Reorganize files, Part 1 (#119)

* delete obselete files

* move files

* build

* update cmake

* update cmake

* fix build

* reorg examples

* update cmake for example and test

profiler/CMakeLists.txt

 include_directories(BEFORE
-    include
-    ${PROJECT_SOURCE_DIR}/host/host_tensor/include
-    ${PROJECT_SOURCE_DIR}/device/include
-    ${PROJECT_SOURCE_DIR}/device_operation/include
-    ${PROJECT_SOURCE_DIR}/reference_operation/include
+    ${PROJECT_SOURCE_DIR}/include/ck
+    ${PROJECT_SOURCE_DIR}/include/ck/utility
+    ${PROJECT_SOURCE_DIR}/include/ck/tensor_description
+    ${PROJECT_SOURCE_DIR}/include/ck/tensor
+    ${PROJECT_SOURCE_DIR}/include/ck/problem_transform
+    ${PROJECT_SOURCE_DIR}/include/ck/tensor_operation/gpu/device
+    ${PROJECT_SOURCE_DIR}/include/ck/tensor_operation/gpu/grid
+    ${PROJECT_SOURCE_DIR}/include/ck/tensor_operation/gpu/block
+    ${PROJECT_SOURCE_DIR}/include/ck/tensor_operation/gpu/warp
+    ${PROJECT_SOURCE_DIR}/include/ck/tensor_operation/gpu/thread
+    ${PROJECT_SOURCE_DIR}/include/ck/tensor_operation/gpu/element
+    ${PROJECT_SOURCE_DIR}/library/include/ck/library/host_tensor
+    ${PROJECT_SOURCE_DIR}/library/include/ck/library/tensor_operation_instance
+    ${PROJECT_SOURCE_DIR}/library/include/ck/library/tensor_operation_instance/gpu/reduce
+    ${PROJECT_SOURCE_DIR}/library/include/ck/library/reference_tensor_operation/cpu
+    ${PROJECT_SOURCE_DIR}/library/include/ck/library/reference_tensor_operation/gpu
     ${PROJECT_SOURCE_DIR}/profiler/include
-    ${PROJECT_SOURCE_DIR}/composable_kernel/include
-    ${PROJECT_SOURCE_DIR}/composable_kernel/include/utility
-    ${PROJECT_SOURCE_DIR}/composable_kernel/include/tensor_description
-    ${PROJECT_SOURCE_DIR}/composable_kernel/include/tensor_operation
-    ${PROJECT_SOURCE_DIR}/composable_kernel/include/problem_transform
-    ${PROJECT_SOURCE_DIR}/external/rocm/include
+    ${PROJECT_SOURCE_DIR}/external/include/half
 )
 
 # ck_profiler
@@ -33,7 +39,7 @@ add_executable(ckProfiler ${PROFILER_SOURCE})
 
 target_link_libraries(ckProfiler PRIVATE host_tensor)
 target_link_libraries(ckProfiler PRIVATE device_gemm_instance)
-target_link_libraries(ckProfiler PRIVATE device_gemm_bias_2d_instance)
+target_link_libraries(ckProfiler PRIVATE device_gemm_bias2d_instance)
 target_link_libraries(ckProfiler PRIVATE device_gemm_bias_relu_instance)
 target_link_libraries(ckProfiler PRIVATE device_gemm_bias_relu_add_instance)
 target_link_libraries(ckProfiler PRIVATE device_batched_gemm_instance)

profiler/src/profile_gemm_bias_2d.cpp

@@ -63,11 +63,6 @@ int profile_gemm_bias_2d(int argc, char* argv[])
     const float alpha = std::stof(argv[14]);
     const float beta  = std::stof(argv[15]);
 
-    int KBatch = 1;
-
-    if(argc == 17)
-        KBatch = std::stoi(argv[16]);
-
     if(data_type == GemmDataType::F32_F32_F32 && layout == GemmMatrixLayout::MK_KN_MN)
     {
         ck::profiler::profile_gemm_bias_2d_impl<float,

profiler/src/profile_gemm_bias_relu.cpp

@@ -58,11 +58,6 @@ int profile_gemm_bias_relu(int argc, char* argv[])
     const int StrideB = std::stoi(argv[12]);
     const int StrideC = std::stoi(argv[13]);
 
-    int KBatch = 1;
-
-    if(argc == 15)
-        KBatch = std::stoi(argv[14]);
-
     if(data_type == GemmDataType::F16_F16_F16 && layout == GemmMatrixLayout::MK_KN_MN)
     {
         ck::profiler::profile_gemm_bias_relu_impl<ck::half_t,

test/CMakeLists.txt

 include_directories(BEFORE
-    include
-    ${PROJECT_SOURCE_DIR}/host/host_tensor/include
-    ${PROJECT_SOURCE_DIR}/host/device/include
-    ${PROJECT_SOURCE_DIR}/device_operation/include
-    ${PROJECT_SOURCE_DIR}/composable_kernel/include
-    ${PROJECT_SOURCE_DIR}/composable_kernel/include/utility
-    ${PROJECT_SOURCE_DIR}/composable_kernel/include/tensor_description
-    ${PROJECT_SOURCE_DIR}/composable_kernel/include/tensor_operation
-    ${PROJECT_SOURCE_DIR}/composable_kernel/include/problem_transform
-    ${PROJECT_SOURCE_DIR}/external/rocm/include
-    ${PROJECT_SOURCE_DIR}/reference_operation/include
+    ${PROJECT_SOURCE_DIR}/include/ck
+    ${PROJECT_SOURCE_DIR}/include/ck/utility
+    ${PROJECT_SOURCE_DIR}/include/ck/tensor_description
+    ${PROJECT_SOURCE_DIR}/include/ck/tensor
+    ${PROJECT_SOURCE_DIR}/include/ck/problem_transform
+    ${PROJECT_SOURCE_DIR}/include/ck/tensor_operation/gpu/device
+    ${PROJECT_SOURCE_DIR}/include/ck/tensor_operation/gpu/grid
+    ${PROJECT_SOURCE_DIR}/include/ck/tensor_operation/gpu/block
+    ${PROJECT_SOURCE_DIR}/include/ck/tensor_operation/gpu/warp
+    ${PROJECT_SOURCE_DIR}/include/ck/tensor_operation/gpu/thread
+    ${PROJECT_SOURCE_DIR}/include/ck/tensor_operation/gpu/element
+    ${PROJECT_SOURCE_DIR}/library/include/ck/library/host_tensor
+    ${PROJECT_SOURCE_DIR}/library/include/ck/library/tensor_operation_instance
+    ${PROJECT_SOURCE_DIR}/library/include/ck/library/tensor_operation_instance/gpu/reduce
+    ${PROJECT_SOURCE_DIR}/library/include/ck/library/reference_tensor_operation/cpu
+    ${PROJECT_SOURCE_DIR}/library/include/ck/library/reference_tensor_operation/gpu
     ${PROJECT_SOURCE_DIR}/test/include
+    ${PROJECT_SOURCE_DIR}/external/include/half
 )
 
 add_custom_target(check COMMAND ${CMAKE_CTEST_COMMAND} --output-on-failure -C ${CMAKE_CFG_INTDIR})
 add_custom_target(tests)
 
-function(add_test_executeable TEST_NAME)
+function(add_test_executable TEST_NAME)
+    message("adding test ${TEST_NAME}")
     add_executable(${TEST_NAME} ${ARGN})
-    target_link_libraries(${TEST_NAME} PRIVATE host_tensor)
-    target_link_libraries(${TEST_NAME} PRIVATE device_gemm_instance)
-    target_link_libraries(${TEST_NAME} PRIVATE device_conv2d_fwd_instance)
-    target_link_libraries(${TEST_NAME} PRIVATE device_conv2d_bwd_data_instance)
     add_test(NAME ${TEST_NAME} COMMAND $<TARGET_FILE:${TEST_NAME}> )
     add_dependencies(tests ${TEST_NAME})
     add_dependencies(check ${TEST_NAME})
-endfunction(add_test_executeable TEST_NAME)
+endfunction(add_test_executable TEST_NAME)
 
-file(GLOB TESTS */*.cpp)
-
-foreach(TEST ${TESTS})
-    get_filename_component(BASE_NAME ${TEST} NAME_WE)
-    message("adding test ${BASE_NAME}")
-    add_test_executeable(test_${BASE_NAME} ${TEST})
-endforeach(TEST ${TESTS})
+add_subdirectory(magic_number_division)
+add_subdirectory(space_filling_curve)
+add_subdirectory(conv_util)
+add_subdirectory(reference_conv_fwd)
+add_subdirectory(gemm)
+add_subdirectory(gemm_split_k)
+add_subdirectory(conv2d_fwd)
+add_subdirectory(convnd_fwd)
+add_subdirectory(conv2d_bwd_data)

test/conv2d_bwd_data/CMakeLists.txt

+add_test_executable(test_conv2d_bwd_data conv2d_bwd_data.cpp)
+target_link_libraries(test_conv2d_bwd_data PRIVATE host_tensor)
+target_link_libraries(test_conv2d_bwd_data PRIVATE device_conv2d_bwd_data_instance)

test/conv2d_fwd/CMakeLists.txt

+add_test_executable(test_conv2d_fwd conv2d_fwd.cpp)
+target_link_libraries(test_conv2d_fwd PRIVATE host_tensor)
+target_link_libraries(test_conv2d_fwd PRIVATE device_conv2d_fwd_instance)

test/conv_util/CMakeLists.txt

+add_test_executable(test_conv_util conv_util.cpp)
+target_link_libraries(test_conv_util PRIVATE host_tensor)

test/convnd_fwd/CMakeLists.txt

+add_test_executable(test_convnd_fwd convnd_fwd.cpp)
+target_link_libraries(test_convnd_fwd PRIVATE host_tensor)

test/gemm/CMakeLists.txt

+add_test_executable(test_gemm_fp32 gemm_fp32.cpp)
+target_link_libraries(test_gemm_fp32 PRIVATE host_tensor)
+target_link_libraries(test_gemm_fp32 PRIVATE device_gemm_instance)
+
+add_test_executable(test_gemm_bf16 gemm_bf16.cpp)
+target_link_libraries(test_gemm_bf16 PRIVATE host_tensor)
+target_link_libraries(test_gemm_bf16 PRIVATE device_gemm_instance)
+
+add_test_executable(test_gemm_int8 gemm_int8.cpp)
+target_link_libraries(test_gemm_int8 PRIVATE host_tensor)
+target_link_libraries(test_gemm_int8 PRIVATE device_gemm_instance)

test/gemm_split_k/CMakeLists.txt

+add_test_executable(test_gemm_split_k gemm_split_k.cpp)
+target_link_libraries(test_gemm_split_k PRIVATE host_tensor)
+target_link_libraries(test_gemm_split_k PRIVATE device_gemm_instance)

test/magic_number_division/CMakeLists.txt

+add_test_executable(test_magic_number_division magic_number_division.cpp)
+target_link_libraries(test_magic_number_division PRIVATE host_tensor)