fixes for gpu module:
- fixed printCudaDeviceInfo for new CC - fixed some compilation errors and warnings - removed unset command from CMake script - removed unused std imports
This commit is contained in:
Vladislav Vinogradov
205
modules/gpu/include/opencv2/gpu/device/block.hpp
Normal file
205
modules/gpu/include/opencv2/gpu/device/block.hpp
Normal file
@@ -0,0 +1,205 @@
|
||||
/*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.
|
||||
// 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*/
|
||||
|
||||
#ifndef __OPENCV_GPU_DEVICE_BLOCK_HPP__
|
||||
#define __OPENCV_GPU_DEVICE_BLOCK_HPP__
|
||||
|
||||
namespace cv { namespace gpu { namespace device
|
||||
{
|
||||
struct Block
|
||||
{
|
||||
static __device__ __forceinline__ unsigned int id()
|
||||
{
|
||||
return blockIdx.x;
|
||||
}
|
||||
|
||||
static __device__ __forceinline__ unsigned int stride()
|
||||
{
|
||||
return blockDim.x * blockDim.y * blockDim.z;
|
||||
}
|
||||
|
||||
static __device__ __forceinline__ void sync()
|
||||
{
|
||||
__syncthreads();
|
||||
}
|
||||
|
||||
static __device__ __forceinline__ int flattenedThreadId()
|
||||
{
|
||||
return threadIdx.z * blockDim.x * blockDim.y + threadIdx.y * blockDim.x + threadIdx.x;
|
||||
}
|
||||
|
||||
template<typename It, typename T>
|
||||
static __device__ __forceinline__ void fill(It beg, It end, const T& value)
|
||||
{
|
||||
int STRIDE = stride();
|
||||
It t = beg + flattenedThreadId();
|
||||
|
||||
for(; t < end; t += STRIDE)
|
||||
*t = value;
|
||||
}
|
||||
|
||||
template<typename OutIt, typename T>
|
||||
static __device__ __forceinline__ void yota(OutIt beg, OutIt end, T value)
|
||||
{
|
||||
int STRIDE = stride();
|
||||
int tid = flattenedThreadId();
|
||||
value += tid;
|
||||
|
||||
for(OutIt t = beg + tid; t < end; t += STRIDE, value += STRIDE)
|
||||
*t = value;
|
||||
}
|
||||
|
||||
template<typename InIt, typename OutIt>
|
||||
static __device__ __forceinline__ void copy(InIt beg, InIt end, OutIt out)
|
||||
{
|
||||
int STRIDE = stride();
|
||||
InIt t = beg + flattenedThreadId();
|
||||
OutIt o = out + (t - beg);
|
||||
|
||||
for(; t < end; t += STRIDE, o += STRIDE)
|
||||
*o = *t;
|
||||
}
|
||||
|
||||
template<typename InIt, typename OutIt, class UnOp>
|
||||
static __device__ __forceinline__ void transfrom(InIt beg, InIt end, OutIt out, UnOp op)
|
||||
{
|
||||
int STRIDE = stride();
|
||||
InIt t = beg + flattenedThreadId();
|
||||
OutIt o = out + (t - beg);
|
||||
|
||||
for(; t < end; t += STRIDE, o += STRIDE)
|
||||
*o = op(*t);
|
||||
}
|
||||
|
||||
template<typename InIt1, typename InIt2, typename OutIt, class BinOp>
|
||||
static __device__ __forceinline__ void transfrom(InIt1 beg1, InIt1 end1, InIt2 beg2, OutIt out, BinOp op)
|
||||
{
|
||||
int STRIDE = stride();
|
||||
InIt1 t1 = beg1 + flattenedThreadId();
|
||||
InIt2 t2 = beg2 + flattenedThreadId();
|
||||
OutIt o = out + (t1 - beg1);
|
||||
|
||||
for(; t1 < end1; t1 += STRIDE, t2 += STRIDE, o += STRIDE)
|
||||
*o = op(*t1, *t2);
|
||||
}
|
||||
|
||||
template<int CTA_SIZE, typename T, class BinOp>
|
||||
static __device__ __forceinline__ void reduce(volatile T* buffer, BinOp op)
|
||||
{
|
||||
int tid = flattenedThreadId();
|
||||
T val = buffer[tid];
|
||||
|
||||
if (CTA_SIZE >= 1024) { if (tid < 512) buffer[tid] = val = op(val, buffer[tid + 512]); __syncthreads(); }
|
||||
if (CTA_SIZE >= 512) { if (tid < 256) buffer[tid] = val = op(val, buffer[tid + 256]); __syncthreads(); }
|
||||
if (CTA_SIZE >= 256) { if (tid < 128) buffer[tid] = val = op(val, buffer[tid + 128]); __syncthreads(); }
|
||||
if (CTA_SIZE >= 128) { if (tid < 64) buffer[tid] = val = op(val, buffer[tid + 64]); __syncthreads(); }
|
||||
|
||||
if (tid < 32)
|
||||
{
|
||||
if (CTA_SIZE >= 64) { buffer[tid] = val = op(val, buffer[tid + 32]); }
|
||||
if (CTA_SIZE >= 32) { buffer[tid] = val = op(val, buffer[tid + 16]); }
|
||||
if (CTA_SIZE >= 16) { buffer[tid] = val = op(val, buffer[tid + 8]); }
|
||||
if (CTA_SIZE >= 8) { buffer[tid] = val = op(val, buffer[tid + 4]); }
|
||||
if (CTA_SIZE >= 4) { buffer[tid] = val = op(val, buffer[tid + 2]); }
|
||||
if (CTA_SIZE >= 2) { buffer[tid] = val = op(val, buffer[tid + 1]); }
|
||||
}
|
||||
}
|
||||
|
||||
template<int CTA_SIZE, typename T, class BinOp>
|
||||
static __device__ __forceinline__ T reduce(volatile T* buffer, T init, BinOp op)
|
||||
{
|
||||
int tid = flattenedThreadId();
|
||||
T val = buffer[tid] = init;
|
||||
__syncthreads();
|
||||
|
||||
if (CTA_SIZE >= 1024) { if (tid < 512) buffer[tid] = val = op(val, buffer[tid + 512]); __syncthreads(); }
|
||||
if (CTA_SIZE >= 512) { if (tid < 256) buffer[tid] = val = op(val, buffer[tid + 256]); __syncthreads(); }
|
||||
if (CTA_SIZE >= 256) { if (tid < 128) buffer[tid] = val = op(val, buffer[tid + 128]); __syncthreads(); }
|
||||
if (CTA_SIZE >= 128) { if (tid < 64) buffer[tid] = val = op(val, buffer[tid + 64]); __syncthreads(); }
|
||||
|
||||
if (tid < 32)
|
||||
{
|
||||
if (CTA_SIZE >= 64) { buffer[tid] = val = op(val, buffer[tid + 32]); }
|
||||
if (CTA_SIZE >= 32) { buffer[tid] = val = op(val, buffer[tid + 16]); }
|
||||
if (CTA_SIZE >= 16) { buffer[tid] = val = op(val, buffer[tid + 8]); }
|
||||
if (CTA_SIZE >= 8) { buffer[tid] = val = op(val, buffer[tid + 4]); }
|
||||
if (CTA_SIZE >= 4) { buffer[tid] = val = op(val, buffer[tid + 2]); }
|
||||
if (CTA_SIZE >= 2) { buffer[tid] = val = op(val, buffer[tid + 1]); }
|
||||
}
|
||||
__syncthreads();
|
||||
return buffer[0];
|
||||
}
|
||||
|
||||
template <typename T, class BinOp>
|
||||
static __device__ __forceinline__ void reduce_n(T* data, unsigned int n, BinOp op)
|
||||
{
|
||||
int ftid = flattenedThreadId();
|
||||
int sft = stride();
|
||||
|
||||
if (sft < n)
|
||||
{
|
||||
for (unsigned int i = sft + ftid; i < n; i += sft)
|
||||
data[ftid] = op(data[ftid], data[i]);
|
||||
|
||||
__syncthreads();
|
||||
|
||||
n = sft;
|
||||
}
|
||||
|
||||
while (n > 1)
|
||||
{
|
||||
unsigned int half = n/2;
|
||||
|
||||
if (ftid < half)
|
||||
data[ftid] = op(data[ftid], data[n - ftid - 1]);
|
||||
|
||||
__syncthreads();
|
||||
|
||||
n = n - half;
|
||||
}
|
||||
}
|
||||
};
|
||||
}}}
|
||||
|
||||
#endif /* __OPENCV_GPU_DEVICE_BLOCK_HPP__ */
|
||||
|
||||
|
||||
@@ -85,8 +85,6 @@ static inline void ___cudaSafeCall(cudaError_t err, const char *file, const int
|
||||
cv::gpu::error(cudaGetErrorString(err), file, line, func);
|
||||
}
|
||||
|
||||
#ifdef __CUDACC__
|
||||
|
||||
namespace cv { namespace gpu
|
||||
{
|
||||
__host__ __device__ __forceinline__ int divUp(int total, int grain)
|
||||
@@ -96,19 +94,25 @@ namespace cv { namespace gpu
|
||||
|
||||
namespace device
|
||||
{
|
||||
using cv::gpu::divUp;
|
||||
|
||||
#ifdef __CUDACC__
|
||||
typedef unsigned char uchar;
|
||||
typedef unsigned short ushort;
|
||||
typedef signed char schar;
|
||||
typedef unsigned int uint;
|
||||
#ifdef _WIN32
|
||||
typedef unsigned int uint;
|
||||
#endif
|
||||
|
||||
template<class T> inline void bindTexture(const textureReference* tex, const PtrStepSz<T>& img)
|
||||
{
|
||||
cudaChannelFormatDesc desc = cudaCreateChannelDesc<T>();
|
||||
cudaSafeCall( cudaBindTexture2D(0, tex, img.ptr(), &desc, img.cols, img.rows, img.step) );
|
||||
}
|
||||
#endif // __CUDACC__
|
||||
}
|
||||
}}
|
||||
|
||||
#endif // __CUDACC__
|
||||
|
||||
|
||||
#endif // __OPENCV_GPU_COMMON_HPP__
|
||||
|
||||
@@ -44,7 +44,6 @@
|
||||
#define OPENCV_GPU_EMULATION_HPP_
|
||||
|
||||
#include "warp_reduce.hpp"
|
||||
#include <stdio.h>
|
||||
|
||||
namespace cv { namespace gpu { namespace device
|
||||
{
|
||||
|
||||
@@ -302,18 +302,18 @@ namespace cv { namespace gpu { namespace device
|
||||
template <> struct name<type> : binary_function<type, type, type> \
|
||||
{ \
|
||||
__device__ __forceinline__ type operator()(type lhs, type rhs) const {return op(lhs, rhs);} \
|
||||
__device__ __forceinline__ name(const name& other):binary_function<type, type, type>(){}\
|
||||
__device__ __forceinline__ name():binary_function<type, type, type>(){}\
|
||||
__device__ __forceinline__ name() {}\
|
||||
__device__ __forceinline__ name(const name&) {}\
|
||||
};
|
||||
|
||||
template <typename T> struct maximum : binary_function<T, T, T>
|
||||
{
|
||||
__device__ __forceinline__ T operator()(typename TypeTraits<T>::ParameterType lhs, typename TypeTraits<T>::ParameterType rhs) const
|
||||
{
|
||||
return lhs < rhs ? rhs : lhs;
|
||||
return max(lhs, rhs);
|
||||
}
|
||||
__device__ __forceinline__ maximum(const maximum& other):binary_function<T, T, T>(){}
|
||||
__device__ __forceinline__ maximum():binary_function<T, T, T>(){}
|
||||
__device__ __forceinline__ maximum() {}
|
||||
__device__ __forceinline__ maximum(const maximum&) {}
|
||||
};
|
||||
|
||||
OPENCV_GPU_IMPLEMENT_MINMAX(maximum, uchar, ::max)
|
||||
@@ -330,10 +330,10 @@ namespace cv { namespace gpu { namespace device
|
||||
{
|
||||
__device__ __forceinline__ T operator()(typename TypeTraits<T>::ParameterType lhs, typename TypeTraits<T>::ParameterType rhs) const
|
||||
{
|
||||
return lhs < rhs ? lhs : rhs;
|
||||
return min(lhs, rhs);
|
||||
}
|
||||
__device__ __forceinline__ minimum(const minimum& other):binary_function<T, T, T>(){}
|
||||
__device__ __forceinline__ minimum():binary_function<T, T, T>(){}
|
||||
__device__ __forceinline__ minimum() {}
|
||||
__device__ __forceinline__ minimum(const minimum&) {}
|
||||
};
|
||||
|
||||
OPENCV_GPU_IMPLEMENT_MINMAX(minimum, uchar, ::min)
|
||||
@@ -350,6 +350,108 @@ namespace cv { namespace gpu { namespace device
|
||||
|
||||
// Math functions
|
||||
///bound=========================================
|
||||
|
||||
template <typename T> struct abs_func : unary_function<T, T>
|
||||
{
|
||||
__device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType x) const
|
||||
{
|
||||
return abs(x);
|
||||
}
|
||||
|
||||
__device__ __forceinline__ abs_func() {}
|
||||
__device__ __forceinline__ abs_func(const abs_func&) {}
|
||||
};
|
||||
template <> struct abs_func<unsigned char> : unary_function<unsigned char, unsigned char>
|
||||
{
|
||||
__device__ __forceinline__ unsigned char operator ()(unsigned char x) const
|
||||
{
|
||||
return x;
|
||||
}
|
||||
|
||||
__device__ __forceinline__ abs_func() {}
|
||||
__device__ __forceinline__ abs_func(const abs_func&) {}
|
||||
};
|
||||
template <> struct abs_func<signed char> : unary_function<signed char, signed char>
|
||||
{
|
||||
__device__ __forceinline__ signed char operator ()(signed char x) const
|
||||
{
|
||||
return ::abs((int)x);
|
||||
}
|
||||
|
||||
__device__ __forceinline__ abs_func() {}
|
||||
__device__ __forceinline__ abs_func(const abs_func&) {}
|
||||
};
|
||||
template <> struct abs_func<char> : unary_function<char, char>
|
||||
{
|
||||
__device__ __forceinline__ char operator ()(char x) const
|
||||
{
|
||||
return ::abs((int)x);
|
||||
}
|
||||
|
||||
__device__ __forceinline__ abs_func() {}
|
||||
__device__ __forceinline__ abs_func(const abs_func&) {}
|
||||
};
|
||||
template <> struct abs_func<unsigned short> : unary_function<unsigned short, unsigned short>
|
||||
{
|
||||
__device__ __forceinline__ unsigned short operator ()(unsigned short x) const
|
||||
{
|
||||
return x;
|
||||
}
|
||||
|
||||
__device__ __forceinline__ abs_func() {}
|
||||
__device__ __forceinline__ abs_func(const abs_func&) {}
|
||||
};
|
||||
template <> struct abs_func<short> : unary_function<short, short>
|
||||
{
|
||||
__device__ __forceinline__ short operator ()(short x) const
|
||||
{
|
||||
return ::abs((int)x);
|
||||
}
|
||||
|
||||
__device__ __forceinline__ abs_func() {}
|
||||
__device__ __forceinline__ abs_func(const abs_func&) {}
|
||||
};
|
||||
template <> struct abs_func<unsigned int> : unary_function<unsigned int, unsigned int>
|
||||
{
|
||||
__device__ __forceinline__ unsigned int operator ()(unsigned int x) const
|
||||
{
|
||||
return x;
|
||||
}
|
||||
|
||||
__device__ __forceinline__ abs_func() {}
|
||||
__device__ __forceinline__ abs_func(const abs_func&) {}
|
||||
};
|
||||
template <> struct abs_func<int> : unary_function<int, int>
|
||||
{
|
||||
__device__ __forceinline__ int operator ()(int x) const
|
||||
{
|
||||
return ::abs(x);
|
||||
}
|
||||
|
||||
__device__ __forceinline__ abs_func() {}
|
||||
__device__ __forceinline__ abs_func(const abs_func&) {}
|
||||
};
|
||||
template <> struct abs_func<float> : unary_function<float, float>
|
||||
{
|
||||
__device__ __forceinline__ float operator ()(float x) const
|
||||
{
|
||||
return ::fabsf(x);
|
||||
}
|
||||
|
||||
__device__ __forceinline__ abs_func() {}
|
||||
__device__ __forceinline__ abs_func(const abs_func&) {}
|
||||
};
|
||||
template <> struct abs_func<double> : unary_function<double, double>
|
||||
{
|
||||
__device__ __forceinline__ double operator ()(double x) const
|
||||
{
|
||||
return ::fabs(x);
|
||||
}
|
||||
|
||||
__device__ __forceinline__ abs_func() {}
|
||||
__device__ __forceinline__ abs_func(const abs_func&) {}
|
||||
};
|
||||
|
||||
#define OPENCV_GPU_IMPLEMENT_UN_FUNCTOR(name, func) \
|
||||
template <typename T> struct name ## _func : unary_function<T, float> \
|
||||
{ \
|
||||
@@ -357,6 +459,8 @@ namespace cv { namespace gpu { namespace device
|
||||
{ \
|
||||
return func ## f(v); \
|
||||
} \
|
||||
__device__ __forceinline__ name ## _func() {} \
|
||||
__device__ __forceinline__ name ## _func(const name ## _func&) {} \
|
||||
}; \
|
||||
template <> struct name ## _func<double> : unary_function<double, double> \
|
||||
{ \
|
||||
@@ -364,6 +468,8 @@ namespace cv { namespace gpu { namespace device
|
||||
{ \
|
||||
return func(v); \
|
||||
} \
|
||||
__device__ __forceinline__ name ## _func() {} \
|
||||
__device__ __forceinline__ name ## _func(const name ## _func&) {} \
|
||||
};
|
||||
|
||||
#define OPENCV_GPU_IMPLEMENT_BIN_FUNCTOR(name, func) \
|
||||
@@ -382,7 +488,6 @@ namespace cv { namespace gpu { namespace device
|
||||
} \
|
||||
};
|
||||
|
||||
OPENCV_GPU_IMPLEMENT_UN_FUNCTOR(fabs, ::fabs)
|
||||
OPENCV_GPU_IMPLEMENT_UN_FUNCTOR(sqrt, ::sqrt)
|
||||
OPENCV_GPU_IMPLEMENT_UN_FUNCTOR(exp, ::exp)
|
||||
OPENCV_GPU_IMPLEMENT_UN_FUNCTOR(exp2, ::exp2)
|
||||
|
||||
@@ -280,7 +280,7 @@ namespace cv { namespace gpu { namespace device
|
||||
OPENCV_GPU_IMPLEMENT_VEC_UNOP (type, operator ! , logical_not) \
|
||||
OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, max, maximum) \
|
||||
OPENCV_GPU_IMPLEMENT_VEC_BINOP(type, min, minimum) \
|
||||
OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, fabs, fabs_func) \
|
||||
OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, abs, abs_func) \
|
||||
OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, sqrt, sqrt_func) \
|
||||
OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, exp, exp_func) \
|
||||
OPENCV_GPU_IMPLEMENT_VEC_UNOP(type, exp2, exp2_func) \
|
||||
@@ -327,4 +327,4 @@ namespace cv { namespace gpu { namespace device
|
||||
#undef OPENCV_GPU_IMPLEMENT_VEC_INT_OP
|
||||
}}} // namespace cv { namespace gpu { namespace device
|
||||
|
||||
#endif // __OPENCV_GPU_VECMATH_HPP__
|
||||
#endif // __OPENCV_GPU_VECMATH_HPP__
|
||||
|
||||
Reference in New Issue
Block a user