Refactor for MIOpen integration (#4)

Refactor, so can bring multi-index transformation and padding support into MIOpen

composable_kernel/include/utility/common_header.hpp

@@ -5,14 +5,12 @@
 #include "utility.hpp"
 #include "integral_constant.hpp"
 #include "number.hpp"
+#include "float_type.hpp"
 #include "type.hpp"
 #include "tuple.hpp"
 #include "math.hpp"
-#include "vector_type.hpp"
 #include "sequence.hpp"
-#include "sequence_helper.hpp"
 #include "array.hpp"
-#include "array_helper.hpp"
 #include "functional.hpp"
 #include "functional2.hpp"
 #include "functional3.hpp"
@@ -22,8 +20,8 @@
 #include "amd_inline_asm.hpp"
 #endif
 
-#if CK_USE_AMD_INTRINSIC
-#include "amd_intrinsic.hpp"
+#if CK_USE_AMD_BUFFER_ADDRESSING
+#include "amd_buffer_addressing.hpp"
 #endif
 
 #endif

composable_kernel/include/utility/array_helper.hpp → composable_kernel/include/utility/print_array.hpp

-#ifndef CK_ARRAY_HELPER_HPP
-#define CK_ARRAY_HELPER_HPP
+#ifndef CK_PRINT_ARRAY_HPP
+#define CK_PRINT_ARRAY_HPP
 
 #include "array.hpp"
 

composable_kernel/include/utility/sequence_helper.hpp → composable_kernel/include/utility/print_sequence.hpp

-#ifndef CK_SEQUENCE_HELPER_HPP
-#define CK_SEQUENCE_HELPER_HPP
+#ifndef CK_PRINT_SEQUENCE_HPP
+#define CK_PRINT_SEQUENCE_HPP
 
 #include "sequence.hpp"
 

driver/include/conv_common.hpp

 #ifndef CONV_COMMON_HPP
 #define CONV_COMMON_HPP
 
-#include "ConstantTensorDescriptor.hpp"
+#include "ConstantTensorDescriptor_deprecated.hpp"
 
 // this is ugly, only for 4d
 template <class InDesc, class WeiDesc>

driver/include/host_conv.hpp

 #pragma once
 #include "tensor.hpp"
 #include "common_header.hpp"
-#include "ConstantTensorDescriptor.hpp"
+#include "ConstantTensorDescriptor_deprecated.hpp"
 
 // this is ugly, only for 4d
 template <class TConstTensorDesc>