fix bug on merge latest develop

CMakeLists.txt

@@ -84,6 +84,7 @@ if( DEFINED CK_OVERRIDE_HIP_VERSION_PATCH )
     message(STATUS "CK_HIP_VERSION_PATCH overriden with ${CK_OVERRIDE_HIP_VERSION_PATCH}")
 endif()
 message(STATUS "Build with HIP ${HIP_VERSION}")
+endif()
 
 ## tidy
 include(EnableCompilerWarnings)
@@ -251,10 +252,12 @@ message("CMAKE_CXX_FLAGS: ${CMAKE_CXX_FLAGS}")
 
 add_custom_target(check COMMAND ${CMAKE_CTEST_COMMAND} --output-on-failure -C ${CMAKE_CFG_INTDIR})
 
+if(NOT CK_NOGPU)
 rocm_package_setup_component(tests
         LIBRARY_NAME composablekernel
         PACKAGE_NAME tests # Prevent -static suffix on package name
 )
+endif()
 
 add_subdirectory(library)
 add_subdirectory(example)
@@ -277,6 +280,7 @@ configure_package_config_file(${CMAKE_CURRENT_SOURCE_DIR}/Config.cmake.in
         NO_CHECK_REQUIRED_COMPONENTS_MACRO
 )
 
+if(NOT CK_NOGPU)
 rocm_install(FILES
     "${CMAKE_CURRENT_BINARY_DIR}/composable_kernelConfig.cmake"
     "${CMAKE_CURRENT_BINARY_DIR}/composable_kernelConfigVersion.cmake"
@@ -293,3 +297,4 @@ rocm_create_package(
     LDCONFIG
     HEADER_ONLY
 )
+endif()

example/cpu_01_conv2d_fwd/cpu_conv2d_fwd.cpp

 #include <sstream>
 #include <iomanip>
-#include "config.hpp"
-#include "device.hpp"
-#include "host_tensor.hpp"
-#include "host_tensor_generator.hpp"
-#include "tensor_layout.hpp"
-#include "device_tensor.hpp"
-#include "device_convnd_fwd_avx2_nhwc_kyxc_nhwk.hpp"
-#include "element_wise_operation_cpu.hpp"
-#include "reference_conv_fwd.hpp"
-#include "element_wise_operation_cpu.hpp"
-#include "dynamic_buffer_cpu.hpp"
-#include "envvar.hpp"
-#include "xdnn_desc.hpp"
+#include "ck/ck.hpp"
+#include "ck/device_utility/kernel_launch.hpp"
+#include "ck/library/host_tensor/device_memory.hpp"
+#include "ck/library/host_tensor/host_tensor.hpp"
+#include "ck/library/host_tensor/host_tensor_generator.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/cpu/device/device_convnd_fwd_avx2_nhwc_kyxc_nhwk.hpp"
+#include "ck/tensor_operation/cpu/element/element_wise_operation_cpu.hpp"
+#include "ck/library/reference_tensor_operation/cpu/reference_conv_fwd.hpp"
+#include "ck/utility/dynamic_buffer_cpu.hpp"
+#include "ck/utility/envvar.hpp"
+#include "ck/device_utility/xdnn_desc.hpp"
 #include <omp.h>
 
 #define AVX2_DATA_ALIGNMENT 32

example/cpu_02_conv2d_fwd_bias_relu_add/cpu_conv2d_fwd_bias_relu_add.cpp

 #include <sstream>
-#include "config.hpp"
-#include "device.hpp"
-#include "host_tensor.hpp"
-#include "host_tensor_generator.hpp"
-#include "tensor_layout.hpp"
-#include "device_tensor.hpp"
-#include "device_convnd_fwd_bias_activation_add_avx2_nhwc_kyxc_nhwk.hpp"
-#include "element_wise_operation_cpu.hpp"
-#include "reference_conv_fwd_bias_activation_add.hpp"
-#include "reference_conv_fwd_bias_activation.hpp"
-#include "element_wise_operation_cpu.hpp"
-#include "dynamic_buffer_cpu.hpp"
-#include "envvar.hpp"
-#include "xdnn_desc.hpp"
+#include "ck/ck.hpp"
+#include "ck/device_utility/kernel_launch.hpp"
+#include "ck/library/host_tensor/device_memory.hpp"
+#include "ck/library/host_tensor/host_tensor.hpp"
+#include "ck/library/host_tensor/host_tensor_generator.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_operation/cpu/device/device_convnd_fwd_bias_activation_add_avx2_nhwc_kyxc_nhwk.hpp"
+#include "ck/tensor_operation/cpu/element/element_wise_operation_cpu.hpp"
+#include "ck/library/reference_tensor_operation/cpu/reference_conv_fwd_bias_activation_add.hpp"
+#include "ck/library/reference_tensor_operation/cpu/reference_conv_fwd_bias_activation.hpp"
+#include "ck/utility/dynamic_buffer_cpu.hpp"
+#include "ck/utility/envvar.hpp"
+#include "ck/device_utility/xdnn_desc.hpp"
 #include <omp.h>
 
 #define AVX2_DATA_ALIGNMENT 32

include/ck/ck.hpp

@@ -3,11 +3,10 @@
 
 #pragma once
 
-#include "ck/options.hpp"
-
 #ifdef CK_NOGPU
 #define __host__
 #define __device__
+#include <stdint.h>
 #else
 #include "hip/hip_runtime.h"
 #include "hip/hip_fp16.h"

include/ck/device_utility/device_prop.hpp

@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
 
 #pragma once
+#ifndef CK_NOGPU
 
 #include <string>
 #include <map>
@@ -52,3 +53,4 @@ inline std::string get_device_name()
 }
 
 } // namespace ck
+#endif

include/ck/device_utility/hip_check_error.hpp

@@ -2,6 +2,7 @@
 // Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
 
 #pragma once
+#ifndef CK_NOGPU
 
 #include <hip/hip_runtime.h>
 
@@ -15,3 +16,4 @@ inline void hip_check_error(hipError_t x)
         throw std::runtime_error(ss.str());
     }
 }
+#endif

include/ck/device_utility/kernel_launch.hpp

@@ -3,6 +3,9 @@
 
 #pragma once
 
+#include <chrono>
+
+#ifndef CK_NOGPU
 #include <hip/hip_runtime.h>
 
 #include "ck/ck.hpp"
@@ -72,3 +75,32 @@ float launch_and_time_kernel(const StreamConfig& stream_config,
     return 0;
 #endif
 }
+#endif
+
+template <typename... Args, typename F>
+void launch_cpu_kernel(F kernel, Args... args)
+{
+    kernel(args...);
+}
+
+template <typename... Args, typename F>
+float launch_and_time_cpu_kernel(F kernel, int nrepeat, Args... args)
+{
+    int nwarmup = 3;
+
+    for(int i = 0; i < nwarmup; i++)
+        kernel(args...);
+
+    auto mStart = std::chrono::high_resolution_clock::now();
+    for(int i = 0; i < nrepeat; i++)
+    {
+        kernel(args...);
+    }
+    auto mStop = std::chrono::high_resolution_clock::now();
+
+    float ms = static_cast<float>(
+                   std::chrono::duration_cast<std::chrono::microseconds>(mStop - mStart).count()) *
+               1e-3;
+
+    return ms / nrepeat;
+}

include/ck/device_utility/xdnn_desc.hpp

@@ -3,6 +3,7 @@
 #include <string>
 #include <vector>
 #include <functional>
+#include <string.h>
 
 #define XDNN_OK 0
 #define XDNN_FAIL 1

include/ck/tensor_operation/cpu/block/blockwise_gemm_avx2.hpp

 #ifndef CK_BLOCKWISE_GEMM_AVX2_HPP
 #define CK_BLOCKWISE_GEMM_AVX2_HPP
 
-#include "common_header.hpp"
-#include "multi_index_transform_helper.hpp"
-#include "tensor_descriptor.hpp"
-#include "tensor_descriptor_helper.hpp"
-#include "threadwise_gemm_avx2.hpp"
+#include "ck/utility/common_header.hpp"
+#include "ck/tensor_description/multi_index_transform_helper.hpp"
+#include "ck/tensor_description/tensor_descriptor.hpp"
+#include "ck/tensor_description/tensor_descriptor_helper.hpp"
+#include "ck/tensor_operation/cpu/thread/threadwise_gemm_avx2.hpp"
 
 namespace ck {
 namespace cpu {

include/ck/tensor_operation/cpu/device/device_base_cpu.hpp

@@ -2,7 +2,7 @@
 #define DEVICE_BASE_CPU_HPP
 
 #include <string>
-#include "stream_config.hpp"
+#include "ck/stream_config.hpp"
 
 namespace ck {
 namespace tensor_operation {

include/ck/tensor_operation/cpu/device/device_conv_fwd_cpu.hpp

@@ -2,8 +2,10 @@
 #define DEVICE_CONV_FWD_CPU_HPP
 
 #include <iostream>
-#include "device_base_cpu.hpp"
-#include "convolution_forward_specialization_cpu.hpp"
+#include <memory>
+#include <vector>
+#include "ck/tensor_operation/cpu/device/device_base_cpu.hpp"
+#include "ck/tensor_operation/cpu/device/convolution_forward_specialization_cpu.hpp"
 
 namespace ck {
 namespace tensor_operation {

include/ck/tensor_operation/cpu/device/device_convnd_direct_fwd_avx2_nhwc_kyxck8_nhwk.hpp

@@ -4,17 +4,20 @@
 #include <iostream>
 #include <sstream>
 #include <numeric>
-#include "device.hpp"
-#include "device_base_cpu.hpp"
-#include "device_conv_fwd_cpu.hpp"
-#include "convolution_forward_specialization_cpu.hpp"
-#include "common_header.hpp"
-#include "../../gpu/device/tensor_layout.hpp"
-#include "tensor_descriptor.hpp"
-#include "tensor_descriptor_helper.hpp"
-#include "gridwise_direct_conv_avx2.hpp"
-#include "threadwise_gemm_avx2.hpp"
-#include "threadwise_tensor_slice_transfer_avx2_specialization.hpp"
+#include <memory>
+#include <vector>
+#include "ck/device_utility/kernel_launch.hpp"
+#include "ck/library/host_tensor/device_memory.hpp"
+#include "ck/tensor_operation/cpu/device/device_base_cpu.hpp"
+#include "ck/tensor_operation/cpu/device/device_conv_fwd_cpu.hpp"
+#include "ck/tensor_operation/cpu/device/convolution_forward_specialization_cpu.hpp"
+#include "ck/utility/common_header.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_description/tensor_descriptor.hpp"
+#include "ck/tensor_description/tensor_descriptor_helper.hpp"
+#include "ck/tensor_operation/cpu/grid/gridwise_direct_conv_avx2.hpp"
+#include "ck/tensor_operation/cpu/thread/threadwise_gemm_avx2.hpp"
+#include "ck/tensor_operation/cpu/thread/threadwise_tensor_slice_transfer_avx2_specialization.hpp"
 
 namespace ck {
 namespace tensor_operation {

include/ck/tensor_operation/cpu/device/device_convnd_fwd_avx2_nhwc_kyxc_nhwk.hpp

@@ -4,17 +4,20 @@
 #include <iostream>
 #include <sstream>
 #include <numeric>
-#include "device.hpp"
-#include "device_base_cpu.hpp"
-#include "device_conv_fwd_cpu.hpp"
-#include "convolution_forward_specialization_cpu.hpp"
-#include "common_header.hpp"
-#include "../../gpu/device/tensor_layout.hpp"
-#include "tensor_descriptor.hpp"
-#include "tensor_descriptor_helper.hpp"
-#include "gridwise_gemm_avx2.hpp"
-#include "threadwise_gemm_avx2.hpp"
-#include "threadwise_tensor_slice_transfer_avx2_specialization.hpp"
+#include <memory>
+#include <vector>
+#include "ck/device_utility/kernel_launch.hpp"
+#include "ck/library/host_tensor/device_memory.hpp"
+#include "ck/tensor_operation/cpu/device/device_base_cpu.hpp"
+#include "ck/tensor_operation/cpu/device/device_conv_fwd_cpu.hpp"
+#include "ck/tensor_operation/cpu/device/convolution_forward_specialization_cpu.hpp"
+#include "ck/utility/common_header.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_description/tensor_descriptor.hpp"
+#include "ck/tensor_description/tensor_descriptor_helper.hpp"
+#include "ck/tensor_operation/cpu/grid/gridwise_gemm_avx2.hpp"
+#include "ck/tensor_operation/cpu/thread/threadwise_gemm_avx2.hpp"
+#include "ck/tensor_operation/cpu/thread/threadwise_tensor_slice_transfer_avx2_specialization.hpp"
 
 namespace ck {
 namespace tensor_operation {

include/ck/tensor_operation/cpu/device/device_convnd_fwd_avx2_nhwc_kyxck8_nhwk.hpp

@@ -4,17 +4,20 @@
 #include <iostream>
 #include <sstream>
 #include <numeric>
-#include "device.hpp"
-#include "device_base_cpu.hpp"
-#include "device_conv_fwd_cpu.hpp"
-#include "convolution_forward_specialization_cpu.hpp"
-#include "common_header.hpp"
-#include "../../gpu/device/tensor_layout.hpp"
-#include "tensor_descriptor.hpp"
-#include "tensor_descriptor_helper.hpp"
-#include "gridwise_gemm_avx2.hpp"
-#include "threadwise_gemm_avx2.hpp"
-#include "threadwise_tensor_slice_transfer_avx2_specialization.hpp"
+#include <memory>
+#include <vector>
+#include "ck/device_utility/kernel_launch.hpp"
+#include "ck/library/host_tensor/device_memory.hpp"
+#include "ck/tensor_operation/cpu/device/device_base_cpu.hpp"
+#include "ck/tensor_operation/cpu/device/device_conv_fwd_cpu.hpp"
+#include "ck/tensor_operation/cpu/device/convolution_forward_specialization_cpu.hpp"
+#include "ck/utility/common_header.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_description/tensor_descriptor.hpp"
+#include "ck/tensor_description/tensor_descriptor_helper.hpp"
+#include "ck/tensor_operation/cpu/grid/gridwise_gemm_avx2.hpp"
+#include "ck/tensor_operation/cpu/thread/threadwise_gemm_avx2.hpp"
+#include "ck/tensor_operation/cpu/thread/threadwise_tensor_slice_transfer_avx2_specialization.hpp"
 
 namespace ck {
 namespace tensor_operation {

include/ck/tensor_operation/cpu/device/device_convnd_fwd_avx2_nhwc_yxck_nhwk.hpp

@@ -4,17 +4,20 @@
 #include <iostream>
 #include <sstream>
 #include <numeric>
-#include "device.hpp"
-#include "device_base_cpu.hpp"
-#include "device_conv_fwd_cpu.hpp"
-#include "convolution_forward_specialization_cpu.hpp"
-#include "common_header.hpp"
-#include "../../gpu/device/tensor_layout.hpp"
-#include "tensor_descriptor.hpp"
-#include "tensor_descriptor_helper.hpp"
-#include "gridwise_gemm_avx2.hpp"
-#include "threadwise_gemm_avx2.hpp"
-#include "threadwise_tensor_slice_transfer_avx2_specialization.hpp"
+#include <memory>
+#include <vector>
+#include "ck/device_utility/kernel_launch.hpp"
+#include "ck/library/host_tensor/device_memory.hpp"
+#include "ck/tensor_operation/cpu/device/device_base_cpu.hpp"
+#include "ck/tensor_operation/cpu/device/device_conv_fwd_cpu.hpp"
+#include "ck/tensor_operation/cpu/device/convolution_forward_specialization_cpu.hpp"
+#include "ck/utility/common_header.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_description/tensor_descriptor.hpp"
+#include "ck/tensor_description/tensor_descriptor_helper.hpp"
+#include "ck/tensor_operation/cpu/grid/gridwise_gemm_avx2.hpp"
+#include "ck/tensor_operation/cpu/thread/threadwise_gemm_avx2.hpp"
+#include "ck/tensor_operation/cpu/thread/threadwise_tensor_slice_transfer_avx2_specialization.hpp"
 
 namespace ck {
 namespace tensor_operation {

include/ck/tensor_operation/cpu/device/device_convnd_fwd_bias_activation_add_avx2_nhwc_kyxc_nhwk.hpp

@@ -4,17 +4,20 @@
 #include <iostream>
 #include <sstream>
 #include <numeric>
-#include "device.hpp"
-#include "device_base_cpu.hpp"
-#include "device_conv_fwd_cpu.hpp"
-#include "convolution_forward_specialization_cpu.hpp"
-#include "common_header.hpp"
-#include "../../gpu/device/tensor_layout.hpp"
-#include "tensor_descriptor.hpp"
-#include "tensor_descriptor_helper.hpp"
-#include "gridwise_gemm_bias_activation_add_avx2.hpp"
-#include "threadwise_gemm_avx2.hpp"
-#include "threadwise_tensor_slice_transfer_avx2_specialization.hpp"
+#include <memory>
+#include <vector>
+#include "ck/device_utility/kernel_launch.hpp"
+#include "ck/library/host_tensor/device_memory.hpp"
+#include "ck/tensor_operation/cpu/device/device_base_cpu.hpp"
+#include "ck/tensor_operation/cpu/device/device_conv_fwd_cpu.hpp"
+#include "ck/tensor_operation/cpu/device/convolution_forward_specialization_cpu.hpp"
+#include "ck/utility/common_header.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_description/tensor_descriptor.hpp"
+#include "ck/tensor_description/tensor_descriptor_helper.hpp"
+#include "ck/tensor_operation/cpu/grid/gridwise_gemm_bias_activation_add_avx2.hpp"
+#include "ck/tensor_operation/cpu/thread/threadwise_gemm_avx2.hpp"
+#include "ck/tensor_operation/cpu/thread/threadwise_tensor_slice_transfer_avx2_specialization.hpp"
 
 namespace ck {
 namespace tensor_operation {

include/ck/tensor_operation/cpu/device/device_convnd_fwd_bias_activation_add_avx2_nhwc_kyxck8_nhwk.hpp

@@ -4,17 +4,20 @@
 #include <iostream>
 #include <sstream>
 #include <numeric>
-#include "device.hpp"
-#include "device_base_cpu.hpp"
-#include "device_conv_fwd_cpu.hpp"
-#include "convolution_forward_specialization_cpu.hpp"
-#include "common_header.hpp"
-#include "../../gpu/device/tensor_layout.hpp"
-#include "tensor_descriptor.hpp"
-#include "tensor_descriptor_helper.hpp"
-#include "gridwise_gemm_bias_activation_add_avx2.hpp"
-#include "threadwise_gemm_avx2.hpp"
-#include "threadwise_tensor_slice_transfer_avx2_specialization.hpp"
+#include <memory>
+#include <vector>
+#include "ck/device_utility/kernel_launch.hpp"
+#include "ck/library/host_tensor/device_memory.hpp"
+#include "ck/tensor_operation/cpu/device/device_base_cpu.hpp"
+#include "ck/tensor_operation/cpu/device/device_conv_fwd_cpu.hpp"
+#include "ck/tensor_operation/cpu/device/convolution_forward_specialization_cpu.hpp"
+#include "ck/utility/common_header.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_description/tensor_descriptor.hpp"
+#include "ck/tensor_description/tensor_descriptor_helper.hpp"
+#include "ck/tensor_operation/cpu/grid/gridwise_gemm_bias_activation_add_avx2.hpp"
+#include "ck/tensor_operation/cpu/thread/threadwise_gemm_avx2.hpp"
+#include "ck/tensor_operation/cpu/thread/threadwise_tensor_slice_transfer_avx2_specialization.hpp"
 
 namespace ck {
 namespace tensor_operation {

include/ck/tensor_operation/cpu/device/device_convnd_fwd_bias_activation_add_avx2_nhwc_yxck_nhwk.hpp

@@ -4,17 +4,20 @@
 #include <iostream>
 #include <sstream>
 #include <numeric>
-#include "device.hpp"
-#include "device_base_cpu.hpp"
-#include "device_conv_fwd_cpu.hpp"
-#include "convolution_forward_specialization_cpu.hpp"
-#include "common_header.hpp"
-#include "../../gpu/device/tensor_layout.hpp"
-#include "tensor_descriptor.hpp"
-#include "tensor_descriptor_helper.hpp"
-#include "gridwise_gemm_bias_activation_add_avx2.hpp"
-#include "threadwise_gemm_avx2.hpp"
-#include "threadwise_tensor_slice_transfer_avx2_specialization.hpp"
+#include <memory>
+#include <vector>
+#include "ck/device_utility/kernel_launch.hpp"
+#include "ck/library/host_tensor/device_memory.hpp"
+#include "ck/tensor_operation/cpu/device/device_base_cpu.hpp"
+#include "ck/tensor_operation/cpu/device/device_conv_fwd_cpu.hpp"
+#include "ck/tensor_operation/cpu/device/convolution_forward_specialization_cpu.hpp"
+#include "ck/utility/common_header.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+#include "ck/tensor_description/tensor_descriptor.hpp"
+#include "ck/tensor_description/tensor_descriptor_helper.hpp"
+#include "ck/tensor_operation/cpu/grid/gridwise_gemm_bias_activation_add_avx2.hpp"
+#include "ck/tensor_operation/cpu/thread/threadwise_gemm_avx2.hpp"
+#include "ck/tensor_operation/cpu/thread/threadwise_tensor_slice_transfer_avx2_specialization.hpp"
 
 namespace ck {
 namespace tensor_operation {

include/ck/tensor_operation/cpu/element/element_wise_operation_cpu.hpp

 #pragma once
-#include "data_type_cpu.hpp"
+#include "ck/utility/data_type_cpu.hpp"
 
 namespace ck {
 namespace tensor_operation {

include/ck/tensor_operation/cpu/grid/gridwise_direct_conv_avx2.hpp

 #ifndef CK_GRIDWISE_DIRECT_CONV_AVX2_HPP
 #define CK_GRIDWISE_DIRECT_CONV_AVX2_HPP
 
-#include "common_header.hpp"
-#include "multi_index_transform_helper.hpp"
-#include "tensor_descriptor.hpp"
-#include "tensor_descriptor_helper.hpp"
-#include "blockwise_gemm_avx2.hpp"
-#include "threadwise_tensor_slice_transfer_avx2.hpp"
-#include "threadwise_tensor_slice_transfer_avx2_specialization.hpp"
-#include "dynamic_buffer_cpu.hpp"
-#include "envvar.hpp"
+#include "ck/utility/common_header.hpp"
+#include "ck/tensor_description/multi_index_transform_helper.hpp"
+#include "ck/tensor_description/tensor_descriptor.hpp"
+#include "ck/tensor_description/tensor_descriptor_helper.hpp"
+#include "ck/tensor_operation/cpu/block/blockwise_gemm_avx2.hpp"
+#include "ck/tensor_operation/cpu/thread/threadwise_tensor_slice_transfer_avx2.hpp"
+#include "ck/tensor_operation/cpu/thread/threadwise_tensor_slice_transfer_avx2_specialization.hpp"
+#include "ck/utility/dynamic_buffer_cpu.hpp"
+#include "ck/utility/envvar.hpp"
 #include <utility>
 #include <unistd.h>
 #include <omp.h>