generic-library/opencv
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

linux compiletion error

Browse Source
This commit is contained in:
Andrey Morozov
2010-07-19 10:49:35 +00:00
parent 07825bad1e
commit ace7c7e93c
5 changed files with 213 additions and 47 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
modules/gpu/src/cuda/cuda_shared.hpp
View File

@@ -44,23 +44,26 @@
#define __OPENCV_CUDA_SHARED_HPP__
#include "opencv2/gpu/devmem2d.hpp"
#include "cuda_runtime_api.h"
#include "cuda_runtime_api.h"
namespace cv
{
namespace gpu
{
{
typedef unsigned char uchar;
typedef unsigned short ushort;
typedef unsigned int uint;
typedef unsigned int uint;
extern "C" void error( const char *error_string, const char *file, const int line, const char *func = "");
namespace impl
{
{
static inline int divUp(int a, int b) { return (a % b == 0) ? a/b : a/b + 1; }
extern "C" void stereoBM_GPU(const DevMem2D& left, const DevMem2D& right, DevMem2D& disp, int maxdisp, DevMem2D_<uint>& minSSD_buf);
extern "C" void set_to_without_mask (const DevMem2D& mat, const double * scalar, int depth, int channels);
extern "C" void set_to_with_mask (const DevMem2D& mat, const double * scalar, const DevMem2D& mask, int depth, int channels);
}
}
}
@@ -68,12 +71,12 @@ namespace cv
#if defined(__GNUC__)
#define cudaSafeCall(expr) ___cudaSafeCall(expr, __FILE__, __LINE__, __func__);
#else /* defined(__CUDACC__) || defined(__MSVC__) */
#define cudaSafeCall(expr) ___cudaSafeCall(expr, __FILE__, __LINE__)
#define cudaSafeCall(expr) ___cudaSafeCall(expr, __FILE__, __LINE__)
#endif
static inline void ___cudaSafeCall(cudaError_t err, const char *file, const int line, const char *func = "")
{
if( cudaSuccess != err)
if( cudaSuccess != err)
cv::gpu::error(cudaGetErrorString(err), __FILE__, __LINE__, func);
}

150
modules/gpu/src/cuda/matrix_operations.cu Normal file
View File

@@ -0,0 +1,150 @@
/*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*/
#include <stddef.h>
#include "cuda_shared.hpp"
#include "cuda_runtime.h"
__constant__ float scalar_d[4];
namespace mat_operators
{
template <typename T, int channels, int count = channels>
struct unroll
{
__device__ static void unroll_set(T * mat, size_t i)
{
mat[i] = static_cast<T>(scalar_d[i % channels]);
unroll<T, channels, count - 1>::unroll_set(mat, i+1);
}
__device__ static void unroll_set_with_mask(T * mat, float mask, size_t i)
{
mat[i] = mask * static_cast<T>(scalar_d[i % channels]);
unroll<T, channels, count - 1>::unroll_set_with_mask(mat, mask, i+1);
}
};
template <typename T, int channels>
struct unroll<T,channels,0>
{
__device__ static void unroll_set(T * , size_t){}
__device__ static void unroll_set_with_mask(T * , float, size_t){}
};
template <typename T, int channels>
__global__ void kernel_set_to_without_mask(T * mat)
{
size_t i = (blockIdx.x * blockDim.x + threadIdx.x) * sizeof(T);
unroll<T, channels>::unroll_set(mat, i);
}
template <typename T, int channels>
__global__ void kernel_set_to_with_mask(T * mat, const float * mask)
{
size_t i = (blockIdx.x * blockDim.x + threadIdx.x) * sizeof(T);
unroll<T, channels>::unroll_set_with_mask(mat, i, mask[i]);
}
}
extern "C" void cv::gpu::impl::set_to_with_mask(const DevMem2D& mat, const double * scalar, const DevMem2D& mask, int depth, int channels)
{
scalar_d[0] = scalar[0];
scalar_d[1] = scalar[1];
scalar_d[2] = scalar[2];
scalar_d[3] = scalar[3];
dim3 numBlocks(mat.rows * mat.step / 256, 1, 1);
dim3 threadsPerBlock(256);
if (channels == 1)
{
if (depth == 1) ::mat_operators::kernel_set_to_with_mask<unsigned char, 1><<<numBlocks,threadsPerBlock>>>(mat.ptr, (float *)mask.ptr);
if (depth == 2) ::mat_operators::kernel_set_to_with_mask<unsigned short, 1><<<numBlocks,threadsPerBlock>>>((unsigned short *)mat.ptr, (float *)mask.ptr);
if (depth == 4) ::mat_operators::kernel_set_to_with_mask<unsigned int, 1><<<numBlocks,threadsPerBlock>>>((unsigned int *)mat.ptr, (float *)mask.ptr);
}
if (channels == 2)
{
if (depth == 1) ::mat_operators::kernel_set_to_with_mask<unsigned char, 2><<<numBlocks,threadsPerBlock>>>(mat.ptr, (float *)mask.ptr);
if (depth == 2) ::mat_operators::kernel_set_to_with_mask<unsigned short, 2><<<numBlocks,threadsPerBlock>>>((unsigned short *)mat.ptr, (float *)mask.ptr);
if (depth == 4) ::mat_operators::kernel_set_to_with_mask<unsigned int, 2><<<numBlocks,threadsPerBlock>>>((unsigned int *)mat.ptr, (float *)mask.ptr);
}
if (channels == 3)
{
if (depth == 1) ::mat_operators::kernel_set_to_with_mask<unsigned char, 3><<<numBlocks,threadsPerBlock>>>(mat.ptr, (float *)mask.ptr);
if (depth == 2) ::mat_operators::kernel_set_to_with_mask<unsigned short, 3><<<numBlocks,threadsPerBlock>>>((unsigned short *)mat.ptr, (float *)mask.ptr);
if (depth == 4) ::mat_operators::kernel_set_to_with_mask<unsigned int, 3><<<numBlocks,threadsPerBlock>>>((unsigned int *)mat.ptr, (float *)mask.ptr);
}
}
extern "C" void cv::gpu::impl::set_to_without_mask(const DevMem2D& mat, const double * scalar, int depth, int channels)
{
scalar_d[0] = scalar[0];
scalar_d[1] = scalar[1];
scalar_d[2] = scalar[2];
scalar_d[3] = scalar[3];
int numBlocks = mat.rows * mat.step / 256;
dim3 threadsPerBlock(256);
if (channels == 1)
{
if (depth == 1) ::mat_operators::kernel_set_to_without_mask<unsigned char, 1><<<numBlocks,threadsPerBlock>>>(mat.ptr);
if (depth == 2) ::mat_operators::kernel_set_to_without_mask<unsigned short, 1><<<numBlocks,threadsPerBlock>>>((unsigned short *)mat.ptr);
if (depth == 4) ::mat_operators::kernel_set_to_without_mask<unsigned int, 1><<<numBlocks,threadsPerBlock>>>((unsigned int *)mat.ptr);
}
if (channels == 2)
{
if (depth == 1) ::mat_operators::kernel_set_to_without_mask<unsigned char, 2><<<numBlocks,threadsPerBlock>>>(mat.ptr);
if (depth == 2) ::mat_operators::kernel_set_to_without_mask<unsigned short, 2><<<numBlocks,threadsPerBlock>>>((unsigned short *)mat.ptr);
if (depth == 4) ::mat_operators::kernel_set_to_without_mask<unsigned int, 2><<<numBlocks,threadsPerBlock>>>((unsigned int *)mat.ptr);
}
if (channels == 3)
{
if (depth == 1) ::mat_operators::kernel_set_to_without_mask<unsigned char, 3><<<numBlocks,threadsPerBlock>>>(mat.ptr);
if (depth == 2) ::mat_operators::kernel_set_to_without_mask<unsigned short, 3><<<numBlocks,threadsPerBlock>>>((unsigned short *)mat.ptr);
if (depth == 4) ::mat_operators::kernel_set_to_without_mask<unsigned int, 3><<<numBlocks,threadsPerBlock>>>((unsigned int *)mat.ptr);
}
}

19
modules/gpu/src/cudastream.cpp
View File

@@ -74,13 +74,13 @@ struct CudaStream::Impl
cudaStream_t stream;
int ref_counter;
};
namespace
namespace
{
template<class S, class D> void devcopy(const S& src, D& dst, cudaStream_t s, cudaMemcpyKind k)
{
dst.create(src.size(), src.type());
size_t bwidth = src.cols * src.elemSize();
cudaSafeCall( cudaMemcpy2DAsync(dst.data, dst.step, src.data, src.step, bwidth, src.rows, k, s) );
cudaSafeCall( cudaMemcpy2DAsync(dst.data, dst.step, src.data, src.step, bwidth, src.rows, k, s) );
};
}
@@ -97,7 +97,7 @@ void cv::gpu::CudaStream::create()
impl = (CudaStream::Impl*)fastMalloc(sizeof(CudaStream::Impl));
impl->stream = stream;
impl->ref_counter = 1;
impl->ref_counter = 1;
}
void cv::gpu::CudaStream::release()
@@ -125,7 +125,7 @@ CudaStream& cv::gpu::CudaStream::operator=(const CudaStream& stream)
CV_XADD(&stream.impl->ref_counter, 1);
release();
impl = stream.impl;
impl = stream.impl;
}
return *this;
}
@@ -138,20 +138,21 @@ bool cv::gpu::CudaStream::queryIfComplete()
return err == cudaSuccess;
cudaSafeCall(err);
return false;
}
void cv::gpu::CudaStream::waitForCompletion() { cudaSafeCall( cudaStreamSynchronize( impl->stream ) ); }
void cv::gpu::CudaStream::enqueueDownload(const GpuMat& src, Mat& dst)
{
void cv::gpu::CudaStream::enqueueDownload(const GpuMat& src, Mat& dst)
{
// if not -> allocation will be done, but after that dst will not point to page locked memory
CV_Assert(src.cols == dst.cols && src.rows == dst.rows && src.type() == dst.type() )
devcopy(src, dst, impl->stream, cudaMemcpyDeviceToHost);
devcopy(src, dst, impl->stream, cudaMemcpyDeviceToHost);
}
void cv::gpu::CudaStream::enqueueDownload(const GpuMat& src, MatPL& dst) { devcopy(src, dst, impl->stream, cudaMemcpyDeviceToHost); }
void cv::gpu::CudaStream::enqueueUpload(const MatPL& src, GpuMat& dst){ devcopy(src, dst, impl->stream, cudaMemcpyHostToDevice); }
void cv::gpu::CudaStream::enqueueUpload(const Mat& src, GpuMat& dst) { devcopy(src, dst, impl->stream, cudaMemcpyHostToDevice); }
void cv::gpu::CudaStream::enqueueUpload(const Mat& src, GpuMat& dst) { devcopy(src, dst, impl->stream, cudaMemcpyHostToDevice); }
void cv::gpu::CudaStream::enqueueCopy(const GpuMat& src, GpuMat& dst) { devcopy(src, dst, impl->stream, cudaMemcpyDeviceToDevice); }
void cv::gpu::CudaStream::enqueueMemSet(const GpuMat& src, Scalar val)
@@ -170,4 +171,4 @@ void cv::gpu::CudaStream::enqueueConvert(const GpuMat& src, GpuMat& dst, int typ
}
#endif /* !defined (HAVE_CUDA) */
#endif /* !defined (HAVE_CUDA) */

36
modules/gpu/src/matrix_operations.cpp
View File

@@ -100,7 +100,7 @@ void cv::gpu::GpuMat::copyTo( GpuMat& m ) const
}
void cv::gpu::GpuMat::copyTo( GpuMat& /*m*/, const GpuMat&/* mask */) const
{
{
CV_Assert(!"Not implemented");
}
@@ -109,15 +109,27 @@ void cv::gpu::GpuMat::convertTo( GpuMat& /*m*/, int /*rtype*/, double /*alpha*/,
CV_Assert(!"Not implemented");
}
GpuMat& cv::gpu::GpuMat::operator = (const Scalar& /*s*/)
GpuMat& GpuMat::operator = (const Scalar& s)
{
CV_Assert(!"Not implemented");
cv::gpu::impl::set_to_without_mask(*this, s.val, this->depth(), this->channels());
return *this;
}
GpuMat& cv::gpu::GpuMat::setTo(const Scalar& /*s*/, const GpuMat& /*mask*/)
GpuMat& GpuMat::setTo(const Scalar& s, const GpuMat& mask)
{
CV_Assert(!"Not implemented");
CV_Assert(mask.type() == CV_8U);
CV_DbgAssert(!this->empty());
if (mask.empty())
{
cv::gpu::impl::set_to_without_mask(*this, s.val, this->depth(), this->channels());
}
else
{
cv::gpu::impl::set_to_with_mask(*this, s.val, mask, this->depth(), this->channels());
}
return *this;
}
@@ -177,7 +189,7 @@ void cv::gpu::GpuMat::create(int _rows, int _cols, int _type)
rows = _rows;
cols = _cols;
size_t esz = elemSize();
size_t esz = elemSize();
void *dev_ptr;
cudaSafeCall( cudaMallocPitch(&dev_ptr, &step, esz * cols, rows) );
@@ -189,7 +201,7 @@ void cv::gpu::GpuMat::create(int _rows, int _cols, int _type)
size_t nettosize = (size_t)_nettosize;
datastart = data = (uchar*)dev_ptr;
dataend = data + nettosize;
dataend = data + nettosize;
refcount = (int*)fastMalloc(sizeof(*refcount));
*refcount = 1;
@@ -201,7 +213,7 @@ void cv::gpu::GpuMat::release()
if( refcount && CV_XADD(refcount, -1) == 1 )
{
fastFree(refcount);
cudaSafeCall( cudaFree(datastart) );
cudaSafeCall( cudaFree(datastart) );
}
data = datastart = dataend = 0;
step = rows = cols = 0;
@@ -233,12 +245,12 @@ void cv::gpu::MatPL::create(int _rows, int _cols, int _type)
CV_Error(CV_StsNoMem, "Too big buffer is allocated");
size_t datasize = alignSize(nettosize, (int)sizeof(*refcount));
//datastart = data = (uchar*)fastMalloc(datasize + sizeof(*refcount));
//datastart = data = (uchar*)fastMalloc(datasize + sizeof(*refcount));
void *ptr;
cudaSafeCall( cudaHostAlloc( &ptr, datasize, cudaHostAllocDefault) );
datastart = data = (uchar*)ptr;
dataend = data + nettosize;
datastart = data = (uchar*)ptr;
dataend = data + nettosize;
refcount = (int*)cv::fastMalloc(sizeof(*refcount));
*refcount = 1;
@@ -257,4 +269,4 @@ void cv::gpu::MatPL::release()
refcount = 0;
}
#endif /* !defined (HAVE_CUDA) */
#endif /* !defined (HAVE_CUDA) */