refactor(dnn): move arch-dependant code to arch.h

GitOrigin-RevId: 52350144b1

dnn/include/megdnn/arch.h

@@ -140,6 +140,16 @@
  #define MEGDNN_DEVICE
 #endif
 
+#if MEGDNN_CC_CUDA
+ #define MEGDNN_FORCE_INLINE __forceinline__
+#else
+#if __GNUC__ || __has_attribute(always_inline)
+ #define MEGDNN_FORCE_INLINE inline __attribute__((always_inline))
+#else
+ #define MEGDNN_FORCE_INLINE inline
+#endif
+#endif
+
 #if defined(_MSC_VER) || defined(WIN32)
  #define ATTR_ALIGNED(v) __declspec(align(v))
 #else

dnn/src/common/resize.cuh

@@ -13,18 +13,10 @@
 
 #include "megdnn/arch.h"
 
-#if MEGDNN_CC_HOST && !defined(__host__)
-#if __GNUC__ || __has_attribute(always_inline)
-#define __forceinline__ inline __attribute__((always_inline))
-#else
-#define __forceinline__ inline
-#endif
-#endif
-
 namespace megdnn {
 namespace resize {
 
-MEGDNN_HOST MEGDNN_DEVICE __forceinline__ void interpolate_cubic(
+MEGDNN_HOST MEGDNN_DEVICE MEGDNN_FORCE_INLINE void interpolate_cubic(
         float x, float* coeffs) {
     const float A = -0.75f;
 

dnn/src/common/rounding_converter.cuh

@@ -12,17 +12,6 @@
 #pragma once
 #include "megdnn/dtype.h"
 
-#if MEGDNN_CC_HOST && !defined(__host__)
-#define MEGDNN_HOST_DEVICE_SELF_DEFINE
-#define __host__
-#define __device__
-#if __GNUC__ || __has_attribute(always_inline)
-#define __forceinline__ inline __attribute__((always_inline))
-#else
-#define __forceinline__ inline
-#endif
-#endif
-
 namespace megdnn {
 namespace rounding {
 
@@ -31,7 +20,8 @@ struct RoundingConverter;
 
 template <>
 struct RoundingConverter<float> {
-    __host__ __device__ __forceinline__ float operator()(float x) const {
+    MEGDNN_HOST MEGDNN_DEVICE MEGDNN_FORCE_INLINE float operator()(
+            float x) const {
         return x;
     }
 };
@@ -40,7 +30,7 @@ struct RoundingConverter<float> {
 
 template <>
 struct RoundingConverter<half_float::half> {
-    __host__ __device__ __forceinline__ half_float::half operator()(
+    MEGDNN_HOST MEGDNN_DEVICE MEGDNN_FORCE_INLINE half_float::half operator()(
             float x) const {
         return static_cast<half_float::half>(x);
     }
@@ -48,8 +38,8 @@ struct RoundingConverter<half_float::half> {
 
 template <>
 struct RoundingConverter<half_bfloat16::bfloat16> {
-    __host__ __device__ __forceinline__ half_bfloat16::bfloat16 operator()(
-            float x) const {
+    MEGDNN_HOST MEGDNN_DEVICE MEGDNN_FORCE_INLINE half_bfloat16::bfloat16
+    operator()(float x) const {
         return static_cast<half_bfloat16::bfloat16>(x);
     }
 };
@@ -58,7 +48,8 @@ struct RoundingConverter<half_bfloat16::bfloat16> {
 
 template <>
 struct RoundingConverter<int8_t> {
-    __host__ __device__ __forceinline__ int8_t operator()(float x) const {
+    MEGDNN_HOST MEGDNN_DEVICE MEGDNN_FORCE_INLINE int8_t
+    operator()(float x) const {
 #if MEGDNN_CC_HOST
         using std::round;
 #endif
@@ -68,11 +59,12 @@ struct RoundingConverter<int8_t> {
 
 template <>
 struct RoundingConverter<uint8_t> {
-    __host__ __device__ __forceinline__ uint8_t operator()(float x) const {
+    MEGDNN_HOST MEGDNN_DEVICE MEGDNN_FORCE_INLINE uint8_t
+    operator()(float x) const {
 #if MEGDNN_CC_HOST
-        using std::round;
         using std::max;
         using std::min;
+        using std::round;
 #endif
         x = min(255.0f, max(0.0f, x));  //! FIXME!!! check other places
         return static_cast<uint8_t>(round(x));
@@ -81,7 +73,8 @@ struct RoundingConverter<uint8_t> {
 
 template <>
 struct RoundingConverter<dt_qint4> {
-    __host__ __device__ __forceinline__ dt_qint4 operator()(float x) const {
+    MEGDNN_HOST MEGDNN_DEVICE MEGDNN_FORCE_INLINE dt_qint4
+    operator()(float x) const {
 #if MEGDNN_CC_HOST
         using std::round;
 #endif
@@ -91,7 +84,8 @@ struct RoundingConverter<dt_qint4> {
 
 template <>
 struct RoundingConverter<dt_quint4> {
-    __host__ __device__ __forceinline__ dt_quint4 operator()(float x) const {
+    MEGDNN_HOST MEGDNN_DEVICE MEGDNN_FORCE_INLINE dt_quint4
+    operator()(float x) const {
 #if MEGDNN_CC_HOST
         using std::round;
 #endif