Skip to content

O
opencv
Commit 3f5dd5f1 authored by Vladislav Vinogradov's avatar Vladislav Vinogradov
Browse files
Options

added implementation GpuMat::convertTo and merged this with matrix_operations.cpp

parent 7bf29e14
Hide whitespace changes
Inline Side-by-side
Showing with 426 additions and 3 deletions
modules/gpu/src/cuda/cuda_shared.hpp
View file @ 3f5dd5f1
...@@ -51,6 +51,7 @@ namespace cv ...@@ -51,6 +51,7 @@ namespace cv
namespace gpu namespace gpu
{ {
typedef unsigned char uchar; typedef unsigned char uchar;
typedef signed char schar;
typedef unsigned short ushort; typedef unsigned short ushort;
typedef unsigned int uint; typedef unsigned int uint;
...@@ -62,6 +63,8 @@ namespace cv ...@@ -62,6 +63,8 @@ namespace cv
extern "C" void set_to_without_mask (const DevMem2D& mat, const double * scalar, int depth, int channels); 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); extern "C" void set_to_with_mask (const DevMem2D& mat, const double * scalar, const DevMem2D& mask, int depth, int channels);
extern "C" void convert_to(const DevMem2D& src, int sdepth, DevMem2D dst, int ddepth, size_t width, size_t height, double alpha, double beta);
} }
} }
} }
......
modules/gpu/src/cuda/matrix_operations.cu
View file @ 3f5dd5f1
...@@ -46,10 +46,17 @@ ...@@ -46,10 +46,17 @@
#include "cuda_shared.hpp" #include "cuda_shared.hpp"
#include "cuda_runtime.h" #include "cuda_runtime.h"
using namespace cv::gpu;
using namespace cv::gpu::impl;
__constant__ __align__(16) float scalar_d[4]; __constant__ __align__(16) float scalar_d[4];
namespace mat_operators namespace mat_operators
{ {
//////////////////////////////////////////////////////////
// SetTo
//////////////////////////////////////////////////////////
template<typename T, int channels> template<typename T, int channels>
__global__ void kernel_set_to_without_mask(T * mat, int cols, int rows, int step) __global__ void kernel_set_to_without_mask(T * mat, int cols, int rows, int step)
{ {
...@@ -76,7 +83,245 @@ namespace mat_operators ...@@ -76,7 +83,245 @@ namespace mat_operators
mat[idx] = scalar_d[ x % channels ]; mat[idx] = scalar_d[ x % channels ];
} }
} }
}
//////////////////////////////////////////////////////////
// ConvertTo
//////////////////////////////////////////////////////////
template <typename T, typename DT, size_t src_elem_size, size_t dst_elem_size>
struct Converter
{
__device__ static void convert(uchar* srcmat, size_t src_step, uchar* dstmat, size_t dst_step, size_t width, size_t height, double alpha, double beta)
{
size_t x = threadIdx.x + blockIdx.x * blockDim.x;
size_t y = threadIdx.y + blockIdx.y * blockDim.y;
if (x < width && y < height)
{
const T* src = (const T*)(srcmat + src_step * y);
DT* dst = (DT*)(dstmat + dst_step * y);
dst[x] = (DT)__double2int_rn(alpha * src[x] + beta);
}
}
__host__ static inline dim3 calcGrid(size_t width, size_t height, dim3 block)
{
return dim3(divUp(width, block.x), divUp(height, block.y));
}
};
template <typename T, typename DT>
struct Converter<T, DT, 1, 1>
{
__device__ static void convert(uchar* srcmat, size_t src_step, uchar* dstmat, size_t dst_step, size_t width, size_t height, double alpha, double beta)
{
size_t x = threadIdx.x + blockIdx.x * blockDim.x;
size_t y = threadIdx.y + blockIdx.y * blockDim.y;
if (y < height)
{
const T* src = (const T*)(srcmat + src_step * y);
DT* dst = (DT*)(dstmat + dst_step * y);
if ((x << 2) + 3 < width)
{
uchar4 src4b = ((const uchar4*)src)[x];
uchar4 dst4b;
const T* src1b = (const T*) &src4b.x;
DT* dst1b = (DT*) &dst4b.x;
dst1b[0] = (DT)__double2int_rn(alpha * src1b[0] + beta);
dst1b[1] = (DT)__double2int_rn(alpha * src1b[1] + beta);
dst1b[2] = (DT)__double2int_rn(alpha * src1b[2] + beta);
dst1b[3] = (DT)__double2int_rn(alpha * src1b[3] + beta);
((uchar4*)dst)[x] = dst4b;
}
else
{
if ((x << 2) + 0 < width)
dst[(x << 2) + 0] = (DT)__double2int_rn(alpha * src[(x << 2) + 0] + beta);
if ((x << 2) + 1 < width)
dst[(x << 2) + 1] = (DT)__double2int_rn(alpha * src[(x << 2) + 1] + beta);
if ((x << 2) + 2 < width)
dst[(x << 2) + 2] = (DT)__double2int_rn(alpha * src[(x << 2) + 2] + beta);
}
}
}
__host__ static inline dim3 calcGrid(size_t width, size_t height, dim3 block)
{
return dim3(divUp(width, block.x << 2), divUp(height, block.y));
}
};/**/
template <typename T, typename DT>
struct Converter<T, DT, 1, 2>
{
__device__ static void convert(uchar* srcmat, size_t src_step, uchar* dstmat, size_t dst_step, size_t width, size_t height, double alpha, double beta)
{
size_t x = threadIdx.x + blockIdx.x * blockDim.x;
size_t y = threadIdx.y + blockIdx.y * blockDim.y;
if (y < height)
{
const T* src = (const T*)(srcmat + src_step * y);
DT* dst = (DT*)(dstmat + dst_step * y);
if ((x << 1) + 1 < width)
{
uchar2 src2b = ((const uchar2*)src)[x];
ushort2 dst2s;
const T* src1b = (const T*) &src2b;
DT* dst1s = (DT*) &dst2s;
dst1s[0] = (DT)__double2int_rn(alpha * src1b[0] + beta);
dst1s[1] = (DT)__double2int_rn(alpha * src1b[1] + beta);
((ushort2*)(dst))[x] = dst2s;
}
else
{
if ((x << 1) < width)
dst[(x << 1)] = (DT)__double2int_rn(alpha * src[(x << 1)] + beta);
}
}
}
__host__ static inline dim3 calcGrid(size_t width, size_t height, dim3 block)
{
return dim3(divUp(width, block.x << 1), divUp(height, block.y));
}
};/**/
template <typename T, typename DT>
struct Converter<T, DT, 2, 1>
{
__device__ static void convert(uchar* srcmat, size_t src_step, uchar* dstmat, size_t dst_step, size_t width, size_t height, double alpha, double beta)
{
size_t x = threadIdx.x + blockIdx.x * blockDim.x;
size_t y = threadIdx.y + blockIdx.y * blockDim.y;
if (y < height)
{
const T* src = (const T*)(srcmat + src_step * y);
DT* dst = (DT*)(dstmat + dst_step * y);
if ((x << 2) + 3 < width)
{
ushort4 src4s = ((const ushort4*)src)[x];
uchar4 dst4b;
const T* src1s = (const T*) &src4s.x;
DT* dst1b = (DT*) &dst4b.x;
dst1b[0] = (DT)__double2int_rn(alpha * src1s[0] + beta);
dst1b[1] = (DT)__double2int_rn(alpha * src1s[1] + beta);
dst1b[2] = (DT)__double2int_rn(alpha * src1s[2] + beta);
dst1b[3] = (DT)__double2int_rn(alpha * src1s[3] + beta);
((uchar4*)(dst))[x] = dst4b;
}
else
{
if ((x << 2) + 0 < width)
dst[(x << 2) + 0] = (DT)__double2int_rn(alpha * src[(x << 2) + 0] + beta);
if ((x << 2) + 1 < width)
dst[(x << 2) + 1] = (DT)__double2int_rn(alpha * src[(x << 2) + 1] + beta);
if ((x << 2) + 2 < width)
dst[(x << 2) + 2] = (DT)__double2int_rn(alpha * src[(x << 2) + 2] + beta);
}
}
}
__host__ static inline dim3 calcGrid(size_t width, size_t height, dim3 block)
{
return dim3(divUp(width, block.x << 2), divUp(height, block.y));
}
};/**/
template <typename T, typename DT>
struct Converter<T, DT, 2, 2>
{
__device__ static void convert(uchar* srcmat, size_t src_step, uchar* dstmat, size_t dst_step, size_t width, size_t height, double alpha, double beta)
{
size_t x = threadIdx.x + blockIdx.x * blockDim.x;
size_t y = threadIdx.y + blockIdx.y * blockDim.y;
if (y < height)
{
const T* src = (const T*)(srcmat + src_step * y);
DT* dst = (DT*)(dstmat + dst_step * y);
if ((x << 1) + 1 < width)
{
ushort2 src2s = ((const ushort2*)src)[x];
ushort2 dst2s;
const T* src1s = (const T*) &src2s.x;
DT* dst1s = (DT*) &dst2s.x;
dst1s[0] = (DT)__double2int_rn(alpha * src1s[0] + beta);
dst1s[1] = (DT)__double2int_rn(alpha * src1s[1] + beta);
((ushort2*)dst)[x] = dst2s;
}
else
{
if ((x << 1) < width)
dst[(x << 1)] = (DT)__double2int_rn(alpha * src[(x << 1)] + beta);
}
}
}
__host__ static inline dim3 calcGrid(size_t width, size_t height, dim3 block)
{
return dim3(divUp(width, block.x << 1), divUp(height, block.y));
}
};/**/
template <typename T, size_t src_elem_size, size_t dst_elem_size>
struct Converter<T, float, src_elem_size, dst_elem_size>
{
__device__ static void convert(uchar* srcmat, size_t src_step, uchar* dstmat, size_t dst_step, size_t width, size_t height, double alpha, double beta)
{
size_t x = threadIdx.x + blockIdx.x * blockDim.x;
size_t y = threadIdx.y + blockIdx.y * blockDim.y;
if (x < width && y < height)
{
const T* src = (const T*)(srcmat + src_step * y);
float* dst = (float*)(dstmat + dst_step * y);
dst[x] = (float)(alpha * src[x] + beta);
}
}
__host__ static inline dim3 calcGrid(size_t width, size_t height, dim3 block)
{
return dim3(divUp(width, block.x), divUp(height, block.y));
}
};
template <typename T, size_t src_elem_size, size_t dst_elem_size>
struct Converter<T, double, src_elem_size, dst_elem_size>
{
__device__ static void convert(uchar* srcmat, size_t src_step, uchar* dstmat, size_t dst_step, size_t width, size_t height, double alpha, double beta)
{
size_t x = threadIdx.x + blockIdx.x * blockDim.x;
size_t y = threadIdx.y + blockIdx.y * blockDim.y;
if (x < width && y < height)
{
const T* src = (const T*)(srcmat + src_step * y);
double* dst = (double*)(dstmat + dst_step * y);
dst[x] = (double)(alpha * src[x] + beta);
}
}
__host__ static inline dim3 calcGrid(size_t width, size_t height, dim3 block)
{
return dim3(divUp(width, block.x), divUp(height, block.y));
}
};
template <typename T, typename DT>
__global__ static void kernel_convert_to(uchar* srcmat, size_t src_step, uchar* dstmat, size_t dst_step, size_t width, size_t height, double alpha, double beta)
{
Converter<T, DT, sizeof(T), sizeof(DT)>::convert(srcmat, src_step, dstmat, dst_step, width, height, alpha, beta);
}
} // namespace mat_operators
//////////////////////////////////////////////////////////////
// SetTo
//////////////////////////////////////////////////////////////
extern "C" void cv::gpu::impl::set_to_without_mask(const DevMem2D& mat, const double * scalar, int elemSize1, int channels) extern "C" void cv::gpu::impl::set_to_without_mask(const DevMem2D& mat, const double * scalar, int elemSize1, int channels)
{ {
...@@ -158,3 +403,66 @@ extern "C" void cv::gpu::impl::set_to_with_mask(const DevMem2D& mat, const doubl ...@@ -158,3 +403,66 @@ extern "C" void cv::gpu::impl::set_to_with_mask(const DevMem2D& mat, const doubl
cudaSafeCall ( cudaThreadSynchronize() ); cudaSafeCall ( cudaThreadSynchronize() );
} }
//////////////////////////////////////////////////////////////
// ConvertTo
//////////////////////////////////////////////////////////////
namespace cv
{
namespace gpu
{
namespace impl
{
typedef void (*CvtFunc)(const DevMem2D& src, DevMem2D& dst, size_t width, size_t height, double alpha, double beta);
//#if !defined(__CUDA_ARCH__) || (__CUDA_ARCH__ >= 130)
template<typename T, typename DT>
void cvt_(const DevMem2D& src, DevMem2D& dst, size_t width, size_t height, double alpha, double beta)
{
dim3 block(32, 8);
dim3 grid = ::mat_operators::Converter<T, DT, sizeof(T), sizeof(DT)>::calcGrid(width, height, block);
::mat_operators::kernel_convert_to<T, DT><<<grid, block>>>(src.ptr, src.step, dst.ptr, dst.step, width, height, alpha, beta);
cudaSafeCall( cudaThreadSynchronize() );
}
//#endif
extern "C" void convert_to(const DevMem2D& src, int sdepth, DevMem2D dst, int ddepth, size_t width, size_t height, double alpha, double beta)
{
static CvtFunc tab[8][8] =
{
{cvt_<uchar, uchar>, cvt_<uchar, schar>, cvt_<uchar, ushort>, cvt_<uchar, short>,
cvt_<uchar, int>, cvt_<uchar, float>, cvt_<uchar, double>, 0},
{cvt_<schar, uchar>, cvt_<schar, schar>, cvt_<schar, ushort>, cvt_<schar, short>,
cvt_<schar, int>, cvt_<schar, float>, cvt_<schar, double>, 0},
{cvt_<ushort, uchar>, cvt_<ushort, schar>, cvt_<ushort, ushort>, cvt_<ushort, short>,
cvt_<ushort, int>, cvt_<ushort, float>, cvt_<ushort, double>, 0},
{cvt_<short, uchar>, cvt_<short, schar>, cvt_<short, ushort>, cvt_<short, short>,
cvt_<short, int>, cvt_<short, float>, cvt_<short, double>, 0},
{cvt_<int, uchar>, cvt_<int, schar>, cvt_<int, ushort>,
cvt_<int, short>, cvt_<int, int>, cvt_<int, float>, cvt_<int, double>, 0},
{cvt_<float, uchar>, cvt_<float, schar>, cvt_<float, ushort>,
cvt_<float, short>, cvt_<float, int>, cvt_<float, float>, cvt_<float, double>, 0},
{cvt_<double, uchar>, cvt_<double, schar>, cvt_<double, ushort>,
cvt_<double, short>, cvt_<double, int>, cvt_<double, float>, cvt_<double, double>, 0},
{0,0,0,0,0,0,0,0}
};
CvtFunc func = tab[sdepth][ddepth];
if (func == 0)
error("Operation \'ConvertTo\' doesn't supported on your GPU model", __FILE__, __LINE__);
func(src, dst, width, height, alpha, beta);
}
}
}
}
modules/gpu/src/matrix_operations.cpp
View file @ 3f5dd5f1
...@@ -104,9 +104,31 @@ void cv::gpu::GpuMat::copyTo( GpuMat& /*m*/, const GpuMat&/* mask */) const ...@@ -104,9 +104,31 @@ void cv::gpu::GpuMat::copyTo( GpuMat& /*m*/, const GpuMat&/* mask */) const
CV_Assert(!"Not implemented"); CV_Assert(!"Not implemented");
} }
void cv::gpu::GpuMat::convertTo( GpuMat& /*m*/, int /*rtype*/, double /*alpha*/, double /*beta*/ ) const void cv::gpu::GpuMat::convertTo( GpuMat& dst, int rtype, double alpha, double beta ) const
{ {
CV_Assert(!"Not implemented"); //CV_Assert(!"Not implemented");
bool noScale = fabs(alpha-1) < std::numeric_limits<double>::epsilon() && fabs(beta) < std::numeric_limits<double>::epsilon();
if( rtype < 0 )
rtype = type();
else
rtype = CV_MAKETYPE(CV_MAT_DEPTH(rtype), channels());
int sdepth = depth(), ddepth = CV_MAT_DEPTH(rtype);
/*if( sdepth == ddepth && noScale )
{
copyTo(dst);
return;
}*/
GpuMat temp;
const GpuMat* psrc = this;
if( sdepth != ddepth && psrc == &dst )
psrc = &(temp = *this);
dst.create( size(), rtype );
impl::convert_to(*psrc, sdepth, dst, ddepth, psrc->cols * psrc->channels(), psrc->rows, alpha, beta);
} }
GpuMat& GpuMat::operator = (const Scalar& s) GpuMat& GpuMat::operator = (const Scalar& s)
......
modules/gpu/src/precomp.hpp
View file @ 3f5dd5f1
...@@ -51,6 +51,7 @@ ...@@ -51,6 +51,7 @@
#endif #endif
#include <iostream> #include <iostream>
#include <limits>
#include "opencv2/gpu/gpu.hpp" #include "opencv2/gpu/gpu.hpp"
......
tests/gpu/src/convert_to.cpp 0 → 100644
View file @ 3f5dd5f1
#include "gputest.hpp"
#include <string>
#include <iostream>
#include <fstream>
#include <iterator>
#include <limits>
#include <numeric>
using namespace cv;
using namespace std;
using namespace gpu;
class CV_GpuMatOpConvertTo : public CvTest
{
public:
CV_GpuMatOpConvertTo();
~CV_GpuMatOpConvertTo();
protected:
void run(int);
};
CV_GpuMatOpConvertTo::CV_GpuMatOpConvertTo(): CvTest( "GpuMatOperatorConvertTo", "convertTo" ) {}
CV_GpuMatOpConvertTo::~CV_GpuMatOpConvertTo() {}
void CV_GpuMatOpConvertTo::run( int /* start_from */)
{
const Size img_size(67, 35);
const int types[] = {CV_8U, CV_8S, CV_16U, CV_16S, CV_32S, CV_32F, CV_64F/**/};
const int types_num = sizeof(types) / sizeof(int);
const char* types_str[] = {"CV_8U", "CV_8S", "CV_16U", "CV_16S", "CV_32S", "CV_32F", "CV_64F"};
bool passed = true;
for (int i = 0; i < types_num && passed; ++i)
{
for (int j = 0; j < types_num && passed; ++j)
{
for (int c = 1; c < 2 && passed; ++c)
{
//if (i == j)
// continue;
const int src_type = CV_MAKETYPE(types[i], c);
const int dst_type = types[j];
const double alpha = (double)rand() / RAND_MAX * 10.0;
const double beta = (double)rand() / RAND_MAX * 10.0;
Mat cpumatsrc(img_size, src_type);
randu(cpumatsrc, Scalar::all(0), Scalar::all(10));
GpuMat gpumatsrc(cpumatsrc);
Mat cpumatdst;
GpuMat gpumatdst;
//double cput = (double)getTickCount();
cpumatsrc.convertTo(cpumatdst, dst_type, alpha, beta);
//cput = ((double)getTickCount() - cput) / getTickFrequency();
//double gput = (double)getTickCount();
gpumatsrc.convertTo(gpumatdst, dst_type, alpha, beta);
//gput = ((double)getTickCount() - gput) / getTickFrequency();
/*cout << "convertTo time: " << endl;
cout << "CPU time: " << cput << endl;
cout << "GPU time: " << gput << endl;/**/
double r = norm(cpumatdst, gpumatdst, NORM_L1);
if (r > 1)
{
/*namedWindow("CPU");
imshow("CPU", cpumatdst);
namedWindow("GPU");
imshow("GPU", gpumatdst);
waitKey();/**/
cout << "Failed:" << endl;
cout << "\tr = " << r << endl;
cout << "\tSRC_TYPE=" << types_str[i] << "C" << c << " DST_TYPE=" << types_str[j] << endl;/**/
passed = false;
}
}
}
}
ts->set_failed_test_info(passed ? CvTS::OK : CvTS::FAIL_GENERIC);
}
CV_GpuMatOpConvertTo CV_GpuMatOpConvertTo_test;
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment