Merge branch 'develop' into amd-develop

test/grouped_convnd_fwd/CMakeLists.txt

-add_gtest_executable(test_grouped_convnd_fwd test_grouped_convnd_fwd.cpp)
-target_link_libraries(test_grouped_convnd_fwd PRIVATE utility device_grouped_conv1d_fwd_instance device_grouped_conv2d_fwd_instance device_grouped_conv3d_fwd_instance)
+add_gtest_executable(test_grouped_convnd_fwd test_grouped_convnd_fwd_xdl_wmma.cpp)
+if(result EQUAL 0)
+    target_link_libraries(test_grouped_convnd_fwd PRIVATE utility device_grouped_conv1d_fwd_instance device_grouped_conv2d_fwd_instance device_grouped_conv3d_fwd_instance)
+endif()
 
 add_gtest_executable(test_grouped_convnd_fwd_multi_ab_interface test_grouped_convnd_fwd_multi_ab_interface.cpp)
-target_link_libraries(test_grouped_convnd_fwd_multi_ab_interface PRIVATE utility)
+if(result EQUAL 0)
+    target_link_libraries(test_grouped_convnd_fwd_multi_ab_interface PRIVATE utility)
+endif()
 
-add_gtest_executable(test_grouped_convnd_fwd_multi_d_interface_compatibility test_grouped_convnd_fwd_multi_d_interface_compatibility.cpp)
-target_link_libraries(test_grouped_convnd_fwd_multi_d_interface_compatibility PRIVATE utility device_grouped_conv3d_fwd_instance)
+add_gtest_executable(test_grouped_convnd_fwd_multi_d_interface_compatibility test_grouped_convnd_fwd_multi_d_interface_compatibility_xdl_wmma.cpp)
+if(result EQUAL 0)
+    target_link_libraries(test_grouped_convnd_fwd_multi_d_interface_compatibility PRIVATE utility device_grouped_conv3d_fwd_instance)
+endif()

test/grouped_gemm/CMakeLists.txt

-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(test_grouped_gemm)
-   add_gtest_executable(test_grouped_gemm_splitk test_grouped_gemm_splitk.cpp)
-   add_gtest_executable(test_grouped_gemm_interface test_grouped_gemm_interface.cpp)
-   target_link_libraries(test_grouped_gemm_splitk PRIVATE utility device_grouped_gemm_instance)
-   target_link_libraries(test_grouped_gemm_interface PRIVATE utility device_grouped_gemm_instance)
-   
-   add_dependencies(test_grouped_gemm test_grouped_gemm_splitk test_grouped_gemm_interface)
-   set(target 1)
- endif()
-endforeach()
+add_custom_target(test_grouped_gemm)
+
+add_gtest_executable(test_grouped_gemm_splitk test_grouped_gemm_splitk_xdl.cpp)
+if(result EQUAL 0)
+    target_link_libraries(test_grouped_gemm_splitk PRIVATE utility device_grouped_gemm_instance)
+    add_dependencies(test_grouped_gemm test_grouped_gemm_splitk)
+endif()
+
+add_gtest_executable(test_grouped_gemm_interface test_grouped_gemm_interface_xdl.cpp)
+if(result EQUAL 0)
+    target_link_libraries(test_grouped_gemm_interface PRIVATE utility device_grouped_gemm_instance)
+    add_dependencies(test_grouped_gemm test_grouped_gemm_interface)
+endif()

test/normalization_bwd_data/CMakeLists.txt

 add_custom_target(test_normalization_bwd_data)
 add_gtest_executable(test_layernorm2d_bwd_data_fp32 test_layernorm2d_bwd_data_fp32.cpp)
-if(result EQUAL 0)
-  target_link_libraries(test_layernorm2d_bwd_data_fp32 PRIVATE utility device_normalization_bwd_data_instance)
-  add_dependencies(test_normalization_bwd_data test_layernorm2d_bwd_data_fp32)
-endif()
+target_link_libraries(test_layernorm2d_bwd_data_fp32 PRIVATE utility device_normalization_bwd_data_instance)
+add_dependencies(test_normalization_bwd_data test_layernorm2d_bwd_data_fp32)
 
 add_gtest_executable(test_groupnorm_bwd_data_fp32 test_groupnorm_bwd_data_fp32.cpp)
-if(result EQUAL 0)
-  target_link_libraries(test_groupnorm_bwd_data_fp32 PRIVATE utility device_normalization_bwd_data_instance)
-  add_dependencies(test_normalization_bwd_data test_groupnorm_bwd_data_fp32)
-endif()
-
+target_link_libraries(test_groupnorm_bwd_data_fp32 PRIVATE utility device_normalization_bwd_data_instance)
+add_dependencies(test_normalization_bwd_data test_groupnorm_bwd_data_fp32)

test/normalization_bwd_gamma_beta/CMakeLists.txt

 add_custom_target(test_normalization_bwd_gamma_beta)
 add_gtest_executable(test_layernorm2d_bwd_gamma_beta_fp32 test_layernorm2d_bwd_gamma_beta_fp32.cpp)
-if(result EQUAL 0)
-  target_link_libraries(test_layernorm2d_bwd_gamma_beta_fp32 PRIVATE utility device_normalization_bwd_gamma_beta_instance)
-  add_dependencies(test_normalization_bwd_gamma_beta test_layernorm2d_bwd_gamma_beta_fp32)
-endif()
+target_link_libraries(test_layernorm2d_bwd_gamma_beta_fp32 PRIVATE utility device_normalization_bwd_gamma_beta_instance)
+add_dependencies(test_normalization_bwd_gamma_beta test_layernorm2d_bwd_gamma_beta_fp32)
 
 add_gtest_executable(test_groupnorm_bwd_gamma_beta_fp32 test_groupnorm_bwd_gamma_beta_fp32.cpp)
-if(result EQUAL 0)
-  target_link_libraries(test_groupnorm_bwd_gamma_beta_fp32 PRIVATE utility device_normalization_bwd_gamma_beta_instance)
-  add_dependencies(test_normalization_bwd_gamma_beta test_groupnorm_bwd_gamma_beta_fp32)
-endif()
-
+target_link_libraries(test_groupnorm_bwd_gamma_beta_fp32 PRIVATE utility device_normalization_bwd_gamma_beta_instance)
+add_dependencies(test_normalization_bwd_gamma_beta test_groupnorm_bwd_gamma_beta_fp32)

test/permute_scale/CMakeLists.txt

 add_custom_target(test_permute)
 add_gtest_executable(test_permute_scale test_permute_scale.cpp)
-if(result EQUAL 0)
-  target_link_libraries(test_permute_scale PRIVATE utility device_permute_scale_instance)
-  add_dependencies(test_permute test_permute_scale)
-endif()
+target_link_libraries(test_permute_scale PRIVATE utility device_permute_scale_instance)
+add_dependencies(test_permute test_permute_scale)

test/permute_scale/test_permute_scale.cpp

@@ -52,40 +52,40 @@ TYPED_TEST_SUITE(TestPermute, KernelTypes);
 TYPED_TEST(TestPermute, Test1D)
 {
     constexpr ck::index_t NumDims = 1;
-    this->template Run<NumDims>({8}, {1}, {2});
-    this->template Run<NumDims>({8}, {2}, {1});
+    this->template Run<NumDims>({16}, {1}, {1});
+    this->template Run<NumDims>({16}, {1}, {2});
     this->template Run<NumDims>({1}, {1}, {1});
 }
 
 TYPED_TEST(TestPermute, Test2D)
 {
     constexpr ck::index_t NumDims = 2;
-    this->template Run<NumDims>({8, 4}, {4, 1}, {1, 8});
-    this->template Run<NumDims>({8, 4}, {1, 8}, {4, 1});
+    this->template Run<NumDims>({8, 16}, {16, 1}, {1, 8});
+    this->template Run<NumDims>({8, 16}, {1, 8}, {16, 1});
     this->template Run<NumDims>({1, 1}, {1, 1}, {1, 1});
 }
 
 TYPED_TEST(TestPermute, Test3D)
 {
     constexpr ck::index_t NumDims = 3;
-    this->template Run<NumDims>({2, 4, 4}, {16, 4, 1}, {1, 2, 8});
-    this->template Run<NumDims>({2, 4, 4}, {1, 2, 8}, {16, 4, 1});
+    this->template Run<NumDims>({8, 2, 8}, {16, 8, 1}, {1, 8, 16});
+    this->template Run<NumDims>({8, 2, 8}, {1, 8, 16}, {16, 8, 1});
     this->template Run<NumDims>({1, 1, 1}, {1, 1, 1}, {1, 1, 1});
 }
 
 TYPED_TEST(TestPermute, Test4D)
 {
     constexpr ck::index_t NumDims = 4;
-    this->template Run<NumDims>({2, 4, 4, 4}, {64, 16, 4, 1}, {1, 2, 8, 32});
-    this->template Run<NumDims>({2, 4, 4, 4}, {1, 2, 8, 32}, {64, 16, 4, 1});
+    this->template Run<NumDims>({8, 2, 3, 8}, {48, 24, 8, 1}, {1, 8, 16, 48});
+    this->template Run<NumDims>({8, 2, 3, 8}, {1, 8, 16, 48}, {48, 24, 8, 1});
     this->template Run<NumDims>({1, 1, 1, 1}, {1, 1, 1, 1}, {1, 1, 1, 1});
 }
 
 TYPED_TEST(TestPermute, Test5D)
 {
     constexpr ck::index_t NumDims = 5;
-    this->template Run<NumDims>({2, 4, 4, 4, 4}, {256, 64, 16, 4, 1}, {1, 2, 8, 32, 128});
-    this->template Run<NumDims>({2, 4, 4, 4, 4}, {1, 2, 8, 32, 128}, {256, 64, 16, 4, 1});
+    this->template Run<NumDims>({8, 2, 3, 4, 8}, {192, 96, 32, 8, 1}, {1, 8, 16, 48, 192});
+    this->template Run<NumDims>({8, 2, 3, 4, 8}, {1, 8, 16, 48, 192}, {192, 96, 32, 8, 1});
     this->template Run<NumDims>({1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}, {1, 1, 1, 1, 1});
 }
 
@@ -93,8 +93,8 @@ TYPED_TEST(TestPermute, Test6D)
 {
     constexpr ck::index_t NumDims = 6;
     this->template Run<NumDims>(
-        {2, 4, 4, 4, 4, 4}, {1024, 256, 64, 16, 4, 1}, {1, 2, 8, 32, 128, 512});
+        {8, 2, 3, 4, 5, 8}, {960, 480, 160, 40, 8, 1}, {1, 8, 16, 48, 192, 960});
     this->template Run<NumDims>(
-        {2, 4, 4, 4, 4, 4}, {1, 2, 8, 32, 128, 512}, {1024, 256, 64, 16, 4, 1});
+        {8, 2, 3, 4, 5, 8}, {1, 8, 16, 48, 192, 960}, {960, 480, 160, 40, 8, 1});
     this->template Run<NumDims>({1, 1, 1, 1, 1, 1}, {1, 1, 1, 1, 1, 1}, {1, 1, 1, 1, 1, 1});
 }

test/transpose/CMakeLists.txt

-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_gtest_executable(test_transpose test_transpose.cpp)
-   target_link_libraries(test_transpose PRIVATE utility device_transpose_instance)
-   set(target 1)
- endif()
-endforeach()
+add_gtest_executable(test_transpose test_transpose_xdl.cpp)
+if(result EQUAL 0)
+    target_link_libraries(test_transpose PRIVATE utility device_transpose_instance)
+endif()

test/wrapper/CMakeLists.txt

@@ -12,10 +12,8 @@ add_dependencies(test_wrapper test_wrapper_copy)
 add_gtest_executable(test_wrapper_partition test_wrapper_partition.cpp)
 target_link_libraries(test_wrapper_partition PRIVATE utility)
 add_dependencies(test_wrapper test_wrapper_partition)
-if(GPU_TARGETS MATCHES "gfx908" OR GPU_TARGETS MATCHES "gfx90a" OR
-   GPU_TARGETS MATCHES "gfx940" OR GPU_TARGETS MATCHES "gfx941" OR
-   GPU_TARGETS MATCHES "gfx942")
-    add_gtest_executable(test_wrapper_gemm test_wrapper_gemm.cpp)
+add_gtest_executable(test_wrapper_gemm test_wrapper_gemm_xdl.cpp)
+if(result EQUAL 0)
     target_link_libraries(test_wrapper_gemm PRIVATE utility)
     add_dependencies(test_wrapper test_wrapper_gemm)
 endif()