temp

example/CMakeLists.txt

@@ -47,7 +47,7 @@ function(add_example_executable EXAMPLE_NAME FILE_NAME)
     #only continue if there are some source files left on the list
     if(FILE_NAME)
         add_executable(${EXAMPLE_NAME} ${FILE_NAME})
-        target_link_libraries(${EXAMPLE_NAME} PRIVATE utility)
+        target_link_libraries(${EXAMPLE_NAME} PRIVATE utility getopt::getopt)
         add_test(NAME ${EXAMPLE_NAME} COMMAND $<TARGET_FILE:${EXAMPLE_NAME}> ${ARGN})
         add_dependencies(examples ${EXAMPLE_NAME})
         add_dependencies(check ${EXAMPLE_NAME})
@@ -56,7 +56,7 @@ function(add_example_executable EXAMPLE_NAME FILE_NAME)
     endif()
     #message("add_example returns ${result}")
     set(result ${result} PARENT_SCOPE)
-endfunction(add_example_executable EXAMPLE_NAME)
+endfunction()
 
 function(add_example_dependencies EXAMPLE_NAME FILE_NAME)
     if(result EQUAL 0)
@@ -106,14 +106,14 @@ function(add_example_executable_no_testing EXAMPLE_NAME FILE_NAME)
     #only continue if there are some source files left on the list
     if(FILE_NAME)
         add_executable(${EXAMPLE_NAME} ${FILE_NAME})
-        target_link_libraries(${EXAMPLE_NAME} PRIVATE utility)
+        target_link_libraries(${EXAMPLE_NAME} PRIVATE utility getopt::getopt)
         add_dependencies(examples ${EXAMPLE_NAME})
         rocm_install(TARGETS ${EXAMPLE_NAME} COMPONENT examples)
         set(result 0)
     endif()
     #message("add_example returns ${result}")
     set(result ${result} PARENT_SCOPE)
-endfunction(add_example_executable_no_testing EXAMPLE_NAME)
+endfunction()
 
 # add all example subdir
 file(GLOB dir_list LIST_DIRECTORIES true *)

include/ck/tensor_operation/gpu/element/unary_element_wise_operation.hpp

@@ -174,6 +174,11 @@ struct PassThrough
     {
         y = x;
     }
+    template <>
+    __host__ __device__ void operator()<int4_t, int>(int4_t& y, const int& x) const
+    {
+        y = type_convert<int4_t>(x);
+    }
 #endif
 
     template <>

include/ck/tensor_operation/gpu/warp/wmma_gemm.hpp

@@ -328,7 +328,7 @@ struct WmmaSelector
     }
 #ifdef CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4
     template <>
-    static constexpr auto GetWmma<int4_t, int, 16, 16>()
+    static constexpr auto GetWmma<int4_t, int4_t, int, 16, 16>()
     {
         return WmmaInstr::wmma_i32_16x16x16_iu4;
     }

library/src/tensor_operation_instance/gpu/CMakeLists.txt

@@ -152,7 +152,6 @@ ENDFOREACH()
 if(CK_DEVICE_OTHER_INSTANCES)
         add_library(device_other_operations STATIC ${CK_DEVICE_OTHER_INSTANCES})
         add_library(composablekernels::device_other_operations ALIAS device_other_operations)
-        target_compile_features(device_other_operations PUBLIC)
         set_target_properties(device_other_operations PROPERTIES POSITION_INDEPENDENT_CODE ON)
         target_include_directories(device_other_operations PUBLIC
             $<INSTALL_INTERFACE:${CMAKE_INSTALL_INCLUDEDIR}/ck>

library/src/tensor_operation_instance/gpu/softmax/CMakeLists.txt

-set(DEVICE_SOFTMAX_INSTANCES)
-list(APPEND DEVICE_SOFTMAX_INSTANCES 
+add_instance_library(device_softmax_instance
     device_softmax_f16_f16_instance_rank3_reduce1.cpp
     device_softmax_f16_f16_instance_rank3_reduce2.cpp
     device_softmax_f16_f16_instance_rank3_reduce3.cpp
@@ -14,4 +13,3 @@ list(APPEND DEVICE_SOFTMAX_INSTANCES
     device_softmax_f32_f32_instance_rank4_reduce2.cpp
     device_softmax_f32_f32_instance_rank4_reduce3.cpp
     device_softmax_f32_f32_instance_rank4_reduce4.cpp)
-add_instance_library(device_softmax_instance ${DEVICE_SOFTMAX_INSTANCES})

library/src/utility/CMakeLists.txt

-## utility
-set(UTILITY_SOURCE
+add_library(utility STATIC
     device_memory.cpp
     host_tensor.cpp
     convolution_parameter.cpp
 )
 
-add_library(utility STATIC ${UTILITY_SOURCE})
 add_library(composable_kernel::utility ALIAS utility)
-
+set_target_properties(utility PROPERTIES POSITION_INDEPENDENT_CODE ON)
+target_compile_options(utility PRIVATE ${CMAKE_COMPILER_WARNINGS})
 target_include_directories(utility PUBLIC
     "$<INSTALL_INTERFACE:${CMAKE_INSTALL_INCLUDEDIR}/ck>"
     "$<INSTALL_INTERFACE:${CMAKE_INSTALL_INCLUDEDIR}/ck/library/utility>"
 )
+if(WIN32)
+    target_compile_definitions(utility PUBLIC NOMINMAX)
+endif()
 
 rocm_install(
     TARGETS utility

profiler/src/CMakeLists.txt

@@ -56,7 +56,7 @@ set(PROFILER_EXECUTABLE ckProfiler)
 add_executable(${PROFILER_EXECUTABLE} ${PROFILER_SOURCES})
 target_compile_options(${PROFILER_EXECUTABLE} PRIVATE -Wno-global-constructors)
 
-target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE utility)
+target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE utility getopt::getopt)
 target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_instance)
 target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_splitk_instance)
 target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_add_multiply_instance)

test/CMakeLists.txt

@@ -3,7 +3,7 @@ include_directories(BEFORE
     ${PROJECT_SOURCE_DIR}/profiler/include
 )
 
-include(googletest)
+include(gtest)
 
 add_custom_target(tests)
 
@@ -50,6 +50,7 @@ function(add_test_executable TEST_NAME)
     #only continue if there are some source files left on the list
     if(ARGN)
         add_executable(${TEST_NAME} ${ARGN})
+        target_link_libraries(${TEST_NAME} PRIVATE getopt::getopt)
         add_test(NAME ${TEST_NAME} COMMAND $<TARGET_FILE:${TEST_NAME}>)
         add_dependencies(tests ${TEST_NAME})
         add_dependencies(check ${TEST_NAME})
@@ -58,9 +59,7 @@ function(add_test_executable TEST_NAME)
     endif()
     #message("add_test returns ${result}")
     set(result ${result} PARENT_SCOPE)
-endfunction(add_test_executable TEST_NAME)
-
-include(GoogleTest)
+endfunction()
 
 function(add_gtest_executable TEST_NAME)
     message("adding gtest ${TEST_NAME}")
@@ -109,14 +108,14 @@ function(add_gtest_executable TEST_NAME)
 
         # suppress gtest warnings
         target_compile_options(${TEST_NAME} PRIVATE -Wno-global-constructors -Wno-undef)
-        target_link_libraries(${TEST_NAME} PRIVATE gtest_main)
+        target_link_libraries(${TEST_NAME} PRIVATE gtest_main getopt::getopt)
         add_test(NAME ${TEST_NAME} COMMAND $<TARGET_FILE:${TEST_NAME}>)
         rocm_install(TARGETS ${TEST_NAME} COMPONENT tests)
         set(result 0)
     endif()
     #message("add_gtest returns ${result}")
     set(result ${result} PARENT_SCOPE)
-endfunction(add_gtest_executable TEST_NAME)
+endfunction()
 
 add_subdirectory(magic_number_division)
 add_subdirectory(space_filling_curve)