mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-14 10:09:41 +00:00
xdlops_v4r4_fwd fp32/fp16 (#34)
* create files for xdlops
* working on blockwise_gemm_xdlops
* add KReduction
* add m/n repeats
* add 2x2 pipeline
* added 128x128 wavegemm
* use StaticBuffer of vector_type
* break vector type to blk_size
* add kpack into xldops_gemm and blockwise_gemm
* abroadcast only
* add fp32 mfma instructions
* adding fp16 mfma
* pack half4_t
* rename kperwave to kpack
* add 32x32x8fp16
* add fp16 mfma
* clean code
* clean code
* V4r4 xdlops kpack (#35)
* add kpack with incorrect results
* bug fix for make_dynamic_naive_tensor_descriptor_aligned_v2
* add 1x1 kernel
* add gridwise_gemm_v2 - single_buffer
* enabled dwordx4 for fp16
Co-authored-by: Chao Liu <chao.liu2@amd.com>
* refactor fwd-v4r4-xdlops
* add v4r4-nhwc-xdlop
* improve some perf of nhwc and nchw by tuning parameters, and change scheuduling in gridwise-gemm loop
* tweak scheduling in gridwise gemm
* add v4r3 with a single output copy
* init commit: output with slice win
* adding sliceWin
* add multiple repeats pattern
* starting adding bwd-v4r1-xdlops
* use tuple as SrcBuffer
* adding bwd-data v4r1 nhwc xdlops
* fix bug in make_dynamic_naive_tensor_descriptor_aligned_v2()
* fix bug in host bwd-data conv
* initial implementation of bwd-data v4r1 nhwc xdlops
* add launch bound flags
* enable launch bound
* add m/nrepeat=4
* tweak bwd-data v4r1 nhwc xdlops
* added bwd-data v4r1 nhwc xlops with output A and weight B
* add fwd-v4r4 nhwc xdlops, A input, B weight, C output
Co-authored-by: Chao Liu <chao.liu2@amd.com>
[ROCm/composable_kernel commit: 3835318cc3]
This commit is contained in:
zjing14
@@ -3,24 +3,46 @@ rm -f CMakeCache.txt
|
||||
rm -f *.cmake
|
||||
rm -rf CMakeFiles
|
||||
|
||||
MY_PROJECT_SOURCE=../
|
||||
MY_PROJECT_SOURCE=../../../
|
||||
MY_PROJECT_INSTALL=../install.dir
|
||||
|
||||
cmake \
|
||||
-D CMAKE_INSTALL_PREFIX=${MY_PROJECT_INSTALL} \
|
||||
-D CMAKE_BUILD_TYPE=Release \
|
||||
-D DEVICE_BACKEND="AMD" \
|
||||
-D CMAKE_CXX_FLAGS="-O3 --amdgpu-target=gfx1030 -gline-tables-only -save-temps=$CWD -ftemplate-backtrace-limit=0" \
|
||||
-D DEVICE_BACKEND=AMD \
|
||||
-D CMAKE_CXX_FLAGS="-O3 --amdgpu-target=gfx908 -mllvm --amdgpu-spill-vgpr-to-agpr=0 -gline-tables-only -save-temps=$CWD" \
|
||||
-D CMAKE_CXX_COMPILER=/opt/rocm/bin/hipcc \
|
||||
-D CMAKE_PREFIX_PATH="/opt/rocm" \
|
||||
-D CMAKE_PREFIX_PATH=/opt/rocm \
|
||||
-D CMAKE_VERBOSE_MAKEFILE:BOOL=ON \
|
||||
${MY_PROJECT_SOURCE}
|
||||
|
||||
#-D CMAKE_CXX_FLAGS="-O3 --amdgpu-target=gfx1030 -gline-tables-only -save-temps=$CWD -ftemplate-backtrace-limit=0" \
|
||||
#-D CMAKE_CXX_FLAGS="-O3 --amdgpu-target=gfx1030 -gline-tables-only -save-temps=$CWD -ftemplate-backtrace-limit=0 -mllvm -print-before=amdgpu-codegenprepare -mllvm -print-module-scope" \
|
||||
#-D CMAKE_CXX_FLAGS="-O3 --amdgpu-target=gfx906 -gline-tables-only -save-temps=$CWD" \
|
||||
#-D CMAKE_CXX_FLAGS="-O3 --amdgpu-target=gfx906 -mllvm --amdgpu-spill-vgpr-to-agpr=0" \
|
||||
#-D CMAKE_CXX_FLAGS="-O3 --amdgpu-target=gfx906 -mllvm --amdgpu-spill-vgpr-to-agpr=0 -save-temps=$CWD" \
|
||||
#-D CMAKE_CXX_FLAGS="-O3 --amdgpu-target=gfx906 -mllvm --amdgpu-enable-global-sgpr-addr -mllvm --amdgpu-spill-vgpr-to-agpr=0" \
|
||||
#-D CMAKE_CXX_FLAGS="-O3 --amdgpu-target=gfx906 -mllvm --amdgpu-enable-global-sgpr-addr -mllvm --amdgpu-spill-vgpr-to-agpr=0 -save-temps=$CWD" \
|
||||
#-D CMAKE_CXX_FLAGS="-O3 --amdgpu-target=gfx906 -mllvm --amdgpu-enable-global-sgpr-addr -mllvm --amdgpu-spill-vgpr-to-agpr=0 -v -gline-tables-only -save-temps=$CWD" \
|
||||
|
||||
#CXX_FLAG_TMP=-Weverything
|
||||
# -Wno-c++98-compat \
|
||||
# -Wno-c++98-compat-pedantic \
|
||||
# -Wno-conversion \
|
||||
# -Wno-double-promotion \
|
||||
# -Wno-exit-time-destructors \
|
||||
# -Wno-extra-semi \
|
||||
# -Wno-float-conversion \
|
||||
# -Wno-gnu-anonymous-struct \
|
||||
# -Wno-gnu-zero-variadic-macro-arguments \
|
||||
# -Wno-missing-noreturn \
|
||||
# -Wno-missing-prototypes \
|
||||
# -Wno-nested-anon-types \
|
||||
# -Wno-padded \
|
||||
# -Wno-return-std-move-in-c++11 \
|
||||
# -Wno-shorten-64-to-32 \
|
||||
# -Wno-sign-conversion \
|
||||
# -Wno-unknown-warning-option \
|
||||
# -Wno-unused-command-line-argument \
|
||||
# -Wno-weak-vtables \
|
||||
# -Wno-covered-switch-default \
|
||||
# -Wno-disabled-macro-expansion \
|
||||
# -Wno-undefined-reinterpret-cast
|
||||
|
||||
|
||||
@@ -1,19 +0,0 @@
|
||||
#!/bin/bash
|
||||
rm -f CMakeCache.txt
|
||||
rm -f *.cmake
|
||||
rm -rf CMakeFiles
|
||||
|
||||
MY_PROJECT_SOURCE=../../../
|
||||
MY_PROJECT_INSTALL=../install.dir
|
||||
|
||||
cmake \
|
||||
-D CMAKE_INSTALL_PREFIX=${MY_PROJECT_INSTALL} \
|
||||
-D CMAKE_BUILD_TYPE=Release \
|
||||
-D DEVICE_BACKEND="AMD" \
|
||||
-D CMAKE_CXX_FLAGS="--amdgpu-target=gfx906" \
|
||||
-D CMAKE_CXX_COMPILER=/opt/rocm/hip/bin/hipcc \
|
||||
-D CMAKE_PREFIX_PATH="/opt/rocm" \
|
||||
-D CMAKE_VERBOSE_MAKEFILE:BOOL=ON \
|
||||
${MY_PROJECT_SOURCE}
|
||||
|
||||
#-D CMAKE_CXX_FLAGS="-gline-tables-only -v --amdgpu-target=gfx906" \
|
||||
@@ -1,8 +0,0 @@
|
||||
#!/bin/bash
|
||||
export KMOPTLLC="-mattr=+enable-ds128 -amdgpu-enable-global-sgpr-addr"
|
||||
export KMDUMPISA=1
|
||||
export KMDUMPLLVM=1
|
||||
export KMDUMPDIR=$PWD
|
||||
|
||||
make -j $1
|
||||
#/opt/rocm/hcc/bin/llvm-objdump -mcpu=gfx906 -source -line-numbers driver/dump-gfx906.isabin > driver/dump-gfx906.isabin.asm
|
||||
2
script/docker-rocm3.7.sh → script/docker-rocm4.1.sh
Normal file → Executable file
2
script/docker-rocm3.7.sh → script/docker-rocm4.1.sh
Normal file → Executable file
@@ -8,7 +8,7 @@ docker run \
|
||||
--group-add sudo \
|
||||
-w /root/workspace \
|
||||
-v $WORKSPACE:/root/workspace \
|
||||
asroy/tensorflow:rocm3.7-tf2.3-dev-omp \
|
||||
rocm/tensorflow:rocm4.1-tf1.15-dev \
|
||||
/bin/bash
|
||||
|
||||
#--network host \
|
||||
@@ -1,14 +1,19 @@
|
||||
#!/bin/bash
|
||||
|
||||
## GPU visibility
|
||||
export ROCR_VISIBLE_DEVICE=0
|
||||
export GPU_DEVICE_ORDINAL=0
|
||||
|
||||
export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH
|
||||
## Boost
|
||||
#export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH
|
||||
|
||||
## Compiling
|
||||
export OLC_DEBUG_HIP_VERBOSE=1
|
||||
export OLC_DEBUG_HIP_DUMP=1
|
||||
export OLC_DEBUG_SAVE_TEMP_DIR=1
|
||||
|
||||
#make -j conv_driver
|
||||
#make -j conv_driver_v2
|
||||
#make -j conv_bwd_data_driver_v2
|
||||
make -j conv_driver_v2_olc
|
||||
|
||||
rm -rf /root/_hip_binary_kernels_/
|
||||
@@ -21,11 +26,21 @@ INIT=$4
|
||||
LOG=$5
|
||||
REPEAT=$6
|
||||
|
||||
###################### layout algo verify init log repeat N__ K__ C__ Y X Hi_ Wi__ Strides Dilations LeftPads RightPads
|
||||
#driver/conv_driver_v2 $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 128 192 3 3 71 71 2 2 1 1 1 1 1 1
|
||||
#driver/conv_driver_v2 $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 384 192 3 3 35 35 2 2 1 1 0 0 0 0
|
||||
#driver/conv_driver_v2 $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 128 128 1 7 17 17 1 1 1 1 0 3 0 3
|
||||
#driver/conv_driver_v2 $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 256 3 3 14 14 1 1 1 1 1 1 1 1
|
||||
################################ layout algo verify init log repeat N__ K___ C___ Y X Hi_ Wi__ Strides Dilations LeftPads RightPads
|
||||
./conv_driver_v2 $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 192 3 3 71 71 2 2 1 1 1 1 1 1
|
||||
#./conv_driver_v2 $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 1024 1 7 17 17 1 1 1 1 0 3 0 3
|
||||
#./conv_driver_v2 $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 256 2048 3 3 14 14 1 1 1 1 1 1 1 1
|
||||
#./conv_driver_v2 $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 256 256 3 3 14 14 1 1 1 1 1 1 1 1
|
||||
#./conv_driver_v2 $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 512 512 3 3 7 7 1 1 1 1 1 1 1 1
|
||||
|
||||
#./conv_driver_v2 $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 128 192 3 3 71 71 2 2 1 1 1 1 1 1
|
||||
./conv_driver_v2_olc $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 128 192 3 3 71 71 2 2 1 1 1 1 1 1
|
||||
#./conv_driver_v2 $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 512 192 3 3 35 35 2 2 1 1 0 0 0 0
|
||||
#./conv_driver_v2 $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 256 256 3 3 30 30 2 2 1 1 0 0 0 0
|
||||
#./conv_driver_v2 $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 512 512 3 3 16 16 2 2 1 1 0 0 0 0
|
||||
|
||||
#./conv_driver_v2 $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 2048 1024 1 1 14 14 2 2 1 1 0 0 0 0
|
||||
#./conv_driver_v2 $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 256 1024 1 1 14 14 1 1 1 1 0 0 0 0
|
||||
#./conv_driver_v2 $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 512 2048 1 1 7 7 1 1 1 1 0 0 0 0
|
||||
|
||||
#./conv_bwd_data_driver_v2 $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 256 256 256 3 3 14 14 1 1 1 1 1 1 1 1
|
||||
|
||||
./conv_driver_v2_olc $LAYOUT $ALGO $VERIFY $INIT $LOG $REPEAT 128 256 192 3 3 71 71 2 2 1 1 1 1 1 1
|
||||
|
||||
Reference in New Issue
Block a user