mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-03-20 15:17:41 +00:00
1fc5a3f3ac
* CMakeLists.txt hack for Windows. * Add Windows build instructions. * Fix type issue with variadic min function. * Use std::common_type to fix the variadic min/max functions. * Enable CPU guard compilation on Windows. * Suppress warnings related to std::getenv on Windows platform. * Git ignore the output directory on Windows platform. * Powershell script for running tests and generating reports. * Improve test logging. * Disable non-conv tests. * Fix Debug build on Windows. * More debug build changes. * Update Windows build instructions. * Enable all tests. * Test fixes. * Suppress not found linker options warning. * Update unsigned long literals and format specifiers to work correctly in Windows * Fix conv 3D bwd weight bilinear tests on Windows. * Revert changes on .gitignore. * Clean-up CMake project file for Windows builds. * clang-format * Fix definition of CMAKE_PREFIX_PATH on both Linux and Windows platforms. * Fix building examples on Windows. * Update Readme. * Remove the suppression of the deprecated warnings. * Remove Windows specific min/max implementations from CK Tile math core. * Remove unnecessary no-op on Windows. --------- Co-authored-by: User <user@example.com> Co-authored-by: Ville Pietilä <none> Co-authored-by: John Afaganis <john.afaganis@amd.com> Co-authored-by: Ville Pietilä <>
273 lines
13 KiB
CMake
273 lines
13 KiB
CMake
# Copyright (c) Advanced Micro Devices, Inc., or its affiliates.
|
|
# SPDX-License-Identifier: MIT
|
|
|
|
include_directories(BEFORE
|
|
${PROJECT_SOURCE_DIR}/include
|
|
${PROJECT_SOURCE_DIR}/library/include
|
|
)
|
|
|
|
if(WIN32)
|
|
# On Windows, HIP uses -nostdlib which prevents C runtime linking
|
|
# We need legacy_stdio_definitions.lib to provide vfprintf and other legacy C functions
|
|
# This is mainly needed for the getopt library.
|
|
set(LEGACY_STDIO_SEARCH_PATHS)
|
|
|
|
# Try to use Visual C++ Tools environment variable (if build executes from Visual Studio Developer Command Prompt)
|
|
if(DEFINED ENV{VCToolsInstallDir})
|
|
list(APPEND LEGACY_STDIO_SEARCH_PATHS "$ENV{VCToolsInstallDir}/lib/x64")
|
|
endif()
|
|
|
|
# Fallback: Search common Visual Studio installation locations
|
|
file(GLOB MSVC_LIB_PATHS "C:/Program Files/Microsoft Visual Studio/*/*/VC/Tools/MSVC/*/lib/x64")
|
|
list(APPEND LEGACY_STDIO_SEARCH_PATHS ${MSVC_LIB_PATHS})
|
|
|
|
# Use find_library to locate the library
|
|
find_library(LEGACY_STDIO_LIB legacy_stdio_definitions
|
|
PATHS ${LEGACY_STDIO_SEARCH_PATHS}
|
|
NO_DEFAULT_PATH
|
|
)
|
|
|
|
if(LEGACY_STDIO_LIB)
|
|
message(STATUS "Found legacy_stdio_definitions.lib: ${LEGACY_STDIO_LIB}")
|
|
add_link_options("SHELL:-Xlinker \"${LEGACY_STDIO_LIB}\"")
|
|
else()
|
|
message(WARNING "Could not find legacy_stdio_definitions.lib - examples may fail to link.")
|
|
endif()
|
|
endif()
|
|
|
|
add_custom_target(examples)
|
|
|
|
|
|
# list of examples that are labelled as REGRESSION_EXAMPLE for make regression (runtime more than 30 seconds)
|
|
# all other tests are labelled as SMOKE_EXAMPLE
|
|
set(REGRESSION_EXAMPLES
|
|
example_sparse_embedding3_forward_layernorm
|
|
)
|
|
|
|
|
|
function(add_example_dependencies EXAMPLE_NAME FILE_NAME)
|
|
if(FILE_NAME)
|
|
add_dependencies(EXAMPLE_NAME FILE_NAME)
|
|
endif()
|
|
endfunction(add_example_dependencies EXAMPLE_NAME)
|
|
|
|
function(add_example_executable EXAMPLE_NAME FILE_NAME)
|
|
message(DEBUG "adding example ${EXAMPLE_NAME}")
|
|
set(result 1)
|
|
if(DEFINED DTYPES)
|
|
foreach(source IN LISTS FILE_NAME)
|
|
get_filename_component(source_name ${source} NAME)
|
|
set(test 0)
|
|
if((source_name MATCHES "_fp16" OR source_name MATCHES "_f16") AND NOT "fp16" IN_LIST DTYPES)
|
|
set(test 1)
|
|
endif()
|
|
if((source_name MATCHES "_fp32" OR source_name MATCHES "_f32") AND NOT "fp32" IN_LIST DTYPES)
|
|
set(test 1)
|
|
endif()
|
|
if((source_name MATCHES "_fp64" OR source_name MATCHES "_f64") AND NOT "fp64" IN_LIST DTYPES)
|
|
set(test 1)
|
|
endif()
|
|
if((source_name MATCHES "_fp8" OR source_name MATCHES "_f8") AND NOT "fp8" IN_LIST DTYPES)
|
|
set(test 1)
|
|
endif()
|
|
if((source_name MATCHES "_bf8" OR source_name MATCHES "_bf8") AND NOT "bf8" IN_LIST DTYPES)
|
|
set(test 1)
|
|
endif()
|
|
if((source_name MATCHES "_bf16" OR source_name MATCHES "_b16") AND NOT "bf16" IN_LIST DTYPES)
|
|
set(test 1)
|
|
endif()
|
|
if((source_name MATCHES "_int8" OR source_name MATCHES "_i8") AND NOT "int8" IN_LIST DTYPES)
|
|
set(test 1)
|
|
endif()
|
|
if(test EQUAL 1)
|
|
message(DEBUG "removing example source file ${source} ")
|
|
list(REMOVE_ITEM FILE_NAME "${source}")
|
|
endif()
|
|
endforeach()
|
|
endif()
|
|
|
|
set(EX_TARGETS ${SUPPORTED_GPU_TARGETS})
|
|
|
|
foreach(source IN LISTS FILE_NAME)
|
|
get_filename_component(source_name ${source} NAME)
|
|
#Do not build any DL examples if DL_KERNELS not set
|
|
if(NOT DEFINED DL_KERNELS AND source_name MATCHES "_dl")
|
|
message(DEBUG "removing dl example ${source} ")
|
|
list(REMOVE_ITEM FILE_NAME "${source}")
|
|
endif()
|
|
#Do not build any DPP examples if DPP_KERNELS not set
|
|
if(NOT DEFINED DPP_KERNELS AND source_name MATCHES "_dpp")
|
|
message(DEBUG "removing dpp example ${source} ")
|
|
list(REMOVE_ITEM FILE_NAME "${source}")
|
|
endif()
|
|
#Do not build any XDL examples if gfx9 targets are not on the list
|
|
if(NOT EX_TARGETS MATCHES "gfx9" AND NOT EX_TARGETS MATCHES "gfx11" AND NOT EX_TARGETS MATCHES "gfx12" AND source_name MATCHES "_xdl")
|
|
message(DEBUG "removing xdl example ${source} ")
|
|
list(REMOVE_ITEM FILE_NAME "${source}")
|
|
endif()
|
|
#Do not build any WMMA examples if gfx11 targets are not on the list
|
|
if(NOT EX_TARGETS MATCHES "gfx11" AND NOT EX_TARGETS MATCHES "gfx12" AND source_name MATCHES "_wmma")
|
|
message(DEBUG "removing wmma example ${source} ")
|
|
list(REMOVE_ITEM FILE_NAME "${source}")
|
|
endif()
|
|
#Do not build any microscaling examples if gfx950 target is not on the list
|
|
if(NOT EX_TARGETS MATCHES "gfx950" AND source_name MATCHES "_mx")
|
|
message(DEBUG "removing microscaling example ${source} ")
|
|
list(REMOVE_ITEM FILE_NAME "${source}")
|
|
endif()
|
|
#Do not build any FP8 examples if CK_ENABLE_FP8 not set
|
|
if(NOT DEFINED CK_ENABLE_FP8 AND source_name MATCHES "_fp8")
|
|
message(DEBUG "removing fp8 example ${source} ")
|
|
list(REMOVE_ITEM FILE_NAME "${source}")
|
|
endif()
|
|
#Do not build any BF8 examples if CK_ENABLE_BF8 not set
|
|
if(NOT DEFINED CK_ENABLE_BF8 AND source_name MATCHES "_bf8")
|
|
message(DEBUG "removing bf8 example ${source} ")
|
|
list(REMOVE_ITEM FILE_NAME "${source}")
|
|
endif()
|
|
# Build fp8 gemm_multiply_multiply and moe only on gfx94/95 and gfx12
|
|
if(NOT EX_TARGETS MATCHES "gfx94" AND NOT EX_TARGETS MATCHES "gfx95" AND NOT EX_TARGETS MATCHES "gfx12")
|
|
if(source_name MATCHES "fp8" AND source_name MATCHES "(gemm_multiply_multiply|moe)")
|
|
message(DEBUG "Skipping ${source} example for current target")
|
|
list(REMOVE_ITEM FILE_NAME "${source}")
|
|
endif()
|
|
endif()
|
|
endforeach()
|
|
#only continue if there are some source files left on the list
|
|
set(source_name_list "")
|
|
foreach(source IN LISTS FILE_NAME)
|
|
get_filename_component(source_name ${source} NAME)
|
|
list(APPEND source_name_list ${source_name})
|
|
endforeach()
|
|
if(FILE_NAME)
|
|
if(source_name_list MATCHES "_xdl" AND NOT source_name_list MATCHES "_pk_i4")
|
|
list(REMOVE_ITEM EX_TARGETS gfx900 gfx906 gfx906:xnack- gfx1030 gfx10-3-generic)
|
|
elseif(source_name_list MATCHES "_wmma")
|
|
list(REMOVE_ITEM EX_TARGETS gfx900 gfx906 gfx906:xnack- gfx908:xnack+ gfx908:xnack- gfx90a:xnack+ gfx90a:xnack- gfx908 gfx90a gfx942 gfx1030 gfx950)
|
|
elseif(source_name_list MATCHES "_mx") #only build mx example for gfx950
|
|
list(REMOVE_ITEM EX_TARGETS gfx900 gfx906 gfx906:xnack- gfx908:xnack+ gfx908:xnack- gfx90a:xnack+ gfx90a:xnack- gfx908 gfx90a gfx942 gfx1030 gfx1100 gfx1101 gfx1102 gfx1103 gfx1150 gfx1151 gfx1152 gfx1153 gfx1200 gfx1201 gfx10-3-generic gfx11-generic gfx12-generic)
|
|
elseif(source_name_list MATCHES "_pk_i4") #only build these examples for gfx942 gfx950 and rdna3/4
|
|
message(DEBUG "trimming targets for ${FILE_NAME}")
|
|
list(REMOVE_ITEM EX_TARGETS gfx900 gfx906 gfx906:xnack- gfx908:xnack+ gfx908:xnack- gfx90a:xnack+ gfx90a:xnack- gfx908 gfx90a gfx1030 gfx10-3-generic)
|
|
endif()
|
|
set_source_files_properties(${FILE_NAME} PROPERTIES LANGUAGE HIP)
|
|
add_executable(${EXAMPLE_NAME} ${FILE_NAME})
|
|
target_link_libraries(${EXAMPLE_NAME} PRIVATE utility)
|
|
target_link_libraries(${EXAMPLE_NAME} PRIVATE getopt::getopt)
|
|
add_test(NAME ${EXAMPLE_NAME} COMMAND $<TARGET_FILE:${EXAMPLE_NAME}> ${ARGN})
|
|
set_property(TARGET ${EXAMPLE_NAME} PROPERTY HIP_ARCHITECTURES ${EX_TARGETS})
|
|
add_dependencies(examples ${EXAMPLE_NAME})
|
|
add_dependencies(check ${EXAMPLE_NAME})
|
|
rocm_install(TARGETS ${EXAMPLE_NAME} COMPONENT examples)
|
|
set(result 0)
|
|
endif()
|
|
message(DEBUG "add_example returns ${result}")
|
|
if(result EQUAL 0 AND NOT "${EXAMPLE_NAME}" IN_LIST REGRESSION_EXAMPLES)
|
|
set_tests_properties(${EXAMPLE_NAME} PROPERTIES LABELS "SMOKE_TEST")
|
|
add_dependencies(smoke ${EXAMPLE_NAME})
|
|
elseif(result EQUAL 0 AND "${EXAMPLE_NAME}" IN_LIST REGRESSION_EXAMPLES)
|
|
set_tests_properties(${EXAMPLE_NAME} PROPERTIES LABELS "REGRESSION_TEST")
|
|
add_dependencies(regression ${EXAMPLE_NAME})
|
|
endif()
|
|
set(result ${result} PARENT_SCOPE)
|
|
endfunction(add_example_executable EXAMPLE_NAME)
|
|
|
|
function(add_example_dependencies EXAMPLE_NAME FILE_NAME)
|
|
if(result EQUAL 0)
|
|
add_dependencies(${EXAMPLE_NAME} ${FILE_NAME})
|
|
endif()
|
|
endfunction(add_example_dependencies EXAMPLE_NAME)
|
|
|
|
function(add_example_executable_no_testing EXAMPLE_NAME FILE_NAME)
|
|
message(DEBUG "adding example ${EXAMPLE_NAME}")
|
|
set(result 1)
|
|
if(DEFINED DTYPES)
|
|
foreach(source IN LISTS FILE_NAME)
|
|
get_filename_component(source_name ${source} NAME)
|
|
set(test 0)
|
|
if((source_name MATCHES "_fp16" OR source_name MATCHES "_f16") AND NOT "fp16" IN_LIST DTYPES)
|
|
set(test 1)
|
|
endif()
|
|
if((source_name MATCHES "_fp32" OR source_name MATCHES "_f32") AND NOT "fp32" IN_LIST DTYPES)
|
|
set(test 1)
|
|
endif()
|
|
if((source_name MATCHES "_fp64" OR source_name MATCHES "_f64") AND NOT "fp64" IN_LIST DTYPES)
|
|
set(test 1)
|
|
endif()
|
|
if((source_name MATCHES "_fp8" OR source_name MATCHES "_f8") AND NOT "fp8" IN_LIST DTYPES)
|
|
set(test 1)
|
|
endif()
|
|
if((source_name MATCHES "_bf8" OR source_name MATCHES "_bf8") AND NOT "bf8" IN_LIST DTYPES)
|
|
set(test 1)
|
|
endif()
|
|
if((source_name MATCHES "_bf16" OR source_name MATCHES "_b16") AND NOT "bf16" IN_LIST DTYPES)
|
|
set(test 1)
|
|
endif()
|
|
if((source_name MATCHES "_int8" OR source_name MATCHES "_i8") AND NOT "int8" IN_LIST DTYPES)
|
|
set(test 1)
|
|
endif()
|
|
if(test EQUAL 1)
|
|
message(DEBUG "removing example ${source} ")
|
|
list(REMOVE_ITEM FILE_NAME "${source}")
|
|
endif()
|
|
endforeach()
|
|
endif()
|
|
|
|
set(EX_TARGETS ${SUPPORTED_GPU_TARGETS})
|
|
|
|
set(source_name_list "")
|
|
foreach(source IN LISTS FILE_NAME)
|
|
get_filename_component(source_name ${source} NAME)
|
|
#Do not build any DL examples if DL_KERNELS not set
|
|
if(NOT DEFINED DL_KERNELS AND source_name MATCHES "_dl")
|
|
message(DEBUG "removing dl example ${source} ")
|
|
list(REMOVE_ITEM FILE_NAME "${source}")
|
|
endif()
|
|
#Do not build any XDL examples if gfx9 targets are not on the list
|
|
if(NOT EX_TARGETS MATCHES "gfx9" AND NOT EX_TARGETS MATCHES "gfx11" AND NOT EX_TARGETS MATCHES "gfx12" AND source_name MATCHES "_xdl")
|
|
message(DEBUG "removing xdl example ${source} ")
|
|
list(REMOVE_ITEM FILE_NAME "${source}")
|
|
endif()
|
|
#Do not build any WMMA examples if gfx11 targets are not on the list
|
|
if(NOT EX_TARGETS MATCHES "gfx11" AND NOT EX_TARGETS MATCHES "gfx12" AND source_name MATCHES "_wmma")
|
|
message(DEBUG "removing wmma example ${source} ")
|
|
list(REMOVE_ITEM FILE_NAME "${source}")
|
|
endif()
|
|
list(APPEND source_name_list ${source_name})
|
|
endforeach()
|
|
#only continue if there are some source files left on the list
|
|
if(FILE_NAME)
|
|
if(source_name_list MATCHES "_xdl")
|
|
list(REMOVE_ITEM EX_TARGETS gfx900 gfx906 gfx906:xnack- gfx1030 gfx10-3-generic)
|
|
elseif(source_name_list MATCHES "_wmma")
|
|
list(REMOVE_ITEM EX_TARGETS gfx900 gfx906 gfx906:xnack- gfx908:xnack+ gfx908:xnack- gfx90a:xnack+ gfx90a:xnack- gfx908 gfx90a gfx942 gfx1030 gfx950)
|
|
endif()
|
|
set_source_files_properties(${FILE_NAME} PROPERTIES LANGUAGE HIP)
|
|
add_executable(${EXAMPLE_NAME} ${FILE_NAME})
|
|
target_link_libraries(${EXAMPLE_NAME} PRIVATE utility)
|
|
target_link_libraries(${EXAMPLE_NAME} PRIVATE getopt::getopt)
|
|
add_dependencies(examples ${EXAMPLE_NAME})
|
|
set_property(TARGET ${EXAMPLE_NAME} PROPERTY HIP_ARCHITECTURES ${EX_TARGETS})
|
|
rocm_install(TARGETS ${EXAMPLE_NAME} COMPONENT examples)
|
|
set(result 0)
|
|
endif()
|
|
|
|
message(DEBUG "add_example returns ${result}")
|
|
set(result ${result} PARENT_SCOPE)
|
|
|
|
endfunction(add_example_executable_no_testing EXAMPLE_NAME)
|
|
|
|
function(example_compile_options EXAMPLE_NAME)
|
|
if(TARGET ${EXAMPLE_NAME})
|
|
target_compile_options(${EXAMPLE_NAME} ${ARGN})
|
|
endif()
|
|
endfunction(example_compile_options)
|
|
|
|
# add all example subdir
|
|
file(GLOB dir_list LIST_DIRECTORIES true *)
|
|
FOREACH(subdir ${dir_list})
|
|
if(IS_DIRECTORY "${subdir}" AND EXISTS "${subdir}/CMakeLists.txt")
|
|
add_subdirectory(${subdir})
|
|
ENDIF()
|
|
ENDFOREACH()
|