColumnPad0(const ck::index_t G,
decltype(thread_layout)>;
const float avg_time = launch_and_time_kernel(StreamConfig{nullptr, true},
kernel,
dim3(grid_size),
dim3(grid_size_x, grid_size_y, 1),
dim3(ck::wrapper::size(thread_layout)),
0,
input_tensor_global,
@@ -178,3 +181,4 @@ int main(int argc, char* argv[])
{1, 1, 1} /*filter_dilations*/);
return 0;
}
// MI100 Perf: 0.255178 ms, 1698.9 GB/s,

307
client_example/25_wrapper/wrapper_optimized_gemm.cpp Normal file
View File

@@ -0,0 +1,307 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
#include <numeric>
#include <cstdlib>
#include <iostream>
#include <initializer_list>
#include <vector>
#include "ck/library/utility/host_tensor.hpp"
#include "ck/host_utility/kernel_launch.hpp"
#include "ck/library/utility/device_memory.hpp"
#include "ck/library/utility/check_err.hpp"
#include "ck/utility/common_header.hpp"
#include "ck/library/utility/fill.hpp"
#include "ck/wrapper/layout.hpp"
#include "ck/wrapper/tensor.hpp"
#include "ck/wrapper/operations/copy.hpp"
#include "ck/wrapper/operations/gemm.hpp"
#include "ck/wrapper/utils/kernel_utils.hpp"
struct SimpleDeviceMem
{
SimpleDeviceMem() = delete;
SimpleDeviceMem(std::size_t mem_size) : p_mem_{}
{
(void)hipMalloc(static_cast<void**>(&p_mem_), mem_size);
}
void* GetDeviceBuffer() { return p_mem_; }
~SimpleDeviceMem() { (void)hipFree(p_mem_); }
void* p_mem_;
};
template <bool DoPad, typename Layout, typename PaddingDims>
__device__ auto ApplyPadding(const Layout& layout, const PaddingDims& padding_dims)
{
if constexpr(DoPad)
{
return ck::wrapper::pad(layout, padding_dims);
}
else
{
return layout;
}
}
template <typename DataType,
typename GemmTraits,
ck::index_t scalar_per_vector,
typename BlockShape,
typename ThreadLayout,
bool DoPadding>
__global__ void __CK_WRAPPER_LAUNCH_BOUNDS__ DeviceGemm(const void* p_a,
const void* p_b,
void* p_c,
const ck::index_t M,
const ck::index_t N,
const ck::index_t K,
const BlockShape tile_shape,
const ThreadLayout thread_layout)
{
constexpr auto MPerBlock = ck::wrapper::size<0>(tile_shape);
constexpr auto NPerBlock = ck::wrapper::size<1>(tile_shape);
constexpr auto KPerBlock = ck::wrapper::size<2>(tile_shape);
constexpr auto K1 = GemmTraits::K1;
constexpr auto K0PerBlock = KPerBlock / K1;
const auto K0 = ck::math::integer_divide_ceil(K, K1);
const auto tile_shape_k0_m_n_k1 = ck::make_tuple(K0PerBlock, MPerBlock, NPerBlock, K1);
// Create layouts for global memory
const auto a_global_layout =
ck::wrapper::make_layout(ck::make_tuple(M, K), ck::make_tuple(K, 1));
const auto b_global_layout =
ck::wrapper::make_layout(ck::make_tuple(N, K), ck::make_tuple(K, 1));
const auto c_global_layout =
ck::wrapper::make_layout(ck::make_tuple(M, N), ck::make_tuple(N, 1));
// Apply padding
auto a_padded_global_layout =
ApplyPadding<DoPadding>(a_global_layout, ck::make_tuple(MPerBlock, KPerBlock));
auto b_padded_global_layout =
ApplyPadding<DoPadding>(b_global_layout, ck::make_tuple(NPerBlock, KPerBlock));
auto c_padded_global_layout =
ApplyPadding<DoPadding>(c_global_layout, ck::make_tuple(MPerBlock, NPerBlock));
// Reshape from M,K to K0,M,K1
const auto reshaped_dims_idxs =
ck::make_tuple(ck::Number<1>{}, ck::make_tuple(ck::Number<0>{}, ck::Number<2>{}));
auto a_padded_unmerged_global_layout =
ck::wrapper::unmerge<1>(a_padded_global_layout, ck::make_tuple(K0, K1), reshaped_dims_idxs);
auto b_padded_unmerged_global_layout =
ck::wrapper::unmerge<1>(b_padded_global_layout, ck::make_tuple(K0, K1), reshaped_dims_idxs);
// Create tensors for global memory
auto a_global_tensor = ck::wrapper::make_tensor<ck::wrapper::MemoryTypeEnum::Global>(
static_cast<const DataType*>(p_a), a_padded_unmerged_global_layout);
auto b_global_tensor = ck::wrapper::make_tensor<ck::wrapper::MemoryTypeEnum::Global>(
static_cast<const DataType*>(p_b), b_padded_unmerged_global_layout);
auto c_global_tensor = ck::wrapper::make_tensor<ck::wrapper::MemoryTypeEnum::Global>(
static_cast<DataType*>(p_c), c_padded_global_layout);
// Create layouts and tensors for lds memory.
constexpr auto a_tile_layout = ck::wrapper::make_layout(
ck::make_tuple(K0PerBlock, MPerBlock, K1),
ck::make_tuple((MPerBlock + ck::Number<1>{}) * K1, K1, ck::Number<1>{}));
constexpr auto b_tile_layout = ck::wrapper::make_layout(
ck::make_tuple(K0PerBlock, NPerBlock, K1),
ck::make_tuple((NPerBlock + ck::Number<1>{}) * K1, K1, ck::Number<1>{}));
__shared__ DataType lds_a[ck::wrapper::size(a_tile_layout) + K0PerBlock];
__shared__ DataType lds_b[ck::wrapper::size(b_tile_layout) + K0PerBlock];
auto a_lds_tensor = ck::wrapper::make_tensor<ck::wrapper::MemoryTypeEnum::Lds>(
static_cast<DataType*>(lds_a), a_tile_layout);
auto b_lds_tensor = ck::wrapper::make_tensor<ck::wrapper::MemoryTypeEnum::Lds>(
static_cast<DataType*>(lds_b), b_tile_layout);
const auto block_idxs = ck::make_tuple(ck::wrapper::slice(),
static_cast<ck::index_t>(blockIdx.x),
static_cast<ck::index_t>(blockIdx.y),
ck::wrapper::slice());
using DimAccessOrder = ck::Tuple<ck::Number<1>, ck::Number<0>, ck::Number<2>>;
constexpr ck::index_t vector_dim = 2;
// Create tile and partition for C global memory. Use specific gemm
// functions to get appropriate layouts.
auto c_global_local_tile =
ck::wrapper::make_local_tile(c_global_tensor,
tile_shape_k0_m_n_k1,
block_idxs,
make_tuple(ck::wrapper::slice(K0PerBlock),
ck::Number<1>{},
ck::Number<1>{},
ck::wrapper::slice(K1)));
auto c_global_local_partition =
ck::wrapper::make_blockwise_gemm_xdl_c_local_partition<DataType,
decltype(a_tile_layout),
decltype(b_tile_layout),
ck::wrapper::size(thread_layout),
GemmTraits>(c_global_local_tile);
// Define and clear c vgpr register
auto c_vgpr_reg = ck::wrapper::make_blockwise_gemm_xdl_c_vgpr<DataType,
decltype(a_tile_layout),
decltype(b_tile_layout),
ck::wrapper::size(thread_layout),
GemmTraits>();
ck::wrapper::clear(c_vgpr_reg);
// Local partitions for lds memory
auto a_lds_tensor_local_partition =
ck::wrapper::make_local_partition(a_lds_tensor, thread_layout, threadIdx.x);
auto b_lds_tensor_local_partition =
ck::wrapper::make_local_partition(b_lds_tensor, thread_layout, threadIdx.x);
// Lamda to slice tensor, then create local tile and partition
auto make_global_partition = [&](auto tensor, auto projection, ck::index_t i) {
const auto k_slice =
ck::make_tuple(ck::wrapper::slice(i * K0PerBlock, (i + 1) * K0PerBlock),
ck::wrapper::slice(),
ck::wrapper::slice());
auto local_tile = ck::wrapper::make_local_tile(
tensor(k_slice), tile_shape_k0_m_n_k1, block_idxs, projection);
return ck::wrapper::make_local_partition(local_tile, thread_layout, threadIdx.x);
};
auto a_global_local_partition = make_global_partition(
a_global_tensor,
make_tuple(ck::Number<1>{}, ck::Number<1>{}, ck::wrapper::slice(N), ck::Number<1>{}),
0);
auto b_global_local_partition = make_global_partition(
b_global_tensor,
make_tuple(ck::Number<1>{}, ck::wrapper::slice(M), ck::Number<1>{}, ck::Number<1>{}),
0);
// (row-major vgpr layout)
auto a_vgpr_tensor =
ck::wrapper::make_register_tensor<ck::wrapper::MemoryTypeEnum::Vgpr, DataType>(
ck::wrapper::make_layout(
shape(a_global_local_partition),
ck::make_tuple(ck::wrapper::size<1>(a_global_local_partition) *
ck::wrapper::size<2>(a_global_local_partition),
ck::wrapper::size<2>(a_global_local_partition),
ck::Number<1>{})));
auto b_vgpr_tensor =
ck::wrapper::make_register_tensor<ck::wrapper::MemoryTypeEnum::Vgpr, DataType>(
ck::wrapper::make_layout(
shape(b_global_local_partition),
ck::make_tuple(ck::wrapper::size<1>(a_global_local_partition) *
ck::wrapper::size<2>(a_global_local_partition),
ck::wrapper::size<2>(a_global_local_partition),
ck::Number<1>{})));
// Copy first values to lds
ck::wrapper::copy<DimAccessOrder, vector_dim, scalar_per_vector>(a_global_local_partition,
a_vgpr_tensor);
ck::wrapper::copy<DimAccessOrder, vector_dim, scalar_per_vector>(b_global_local_partition,
b_vgpr_tensor);
ck::wrapper::copy<DimAccessOrder, vector_dim, scalar_per_vector>(a_vgpr_tensor,
a_lds_tensor_local_partition);
ck::wrapper::copy<DimAccessOrder, vector_dim, scalar_per_vector>(b_vgpr_tensor,
b_lds_tensor_local_partition);
// Pipeline loop
const ck::index_t num_loop =
__builtin_amdgcn_readfirstlane(ck::math::integer_divide_ceil(K, KPerBlock));
// Skip if only tile should be processed
if(num_loop > 1)
{
ck::index_t i = 0;
do
{
auto a_global_local_partition_i = make_global_partition(
a_global_tensor,
make_tuple(
ck::Number<1>{}, ck::Number<1>{}, ck::wrapper::slice(N), ck::Number<1>{}),
i + 1);
auto b_global_local_partition_i = make_global_partition(
b_global_tensor,
make_tuple(
ck::Number<1>{}, ck::wrapper::slice(M), ck::Number<1>{}, ck::Number<1>{}),
i + 1);
// Copy data to A vgpr.
ck::wrapper::copy<DimAccessOrder, vector_dim, scalar_per_vector>(
a_global_local_partition_i, a_vgpr_tensor);
// Synchronize.
ck::block_sync_lds();
// Copy data to B vgpr.
ck::wrapper::copy<DimAccessOrder, vector_dim, scalar_per_vector>(
b_global_local_partition_i, b_vgpr_tensor);
// Perform gemm.
ck::wrapper::blockwise_gemm_xdl<DataType, ck::wrapper::size(thread_layout), GemmTraits>(
a_lds_tensor, b_lds_tensor, c_vgpr_reg);
// Synchronize
ck::block_sync_lds();
// Copy data to A and B lds tiles.
ck::wrapper::copy<DimAccessOrder, vector_dim, scalar_per_vector>(
a_vgpr_tensor, a_lds_tensor_local_partition);
ck::wrapper::copy<DimAccessOrder, vector_dim, scalar_per_vector>(
b_vgpr_tensor, b_lds_tensor_local_partition);
++i;
} while(i < (num_loop - 1));
}
// Handle tail.
ck::block_sync_lds();
ck::wrapper::blockwise_gemm_xdl<DataType, ck::wrapper::size(thread_layout), GemmTraits>(
a_lds_tensor, b_lds_tensor, c_vgpr_reg);
// Store data from C vgpr to C global memory.
ck::wrapper::copy(c_vgpr_reg, c_global_local_partition);
}
template <typename DataType,
typename GemmTraits,
ck::index_t scalar_per_vector,
bool DoPadding,
typename BlockShape,
typename ThreadLayout>
void PerformGemm(const ck::index_t M,
const ck::index_t N,
const ck::index_t K,
const BlockShape& tile_shape,
const ThreadLayout& thread_layout)
{
// Global memory buffers
SimpleDeviceMem a_mem(M * K * sizeof(DataType));
SimpleDeviceMem b_mem(K * N * sizeof(DataType));
SimpleDeviceMem c_mem(M * N * sizeof(DataType));
const ck::index_t grid_size_x =
ck::math::integer_divide_ceil(M, ck::wrapper::size<0>(tile_shape));
const ck::index_t grid_size_y =
ck::math::integer_divide_ceil(N, ck::wrapper::size<1>(tile_shape));
const auto kernel =
DeviceGemm<DataType, GemmTraits, scalar_per_vector, BlockShape, ThreadLayout, DoPadding>;
const float avg_time = launch_and_time_kernel(StreamConfig{nullptr, true},
kernel,
dim3(grid_size_x, grid_size_y, 1),
dim3(ck::wrapper::size(thread_layout)),
0,
a_mem.GetDeviceBuffer(),
b_mem.GetDeviceBuffer(),
c_mem.GetDeviceBuffer(),
M,
N,
K,
tile_shape,
thread_layout);
std::size_t flop = std::size_t(2) * M * N * K;
std::size_t num_btype =
sizeof(DataType) * M * K + sizeof(DataType) * K * N + sizeof(DataType) * M * N;
float tflops = static_cast<float>(flop) / 1.E9 / avg_time;
float gb_per_sec = num_btype / 1.E6 / avg_time;
std::cout << "Perf: " << std::setw(10) << avg_time << " ms, " << tflops << " TFlops, "
<< gb_per_sec << " GB/s, " << std::endl;
}
int main(int argc, char* argv[])
{
using DataType = ck::half_t;
const auto thread_layout =
ck::wrapper::make_layout(ck::make_tuple(ck::Number<4>{}, ck::Number<64>{}, ck::Number<1>{}),
ck::make_tuple(ck::Number<1>{}, ck::Number<4>{}, ck::Number<1>{}));
const auto tile_shape = ck::make_tuple(ck::Number<256>{}, ck::Number<128>{}, ck::Number<32>{});
PerformGemm<DataType, ck::wrapper::BlockwisGemmXdlTraits_32x32Xdl_4x2XdlPerWave_8K1, 8, false>(
3840, 4096, 4096, tile_shape, thread_layout);
return 0;
}

2
dev-requirements.txt
View File

@@ -1,3 +1,3 @@
ROCm/rocm-recipes
RadeonOpenCompute/rocm-cmake@04f694df2a8dc9d7e35fa4dee4ba5fa407ec04f8 --build
ROCm/rocm-cmake@04f694df2a8dc9d7e35fa4dee4ba5fa407ec04f8 --build
danmar/cppcheck@2.9

8
docs/Contributors_Guide.rst
View File

@@ -17,7 +17,7 @@ Getting started
`Composable Kernel User Guide <https://rocm.docs.amd.com/projects/composable_kernel/en/latest/>`_.
It provides insight into the core concepts, environment configuration, and steps to obtain or
build the library. You can also find some of this information in the
`README file <https://github.com/ROCmSoftwarePlatform/composable_kernel/blob/develop/README.md>`_
`README file <https://github.com/ROCm/composable_kernel/blob/develop/README.md>`_
on the project's GitHub page.
#. **Additional reading:** The blog post `AMD Composable Kernel library: efficient fused kernels for AI apps with just a few lines of code <https://community.amd.com/t5/instinct-accelerators/amd-composable-kernel-library-efficient-fused-kernels-for-ai/ba-p/553224>`_ provides a deeper understanding of the CK library and showcases its performance capabilities.
<https://community.amd.com/t5/instinct-accelerators/amd-composable-kernel-library-efficient-fused-kernels-for-ai/ba-p/553224>`_
@@ -33,7 +33,7 @@ You can make an impact by reporting issues or proposing code enhancements throug
Reporting issues
----------------
Use `Github issues <https://github.com/ROCmSoftwarePlatform/composable_kernel/issues>`_
Use `Github issues <https://github.com/ROCm/composable_kernel/issues>`_
to track public bugs and enhancement requests.
If you encounter an issue with the library, please check if the problem has already been
@@ -68,7 +68,7 @@ Creating Pull Requests
----------------------
You can submit `Pull Requests (PR) on GitHub
<https://github.com/ROCmSoftwarePlatform/composable_kernel/pulls>`_.
<https://github.com/ROCm/composable_kernel/pulls>`_.
All contributors are required to develop their changes on a separate branch and then create a
pull request to merge their changes into the `develop` branch, which is the default
@@ -89,7 +89,7 @@ When submitting a Pull Request you should:
the project's root directory. We leverage `pre-commit` to run `clang-format` automatically. We
highly recommend contributors utilize this method to maintain consistent code formatting.
Instructions on setting up `pre-commit` can be found in the project's
`README file <https://github.com/ROCmSoftwarePlatform/composable_kernel/blob/develop/README.md>`_
`README file <https://github.com/ROCm/composable_kernel/blob/develop/README.md>`_
* Link your PR to any related issues:

4
docs/dockerhub.rst
View File

@@ -38,7 +38,7 @@ The docker images have everything you need for running CK including:
* `ROCm <https://www.amd.com/en/graphics/servers-solutions-rocm>`_
* `CMake <https://cmake.org/getting-started/>`_
* `Compiler <https://github.com/RadeonOpenCompute/llvm-project>`_
* `Compiler <https://github.com/ROCm/llvm-project>`_
* `Composable Kernel library <https://github.com/ROCm/composable_kernel>`_
Running the docker container
@@ -97,5 +97,5 @@ Editing the docker image
=======================
If you want to customize the docker image, edit the
`Dockerfile <https://github.com/ROCmSoftwarePlatform/composable_kernel/blob/develop/Dockerfile>`_
`Dockerfile <https://github.com/ROCm/composable_kernel/blob/develop/Dockerfile>`_
from the GitHub repository to suit your needs.

2
docs/sphinx/requirements.in
View File

@@ -1,2 +1,2 @@
rocm-docs-core==0.34.0
rocm-docs-core==0.35.0
sphinxcontrib-bibtex==2.6.2

2
docs/sphinx/requirements.txt
View File

@@ -113,7 +113,7 @@ requests==2.31.0
# via
# pygithub
# sphinx
rocm-docs-core==0.34.0
rocm-docs-core==0.35.0
# via -r requirements.in
six==1.16.0
# via

4
docs/tutorial_hello_world.rst
View File

@@ -32,7 +32,7 @@ CK library acceleration features are based on:
If you need more technical details and benchmarking results read the following
`blog post <https://community.amd.com/t5/instinct-accelerators/amd-composable-kernel-library-efficient-fused-kernels-for-ai/ba-p/553224>`_.
To download the library visit the `composable_kernel repository <https://github.com/ROCmSoftwarePlatform/composable_kernel>`_.
To download the library visit the `composable_kernel repository <https://github.com/ROCm/composable_kernel>`_.
Hardware targets
================
@@ -58,7 +58,7 @@ This tutorial is based on the use of docker images as explained in :ref:`docker-
.. note::
You can also `install ROCm <https://rocm.docs.amd.com/projects/install-on-linux/en/latest/>`_ on your system, clone the `Composable Kernel repository <https://github.com/ROCmSoftwarePlatform/composable_kernel.git>`_ on GitHub, and use that to build and run the examples using the commands described below.
You can also `install ROCm <https://rocm.docs.amd.com/projects/install-on-linux/en/latest/>`_ on your system, clone the `Composable Kernel repository <https://github.com/ROCm/composable_kernel.git>`_ on GitHub, and use that to build and run the examples using the commands described below.
Both the docker container and GitHub repository include the Composable Kernel library. Navigate to the library::

10
docs/wrapper.rst
View File

@@ -12,10 +12,6 @@ Wrapper
Description
-------------------------------------
.. note::
The wrapper is under development and its functionality is limited.
The CK library provides a lightweight wrapper for more complex operations implemented in
the library.
@@ -54,9 +50,15 @@ Output::
2 6 10 14 18 22 26 30
Tutorials:
* `GEMM tutorial <https://github.com/ROCm/composable_kernel/blob/develop/client_example/25_wrapper/README.md>`_
Advanced examples:
* `Image to column <https://github.com/ROCm/composable_kernel/blob/develop/client_example/25_wrapper/wrapper_img2col.cpp>`_
* `Basic gemm <https://github.com/ROCm/composable_kernel/blob/develop/client_example/25_wrapper/wrapper_basic_gemm.cpp>`_
* `Optimized gemm <https://github.com/ROCm/composable_kernel/blob/develop/client_example/25_wrapper/wrapper_optimized_gemm.cpp>`_
-------------------------------------
Layout

2
example/01_gemm/common.hpp
View File

@@ -49,7 +49,7 @@ struct ProblemSizeStreamK final
struct ExecutionConfig final
{
bool do_verification = true;
int init_method = 1;
int init_method = 2;
bool time_kernel = false;
};

7
example/01_gemm/run_gemm_example.inc
View File

@@ -69,8 +69,8 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
ck::utils::FillUniformDistributionIntegerValue<BDataType>{-5.f, 5.f}(b_k_n);
break;
default:
ck::utils::FillUniformDistribution<ADataType>{-1.f, 1.f}(a_m_k);
ck::utils::FillUniformDistribution<BDataType>{-1.f, 1.f}(b_k_n);
ck::utils::FillUniformDistribution<ADataType>{-0.1f, 0.1f}(a_m_k);
ck::utils::FillUniformDistribution<BDataType>{-0.1f, 0.1f}(b_k_n);
}
Tensor<CDataType> c_m_n_host_result(f_host_tensor_descriptor(M, N, StrideC, CLayout{}));
@@ -240,7 +240,8 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
#else
c_m_n_device_buf.FromDevice(c_m_n_device_result.mData.data());
return ck::utils::check_err(c_m_n_device_result, c_m_n_host_result);
return ck::utils::check_err(
c_m_n_device_result, c_m_n_host_result, "Error: Incorrect results!", 1e-1, 1e-1);
#endif
}

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); }

Some files were not shown because too many files have changed in this diff Show More