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

fixed compilation without CUBLAS and CUFFT

Browse Source
This commit is contained in:
Vladislav Vinogradov
2013-07-18 11:24:34 +04:00
parent ee13ee3c6d
commit ebe7ff991f
4 changed files with 115 additions and 29 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
modules/gpu/src/cuda/imgproc.cu
View File

@@ -619,6 +619,7 @@ namespace cv { namespace gpu { namespace device
//////////////////////////////////////////////////////////////////////////
// mulSpectrums
#ifdef HAVE_CUFFT
__global__ void mulSpectrumsKernel(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, PtrStepSz<cufftComplex> c)
{
const int x = blockIdx.x * blockDim.x + threadIdx.x;
@@ -642,11 +643,13 @@ namespace cv { namespace gpu { namespace device
if (stream == 0)
cudaSafeCall( cudaDeviceSynchronize() );
}
#endif
//////////////////////////////////////////////////////////////////////////
// mulSpectrums_CONJ
#ifdef HAVE_CUFFT
__global__ void mulSpectrumsKernel_CONJ(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, PtrStepSz<cufftComplex> c)
{
const int x = blockIdx.x * blockDim.x + threadIdx.x;
@@ -670,11 +673,13 @@ namespace cv { namespace gpu { namespace device
if (stream == 0)
cudaSafeCall( cudaDeviceSynchronize() );
}
#endif
//////////////////////////////////////////////////////////////////////////
// mulAndScaleSpectrums
#ifdef HAVE_CUFFT
__global__ void mulAndScaleSpectrumsKernel(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, float scale, PtrStepSz<cufftComplex> c)
{
const int x = blockIdx.x * blockDim.x + threadIdx.x;
@@ -699,11 +704,13 @@ namespace cv { namespace gpu { namespace device
if (stream)
cudaSafeCall( cudaDeviceSynchronize() );
}
#endif
//////////////////////////////////////////////////////////////////////////
// mulAndScaleSpectrums_CONJ
#ifdef HAVE_CUFFT
__global__ void mulAndScaleSpectrumsKernel_CONJ(const PtrStep<cufftComplex> a, const PtrStep<cufftComplex> b, float scale, PtrStepSz<cufftComplex> c)
{
const int x = blockIdx.x * blockDim.x + threadIdx.x;
@@ -728,6 +735,7 @@ namespace cv { namespace gpu { namespace device
if (stream == 0)
cudaSafeCall( cudaDeviceSynchronize() );
}
#endif
//////////////////////////////////////////////////////////////////////////
// buildWarpMaps

97
modules/gpu/src/cuda/safe_call.hpp
View File

@@ -43,53 +43,94 @@
#ifndef __OPENCV_CUDA_SAFE_CALL_HPP__
#define __OPENCV_CUDA_SAFE_CALL_HPP__
#include <cuda_runtime_api.h>
#include <cufft.h>
#include <cublas.h>
#include "NCV.hpp"
#include "cvconfig.h"
#if defined(__GNUC__)
#define nppSafeCall(expr) ___nppSafeCall(expr, __FILE__, __LINE__, __func__)
#define ncvSafeCall(expr) ___ncvSafeCall(expr, __FILE__, __LINE__, __func__)
#define cufftSafeCall(expr) ___cufftSafeCall(expr, __FILE__, __LINE__, __func__)
#define cublasSafeCall(expr) ___cublasSafeCall(expr, __FILE__, __LINE__, __func__)
#else /* defined(__CUDACC__) || defined(__MSVC__) */
#define nppSafeCall(expr) ___nppSafeCall(expr, __FILE__, __LINE__)
#define ncvSafeCall(expr) ___ncvSafeCall(expr, __FILE__, __LINE__)
#define cufftSafeCall(expr) ___cufftSafeCall(expr, __FILE__, __LINE__)
#define cublasSafeCall(expr) ___cublasSafeCall(expr, __FILE__, __LINE__)
#include <cuda_runtime_api.h>
#ifdef HAVE_CUFFT
# include <cufft.h>
#endif
namespace cv { namespace gpu
{
void nppError(int err, const char *file, const int line, const char *func = "");
void ncvError(int err, const char *file, const int line, const char *func = "");
#ifdef HAVE_CUBLAS
# include <cublas.h>
#endif
#include "NCV.hpp"
namespace cv { namespace gpu {
void nppError(int err, const char *file, const int line, const char *func = "");
void ncvError(int err, const char *file, const int line, const char *func = "");
#ifdef HAVE_CUFFT
void cufftError(int err, const char *file, const int line, const char *func = "");
#endif
#ifdef HAVE_CUBLAS
void cublasError(int err, const char *file, const int line, const char *func = "");
#endif
}}
// nppSafeCall
static inline void ___nppSafeCall(int err, const char *file, const int line, const char *func = "")
{
if (err < 0)
cv::gpu::nppError(err, file, line, func);
}
#if defined(__GNUC__)
#define nppSafeCall(expr) ___nppSafeCall(expr, __FILE__, __LINE__, __func__)
#else
#define nppSafeCall(expr) ___nppSafeCall(expr, __FILE__, __LINE__)
#endif
// ncvSafeCall
static inline void ___ncvSafeCall(int err, const char *file, const int line, const char *func = "")
{
if (NCV_SUCCESS != err)
cv::gpu::ncvError(err, file, line, func);
}
static inline void ___cufftSafeCall(cufftResult_t err, const char *file, const int line, const char *func = "")
{
if (CUFFT_SUCCESS != err)
cv::gpu::cufftError(err, file, line, func);
}
#if defined(__GNUC__)
#define ncvSafeCall(expr) ___ncvSafeCall(expr, __FILE__, __LINE__, __func__)
#else
#define ncvSafeCall(expr) ___ncvSafeCall(expr, __FILE__, __LINE__)
#endif
static inline void ___cublasSafeCall(cublasStatus_t err, const char *file, const int line, const char *func = "")
{
if (CUBLAS_STATUS_SUCCESS != err)
cv::gpu::cublasError(err, file, line, func);
}
// cufftSafeCall
#ifdef HAVE_CUFFT
static inline void ___cufftSafeCall(cufftResult_t err, const char *file, const int line, const char *func = "")
{
if (CUFFT_SUCCESS != err)
cv::gpu::cufftError(err, file, line, func);
}
#if defined(__GNUC__)
#define cufftSafeCall(expr) ___cufftSafeCall(expr, __FILE__, __LINE__, __func__)
#else
#define cufftSafeCall(expr) ___cufftSafeCall(expr, __FILE__, __LINE__)
#endif
#endif
// cublasSafeCall
#ifdef HAVE_CUBLAS
static inline void ___cublasSafeCall(cublasStatus_t err, const char *file, const int line, const char *func = "")
{
if (CUBLAS_STATUS_SUCCESS != err)
cv::gpu::cublasError(err, file, line, func);
}
#if defined(__GNUC__)
#define cublasSafeCall(expr) ___cublasSafeCall(expr, __FILE__, __LINE__, __func__)
#else
#define cublasSafeCall(expr) ___cublasSafeCall(expr, __FILE__, __LINE__)
#endif
#endif
#endif /* __OPENCV_CUDA_SAFE_CALL_HPP__ */