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

refactoring: moved gpu reduction-based functions into separated file

Browse Source
This commit is contained in:
Alexey Spizhevoy
2010-12-20 09:51:25 +00:00
parent 1922e50f19
commit df8529377b
7 changed files with 2377 additions and 2260 deletions
Download Patch File Download Diff File Expand all files Collapse all files

152
modules/gpu/src/element_operations.cpp
View File

@@ -66,10 +66,14 @@ void cv::gpu::bitwise_and(const GpuMat&, const GpuMat&, GpuMat&, const GpuMat&)
void cv::gpu::bitwise_and(const GpuMat&, const GpuMat&, GpuMat&, const GpuMat&, const Stream&) { throw_nogpu(); }
void cv::gpu::bitwise_xor(const GpuMat&, const GpuMat&, GpuMat&, const GpuMat&) { throw_nogpu(); }
void cv::gpu::bitwise_xor(const GpuMat&, const GpuMat&, GpuMat&, const GpuMat&, const Stream&) { throw_nogpu(); }
cv::gpu::GpuMat cv::gpu::operator ~ (const GpuMat&) { throw_nogpu(); return GpuMat(); }
cv::gpu::GpuMat cv::gpu::operator | (const GpuMat&, const GpuMat&) { throw_nogpu(); return GpuMat(); }
cv::gpu::GpuMat cv::gpu::operator & (const GpuMat&, const GpuMat&) { throw_nogpu(); return GpuMat(); }
cv::gpu::GpuMat cv::gpu::operator ^ (const GpuMat&, const GpuMat&) { throw_nogpu(); return GpuMat(); }
void cv::gpu::min(const GpuMat&, const GpuMat&, GpuMat&) { throw_nogpu(); }
void cv::gpu::min(const GpuMat&, const GpuMat&, GpuMat&, const Stream&) { throw_nogpu(); }
void cv::gpu::min(const GpuMat&, double, GpuMat&) { throw_nogpu(); }
void cv::gpu::min(const GpuMat&, double, GpuMat&, const Stream&) { throw_nogpu(); }
void cv::gpu::max(const GpuMat&, const GpuMat&, GpuMat&) { throw_nogpu(); }
void cv::gpu::max(const GpuMat&, const GpuMat&, GpuMat&, const Stream&) { throw_nogpu(); }
void cv::gpu::max(const GpuMat&, double, GpuMat&) { throw_nogpu(); }
void cv::gpu::max(const GpuMat&, double, GpuMat&, const Stream&) { throw_nogpu(); }
#else
@@ -574,4 +578,144 @@ void cv::gpu::bitwise_xor(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, c
::bitwiseXorCaller(src1, src2, dst, mask, StreamAccessor::getStream(stream));
}
//////////////////////////////////////////////////////////////////////////////
// Minimum and maximum operations
namespace cv { namespace gpu { namespace mathfunc
{
template <typename T>
void min_gpu(const DevMem2D_<T>& src1, const DevMem2D_<T>& src2, const DevMem2D_<T>& dst, cudaStream_t stream);
template <typename T>
void max_gpu(const DevMem2D_<T>& src1, const DevMem2D_<T>& src2, const DevMem2D_<T>& dst, cudaStream_t stream);
template <typename T>
void min_gpu(const DevMem2D_<T>& src1, double src2, const DevMem2D_<T>& dst, cudaStream_t stream);
template <typename T>
void max_gpu(const DevMem2D_<T>& src1, double src2, const DevMem2D_<T>& dst, cudaStream_t stream);
}}}
namespace
{
template <typename T>
void min_caller(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, cudaStream_t stream)
{
CV_Assert(src1.size() == src2.size() && src1.type() == src2.type());
dst.create(src1.size(), src1.type());
mathfunc::min_gpu<T>(src1.reshape(1), src2.reshape(1), dst.reshape(1), stream);
}
template <typename T>
void min_caller(const GpuMat& src1, double src2, GpuMat& dst, cudaStream_t stream)
{
dst.create(src1.size(), src1.type());
mathfunc::min_gpu<T>(src1.reshape(1), src2, dst.reshape(1), stream);
}
template <typename T>
void max_caller(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, cudaStream_t stream)
{
CV_Assert(src1.size() == src2.size() && src1.type() == src2.type());
dst.create(src1.size(), src1.type());
mathfunc::max_gpu<T>(src1.reshape(1), src2.reshape(1), dst.reshape(1), stream);
}
template <typename T>
void max_caller(const GpuMat& src1, double src2, GpuMat& dst, cudaStream_t stream)
{
dst.create(src1.size(), src1.type());
mathfunc::max_gpu<T>(src1.reshape(1), src2, dst.reshape(1), stream);
}
}
void cv::gpu::min(const GpuMat& src1, const GpuMat& src2, GpuMat& dst)
{
typedef void (*func_t)(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, cudaStream_t stream);
static const func_t funcs[] =
{
min_caller<uchar>, min_caller<char>, min_caller<ushort>, min_caller<short>, min_caller<int>,
min_caller<float>, min_caller<double>
};
funcs[src1.depth()](src1, src2, dst, 0);
}
void cv::gpu::min(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const Stream& stream)
{
typedef void (*func_t)(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, cudaStream_t stream);
static const func_t funcs[] =
{
min_caller<uchar>, min_caller<char>, min_caller<ushort>, min_caller<short>, min_caller<int>,
min_caller<float>, min_caller<double>
};
funcs[src1.depth()](src1, src2, dst, StreamAccessor::getStream(stream));
}
void cv::gpu::min(const GpuMat& src1, double src2, GpuMat& dst)
{
typedef void (*func_t)(const GpuMat& src1, double src2, GpuMat& dst, cudaStream_t stream);
static const func_t funcs[] =
{
min_caller<uchar>, min_caller<char>, min_caller<ushort>, min_caller<short>, min_caller<int>,
min_caller<float>, min_caller<double>
};
funcs[src1.depth()](src1, src2, dst, 0);
}
void cv::gpu::min(const GpuMat& src1, double src2, GpuMat& dst, const Stream& stream)
{
typedef void (*func_t)(const GpuMat& src1, double src2, GpuMat& dst, cudaStream_t stream);
static const func_t funcs[] =
{
min_caller<uchar>, min_caller<char>, min_caller<ushort>, min_caller<short>, min_caller<int>,
min_caller<float>, min_caller<double>
};
funcs[src1.depth()](src1, src2, dst, StreamAccessor::getStream(stream));
}
void cv::gpu::max(const GpuMat& src1, const GpuMat& src2, GpuMat& dst)
{
typedef void (*func_t)(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, cudaStream_t stream);
static const func_t funcs[] =
{
max_caller<uchar>, max_caller<char>, max_caller<ushort>, max_caller<short>, max_caller<int>,
max_caller<float>, max_caller<double>
};
funcs[src1.depth()](src1, src2, dst, 0);
}
void cv::gpu::max(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const Stream& stream)
{
typedef void (*func_t)(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, cudaStream_t stream);
static const func_t funcs[] =
{
max_caller<uchar>, max_caller<char>, max_caller<ushort>, max_caller<short>, max_caller<int>,
max_caller<float>, max_caller<double>
};
funcs[src1.depth()](src1, src2, dst, StreamAccessor::getStream(stream));
}
void cv::gpu::max(const GpuMat& src1, double src2, GpuMat& dst)
{
typedef void (*func_t)(const GpuMat& src1, double src2, GpuMat& dst, cudaStream_t stream);
static const func_t funcs[] =
{
max_caller<uchar>, max_caller<char>, max_caller<ushort>, max_caller<short>, max_caller<int>,
max_caller<float>, max_caller<double>
};
funcs[src1.depth()](src1, src2, dst, 0);
}
void cv::gpu::max(const GpuMat& src1, double src2, GpuMat& dst, const Stream& stream)
{
typedef void (*func_t)(const GpuMat& src1, double src2, GpuMat& dst, cudaStream_t stream);
static const func_t funcs[] =
{
max_caller<uchar>, max_caller<char>, max_caller<ushort>, max_caller<short>, max_caller<int>,
max_caller<float>, max_caller<double>
};
funcs[src1.depth()](src1, src2, dst, StreamAccessor::getStream(stream));
}
#endif