Skip to content

[Paddle Inference]cutlass kernel compile optimization #64641

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
zhoutianzi666 merged 9 commits into from
May 31, 2024
Merged

[Paddle Inference]cutlass kernel compile optimization #64641

Show file tree
Hide file tree
Changes from 5 commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
b7a05f0
cutlass compile optimization
YKTian-x2b May 27, 2024
15bca85
compile cutlass only once
YKTian-x2b May 28, 2024
ab440cb
with default params
YKTian-x2b May 28, 2024
7619c71
CMake version=3.18
YKTian-x2b May 29, 2024
669cdbf
bf16 headfile miss
YKTian-x2b May 29, 2024
364f6cc
use paddle cmake
YKTian-x2b May 29, 2024
3138fe5
Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into…
YKTian-x2b May 30, 2024
cb218b1
restrict gpu arch
YKTian-x2b May 30, 2024
253b4e7
for case 80;86
YKTian-x2b May 31, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 15 additions & 0 deletions paddle/phi/kernels/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -55,6 +55,21 @@ if(DEFINED REDUCE_INFERENCE_LIB_SIZE)
endif()

if(WITH_CUTLASS)
add_custom_target(
gemm_epilogue_compile_script ALL
COMMAND bash compile.sh "${PYTHON_EXECUTABLE}" "${CUDA_TOOLKIT_ROOT_DIR}"
"${NVCC_ARCH_BIN}"
WORKING_DIRECTORY
${PADDLE_SOURCE_DIR}/paddle/phi/kernels/fusion/cutlass/gemm_epilogue
COMMENT "GemmEpilogue compile script")
add_custom_target(
fused_conv2d_add_act_compile_script ALL
COMMAND bash compile.sh "${PYTHON_EXECUTABLE}" "${CUDA_TOOLKIT_ROOT_DIR}"
"${NVCC_ARCH_BIN}"
WORKING_DIRECTORY
${PADDLE_SOURCE_DIR}/paddle/phi/kernels/fusion/cutlass/conv2d
COMMENT "FusedConv2dAddAct compile script")

execute_process(
COMMAND
${PYTHON_EXECUTABLE}
Expand Down
2 changes: 1 addition & 1 deletion paddle/phi/kernels/fusion/cutlass/conv2d/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
cmake_minimum_required(VERSION 3.23)
cmake_minimum_required(VERSION 3.18)

if(NOT DEFINED PYTHON_EXECUTABLE)
message(
Expand Down
26 changes: 18 additions & 8 deletions paddle/phi/kernels/fusion/cutlass/conv2d/compile.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,21 +13,31 @@
# limitations under the License.
set -e

build_directory="build"
if [ ! -d "$build_directory" ]; then
mkdir $build_directory
fi

libname="$build_directory/libCutlassConv2d.so"
if [ -e "$libname" ]; then
exit 0
fi

cutlass_repo_directory="cutlass"
if [ ! -d "$cutlass_repo_directory" ]; then
git clone --branch v3.0.0 https://github.com/NVIDIA/cutlass
fi

build_directory="build"
if [ ! -d "$build_directory" ]; then
mkdir $build_directory
fi

python_exe_path="python"
cuda_root_path="/usr/local/cuda"
gpu_cc="80"
default_python_exe_path="/usr/bin/python"
default_cuda_root_path="/usr/local/cuda"
default_gpu_cc="80"

python_exe_path="${1:-$default_python_exe_path}"
cuda_root_path="${2:-$default_cuda_root_path}"
gpu_cc="${3:-$default_gpu_cc}"

cd $build_directory
cmake .. -DPYTHON_EXECUTABLE=$python_exe_path -DCUDA_TOOLKIT_ROOT_DIR=$cuda_root_path -DCOMPUTE_CAPABILITY=$gpu_cc
Copy link
Contributor

@vivienfanghuagood vivienfanghuagood May 29, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

这里是否应该用paddle里默认的cmake

Copy link
Contributor Author

@YKTian-x2b YKTian-x2b May 29, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

get！马上改。

make -j
make -j8
cd -
2 changes: 1 addition & 1 deletion paddle/phi/kernels/fusion/cutlass/gemm_epilogue/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
cmake_minimum_required(VERSION 3.23)
cmake_minimum_required(VERSION 3.18)

if(NOT DEFINED PYTHON_EXECUTABLE)
message(
Expand Down
26 changes: 18 additions & 8 deletions paddle/phi/kernels/fusion/cutlass/gemm_epilogue/compile.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,21 +13,31 @@
# limitations under the License.
set -e

build_directory="build"
if [ ! -d "$build_directory" ]; then
mkdir $build_directory
fi

libname="$build_directory/libCutlassGemmEpilogue.so"
if [ -e "$libname" ]; then
exit 0
fi

cutlass_repo_directory="cutlass"
if [ ! -d "$cutlass_repo_directory" ]; then
git clone --branch v2.11.0 https://github.com/NVIDIA/cutlass
fi

build_directory="build"
if [ ! -d "$build_directory" ]; then
mkdir $build_directory
fi

python_exe_path="/usr/bin/python"
cuda_root_path="/usr/local/cuda"
gpu_cc="80"
default_python_exe_path="/usr/bin/python"
default_cuda_root_path="/usr/local/cuda"
default_gpu_cc="80"

python_exe_path="${1:-$default_python_exe_path}"
cuda_root_path="${2:-$default_cuda_root_path}"
gpu_cc="${3:-$default_gpu_cc}"

cd $build_directory
cmake .. -DPYTHON_EXECUTABLE=$python_exe_path -DCUDA_TOOLKIT_ROOT_DIR=$cuda_root_path -DCOMPUTE_CAPABILITY=$gpu_cc
make -j
make -j8
cd -
2 changes: 2 additions & 0 deletions paddle/phi/kernels/fusion/cutlass/gemm_epilogue/gemm_epilogue_util.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,6 +13,8 @@
// limitations under the License.

#pragma once
#include <cuda_bf16.h>
#include <cuda_fp16.h>
#include <vector>

#include "paddle/phi/kernels/fusion/cutlass/gemm_epilogue/gemm_epilogue_decl.h"
Expand Down