merge from develop and revisison for pr#881

example/41_grouped_conv_conv_fwd/CMakeLists.txt

@@ -3,9 +3,15 @@ list(APPEND gpu_list2 gfx908 gfx90a)
 set(target 0)
 foreach(gpu IN LISTS GPU_TARGETS)
  if(gpu IN_LIST gpu_list1 AND target EQUAL 0)
-   add_example_executable(example_grouped_conv_conv_fwd_xdl_fp32 grouped_conv_conv_fwd_xdl_fp32.cpp)
-   add_example_executable(example_grouped_conv_conv_fwd_xdl_fp16 grouped_conv_conv_fwd_xdl_fp16.cpp)
-   add_example_executable(example_grouped_conv_conv_fwd_xdl_bf16 grouped_conv_conv_fwd_xdl_bf16.cpp)
+   if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
+      add_example_executable(example_grouped_conv_conv_fwd_xdl_fp32 grouped_conv_conv_fwd_xdl_fp32.cpp)
+   endif()
+   if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
+      add_example_executable(example_grouped_conv_conv_fwd_xdl_fp16 grouped_conv_conv_fwd_xdl_fp16.cpp)
+   endif()
+   if(DTYPES MATCHES "bf16" OR NOT DEFINED DTYPES)
+      add_example_executable(example_grouped_conv_conv_fwd_xdl_bf16 grouped_conv_conv_fwd_xdl_bf16.cpp)
+   endif()
    if(USE_BITINT_EXTENSION_INT4)
       add_example_executable(example_grouped_conv_conv_fwd_xdl_int4 grouped_conv_conv_fwd_xdl_int4.cpp)
    endif(USE_BITINT_EXTENSION_INT4)
@@ -13,10 +19,8 @@ foreach(gpu IN LISTS GPU_TARGETS)
  endif()
 endforeach()
 
-set(target 0)
-foreach(gpu IN LISTS GPU_TARGETS)
- if(gpu IN_LIST gpu_list2 AND target EQUAL 0)
-   add_example_executable(example_grouped_conv_conv_fwd_xdl_int8 grouped_conv_conv_fwd_xdl_int8.cpp)
-   set(target 1)
- endif()
-endforeach()
+if(NOT GPU_TARGETS MATCHES "gfx94" AND NOT GPU_TARGETS MATCHES "gfx1")
+   if(DTYPES MATCHES "int8" OR NOT DEFINED DTYPES)
+      add_example_executable(example_grouped_conv_conv_fwd_xdl_int8 grouped_conv_conv_fwd_xdl_int8.cpp)
+   endif()
+endif()

example/42_groupnorm/CMakeLists.txt

-add_example_executable(example_groupnorm_sigmoid_mul_fp16 groupnorm_sigmoid_mul_fp16.cpp)
-add_example_executable(example_groupnorm_splitk_fp16 groupnorm_splitk_fp16.cpp)
-add_example_executable(example_groupnorm_swish_fp16 groupnorm_swish_fp16.cpp)
+if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
+    add_example_executable(example_groupnorm_sigmoid_mul_fp16 groupnorm_sigmoid_mul_fp16.cpp)
+    add_example_executable(example_groupnorm_splitk_fp16 groupnorm_splitk_fp16.cpp)
+    add_example_executable(example_groupnorm_swish_fp16 groupnorm_swish_fp16.cpp)
+endif()

example/43_splitk_gemm_bias_e_permute/CMakeLists.txt

-add_example_executable(example_splitk_gemm_bias_e_permute_xdl_fp16 splitk_gemm_bias_e_permute_xdl_fp16.cpp)
-add_example_executable(example_splitk_gemm_bias_e_permute_xdl_fp32 splitk_gemm_bias_e_permute_xdl_fp32.cpp)
+if(DTYPES MATCHES "fp16" OR NOT DEFINED DTYPES)
+    add_example_executable(example_splitk_gemm_bias_e_permute_xdl_fp16 splitk_gemm_bias_e_permute_xdl_fp16.cpp)
+endif()
+if(DTYPES MATCHES "fp32" OR NOT DEFINED DTYPES)
+    add_example_executable(example_splitk_gemm_bias_e_permute_xdl_fp32 splitk_gemm_bias_e_permute_xdl_fp32.cpp)
+endif()

example/43_splitk_gemm_bias_e_permute/splitk_gemm_bias_e_permute_xdl_fp16.cpp

@@ -8,7 +8,7 @@
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
-#include "ck/tensor_operation/gpu/device/device_splitk_contraction_multiple_d_xdl_cshuffle.hpp"
+#include "ck/tensor_operation/gpu/device/impl/device_splitk_contraction_multiple_d_xdl_cshuffle.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/utility/check_err.hpp"