Files
composable_kernel/device_operation/CMakeLists.txt
ltqin 0180bb825c NHWC conv 2d: bwd fp32/fp16/bfp16/int8, Device level tuning and host API (#92)
* start conv2d bwd api

* kernel running

* add bwd reference

* change to no shuffle

* fix bwd reference

* pass verification

* add Filter1x1Stride1Pad0 and start testing

* change some tuning parameter

* fix test error

* add fp16 tuning parameter

* add bf16 tuning parameter

* add int8 tuning parameters

* change fp32 tuning parameter

* add bwd to profiler

* fix bug for bwd profiler

* fix ckProfiler bug

* change conv2d_bwd_xdl to fp16

* fix bug in comments

* fix precompile id

* fix enum conv name

* chage _bwd_ to _bwd_data_

* change conv2d_bwd example id

* bwd to bwd data

* fix prehead

* fix MakeDefaultBlock2CTileMap ,import form merge develop

* format bwd instance

* bwd to bwd data

* change name bwd to bwd data

* change name bwd to bwd data in example

* formate code

* change conv2d bwd data id in example

* rewrite readme for example

* fix CalculateMagicNumbers about div zero

* add workaround CK_WORKAROUND_SWDEV_325164

* change test_conf2d_bwd_data show info

* format

* fix bug for workaround:CK_WORKAROUND_SWDEV_325164

* formate tuning parameters

* formate tuning parameters again

* formate tuning parameters 3

* formate tuning parameters 4

* remove add function template

* format

* update comment

Co-authored-by: ltqin <letaoqin@amd.com>
Co-authored-by: Chao Liu <chao.liu2@amd.com>

[ROCm/composable_kernel commit: c254e5abd2]
2022-03-04 00:08:26 -06:00

171 lines
12 KiB
CMake

include_directories(BEFORE
include
${PROJECT_SOURCE_DIR}/host/host_tensor/include
${PROJECT_SOURCE_DIR}/device/include
${PROJECT_SOURCE_DIR}/device_operation/include
${PROJECT_SOURCE_DIR}/profiler/include
${PROJECT_SOURCE_DIR}/composable_kernel/include
${PROJECT_SOURCE_DIR}/composable_kernel/include/utility
${PROJECT_SOURCE_DIR}/composable_kernel/include/tensor_description
${PROJECT_SOURCE_DIR}/composable_kernel/include/tensor_operation
${PROJECT_SOURCE_DIR}/composable_kernel/include/problem_transform
${PROJECT_SOURCE_DIR}/external/rocm/include
)
# device_gemm_instance
set(DEVICE_GEMM_INSTANCE_SOURCE
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_f32_f32_f32_mk_kn_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_f32_f32_f32_mk_nk_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_f32_f32_f32_km_kn_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_f32_f32_f32_km_nk_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_f16_f16_f16_mk_kn_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_f16_f16_f16_mk_nk_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_f16_f16_f16_km_kn_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_f16_f16_f16_km_nk_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_mk_nk_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_splitk_f32_f32_f32_mk_kn_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_splitk_f32_f32_f32_mk_nk_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_splitk_f32_f32_f32_km_kn_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_splitk_f32_f32_f32_km_nk_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_splitk_f16_f16_f16_mk_nk_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_splitk_f16_f16_f16_km_kn_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_splitk_f16_f16_f16_km_nk_mn_instance.cpp;
)
# device_gemm_bias_2d_instance
set(DEVICE_GEMM_BIAS_2D_INSTANCE_SOURCE
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_bias_2d_f32_f32_f32_km_kn_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_bias_2d_f32_f32_f32_km_nk_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_bias_2d_f32_f32_f32_mk_kn_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_bias_2d_f32_f32_f32_mk_nk_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_bias_2d_f16_f16_f16_km_kn_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_bias_2d_f16_f16_f16_km_nk_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_bias_2d_f16_f16_f16_mk_kn_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_bias_2d_f16_f16_f16_mk_nk_mn_instance.cpp;
)
# device_gemm_bias_relu_instance
set(DEVICE_GEMM_BIAS_RELU_INSTANCE_SOURCE
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_bias_relu_f16_f16_f16_mk_kn_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_bias_relu_f16_f16_f16_mk_nk_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_bias_relu_f16_f16_f16_km_kn_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_bias_relu_f16_f16_f16_km_nk_mn_instance.cpp;
)
# device_gemm_bias_relu_add_instance
set(DEVICE_GEMM_BIAS_RELU_ADD_INSTANCE_SOURCE
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_bias_relu_add_f16_f16_f16_mk_kn_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_bias_relu_add_f16_f16_f16_mk_nk_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_bias_relu_add_f16_f16_f16_km_kn_mn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_gemm_xdl_c_shuffle_bias_relu_add_f16_f16_f16_km_nk_mn_instance.cpp;
)
set(DEVICE_BATCHED_GEMM_INSTANCE_SOURCE
${PROJECT_SOURCE_DIR}/device_operation/src/device_batched_gemm_xdl_f16_f16_f16_gmk_gkn_gmn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_batched_gemm_xdl_f16_f16_f16_gmk_gnk_gmn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_batched_gemm_xdl_f16_f16_f16_gkm_gkn_gmn_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_batched_gemm_xdl_f16_f16_f16_gkm_gnk_gmn_instance.cpp;
)
# device_conv2d_fwd_instance
set(DEVICE_CONV2D_FWD_INSTANCE_SOURCE
${PROJECT_SOURCE_DIR}/device_operation/src/device_conv2d_fwd_xdl_nhwc_kyxc_nhwk_f32_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_conv2d_fwd_xdl_nhwc_kyxc_nhwk_f16_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_conv2d_fwd_xdl_nhwc_kyxc_nhwk_bf16_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_conv2d_fwd_xdl_nhwc_kyxc_nhwk_int8_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk_f16_instance.cpp;
)
# device_conv1d_fwd_instance
set(DEVICE_CONV1D_FWD_INSTANCE_SOURCE
${PROJECT_SOURCE_DIR}/device_operation/src/device_conv1d_fwd_xdl_nwc_kxc_nwk_f32_instance.cpp;
)
# device_conv2d_fwd_bias_relu_instance
set(DEVICE_CONV2D_FWD_BIAS_RELU_INSTANCE_SOURCE
${PROJECT_SOURCE_DIR}/device_operation/src/device_conv2d_fwd_xdl_c_shuffle_bias_relu_nhwc_kyxc_nhwk_f16_instance.cpp;
)
# device_conv2d_fwd_bias_relu_add_instance
set(DEVICE_CONV2D_FWD_BIAS_RELU_ADD_INSTANCE_SOURCE
${PROJECT_SOURCE_DIR}/device_operation/src/device_conv2d_fwd_xdl_c_shuffle_bias_relu_add_nhwc_kyxc_nhwk_f16_instance.cpp;
)
# device_conv2d_fwd_bias_relu_atomic_add_instance
set(DEVICE_CONV2D_FWD_BIAS_RELU_ATOMIC_ADD_INSTANCE_SOURCE
${PROJECT_SOURCE_DIR}/device_operation/src/device_conv2d_fwd_xdl_c_shuffle_bias_relu_atomic_add_nhwc_kyxc_nhwk_f16_instance.cpp;
)
# device_conv2d_bwd_data_instance
set(DEVICE_CONV2D_BWD_DATA_INSTANCE_SOURCE
${PROJECT_SOURCE_DIR}/device_operation/src/device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_f32_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_f16_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_bf16_instance.cpp;
${PROJECT_SOURCE_DIR}/device_operation/src/device_conv2d_bwd_data_xdl_nhwc_kyxc_nhwk_int8_instance.cpp;
)
add_library(device_gemm_instance SHARED ${DEVICE_GEMM_INSTANCE_SOURCE})
add_library(device_gemm_bias_relu_instance SHARED ${DEVICE_GEMM_BIAS_RELU_INSTANCE_SOURCE})
add_library(device_gemm_bias_relu_add_instance SHARED ${DEVICE_GEMM_BIAS_RELU_ADD_INSTANCE_SOURCE})
add_library(device_batched_gemm_instance SHARED ${DEVICE_BATCHED_GEMM_INSTANCE_SOURCE})
add_library(device_conv1d_fwd_instance SHARED ${DEVICE_CONV1D_FWD_INSTANCE_SOURCE})
add_library(device_conv2d_fwd_instance SHARED ${DEVICE_CONV2D_FWD_INSTANCE_SOURCE})
add_library(device_conv2d_fwd_bias_relu_instance SHARED ${DEVICE_CONV2D_FWD_BIAS_RELU_INSTANCE_SOURCE})
add_library(device_conv2d_fwd_bias_relu_add_instance SHARED ${DEVICE_CONV2D_FWD_BIAS_RELU_ADD_INSTANCE_SOURCE})
add_library(device_conv2d_fwd_bias_relu_atomic_add_instance SHARED ${DEVICE_CONV2D_FWD_BIAS_RELU_ATOMIC_ADD_INSTANCE_SOURCE})
add_library(device_gemm_bias_2d_instance SHARED ${DEVICE_GEMM_BIAS_2D_INSTANCE_SOURCE})
add_library(device_conv2d_bwd_data_instance SHARED ${DEVICE_CONV2D_BWD_DATA_INSTANCE_SOURCE})
target_include_directories(device_gemm_instance SYSTEM PUBLIC $<BUILD_INTERFACE:${HALF_INCLUDE_DIR}>)
target_include_directories(device_gemm_bias_2d_instance SYSTEM PUBLIC $<BUILD_INTERFACE:${HALF_INCLUDE_DIR}>)
target_include_directories(device_gemm_bias_relu_instance SYSTEM PUBLIC $<BUILD_INTERFACE:${HALF_INCLUDE_DIR}>)
target_include_directories(device_gemm_bias_relu_add_instance SYSTEM PUBLIC $<BUILD_INTERFACE:${HALF_INCLUDE_DIR}>)
target_include_directories(device_batched_gemm_instance SYSTEM PUBLIC $<BUILD_INTERFACE:${HALF_INCLUDE_DIR}>)
target_include_directories(device_conv1d_fwd_instance SYSTEM PUBLIC $<BUILD_INTERFACE:${HALF_INCLUDE_DIR}>)
target_include_directories(device_conv2d_fwd_instance SYSTEM PUBLIC $<BUILD_INTERFACE:${HALF_INCLUDE_DIR}>)
target_include_directories(device_conv2d_fwd_bias_relu_instance SYSTEM PUBLIC $<BUILD_INTERFACE:${HALF_INCLUDE_DIR}>)
target_include_directories(device_conv2d_fwd_bias_relu_add_instance SYSTEM PUBLIC $<BUILD_INTERFACE:${HALF_INCLUDE_DIR}>)
target_include_directories(device_conv2d_fwd_bias_relu_atomic_add_instance SYSTEM PUBLIC $<BUILD_INTERFACE:${HALF_INCLUDE_DIR}>)
target_include_directories(device_conv2d_bwd_data_instance SYSTEM PUBLIC $<BUILD_INTERFACE:${HALF_INCLUDE_DIR}>)
target_compile_features(device_gemm_instance PUBLIC)
target_compile_features(device_gemm_bias_2d_instance PUBLIC)
target_compile_features(device_gemm_bias_relu_instance PUBLIC)
target_compile_features(device_gemm_bias_relu_add_instance PUBLIC)
target_compile_features(device_batched_gemm_instance PUBLIC)
target_compile_features(device_conv1d_fwd_instance PUBLIC)
target_compile_features(device_conv2d_fwd_instance PUBLIC)
target_compile_features(device_conv2d_fwd_bias_relu_instance PUBLIC)
target_compile_features(device_conv2d_fwd_bias_relu_add_instance PUBLIC)
target_compile_features(device_conv2d_fwd_bias_relu_atomic_add_instance PUBLIC)
target_compile_features(device_conv2d_bwd_data_instance PUBLIC)
set_target_properties(device_gemm_instance PROPERTIES POSITION_INDEPENDENT_CODE ON)
set_target_properties(device_gemm_bias_2d_instance PROPERTIES POSITION_INDEPENDENT_CODE ON)
set_target_properties(device_gemm_bias_relu_instance PROPERTIES POSITION_INDEPENDENT_CODE ON)
set_target_properties(device_gemm_bias_relu_add_instance PROPERTIES POSITION_INDEPENDENT_CODE ON)
set_target_properties(device_batched_gemm_instance PROPERTIES POSITION_INDEPENDENT_CODE ON)
set_target_properties(device_conv1d_fwd_instance PROPERTIES POSITION_INDEPENDENT_CODE ON)
set_target_properties(device_conv2d_fwd_instance PROPERTIES POSITION_INDEPENDENT_CODE ON)
set_target_properties(device_conv2d_fwd_bias_relu_instance PROPERTIES POSITION_INDEPENDENT_CODE ON)
set_target_properties(device_conv2d_fwd_bias_relu_add_instance PROPERTIES POSITION_INDEPENDENT_CODE ON)
set_target_properties(device_conv2d_fwd_bias_relu_atomic_add_instance PROPERTIES POSITION_INDEPENDENT_CODE ON)
set_target_properties(device_conv2d_bwd_data_instance PROPERTIES POSITION_INDEPENDENT_CODE ON)
install(TARGETS device_gemm_instance LIBRARY DESTINATION lib)
install(TARGETS device_gemm_bias_2d_instance LIBRARY DESTINATION lib)
install(TARGETS device_gemm_bias_relu_instance LIBRARY DESTINATION lib)
install(TARGETS device_gemm_bias_relu_add_instance LIBRARY DESTINATION lib)
install(TARGETS device_batched_gemm_instance LIBRARY DESTINATION lib)
install(TARGETS device_conv1d_fwd_instance LIBRARY DESTINATION lib)
install(TARGETS device_conv2d_fwd_instance LIBRARY DESTINATION lib)
install(TARGETS device_conv2d_fwd_bias_relu_instance LIBRARY DESTINATION lib)
install(TARGETS device_conv2d_fwd_bias_relu_add_instance LIBRARY DESTINATION lib)
install(TARGETS device_conv2d_fwd_bias_relu_atomic_add_instance LIBRARY DESTINATION lib)
install(TARGETS device_conv2d_bwd_data_instance LIBRARY DESTINATION lib)