"vscode:/vscode.git/clone" did not exist on "23618455489abd04a084e24d5b96b3ab5d7b12bc"
Unverified Commit 5c75ff5c authored by zjing14's avatar zjing14 Committed by GitHub
Browse files

Merge branch 'develop' into ab_blk_copy_multi_d

parents 96417775 48ba6e8a
cmake_minimum_required(VERSION 3.14)
if(POLICY CMP0140)
# policies CMP0140 not known to CMake until 3.25
cmake_policy(SET CMP0140 NEW)
endif()
# This has to be initialized before the project() command appears
# Set the default of CMAKE_BUILD_TYPE to be release, unless user specifies with -D. MSVC_IDE does not use CMAKE_BUILD_TYPE
......@@ -107,16 +111,18 @@ else()
add_definitions(-DPROFILER_ONLY)
set(GPU_TARGETS "" CACHE STRING "" FORCE)
if(GPU_TARGETS)
message(FATAL_ERROR "For PROFILE_ONLY build, please do not set GPU_TARGETS, use GPU_ARCH = gfx9, gfx10, or gfx11")
message(FATAL_ERROR "For PROFILE_ONLY build, please do not set GPU_TARGETS, use GPU_ARCH = gfx90, gfx94, gfx10, or gfx11")
endif()
if(GPU_ARCH MATCHES "gfx9")
rocm_check_target_ids(DEFAULT_GPU_TARGETS TARGETS "gfx900;gfx906;gfx908;gfx90a;gfx940;gfx941;gfx942")
if(GPU_ARCH MATCHES "gfx90")
rocm_check_target_ids(DEFAULT_GPU_TARGETS TARGETS "gfx900;gfx906;gfx908;gfx90a")
elseif(GPU_ARCH MATCHES "gfx94")
rocm_check_target_ids(DEFAULT_GPU_TARGETS TARGETS "gfx940;gfx941;gfx942")
elseif(GPU_ARCH MATCHES "gfx10")
rocm_check_target_ids(DEFAULT_GPU_TARGETS TARGETS "gfx1030")
elseif(GPU_ARCH MATCHES "gfx11")
rocm_check_target_ids(DEFAULT_GPU_TARGETS TARGETS "gfx1100;gfx1101;gfx1102")
else()
message(FATAL_ERROR "For PROFILE_ONLY build, please specify GPU_ARCH as gfx9, gfx10, or gfx11")
message(FATAL_ERROR "For PROFILE_ONLY build, please specify GPU_ARCH as gfx90, gfx94, gfx10, or gfx11")
endif()
set(GPU_TARGETS "${DEFAULT_GPU_TARGETS}" CACHE STRING " " FORCE)
endif()
......@@ -383,31 +389,31 @@ IF(IS_DIRECTORY "${PROJECT_SOURCE_DIR}/library/src/tensor_operation_instance/gpu
set(cmake_instance)
file(READ "${PROJECT_SOURCE_DIR}/library/src/tensor_operation_instance/gpu/${subdir_path}/CMakeLists.txt" cmake_instance)
set(add_inst 0)
if("${cmake_instance}" MATCHES "DTYPES MATCHES \"fp8\" " AND DTYPES MATCHES "fp8")
if(("${cmake_instance}" MATCHES "fp8" OR "${cmake_instance}" MATCHES "_f8") AND DTYPES MATCHES "fp8")
#message("fp8 instance found!")
set(add_inst 1)
endif()
if("${cmake_instance}" MATCHES "DTYPES MATCHES \"bf8\" " AND DTYPES MATCHES "bf8")
if(("${cmake_instance}" MATCHES "bf8" OR "${cmake_instance}" MATCHES "_b8") AND DTYPES MATCHES "bf8")
#message("bf8 instance found!")
set(add_inst 1)
endif()
if("${cmake_instance}" MATCHES "DTYPES MATCHES \"fp16\"" AND DTYPES MATCHES "fp16")
if(("${cmake_instance}" MATCHES "fp16" OR "${cmake_instance}" MATCHES "_f16") AND DTYPES MATCHES "fp16")
#message("fp16 instance found!")
set(add_inst 1)
endif()
if("${cmake_instance}" MATCHES "DTYPES MATCHES \"fp32\"" AND DTYPES MATCHES "fp32")
if(("${cmake_instance}" MATCHES "fp32" OR "${cmake_instance}" MATCHES "_f32") AND DTYPES MATCHES "fp32")
#message("fp32 instance found!")
set(add_inst 1)
endif()
if("${cmake_instance}" MATCHES "DTYPES MATCHES \"fp64\"" AND DTYPES MATCHES "fp64")
if(("${cmake_instance}" MATCHES "fp64" OR "${cmake_instance}" MATCHES "_f64") AND DTYPES MATCHES "fp64")
#message("fp64 instance found!")
set(add_inst 1)
endif()
if("${cmake_instance}" MATCHES "DTYPES MATCHES \"bf16\"" AND DTYPES MATCHES "bf16")
if(("${cmake_instance}" MATCHES "bf16" OR "${cmake_instance}" MATCHES "_b16") AND DTYPES MATCHES "bf16")
#message("bf16 instance found!")
set(add_inst 1)
endif()
if("${cmake_instance}" MATCHES "DTYPES MATCHES \"int8\"" AND DTYPES MATCHES "int8")
if(("${cmake_instance}" MATCHES "int8" OR "${cmake_instance}" MATCHES "_i8") AND DTYPES MATCHES "int8")
#message("int8 instance found!")
set(add_inst 1)
endif()
......
if(DL_KERNELS)
add_custom_target(example_gemm_dl)
add_custom_target(example_gemm_dl)
add_example_executable(example_gemm_dl_fp32 gemm_dl_fp32.cpp)
add_dependencies(example_gemm_dl example_gemm_dl_fp32)
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
add_example_executable(example_gemm_dl_fp16 gemm_dl_fp16.cpp)
add_example_executable(example_gemm_dl_fp32 gemm_dl_fp32.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_dl example_gemm_dl_fp32)
endif()
add_example_executable(example_gemm_dl_fp16 gemm_dl_fp16.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_dl example_gemm_dl_fp16)
add_example_executable(example_gemm_dpp_fp16 gemm_dpp_fp16.cpp)
endif()
if(DTYPES MATCHES "int8" OR NOT DEFINED DTYPES)
add_example_executable(example_gemm_dl_int8 gemm_dl_int8.cpp)
endif()
add_example_executable(example_gemm_dpp_fp16 gemm_dpp_fp16.cpp)
add_example_executable(example_gemm_dl_int8 gemm_dl_int8.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_dl example_gemm_dl_int8)
endif()
if(USE_BITINT_EXTENSION_INT4)
endif()
if(USE_BITINT_EXTENSION_INT4)
add_example_executable(example_gemm_dl_int4 gemm_dl_int4.cpp)
add_dependencies(example_gemm_dl example_gemm_dl_int4)
endif(USE_BITINT_EXTENSION_INT4)
endif()
endif(USE_BITINT_EXTENSION_INT4)
add_custom_target(example_gemm_xdl)
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
add_example_executable(example_gemm_xdl_fp16 gemm_xdl_fp16.cpp)
add_example_executable(example_gemm_xdl_wavelet_fp16 gemm_xdl_wavelet_fp16.cpp)
add_dependencies(example_gemm_xdl example_gemm_xdl_fp16)
add_dependencies(example_gemm_xdl example_gemm_xdl_wavelet_fp16)
add_example_executable(example_gemm_xdl_skip_b_lds_fp16 gemm_xdl_skip_b_lds_fp16.cpp)
add_dependencies(example_gemm_xdl example_gemm_xdl_skip_b_lds_fp16)
if(GPU_TARGETS MATCHES "gfx1100" OR GPU_TARGETS MATCHES "gfx1101" OR GPU_TARGETS MATCHES "gfx1102")
add_example_executable(example_gemm_xdl_fp16 gemm_xdl_fp16.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_xdl example_gemm_xdl_fp16)
endif()
add_example_executable(example_gemm_xdl_wavelet_fp16 gemm_xdl_wavelet_fp16.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_xdl example_gemm_xdl_wavelet_fp16)
endif()
add_example_executable(example_gemm_xdl_skip_b_lds_fp16 gemm_xdl_skip_b_lds_fp16.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_xdl example_gemm_xdl_skip_b_lds_fp16)
endif()
if(GPU_TARGETS MATCHES "gfx1100" OR GPU_TARGETS MATCHES "gfx1101" OR GPU_TARGETS MATCHES "gfx1102")
add_custom_target(example_gemm_wmma)
add_example_executable(example_gemm_wmma_fp16 gemm_wmma_fp16.cpp)
add_dependencies(example_gemm_wmma example_gemm_wmma_fp16)
endif()
if(result EQUAL 0)
add_dependencies(example_gemm_wmma example_gemm_wmma_fp16)
endif()
endif()
if(DTYPES MATCHES "bf16" OR NOT DEFINED DTYPES)
add_example_executable(example_gemm_xdl_bf16 gemm_xdl_bf16.cpp)
add_example_executable(example_gemm_xdl_bf16 gemm_xdl_bf16.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_xdl example_gemm_xdl_bf16)
add_example_executable(example_gemm_xdl_bf16_rtn gemm_xdl_bf16_rtn.cpp)
add_dependencies(example_gemm_xdl example_gemm_xdl_bf16_rtn)
endif()
if(DTYPES MATCHES "int8" OR NOT DEFINED DTYPES)
add_example_executable(example_gemm_xdl_int8 gemm_xdl_int8.cpp)
add_example_executable(example_gemm_xdl_int8 gemm_xdl_int8.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_xdl example_gemm_xdl_int8)
endif()
......@@ -54,22 +57,23 @@ if(USE_BITINT_EXTENSION_INT4)
add_dependencies(example_gemm_xdl example_gemm_xdl_int4)
endif(USE_BITINT_EXTENSION_INT4)
if(DTYPES MATCHES "fp64" OR NOT DEFINED DTYPES)
# FIXME: re-enable this exampe as test when SWDEV-335738 is fixed
add_example_executable_no_testing(example_gemm_xdl_fp64 gemm_xdl_fp64.cpp)
# FIXME: re-enable this exampe as test when SWDEV-335738 is fixed
add_example_executable_no_testing(example_gemm_xdl_fp64 gemm_xdl_fp64.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_xdl example_gemm_xdl_fp64)
endif()
add_example_executable(example_gemm_xdl_streamk gemm_xdl_streamk.cpp)
if(DTYPES MATCHES "fp8" OR NOT DEFINED DTYPES)
if(GPU_TARGETS MATCHES "gfx940" OR GPU_TARGETS MATCHES "gfx941" OR GPU_TARGETS MATCHES "gfx942")
add_example_executable(example_gemm_xdl_f8 gemm_xdl_f8.cpp)
if(GPU_TARGETS MATCHES "gfx940" OR GPU_TARGETS MATCHES "gfx941" OR GPU_TARGETS MATCHES "gfx942")
add_example_executable(example_gemm_xdl_f8 gemm_xdl_f8.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_xdl example_gemm_xdl_f8)
endif()
endif()
if((DTYPES MATCHES "fp8" AND DTYPES MATCHES "fp16") OR NOT DEFINED DTYPES)
add_example_executable(example_gemm_xdl_fp16_f8 gemm_xdl_fp16_f8.cpp)
add_example_executable(example_gemm_xdl_fp16_f8 gemm_xdl_fp16_f8.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_xdl example_gemm_xdl_fp16_f8)
endif()
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
list(APPEND gpu_list1 gfx1100 gfx1101 gfx1102)
list(APPEND gpu_list2 gfx908 gfx90a gfx940 gfx941 gfx942)
set(target 0)
......@@ -19,4 +18,3 @@ foreach(gpu IN LISTS GPU_TARGETS)
set(target 1)
endif()
endforeach()
endif()
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
list(APPEND gpu_list gfx908 gfx90a gfx940 gfx941 gfx942)
set(target 0)
foreach(gpu IN LISTS GPU_TARGETS)
......@@ -7,4 +6,3 @@ foreach(gpu IN LISTS GPU_TARGETS)
set(target 1)
endif()
endforeach()
endif()
......@@ -3,24 +3,24 @@ set(target 0)
foreach(gpu IN LISTS GPU_TARGETS)
if(gpu IN_LIST gpu_list AND target EQUAL 0)
add_custom_target(example_gemm_add_add_fastgelu_xdl)
if(DTYPES MATCHES "bf16" OR NOT DEFINED DTYPES)
add_example_executable(example_gemm_add_add_fastgelu_xdl_bf16 gemm_add_add_fastgelu_xdl_bf16.cpp)
add_example_executable(example_gemm_add_add_fastgelu_xdl_bf16 gemm_add_add_fastgelu_xdl_bf16.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_add_add_fastgelu_xdl example_gemm_add_add_fastgelu_xdl_bf16)
endif()
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
add_example_executable(example_gemm_add_add_fastgelu_xdl_fp16 gemm_add_add_fastgelu_xdl_fp16.cpp)
add_example_executable(example_gemm_add_add_fastgelu_xdl_fp16 gemm_add_add_fastgelu_xdl_fp16.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_add_add_fastgelu_xdl example_gemm_add_add_fastgelu_xdl_fp16)
endif()
if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
add_example_executable(example_gemm_add_add_fastgelu_xdl_fp32 gemm_add_add_fastgelu_xdl_fp32.cpp)
add_example_executable(example_gemm_add_add_fastgelu_xdl_fp32 gemm_add_add_fastgelu_xdl_fp32.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_add_add_fastgelu_xdl example_gemm_add_add_fastgelu_xdl_fp32)
endif()
if(USE_BITINT_EXTENSION_INT4)
add_example_executable(example_gemm_add_add_fastgelu_xdl_int4 gemm_add_add_fastgelu_xdl_int4.cpp)
add_dependencies(example_gemm_add_add_fastgelu_xdl example_gemm_add_add_fastgelu_xdl_int4)
endif(USE_BITINT_EXTENSION_INT4)
if(DTYPES MATCHES "int8" OR NOT DEFINED DTYPES)
add_example_executable(example_gemm_add_add_fastgelu_xdl_int8 gemm_add_add_fastgelu_xdl_int8.cpp)
add_example_executable(example_gemm_add_add_fastgelu_xdl_int8 gemm_add_add_fastgelu_xdl_int8.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_add_add_fastgelu_xdl example_gemm_add_add_fastgelu_xdl_int8)
endif()
set(target 1)
......
......@@ -2,34 +2,16 @@ list(APPEND gpu_list gfx908 gfx90a gfx940 gfx941 gfx942)
set(target 0)
foreach(gpu IN LISTS GPU_TARGETS)
if(gpu IN_LIST gpu_list AND target EQUAL 0)
if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
add_example_executable(example_convnd_fwd_xdl_fp32 convnd_fwd_xdl_fp32.cpp)
endif()
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
add_example_executable(example_convnd_fwd_xdl_fp16 convnd_fwd_xdl_fp16.cpp)
endif()
if(DTYPES MATCHES "bf16" OR NOT DEFINED DTYPES)
add_example_executable(example_convnd_fwd_xdl_bf16 convnd_fwd_xdl_bf16.cpp)
endif()
if(DTYPES MATCHES "int8" OR NOT DEFINED DTYPES)
add_example_executable(example_convnd_fwd_xdl_int8 convnd_fwd_xdl_int8.cpp)
endif()
# FIXME: re-enable this exampe as test when SWDEV-335738 is fixed
if(DTYPES MATCHES "fp64" OR NOT DEFINED DTYPES)
# FIXME: re-enable this exampe as test when SWDEV-335738 is fixed
add_example_executable_no_testing(example_convnd_fwd_xdl_fp64 convnd_fwd_xdl_fp64.cpp)
endif()
set(target 1)
endif()
endforeach()
if(DL_KERNELS)
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
add_example_executable(example_convnd_fwd_dl_fp16 convnd_fwd_dl_fp16.cpp)
endif()
if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
add_example_executable(example_convnd_fwd_dl_fp32 convnd_fwd_dl_fp32.cpp)
endif()
if(DTYPES MATCHES "int8" OR NOT DEFINED DTYPES)
add_example_executable(example_convnd_fwd_dl_int8 convnd_fwd_dl_int8.cpp)
endif()
endif()
add_example_executable(example_convnd_fwd_dl_fp16 convnd_fwd_dl_fp16.cpp)
add_example_executable(example_convnd_fwd_dl_fp32 convnd_fwd_dl_fp32.cpp)
add_example_executable(example_convnd_fwd_dl_int8 convnd_fwd_dl_int8.cpp)
......@@ -2,27 +2,27 @@ list(APPEND gpu_list gfx908 gfx90a gfx940 gfx941 gfx942)
set(target 0)
foreach(gpu IN LISTS GPU_TARGETS)
if(gpu IN_LIST gpu_list AND target EQUAL 0)
add_custom_target(example_convnd_fwd_reduce_xdl)
if(DTYPES MATCHES "int8" OR NOT DEFINED DTYPES)
add_example_executable(example_convnd_fwd_max_xdl_int8 convnd_fwd_max_xdl_int8.cpp)
add_custom_target(example_convnd_fwd_reduce_xdl)
add_example_executable(example_convnd_fwd_max_xdl_int8 convnd_fwd_max_xdl_int8.cpp)
if(result EQUAL 0)
add_dependencies(example_convnd_fwd_reduce_xdl example_convnd_fwd_max_xdl_int8)
endif()
if(DTYPES MATCHES "bf16" OR NOT DEFINED DTYPES)
add_example_executable_no_testing(example_convnd_fwd_max_xdl_bf16 convnd_fwd_max_xdl_bf16.cpp)
endif()
add_example_executable_no_testing(example_convnd_fwd_max_xdl_bf16 convnd_fwd_max_xdl_bf16.cpp)
if(result EQUAL 0)
add_dependencies(example_convnd_fwd_reduce_xdl example_convnd_fwd_max_xdl_bf16)
endif()
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
add_example_executable_no_testing(example_convnd_fwd_max_xdl_fp16 convnd_fwd_max_xdl_fp16.cpp)
endif()
add_example_executable_no_testing(example_convnd_fwd_max_xdl_fp16 convnd_fwd_max_xdl_fp16.cpp)
if(result EQUAL 0)
add_dependencies(example_convnd_fwd_reduce_xdl example_convnd_fwd_max_xdl_fp16)
endif()
if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
add_example_executable(example_convnd_fwd_max_xdl_fp32 convnd_fwd_max_xdl_fp32.cpp)
endif()
add_example_executable(example_convnd_fwd_max_xdl_fp32 convnd_fwd_max_xdl_fp32.cpp)
if(result EQUAL 0)
add_dependencies(example_convnd_fwd_reduce_xdl example_convnd_fwd_max_xdl_fp32)
endif()
if(USE_BITINT_EXTENSION_INT4)
endif()
if(USE_BITINT_EXTENSION_INT4)
add_example_executable(example_convnd_fwd_max_xdl_int4 convnd_fwd_max_xdl_int4.cpp)
add_dependencies(example_convnd_fwd_reduce_xdl example_convnd_fwd_max_xdl_int4)
endif(USE_BITINT_EXTENSION_INT4)
set(target 1)
endif(USE_BITINT_EXTENSION_INT4)
set(target 1)
endif()
endforeach()
\ No newline at end of file
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
add_example_executable(example_pool2d_fwd_fp16 pool2d_fwd_fp16.cpp)
endif()
if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
add_example_executable(example_pool2d_fwd_fp32 pool2d_fwd_fp32.cpp)
endif()
add_example_executable(example_pool2d_fwd_fp16 pool2d_fwd_fp16.cpp)
add_example_executable(example_pool2d_fwd_fp32 pool2d_fwd_fp32.cpp)
if(DTYPES MATCHES "int8" OR NOT DEFINED DTYPES)
# dlops
if(DL_KERNELS)
add_example_executable(example_gemm_dl_quantization_int8 gemm_dl_quantization_int8.cpp)
endif()
add_example_executable(example_gemm_dl_quantization_int8 gemm_dl_quantization_int8.cpp)
# xdlops
list(APPEND gpu_list gfx908 gfx90a gfx940 gfx941 gfx942)
set(target 0)
......@@ -14,4 +10,3 @@ foreach(gpu IN LISTS GPU_TARGETS)
set(target 1)
endif()
endforeach()
endif()
\ No newline at end of file
add_custom_target(example_grouped_gemm_xdl)
if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
add_example_executable(example_grouped_gemm_xdl_fp32 grouped_gemm_xdl_fp32.cpp)
add_example_executable(example_grouped_gemm_xdl_fp32 grouped_gemm_xdl_fp32.cpp)
if(result EQUAL 0)
add_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_fp32)
endif()
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
add_example_executable(example_grouped_gemm_xdl_fp16 grouped_gemm_xdl_fp16.cpp)
add_example_executable(example_grouped_gemm_multiple_d_dl_fp16 grouped_gemm_multiple_d_dl_fp16.cpp)
add_example_executable(example_grouped_gemm_xdl_splitk_fp16 grouped_gemm_xdl_splitk_fp16.cpp)
add_example_executable(example_grouped_gemm_xdl_fixed_nk_fp16 grouped_gemm_xdl_fixed_nk_fp16.cpp)
add_example_executable(example_grouped_gemm_xdl_fixed_nk_bias_fp16 grouped_gemm_xdl_fixed_nk_bias_fp16.cpp)
add_dependencies(example_grouped_gemm_xdl
example_grouped_gemm_xdl_fp16
example_grouped_gemm_multiple_d_dl_fp16
example_grouped_gemm_xdl_splitk_fp16
example_grouped_gemm_xdl_fixed_nk_fp16
example_grouped_gemm_xdl_fixed_nk_bias_fp16)
endif()
if(DTYPES MATCHES "bf16" OR NOT DEFINED DTYPES)
add_example_executable(example_grouped_gemm_xdl_bfp16 grouped_gemm_xdl_bfp16.cpp)
add_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_bfp16)
endif()
if(DTYPES MATCHES "int8" OR NOT DEFINED DTYPES)
add_example_executable(example_grouped_gemm_xdl_int8 grouped_gemm_xdl_int8.cpp)
add_example_executable(example_grouped_gemm_xdl_fp16 grouped_gemm_xdl_fp16.cpp)
if(result EQUAL 0)
add_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_fp16)
endif()
add_example_executable(example_grouped_gemm_multiple_d_dl_fp16 grouped_gemm_multiple_d_dl_fp16.cpp)
if(result EQUAL 0)
add_dependencies(example_grouped_gemm_xdl example_grouped_gemm_multiple_d_dl_fp16)
endif()
add_example_executable(example_grouped_gemm_xdl_splitk_fp16 grouped_gemm_xdl_splitk_fp16.cpp)
if(result EQUAL 0)
add_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_splitk_fp16)
endif()
add_example_executable(example_grouped_gemm_xdl_fixed_nk_fp16 grouped_gemm_xdl_fixed_nk_fp16.cpp)
if(result EQUAL 0)
add_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_fixed_nk_fp16)
endif()
add_example_executable(example_grouped_gemm_xdl_fixed_nk_bias_fp16 grouped_gemm_xdl_fixed_nk_bias_fp16.cpp)
if(result EQUAL 0)
add_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_fixed_nk_bias_fp16)
endif()
add_example_executable(example_grouped_gemm_xdl_bf16 grouped_gemm_xdl_bf16.cpp)
if(result EQUAL 0)
add_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_bf16)
endif()
add_example_executable(example_grouped_gemm_xdl_int8 grouped_gemm_xdl_int8.cpp)
if(result EQUAL 0)
add_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_int8)
endif()
if(DTYPES MATCHES "f8" OR NOT DEFINED DTYPES)
add_example_executable(example_grouped_gemm_xdl_fixed_nk_fp8 grouped_gemm_xdl_fixed_nk_fp8.cpp)
add_example_executable(example_grouped_gemm_xdl_fixed_nk_fp8 grouped_gemm_xdl_fixed_nk_fp8.cpp)
if(result EQUAL 0)
add_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_fixed_nk_fp8)
endif()
if(USE_BITINT_EXTENSION_INT4)
add_example_executable(example_grouped_gemm_xdl_int4 grouped_gemm_xdl_int4.cpp)
add_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_int4)
if(result EQUAL 0)
add_dependencies(example_grouped_gemm_xdl example_grouped_gemm_xdl_int4)
endif()
endif()
......@@ -6,30 +6,43 @@ foreach(gpu IN LISTS GPU_TARGETS)
add_custom_target(example_gemm_reduce_xdl_max)
add_custom_target(example_gemm_reduce_xdl_mean_meansquare)
add_custom_target(example_gemm_add_add_mean_meansquare_xdl)
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
add_example_executable(example_gemm_max_xdl_fp16 gemm_max_xdl_fp16.cpp)
add_example_executable(example_gemm_add_add_mean_meansquare_xdl_fp16 gemm_add_add_mean_meansquare_xdl_fp16.cpp)
add_example_executable(example_gemm_mean_meansquare_xdl_fp16 gemm_mean_meansquare_xdl_fp16.cpp)
add_example_executable(example_gemm_max_xdl_fp16 gemm_max_xdl_fp16.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_reduce_xdl_max example_gemm_max_xdl_fp16)
endif()
add_example_executable(example_gemm_add_add_mean_meansquare_xdl_fp16 gemm_add_add_mean_meansquare_xdl_fp16.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_add_add_mean_meansquare_xdl example_gemm_add_add_mean_meansquare_xdl_fp16)
endif()
add_example_executable(example_gemm_mean_meansquare_xdl_fp16 gemm_mean_meansquare_xdl_fp16.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_reduce_xdl_mean_meansquare example_gemm_mean_meansquare_xdl_fp16)
endif()
if(DTYPES MATCHES "int8" OR NOT DEFINED DTYPES)
add_example_executable(example_gemm_max_xdl_int8 gemm_max_xdl_int8.cpp)
add_example_executable(example_gemm_add_addsquare_xdl_int8 gemm_add_addsquare_xdl_int8.cpp)
add_example_executable(example_gemm_max_xdl_int8 gemm_max_xdl_int8.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_reduce_xdl_max example_gemm_max_xdl_int8)
endif()
add_example_executable(example_gemm_add_addsquare_xdl_int8 gemm_add_addsquare_xdl_int8.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_reduce_xdl_mean_meansquare example_gemm_add_addsquare_xdl_int8)
endif()
if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
add_example_executable(example_gemm_max_xdl_fp32 gemm_max_xdl_fp32.cpp)
add_example_executable(example_gemm_mean_meansquare_xdl_fp32 gemm_mean_meansquare_xdl_fp32.cpp)
add_example_executable(example_gemm_max_xdl_fp32 gemm_max_xdl_fp32.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_reduce_xdl_max example_gemm_max_xdl_fp32)
endif()
add_example_executable(example_gemm_mean_meansquare_xdl_fp32 gemm_mean_meansquare_xdl_fp32.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_reduce_xdl_mean_meansquare example_gemm_mean_meansquare_xdl_fp32)
endif()
if(DTYPES MATCHES "bf16" OR NOT DEFINED DTYPES)
add_example_executable(example_gemm_max_xdl_bf16 gemm_max_xdl_bf16.cpp)
add_example_executable(example_gemm_mean_meansquare_xdl_bf16 gemm_mean_meansquare_xdl_bf16.cpp)
add_example_executable(example_gemm_max_xdl_bf16 gemm_max_xdl_bf16.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_reduce_xdl_max example_gemm_max_xdl_bf16)
endif()
add_example_executable(example_gemm_mean_meansquare_xdl_bf16 gemm_mean_meansquare_xdl_bf16.cpp)
if(result EQUAL 0)
add_dependencies(example_gemm_reduce_xdl_mean_meansquare example_gemm_mean_meansquare_xdl_bf16)
endif()
......@@ -40,7 +53,9 @@ foreach(gpu IN LISTS GPU_TARGETS)
if(USE_BITINT_EXTENSION_INT4)
add_example_executable(example_gemm_max_xdl_int4 gemm_max_xdl_int4.cpp)
add_dependencies(example_gemm_reduce_xdl_max example_gemm_max_xdl_int4)
if(result EQUAL 0)
add_dependencies(example_gemm_reduce_xdl_max example_gemm_max_xdl_int4)
endif()
endif()
set(target 1)
endif()
......
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
list(APPEND gpu_list gfx908 gfx90a gfx940 gfx941 gfx942)
set(target 0)
foreach(gpu IN LISTS GPU_TARGETS)
if(gpu IN_LIST gpu_list AND target EQUAL 0)
add_example_executable(example_convnd_bwd_data_xdl_fp16 convnd_bwd_data_xdl_fp16.cpp)
target_link_libraries(example_convnd_bwd_data_xdl_fp16 PRIVATE utility)
if(result EQUAL 0)
target_link_libraries(example_convnd_bwd_data_xdl_fp16 PRIVATE utility)
endif()
set(target 1)
endif()
endforeach()
if(DL_KERNELS)
add_example_executable(example_convnd_bwd_data_dl_fp16 convnd_bwd_data_dl_fp16.cpp)
target_link_libraries(example_convnd_bwd_data_dl_fp16 PRIVATE utility)
endif()
add_example_executable(example_convnd_bwd_data_dl_fp16 convnd_bwd_data_dl_fp16.cpp)
if(result EQUAL 0)
target_link_libraries(example_convnd_bwd_data_dl_fp16 PRIVATE utility)
endif()
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
list(APPEND gpu_list gfx908 gfx90a gfx940 gfx941 gfx942)
set(target 0)
foreach(gpu IN LISTS GPU_TARGETS)
......@@ -7,4 +6,3 @@ foreach(gpu IN LISTS GPU_TARGETS)
set(target 1)
endif()
endforeach()
endif()
......@@ -3,22 +3,20 @@ set(target 0)
foreach(gpu IN LISTS GPU_TARGETS)
if(gpu IN_LIST gpu_list AND target EQUAL 0)
add_custom_target(example_grouped_conv_bwd_weight)
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
add_example_executable(example_grouped_conv_bwd_weight_xdl_fp16 grouped_conv_bwd_weight_xdl_fp16.cpp)
add_example_executable(example_grouped_conv_bwd_weight_xdl_fp16 grouped_conv_bwd_weight_xdl_fp16.cpp)
if(result EQUAL 0)
add_dependencies(example_grouped_conv_bwd_weight example_grouped_conv_bwd_weight_xdl_fp16)
endif()
if(DTYPES MATCHES "bf16" OR NOT DEFINED DTYPES)
add_example_executable(example_grouped_conv_bwd_weight_xdl_bf16 grouped_conv_bwd_weight_xdl_bf16.cpp)
add_example_executable(example_grouped_conv_bwd_weight_xdl_bf16 grouped_conv_bwd_weight_xdl_bf16.cpp)
if(result EQUAL 0)
add_dependencies(example_grouped_conv_bwd_weight example_grouped_conv_bwd_weight_xdl_bf16)
endif()
set(target 1)
endif()
endforeach()
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
if(DL_KERNELS)
add_custom_target(example_grouped_conv_bwd_weight_dl)
add_example_executable(example_grouped_conv_bwd_weight_dl_fp16 grouped_conv_bwd_weight_dl_fp16.cpp)
add_dependencies(example_grouped_conv_bwd_weight_dl example_grouped_conv_bwd_weight_dl_fp16)
endif()
endif()
\ No newline at end of file
add_custom_target(example_grouped_conv_bwd_weight_dl)
add_example_executable(example_grouped_conv_bwd_weight_dl_fp16 grouped_conv_bwd_weight_dl_fp16.cpp)
if(result EQUAL 0)
add_dependencies(example_grouped_conv_bwd_weight_dl example_grouped_conv_bwd_weight_dl_fp16)
endif()
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
list(APPEND gpu_list gfx908 gfx90a gfx940 gfx941 gfx942)
set(target 0)
foreach(gpu IN LISTS GPU_TARGETS)
......@@ -10,4 +9,4 @@ foreach(gpu IN LISTS GPU_TARGETS)
set(target 1)
endif()
endforeach()
endif()
add_custom_target(example_cgemm_xdl)
if(DTYPES MATCHES "bf16" OR NOT DEFINED DTYPES)
add_example_executable(example_cgemm_xdl_bf16 cgemm_xdl_bf16.cpp)
add_example_executable(example_cgemm_xdl_bf16 cgemm_xdl_bf16.cpp)
if(result EQUAL 0)
add_dependencies(example_cgemm_xdl example_cgemm_xdl_bf16)
endif()
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
add_example_executable(example_cgemm_xdl_fp16 cgemm_xdl_fp16.cpp)
add_example_executable(example_cgemm_xdl_fp16 cgemm_xdl_fp16.cpp)
if(result EQUAL 0)
add_dependencies(example_cgemm_xdl example_cgemm_xdl_fp16)
endif()
if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
add_example_executable(example_cgemm_xdl_fp32 cgemm_xdl_fp32.cpp)
add_dependencies(example_cgemm_xdl example_cgemm_xdl_fp32)
if(result EQUAL 0)
add_dependencies(example_cgemm_xdl example_cgemm_xdl_fp32)
endif()
if(DTYPES MATCHES "int8" OR NOT DEFINED DTYPES)
add_example_executable(example_cgemm_xdl_int8 cgemm_xdl_int8.cpp)
add_example_executable(example_cgemm_xdl_int8 cgemm_xdl_int8.cpp)
if(result EQUAL 0)
add_dependencies(example_cgemm_xdl example_cgemm_xdl_int8)
endif()
if(USE_BITINT_EXTENSION_INT4)
......
add_custom_target(example_batched_gemm_xdl)
if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
add_example_executable(example_batched_gemm_xdl_fp32 batched_gemm_xdl_fp32.cpp)
add_example_executable(example_batched_gemm_xdl_fp32 batched_gemm_xdl_fp32.cpp)
if(result EQUAL 0)
add_dependencies(example_batched_gemm_xdl example_batched_gemm_xdl_fp32)
endif()
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
add_example_executable(example_batched_gemm_xdl_fp16 batched_gemm_xdl_fp16.cpp)
add_example_executable(example_batched_gemm_xdl_fp16 batched_gemm_xdl_fp16.cpp)
if(result EQUAL 0)
add_dependencies(example_batched_gemm_xdl example_batched_gemm_xdl_fp16)
endif()
if(DTYPES MATCHES "bf16" OR NOT DEFINED DTYPES)
add_example_executable(example_batched_gemm_xdl_bfp16 batched_gemm_xdl_bfp16.cpp)
add_dependencies(example_batched_gemm_xdl example_batched_gemm_xdl_bfp16)
add_example_executable(example_batched_gemm_xdl_bf16 batched_gemm_xdl_bf16.cpp)
if(result EQUAL 0)
add_dependencies(example_batched_gemm_xdl example_batched_gemm_xdl_bf16)
endif()
if(DTYPES MATCHES "int8" OR NOT DEFINED DTYPES)
add_example_executable(example_batched_gemm_xdl_int8 batched_gemm_xdl_int8.cpp)
add_example_executable(example_batched_gemm_xdl_int8 batched_gemm_xdl_int8.cpp)
if(result EQUAL 0)
add_dependencies(example_batched_gemm_xdl example_batched_gemm_xdl_int8)
endif()
if(USE_BITINT_EXTENSION_INT4)
add_example_executable(example_batched_gemm_xdl_int4 batched_gemm_xdl_int4.cpp)
add_dependencies(example_batched_gemm_xdl example_batched_gemm_xdl_int4)
if(result EQUAL 0)
add_dependencies(example_batched_gemm_xdl example_batched_gemm_xdl_int4)
endif()
endif()
if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
add_example_executable(example_gemm_bias_e_permute_g1m3n2k1_xdl_fp16 gemm_bias_e_permute_g1m3n2k1_xdl_fp16.cpp)
add_example_executable(example_gemm_bias_e_permute_g1m2n3k1_xdl_fp16 gemm_bias_e_permute_g1m2n3k1_xdl_fp16.cpp)
endif()
add_example_executable(example_gemm_bias_e_permute_g1m3n2k1_xdl_fp16 gemm_bias_e_permute_g1m3n2k1_xdl_fp16.cpp)
add_example_executable(example_gemm_bias_e_permute_g1m2n3k1_xdl_fp16 gemm_bias_e_permute_g1m2n3k1_xdl_fp16.cpp)
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment