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

reduce remap instantiates for tiny build

Browse Source
This commit is contained in:
Vladislav Vinogradov 2015-03-04 16:18:57 +03:00
parent e7b55f4982
commit 612888c066
2 changed files with 49 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

41
modules/gpu/src/cuda/remap.cu
View File

@ -209,6 +209,7 @@ namespace cv { namespace gpu { namespace device
typedef void (*caller_t)(PtrStepSz<T> src, PtrStepSz<T> srcWhole, int xoff, int yoff, PtrStepSzf xmap, PtrStepSzf ymap,
PtrStepSz<T> dst, const float* borderValue, cudaStream_t stream, bool cc20);
#ifdef OPENCV_TINY_GPU_MODULE
static const caller_t callers[3][5] =
{
{
@ -216,25 +217,55 @@ namespace cv { namespace gpu { namespace device
RemapDispatcher<PointFilter, BrdReplicate, T>::call,
RemapDispatcher<PointFilter, BrdConstant, T>::call,
RemapDispatcher<PointFilter, BrdReflect, T>::call,
RemapDispatcher<PointFilter, BrdWrap, T>::call
0/*RemapDispatcher<PointFilter, BrdWrap, T>::call*/,
},
{
RemapDispatcher<LinearFilter, BrdReflect101, T>::call,
RemapDispatcher<LinearFilter, BrdReplicate, T>::call,
RemapDispatcher<LinearFilter, BrdConstant, T>::call,
RemapDispatcher<LinearFilter, BrdReflect, T>::call,
RemapDispatcher<LinearFilter, BrdWrap, T>::call
0/*RemapDispatcher<LinearFilter, BrdWrap, T>::call*/,
},
{
0/*RemapDispatcher<CubicFilter, BrdReflect101, T>::call*/,
0/*RemapDispatcher<CubicFilter, BrdReplicate, T>::call*/,
0/*RemapDispatcher<CubicFilter, BrdConstant, T>::call*/,
0/*RemapDispatcher<CubicFilter, BrdReflect, T>::call*/,
0/*RemapDispatcher<CubicFilter, BrdWrap, T>::call*/,
}
};
#else
static const caller_t callers[3][5] =
{
{
RemapDispatcher<PointFilter, BrdReflect101, T>::call,
RemapDispatcher<PointFilter, BrdReplicate, T>::call,
RemapDispatcher<PointFilter, BrdConstant, T>::call,
RemapDispatcher<PointFilter, BrdReflect, T>::call,
RemapDispatcher<PointFilter, BrdWrap, T>::call,
},
{
RemapDispatcher<LinearFilter, BrdReflect101, T>::call,
RemapDispatcher<LinearFilter, BrdReplicate, T>::call,
RemapDispatcher<LinearFilter, BrdConstant, T>::call,
RemapDispatcher<LinearFilter, BrdReflect, T>::call,
RemapDispatcher<LinearFilter, BrdWrap, T>::call,
},
{
RemapDispatcher<CubicFilter, BrdReflect101, T>::call,
RemapDispatcher<CubicFilter, BrdReplicate, T>::call,
RemapDispatcher<CubicFilter, BrdConstant, T>::call,
RemapDispatcher<CubicFilter, BrdReflect, T>::call,
RemapDispatcher<CubicFilter, BrdWrap, T>::call
RemapDispatcher<CubicFilter, BrdWrap, T>::call,
}
};
#endif
callers[interpolation][borderMode](static_cast< PtrStepSz<T> >(src), static_cast< PtrStepSz<T> >(srcWhole), xoff, yoff, xmap, ymap,
const caller_t caller = callers[interpolation][borderMode];
if (!caller)
cv::gpu::error("Unsupported input parameters for remap", __FILE__, __LINE__, "");
caller(static_cast< PtrStepSz<T> >(src), static_cast< PtrStepSz<T> >(srcWhole), xoff, yoff, xmap, ymap,
static_cast< PtrStepSz<T> >(dst), borderValue, stream, cc20);
}
@ -243,6 +274,7 @@ namespace cv { namespace gpu { namespace device
template void remap_gpu<uchar3>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, PtrStepSzf xmap, PtrStepSzf ymap, PtrStepSzb dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);
template void remap_gpu<uchar4>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, PtrStepSzf xmap, PtrStepSzf ymap, PtrStepSzb dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);
#ifndef OPENCV_TINY_GPU_MODULE
//template void remap_gpu<schar>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, PtrStepSzf xmap, PtrStepSzf ymap, PtrStepSzb dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);
//template void remap_gpu<char2>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, PtrStepSzf xmap, PtrStepSzf ymap, PtrStepSzb dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);
//template void remap_gpu<char3>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, PtrStepSzf xmap, PtrStepSzf ymap, PtrStepSzb dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);
@ -262,6 +294,7 @@ namespace cv { namespace gpu { namespace device
//template void remap_gpu<int2>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, PtrStepSzf xmap, PtrStepSzf ymap, PtrStepSzb dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);
//template void remap_gpu<int3>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, PtrStepSzf xmap, PtrStepSzf ymap, PtrStepSzb dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);
//template void remap_gpu<int4>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, PtrStepSzf xmap, PtrStepSzf ymap, PtrStepSzb dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);
#endif
template void remap_gpu<float >(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, PtrStepSzf xmap, PtrStepSzf ymap, PtrStepSzb dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);
//template void remap_gpu<float2>(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, PtrStepSzf xmap, PtrStepSzf ymap, PtrStepSzb dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);

12
modules/gpu/src/remap.cpp
View File

@ -65,6 +65,17 @@ void cv::gpu::remap(const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const Gp
typedef void (*func_t)(PtrStepSzb src, PtrStepSzb srcWhole, int xoff, int yoff, PtrStepSzf xmap, PtrStepSzf ymap, PtrStepSzb dst, int interpolation,
int borderMode, const float* borderValue, cudaStream_t stream, bool cc20);
#ifdef OPENCV_TINY_GPU_MODULE
static const func_t funcs[6][4] =
{
{remap_gpu<uchar> , 0 /*remap_gpu<uchar2>*/ , remap_gpu<uchar3> , remap_gpu<uchar4> },
{0, 0, 0, 0},
{0, 0, 0, 0},
{0, 0, 0, 0},
{0, 0, 0, 0},
{remap_gpu<float> , 0 /*remap_gpu<float2>*/ , remap_gpu<float3> , remap_gpu<float4> }
};
#else
static const func_t funcs[6][4] =
{
{remap_gpu<uchar> , 0 /*remap_gpu<uchar2>*/ , remap_gpu<uchar3> , remap_gpu<uchar4> },
@ -74,6 +85,7 @@ void cv::gpu::remap(const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const Gp
{0 /*remap_gpu<int>*/ , 0 /*remap_gpu<int2>*/ , 0 /*remap_gpu<int3>*/ , 0 /*remap_gpu<int4>*/ },
{remap_gpu<float> , 0 /*remap_gpu<float2>*/ , remap_gpu<float3> , remap_gpu<float4> }
};
#endif
CV_Assert(src.depth() <= CV_32F && src.channels() <= 4);
CV_Assert(xmap.type() == CV_32F && ymap.type() == CV_32F && xmap.size() == ymap.size());