some small changes

CMakeLists.txt

@@ -113,7 +113,7 @@ message("checking which targets are supported")
 #Setting GPU_TARGETS on command line will override this list
 if(NOT PROFILER_ONLY)
     rocm_check_target_ids(DEFAULT_GPU_TARGETS
-        TARGETS "gfx908;gfx90a;gfx940;gfx941;gfx942;gfx1030;gfx1100;gfx1101;gfx1102;gfx1200")
+        TARGETS "gfx908;gfx90a;gfx940;gfx941;gfx942;gfx1030;gfx1100;gfx1101;gfx1102")
 else()
     add_definitions(-DPROFILER_ONLY)
     set(GPU_TARGETS "" CACHE STRING "" FORCE)
@@ -129,7 +129,7 @@ else()
     elseif(GPU_ARCH MATCHES "gfx11")
         rocm_check_target_ids(DEFAULT_GPU_TARGETS TARGETS "gfx1100;gfx1101;gfx1102")
     elseif(GPU_ARCH MATCHES "gfx12")
-        rocm_check_target_ids(DEFAULT_GPU_TARGETS TARGETS "gfx1200")
+        rocm_check_target_ids(DEFAULT_GPU_TARGETS TARGETS "gfx1200;gfx1201")
     else()
         message(FATAL_ERROR "For PROFILE_ONLY build, please specify GPU_ARCH as gfx90, gfx94, gfx10, gfx11 or gfx12")
     endif()

cmake/EnableCompilerWarnings.cmake

@@ -66,7 +66,7 @@ else()
             -Wunreachable-code
             -Wunused
             -Wno-reserved-identifier
-	    #-Werror
+	    -Werror
             -Wno-option-ignored
             -Wsign-compare
             -Wno-extra-semi-stmt

example/02_gemm_bilinear/CMakeLists.txt

-list(APPEND gpu_list1 gfx1100 gfx1101 gfx1102 gfx1200)
+list(APPEND gpu_list1 gfx1100 gfx1101 gfx1102 gfx1103 gfx1200 gfx1201)
 list(APPEND gpu_list2 gfx908 gfx90a gfx940 gfx941 gfx942 gfx950)
 set(target 0)
 foreach(gpu IN LISTS GPU_TARGETS)

include/ck/ck.hpp

@@ -58,7 +58,7 @@
 #if defined(__gfx1100__) || defined(__gfx1101__) || defined(__gfx1102__) || defined(__gfx1103__)
 #define __gfx11__
 #endif
-#if defined(__gfx1200__)
+#if defined(__gfx1200__) || defined(__gfx1201__)
 #define __gfx12__
 #endif
 
@@ -107,7 +107,7 @@
 // WMMA instruction
 #ifndef __HIP_DEVICE_COMPILE__ // for host code
 #define CK_USE_AMD_WMMA
-#elif defined(__gfx11__) // for GPU code
+#elif defined(__gfx11__) || defined(__gfx12__) // for GPU code
 #define CK_USE_AMD_WMMA
 #endif
 

library/include/ck/library/utility/check_err.hpp

@@ -156,7 +156,7 @@ check_err(const Range& out,
         {
             max_err = err > max_err ? err : max_err;
             err_count++;
-            // if(err_count < 5)
+            if(err_count < 5)
             {
                 std::cerr << msg << std::setw(12) << std::setprecision(7) << " out[" << i
                           << "] != ref[" << i << "]: " << o << " != " << r << std::endl;

profiler/src/CMakeLists.txt

@@ -68,7 +68,6 @@ target_compile_options(${PROFILER_EXECUTABLE} PRIVATE -Wno-global-constructors)
 
 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)
 target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_multiply_add_instance)

script/cmake-ck-dev.sh

@@ -11,7 +11,7 @@ cmake
 -D CMAKE_CXX_FLAGS="-std=c++17 -O3 -ftemplate-backtrace-limit=0  -fPIE  -Wno-gnu-line-marker"     \
 -D CMAKE_BUILD_TYPE=Release                                                                       \
 -D BUILD_DEV=ON                                                                                   \
--D GPU_TARGETS="gfx90a"                                                             \
+-D GPU_TARGETS="gfx908;gfx90a;gfx940"                                                             \
 -D CMAKE_VERBOSE_MAKEFILE:BOOL=ON                                                                 \
 -D USE_BITINT_EXTENSION_INT4=OFF                                                                  \
 ${MY_PROJECT_SOURCE}