496 lines
16 KiB
C++
496 lines
16 KiB
C++
/*M///////////////////////////////////////////////////////////////////////////////////////
|
|
//
|
|
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
|
|
//
|
|
// By downloading, copying, installing or using the software you agree to this license.
|
|
// If you do not agree to this license, do not download, install,
|
|
// copy or use the software.
|
|
//
|
|
//
|
|
// License Agreement
|
|
// For Open Source Computer Vision Library
|
|
//
|
|
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
|
|
// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
|
|
// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
|
|
// Third party copyrights are property of their respective owners.
|
|
//
|
|
// Redistribution and use in source and binary forms, with or without modification,
|
|
// are permitted provided that the following conditions are met:
|
|
//
|
|
// * Redistribution's of source code must retain the above copyright notice,
|
|
// this list of conditions and the following disclaimer.
|
|
//
|
|
// * Redistribution's in binary form must reproduce the above copyright notice,
|
|
// this list of conditions and the following disclaimer in the documentation
|
|
// and/or other materials provided with the distribution.
|
|
//
|
|
// * The name of the copyright holders may not be used to endorse or promote products
|
|
// derived from this software without specific prior written permission.
|
|
//
|
|
// This software is provided by the copyright holders and contributors "as is" and
|
|
// any express or implied warranties, including, but not limited to, the implied
|
|
// warranties of merchantability and fitness for a particular purpose are disclaimed.
|
|
// In no event shall the Intel Corporation or contributors be liable for any direct,
|
|
// indirect, incidental, special, exemplary, or consequential damages
|
|
// (including, but not limited to, procurement of substitute goods or services;
|
|
// loss of use, data, or profits; or business interruption) however caused
|
|
// and on any theory of liability, whether in contract, strict liability,
|
|
// or tort (including negligence or otherwise) arising in any way out of
|
|
// the use of this software, even if advised of the possibility of such damage.
|
|
//
|
|
//M*/
|
|
|
|
#pragma once
|
|
|
|
#ifndef OPENCV_CUDEV_WARP_SHUFFLE_HPP
|
|
#define OPENCV_CUDEV_WARP_SHUFFLE_HPP
|
|
|
|
#include "../common.hpp"
|
|
#include "../util/vec_traits.hpp"
|
|
#include "../block/block.hpp"
|
|
#include "warp.hpp"
|
|
|
|
namespace cv { namespace cudev {
|
|
|
|
//! @addtogroup cudev
|
|
//! @{
|
|
|
|
#if CV_CUDEV_ARCH >= 300
|
|
|
|
#if __CUDACC_VER_MAJOR__ >= 9
|
|
# define __shfl(x, y, z) __shfl_sync(0xFFFFFFFFU, x, y, z)
|
|
# define __shfl_xor(x, y, z) __shfl_xor_sync(0xFFFFFFFFU, x, y, z)
|
|
//# define __shfl_up(x, y, z) __shfl_up_sync(0xFFFFFFFFU, x, y, z)
|
|
# define __shfl_down(x, y, z) __shfl_down_sync(0xFFFFFFFFU, x, y, z)
|
|
#endif
|
|
|
|
// shfl
|
|
__device__ __forceinline__ uchar shfl(uchar val, int srcLane, int width = warpSize)
|
|
{
|
|
return (uchar) __shfl((int) val, srcLane, width);
|
|
}
|
|
|
|
__device__ __forceinline__ schar shfl(schar val, int srcLane, int width = warpSize)
|
|
{
|
|
return (schar) __shfl((int) val, srcLane, width);
|
|
}
|
|
|
|
__device__ __forceinline__ ushort shfl(ushort val, int srcLane, int width = warpSize)
|
|
{
|
|
return (ushort) __shfl((int) val, srcLane, width);
|
|
}
|
|
|
|
__device__ __forceinline__ short shfl(short val, int srcLane, int width = warpSize)
|
|
{
|
|
return (short) __shfl((int) val, srcLane, width);
|
|
}
|
|
|
|
__device__ __forceinline__ int shfl(int val, int srcLane, int width = warpSize)
|
|
{
|
|
return __shfl(val, srcLane, width);
|
|
}
|
|
|
|
__device__ __forceinline__ uint shfl(uint val, int srcLane, int width = warpSize)
|
|
{
|
|
return (uint) __shfl((int) val, srcLane, width);
|
|
}
|
|
|
|
__device__ __forceinline__ float shfl(float val, int srcLane, int width = warpSize)
|
|
{
|
|
return __shfl(val, srcLane, width);
|
|
}
|
|
|
|
__device__ double shfl(double val, int srcLane, int width = warpSize)
|
|
{
|
|
int lo = __double2loint(val);
|
|
int hi = __double2hiint(val);
|
|
|
|
lo = __shfl(lo, srcLane, width);
|
|
hi = __shfl(hi, srcLane, width);
|
|
|
|
return __hiloint2double(hi, lo);
|
|
}
|
|
|
|
#define CV_CUDEV_SHFL_VEC_INST(input_type) \
|
|
__device__ __forceinline__ input_type ## 1 shfl(const input_type ## 1 & val, int srcLane, int width = warpSize) \
|
|
{ \
|
|
return VecTraits<input_type ## 1>::make( \
|
|
shfl(val.x, srcLane, width) \
|
|
); \
|
|
} \
|
|
__device__ __forceinline__ input_type ## 2 shfl(const input_type ## 2 & val, int srcLane, int width = warpSize) \
|
|
{ \
|
|
return VecTraits<input_type ## 2>::make( \
|
|
shfl(val.x, srcLane, width), \
|
|
shfl(val.y, srcLane, width) \
|
|
); \
|
|
} \
|
|
__device__ __forceinline__ input_type ## 3 shfl(const input_type ## 3 & val, int srcLane, int width = warpSize) \
|
|
{ \
|
|
return VecTraits<input_type ## 3>::make( \
|
|
shfl(val.x, srcLane, width), \
|
|
shfl(val.y, srcLane, width), \
|
|
shfl(val.z, srcLane, width) \
|
|
); \
|
|
} \
|
|
__device__ __forceinline__ input_type ## 4 shfl(const input_type ## 4 & val, int srcLane, int width = warpSize) \
|
|
{ \
|
|
return VecTraits<input_type ## 4>::make( \
|
|
shfl(val.x, srcLane, width), \
|
|
shfl(val.y, srcLane, width), \
|
|
shfl(val.z, srcLane, width), \
|
|
shfl(val.w, srcLane, width) \
|
|
); \
|
|
}
|
|
|
|
CV_CUDEV_SHFL_VEC_INST(uchar)
|
|
CV_CUDEV_SHFL_VEC_INST(char)
|
|
CV_CUDEV_SHFL_VEC_INST(ushort)
|
|
CV_CUDEV_SHFL_VEC_INST(short)
|
|
CV_CUDEV_SHFL_VEC_INST(uint)
|
|
CV_CUDEV_SHFL_VEC_INST(int)
|
|
CV_CUDEV_SHFL_VEC_INST(float)
|
|
CV_CUDEV_SHFL_VEC_INST(double)
|
|
|
|
#undef CV_CUDEV_SHFL_VEC_INST
|
|
|
|
// shfl_up
|
|
|
|
#if __CUDACC_VER_MAJOR__ >= 9
|
|
|
|
template <typename T>
|
|
__device__ __forceinline__ T shfl_up_sync(uint mask, T val, uint delta, int width = warpSize)
|
|
{
|
|
return (T) __shfl_up_sync(mask, val, delta, width);
|
|
}
|
|
|
|
#else
|
|
|
|
__device__ __forceinline__ uchar shfl_up(uchar val, uint delta, int width = warpSize)
|
|
{
|
|
return (uchar) __shfl_up((int) val, delta, width);
|
|
}
|
|
|
|
__device__ __forceinline__ schar shfl_up(schar val, uint delta, int width = warpSize)
|
|
{
|
|
return (schar) __shfl_up((int) val, delta, width);
|
|
}
|
|
|
|
__device__ __forceinline__ ushort shfl_up(ushort val, uint delta, int width = warpSize)
|
|
{
|
|
return (ushort) __shfl_up((int) val, delta, width);
|
|
}
|
|
|
|
__device__ __forceinline__ short shfl_up(short val, uint delta, int width = warpSize)
|
|
{
|
|
return (short) __shfl_up((int) val, delta, width);
|
|
}
|
|
|
|
__device__ __forceinline__ int shfl_up(int val, uint delta, int width = warpSize)
|
|
{
|
|
return __shfl_up(val, delta, width);
|
|
}
|
|
|
|
__device__ __forceinline__ uint shfl_up(uint val, uint delta, int width = warpSize)
|
|
{
|
|
return (uint) __shfl_up((int) val, delta, width);
|
|
}
|
|
|
|
__device__ __forceinline__ float shfl_up(float val, uint delta, int width = warpSize)
|
|
{
|
|
return __shfl_up(val, delta, width);
|
|
}
|
|
|
|
__device__ double shfl_up(double val, uint delta, int width = warpSize)
|
|
{
|
|
int lo = __double2loint(val);
|
|
int hi = __double2hiint(val);
|
|
|
|
lo = __shfl_up(lo, delta, width);
|
|
hi = __shfl_up(hi, delta, width);
|
|
|
|
return __hiloint2double(hi, lo);
|
|
}
|
|
|
|
__device__ __forceinline__ uint64 shfl_up(uint64 val, uint delta, int width = warpSize)
|
|
{
|
|
return __shfl_up(val, delta, width);
|
|
}
|
|
|
|
#define CV_CUDEV_SHFL_UP_VEC_INST(input_type) \
|
|
__device__ __forceinline__ input_type ## 1 shfl_up(const input_type ## 1 & val, uint delta, int width = warpSize) \
|
|
{ \
|
|
return VecTraits<input_type ## 1>::make( \
|
|
shfl_up(val.x, delta, width) \
|
|
); \
|
|
} \
|
|
__device__ __forceinline__ input_type ## 2 shfl_up(const input_type ## 2 & val, uint delta, int width = warpSize) \
|
|
{ \
|
|
return VecTraits<input_type ## 2>::make( \
|
|
shfl_up(val.x, delta, width), \
|
|
shfl_up(val.y, delta, width) \
|
|
); \
|
|
} \
|
|
__device__ __forceinline__ input_type ## 3 shfl_up(const input_type ## 3 & val, uint delta, int width = warpSize) \
|
|
{ \
|
|
return VecTraits<input_type ## 3>::make( \
|
|
shfl_up(val.x, delta, width), \
|
|
shfl_up(val.y, delta, width), \
|
|
shfl_up(val.z, delta, width) \
|
|
); \
|
|
} \
|
|
__device__ __forceinline__ input_type ## 4 shfl_up(const input_type ## 4 & val, uint delta, int width = warpSize) \
|
|
{ \
|
|
return VecTraits<input_type ## 4>::make( \
|
|
shfl_up(val.x, delta, width), \
|
|
shfl_up(val.y, delta, width), \
|
|
shfl_up(val.z, delta, width), \
|
|
shfl_up(val.w, delta, width) \
|
|
); \
|
|
}
|
|
|
|
CV_CUDEV_SHFL_UP_VEC_INST(uchar)
|
|
CV_CUDEV_SHFL_UP_VEC_INST(char)
|
|
CV_CUDEV_SHFL_UP_VEC_INST(ushort)
|
|
CV_CUDEV_SHFL_UP_VEC_INST(short)
|
|
CV_CUDEV_SHFL_UP_VEC_INST(uint)
|
|
CV_CUDEV_SHFL_UP_VEC_INST(int)
|
|
CV_CUDEV_SHFL_UP_VEC_INST(float)
|
|
CV_CUDEV_SHFL_UP_VEC_INST(double)
|
|
|
|
#undef CV_CUDEV_SHFL_UP_VEC_INST
|
|
|
|
#endif
|
|
|
|
template <typename T>
|
|
__device__ __forceinline__ T compatible_shfl_up(T val, uint delta, int width = warpSize)
|
|
{
|
|
#if __CUDACC_VER_MAJOR__ < 9
|
|
|
|
return shfl_up(val, delta, width);
|
|
|
|
#else // __CUDACC_VER_MAJOR__ < 9
|
|
|
|
#if CV_CUDEV_ARCH >= 700
|
|
return shfl_up_sync(0xFFFFFFFFU, val, delta, width);
|
|
#else
|
|
const int block_size = Block::blockSize();
|
|
const int residual = block_size & (warpSize - 1);
|
|
|
|
if (0 == residual)
|
|
return shfl_up_sync(0xFFFFFFFFU, val, delta, width);
|
|
else
|
|
{
|
|
const int n_warps = divUp(block_size, warpSize);
|
|
const int warp_id = Warp::warpId();
|
|
|
|
if (warp_id < n_warps - 1)
|
|
return shfl_up_sync(0xFFFFFFFFU, val, delta, width);
|
|
else
|
|
{
|
|
// We are at the last threads of a block whose number of threads
|
|
// is not a multiple of the warp size
|
|
uint mask = (1LU << residual) - 1;
|
|
return shfl_up_sync(mask, val, delta, width);
|
|
}
|
|
}
|
|
#endif
|
|
|
|
#endif // __CUDACC_VER_MAJOR__ < 9
|
|
}
|
|
|
|
// shfl_down
|
|
|
|
__device__ __forceinline__ uchar shfl_down(uchar val, uint delta, int width = warpSize)
|
|
{
|
|
return (uchar) __shfl_down((int) val, delta, width);
|
|
}
|
|
|
|
__device__ __forceinline__ schar shfl_down(schar val, uint delta, int width = warpSize)
|
|
{
|
|
return (schar) __shfl_down((int) val, delta, width);
|
|
}
|
|
|
|
__device__ __forceinline__ ushort shfl_down(ushort val, uint delta, int width = warpSize)
|
|
{
|
|
return (ushort) __shfl_down((int) val, delta, width);
|
|
}
|
|
|
|
__device__ __forceinline__ short shfl_down(short val, uint delta, int width = warpSize)
|
|
{
|
|
return (short) __shfl_down((int) val, delta, width);
|
|
}
|
|
|
|
__device__ __forceinline__ int shfl_down(int val, uint delta, int width = warpSize)
|
|
{
|
|
return __shfl_down(val, delta, width);
|
|
}
|
|
|
|
__device__ __forceinline__ uint shfl_down(uint val, uint delta, int width = warpSize)
|
|
{
|
|
return (uint) __shfl_down((int) val, delta, width);
|
|
}
|
|
|
|
__device__ __forceinline__ float shfl_down(float val, uint delta, int width = warpSize)
|
|
{
|
|
return __shfl_down(val, delta, width);
|
|
}
|
|
|
|
__device__ double shfl_down(double val, uint delta, int width = warpSize)
|
|
{
|
|
int lo = __double2loint(val);
|
|
int hi = __double2hiint(val);
|
|
|
|
lo = __shfl_down(lo, delta, width);
|
|
hi = __shfl_down(hi, delta, width);
|
|
|
|
return __hiloint2double(hi, lo);
|
|
}
|
|
|
|
#define CV_CUDEV_SHFL_DOWN_VEC_INST(input_type) \
|
|
__device__ __forceinline__ input_type ## 1 shfl_down(const input_type ## 1 & val, uint delta, int width = warpSize) \
|
|
{ \
|
|
return VecTraits<input_type ## 1>::make( \
|
|
shfl_down(val.x, delta, width) \
|
|
); \
|
|
} \
|
|
__device__ __forceinline__ input_type ## 2 shfl_down(const input_type ## 2 & val, uint delta, int width = warpSize) \
|
|
{ \
|
|
return VecTraits<input_type ## 2>::make( \
|
|
shfl_down(val.x, delta, width), \
|
|
shfl_down(val.y, delta, width) \
|
|
); \
|
|
} \
|
|
__device__ __forceinline__ input_type ## 3 shfl_down(const input_type ## 3 & val, uint delta, int width = warpSize) \
|
|
{ \
|
|
return VecTraits<input_type ## 3>::make( \
|
|
shfl_down(val.x, delta, width), \
|
|
shfl_down(val.y, delta, width), \
|
|
shfl_down(val.z, delta, width) \
|
|
); \
|
|
} \
|
|
__device__ __forceinline__ input_type ## 4 shfl_down(const input_type ## 4 & val, uint delta, int width = warpSize) \
|
|
{ \
|
|
return VecTraits<input_type ## 4>::make( \
|
|
shfl_down(val.x, delta, width), \
|
|
shfl_down(val.y, delta, width), \
|
|
shfl_down(val.z, delta, width), \
|
|
shfl_down(val.w, delta, width) \
|
|
); \
|
|
}
|
|
|
|
CV_CUDEV_SHFL_DOWN_VEC_INST(uchar)
|
|
CV_CUDEV_SHFL_DOWN_VEC_INST(char)
|
|
CV_CUDEV_SHFL_DOWN_VEC_INST(ushort)
|
|
CV_CUDEV_SHFL_DOWN_VEC_INST(short)
|
|
CV_CUDEV_SHFL_DOWN_VEC_INST(uint)
|
|
CV_CUDEV_SHFL_DOWN_VEC_INST(int)
|
|
CV_CUDEV_SHFL_DOWN_VEC_INST(float)
|
|
CV_CUDEV_SHFL_DOWN_VEC_INST(double)
|
|
|
|
#undef CV_CUDEV_SHFL_DOWN_VEC_INST
|
|
|
|
// shfl_xor
|
|
|
|
__device__ __forceinline__ uchar shfl_xor(uchar val, int laneMask, int width = warpSize)
|
|
{
|
|
return (uchar) __shfl_xor((int) val, laneMask, width);
|
|
}
|
|
|
|
__device__ __forceinline__ schar shfl_xor(schar val, int laneMask, int width = warpSize)
|
|
{
|
|
return (schar) __shfl_xor((int) val, laneMask, width);
|
|
}
|
|
|
|
__device__ __forceinline__ ushort shfl_xor(ushort val, int laneMask, int width = warpSize)
|
|
{
|
|
return (ushort) __shfl_xor((int) val, laneMask, width);
|
|
}
|
|
|
|
__device__ __forceinline__ short shfl_xor(short val, int laneMask, int width = warpSize)
|
|
{
|
|
return (short) __shfl_xor((int) val, laneMask, width);
|
|
}
|
|
|
|
__device__ __forceinline__ int shfl_xor(int val, int laneMask, int width = warpSize)
|
|
{
|
|
return __shfl_xor(val, laneMask, width);
|
|
}
|
|
|
|
__device__ __forceinline__ uint shfl_xor(uint val, int laneMask, int width = warpSize)
|
|
{
|
|
return (uint) __shfl_xor((int) val, laneMask, width);
|
|
}
|
|
|
|
__device__ __forceinline__ float shfl_xor(float val, int laneMask, int width = warpSize)
|
|
{
|
|
return __shfl_xor(val, laneMask, width);
|
|
}
|
|
|
|
__device__ double shfl_xor(double val, int laneMask, int width = warpSize)
|
|
{
|
|
int lo = __double2loint(val);
|
|
int hi = __double2hiint(val);
|
|
|
|
lo = __shfl_xor(lo, laneMask, width);
|
|
hi = __shfl_xor(hi, laneMask, width);
|
|
|
|
return __hiloint2double(hi, lo);
|
|
}
|
|
|
|
#define CV_CUDEV_SHFL_XOR_VEC_INST(input_type) \
|
|
__device__ __forceinline__ input_type ## 1 shfl_xor(const input_type ## 1 & val, int laneMask, int width = warpSize) \
|
|
{ \
|
|
return VecTraits<input_type ## 1>::make( \
|
|
shfl_xor(val.x, laneMask, width) \
|
|
); \
|
|
} \
|
|
__device__ __forceinline__ input_type ## 2 shfl_xor(const input_type ## 2 & val, int laneMask, int width = warpSize) \
|
|
{ \
|
|
return VecTraits<input_type ## 2>::make( \
|
|
shfl_xor(val.x, laneMask, width), \
|
|
shfl_xor(val.y, laneMask, width) \
|
|
); \
|
|
} \
|
|
__device__ __forceinline__ input_type ## 3 shfl_xor(const input_type ## 3 & val, int laneMask, int width = warpSize) \
|
|
{ \
|
|
return VecTraits<input_type ## 3>::make( \
|
|
shfl_xor(val.x, laneMask, width), \
|
|
shfl_xor(val.y, laneMask, width), \
|
|
shfl_xor(val.z, laneMask, width) \
|
|
); \
|
|
} \
|
|
__device__ __forceinline__ input_type ## 4 shfl_xor(const input_type ## 4 & val, int laneMask, int width = warpSize) \
|
|
{ \
|
|
return VecTraits<input_type ## 4>::make( \
|
|
shfl_xor(val.x, laneMask, width), \
|
|
shfl_xor(val.y, laneMask, width), \
|
|
shfl_xor(val.z, laneMask, width), \
|
|
shfl_xor(val.w, laneMask, width) \
|
|
); \
|
|
}
|
|
|
|
CV_CUDEV_SHFL_XOR_VEC_INST(uchar)
|
|
CV_CUDEV_SHFL_XOR_VEC_INST(char)
|
|
CV_CUDEV_SHFL_XOR_VEC_INST(ushort)
|
|
CV_CUDEV_SHFL_XOR_VEC_INST(short)
|
|
CV_CUDEV_SHFL_XOR_VEC_INST(uint)
|
|
CV_CUDEV_SHFL_XOR_VEC_INST(int)
|
|
CV_CUDEV_SHFL_XOR_VEC_INST(float)
|
|
CV_CUDEV_SHFL_XOR_VEC_INST(double)
|
|
|
|
#undef CV_CUDEV_SHFL_XOR_VEC_INST
|
|
#undef __shfl
|
|
#undef __shfl_xor
|
|
#undef __shfl_up
|
|
#undef __shfl_down
|
|
|
|
#endif // CV_CUDEV_ARCH >= 300
|
|
|
|
//! @}
|
|
|
|
}}
|
|
|
|
#endif
|