gpu module refactoring: moved per-element operations into separated file
This commit is contained in:
Alexey Spizhevoy
@@ -48,18 +48,7 @@ using namespace std;
|
||||
|
||||
#if !defined (HAVE_CUDA)
|
||||
|
||||
void cv::gpu::add(const GpuMat&, const GpuMat&, GpuMat&) { throw_nogpu(); }
|
||||
void cv::gpu::add(const GpuMat&, const Scalar&, GpuMat&) { throw_nogpu(); }
|
||||
void cv::gpu::subtract(const GpuMat&, const GpuMat&, GpuMat&) { throw_nogpu(); }
|
||||
void cv::gpu::subtract(const GpuMat&, const Scalar&, GpuMat&) { throw_nogpu(); }
|
||||
void cv::gpu::multiply(const GpuMat&, const GpuMat&, GpuMat&) { throw_nogpu(); }
|
||||
void cv::gpu::multiply(const GpuMat&, const Scalar&, GpuMat&) { throw_nogpu(); }
|
||||
void cv::gpu::divide(const GpuMat&, const GpuMat&, GpuMat&) { throw_nogpu(); }
|
||||
void cv::gpu::divide(const GpuMat&, const Scalar&, GpuMat&) { throw_nogpu(); }
|
||||
void cv::gpu::transpose(const GpuMat&, GpuMat&) { throw_nogpu(); }
|
||||
void cv::gpu::absdiff(const GpuMat&, const GpuMat&, GpuMat&) { throw_nogpu(); }
|
||||
void cv::gpu::absdiff(const GpuMat&, const Scalar&, GpuMat&) { throw_nogpu(); }
|
||||
void cv::gpu::compare(const GpuMat&, const GpuMat&, GpuMat&, int) { throw_nogpu(); }
|
||||
void cv::gpu::meanStdDev(const GpuMat&, Scalar&, Scalar&) { throw_nogpu(); }
|
||||
double cv::gpu::norm(const GpuMat&, int) { throw_nogpu(); return 0.0; }
|
||||
double cv::gpu::norm(const GpuMat&, const GpuMat&, int) { throw_nogpu(); return 0.0; }
|
||||
@@ -89,18 +78,6 @@ void cv::gpu::cartToPolar(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, bool)
|
||||
void cv::gpu::cartToPolar(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, bool, const Stream&) { throw_nogpu(); }
|
||||
void cv::gpu::polarToCart(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, bool) { throw_nogpu(); }
|
||||
void cv::gpu::polarToCart(const GpuMat&, const GpuMat&, GpuMat&, GpuMat&, bool, const Stream&) { throw_nogpu(); }
|
||||
void cv::gpu::bitwise_not(const GpuMat&, GpuMat&, const GpuMat&) { throw_nogpu(); }
|
||||
void cv::gpu::bitwise_not(const GpuMat&, GpuMat&, const GpuMat&, const Stream&) { throw_nogpu(); }
|
||||
void cv::gpu::bitwise_or(const GpuMat&, const GpuMat&, GpuMat&, const GpuMat&) { throw_nogpu(); }
|
||||
void cv::gpu::bitwise_or(const GpuMat&, const GpuMat&, GpuMat&, const GpuMat&, const Stream&) { throw_nogpu(); }
|
||||
void cv::gpu::bitwise_and(const GpuMat&, const GpuMat&, GpuMat&, const GpuMat&) { throw_nogpu(); }
|
||||
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(); }
|
||||
@@ -112,164 +89,6 @@ void cv::gpu::max(const GpuMat&, double, GpuMat&, const Stream&) { throw_nogpu()
|
||||
|
||||
#else /* !defined (HAVE_CUDA) */
|
||||
|
||||
////////////////////////////////////////////////////////////////////////
|
||||
// add subtract multiply divide
|
||||
|
||||
namespace
|
||||
{
|
||||
typedef NppStatus (*npp_arithm_8u_t)(const Npp8u* pSrc1, int nSrc1Step, const Npp8u* pSrc2, int nSrc2Step, Npp8u* pDst, int nDstStep,
|
||||
NppiSize oSizeROI, int nScaleFactor);
|
||||
typedef NppStatus (*npp_arithm_32s_t)(const Npp32s* pSrc1, int nSrc1Step, const Npp32s* pSrc2, int nSrc2Step, Npp32s* pDst,
|
||||
int nDstStep, NppiSize oSizeROI);
|
||||
typedef NppStatus (*npp_arithm_32f_t)(const Npp32f* pSrc1, int nSrc1Step, const Npp32f* pSrc2, int nSrc2Step, Npp32f* pDst,
|
||||
int nDstStep, NppiSize oSizeROI);
|
||||
|
||||
void nppArithmCaller(const GpuMat& src1, const GpuMat& src2, GpuMat& dst,
|
||||
npp_arithm_8u_t npp_func_8uc1, npp_arithm_8u_t npp_func_8uc4,
|
||||
npp_arithm_32s_t npp_func_32sc1, npp_arithm_32f_t npp_func_32fc1)
|
||||
{
|
||||
CV_DbgAssert(src1.size() == src2.size() && src1.type() == src2.type());
|
||||
|
||||
CV_Assert(src1.type() == CV_8UC1 || src1.type() == CV_8UC4 || src1.type() == CV_32SC1 || src1.type() == CV_32FC1);
|
||||
|
||||
dst.create( src1.size(), src1.type() );
|
||||
|
||||
NppiSize sz;
|
||||
sz.width = src1.cols;
|
||||
sz.height = src1.rows;
|
||||
|
||||
switch (src1.type())
|
||||
{
|
||||
case CV_8UC1:
|
||||
nppSafeCall( npp_func_8uc1(src1.ptr<Npp8u>(), src1.step,
|
||||
src2.ptr<Npp8u>(), src2.step,
|
||||
dst.ptr<Npp8u>(), dst.step, sz, 0) );
|
||||
break;
|
||||
case CV_8UC4:
|
||||
nppSafeCall( npp_func_8uc4(src1.ptr<Npp8u>(), src1.step,
|
||||
src2.ptr<Npp8u>(), src2.step,
|
||||
dst.ptr<Npp8u>(), dst.step, sz, 0) );
|
||||
break;
|
||||
case CV_32SC1:
|
||||
nppSafeCall( npp_func_32sc1(src1.ptr<Npp32s>(), src1.step,
|
||||
src2.ptr<Npp32s>(), src2.step,
|
||||
dst.ptr<Npp32s>(), dst.step, sz) );
|
||||
break;
|
||||
case CV_32FC1:
|
||||
nppSafeCall( npp_func_32fc1(src1.ptr<Npp32f>(), src1.step,
|
||||
src2.ptr<Npp32f>(), src2.step,
|
||||
dst.ptr<Npp32f>(), dst.step, sz) );
|
||||
break;
|
||||
default:
|
||||
CV_Assert(!"Unsupported source type");
|
||||
}
|
||||
}
|
||||
|
||||
template<int SCN> struct NppArithmScalarFunc;
|
||||
template<> struct NppArithmScalarFunc<1>
|
||||
{
|
||||
typedef NppStatus (*func_ptr)(const Npp32f *pSrc, int nSrcStep, Npp32f nValue, Npp32f *pDst,
|
||||
int nDstStep, NppiSize oSizeROI);
|
||||
};
|
||||
template<> struct NppArithmScalarFunc<2>
|
||||
{
|
||||
typedef NppStatus (*func_ptr)(const Npp32fc *pSrc, int nSrcStep, Npp32fc nValue, Npp32fc *pDst,
|
||||
int nDstStep, NppiSize oSizeROI);
|
||||
};
|
||||
|
||||
template<int SCN, typename NppArithmScalarFunc<SCN>::func_ptr func> struct NppArithmScalar;
|
||||
template<typename NppArithmScalarFunc<1>::func_ptr func> struct NppArithmScalar<1, func>
|
||||
{
|
||||
static void calc(const GpuMat& src, const Scalar& sc, GpuMat& dst)
|
||||
{
|
||||
dst.create(src.size(), src.type());
|
||||
|
||||
NppiSize sz;
|
||||
sz.width = src.cols;
|
||||
sz.height = src.rows;
|
||||
|
||||
nppSafeCall( func(src.ptr<Npp32f>(), src.step, (Npp32f)sc[0], dst.ptr<Npp32f>(), dst.step, sz) );
|
||||
}
|
||||
};
|
||||
template<typename NppArithmScalarFunc<2>::func_ptr func> struct NppArithmScalar<2, func>
|
||||
{
|
||||
static void calc(const GpuMat& src, const Scalar& sc, GpuMat& dst)
|
||||
{
|
||||
dst.create(src.size(), src.type());
|
||||
|
||||
NppiSize sz;
|
||||
sz.width = src.cols;
|
||||
sz.height = src.rows;
|
||||
|
||||
Npp32fc nValue;
|
||||
nValue.re = (Npp32f)sc[0];
|
||||
nValue.im = (Npp32f)sc[1];
|
||||
|
||||
nppSafeCall( func(src.ptr<Npp32fc>(), src.step, nValue, dst.ptr<Npp32fc>(), dst.step, sz) );
|
||||
}
|
||||
};
|
||||
}
|
||||
|
||||
void cv::gpu::add(const GpuMat& src1, const GpuMat& src2, GpuMat& dst)
|
||||
{
|
||||
nppArithmCaller(src1, src2, dst, nppiAdd_8u_C1RSfs, nppiAdd_8u_C4RSfs, nppiAdd_32s_C1R, nppiAdd_32f_C1R);
|
||||
}
|
||||
|
||||
void cv::gpu::subtract(const GpuMat& src1, const GpuMat& src2, GpuMat& dst)
|
||||
{
|
||||
nppArithmCaller(src2, src1, dst, nppiSub_8u_C1RSfs, nppiSub_8u_C4RSfs, nppiSub_32s_C1R, nppiSub_32f_C1R);
|
||||
}
|
||||
|
||||
void cv::gpu::multiply(const GpuMat& src1, const GpuMat& src2, GpuMat& dst)
|
||||
{
|
||||
nppArithmCaller(src1, src2, dst, nppiMul_8u_C1RSfs, nppiMul_8u_C4RSfs, nppiMul_32s_C1R, nppiMul_32f_C1R);
|
||||
}
|
||||
|
||||
void cv::gpu::divide(const GpuMat& src1, const GpuMat& src2, GpuMat& dst)
|
||||
{
|
||||
nppArithmCaller(src2, src1, dst, nppiDiv_8u_C1RSfs, nppiDiv_8u_C4RSfs, nppiDiv_32s_C1R, nppiDiv_32f_C1R);
|
||||
}
|
||||
|
||||
void cv::gpu::add(const GpuMat& src, const Scalar& sc, GpuMat& dst)
|
||||
{
|
||||
typedef void (*caller_t)(const GpuMat& src, const Scalar& sc, GpuMat& dst);
|
||||
static const caller_t callers[] = {0, NppArithmScalar<1, nppiAddC_32f_C1R>::calc, NppArithmScalar<2, nppiAddC_32fc_C1R>::calc};
|
||||
|
||||
CV_Assert(src.type() == CV_32FC1 || src.type() == CV_32FC2);
|
||||
|
||||
callers[src.channels()](src, sc, dst);
|
||||
}
|
||||
|
||||
void cv::gpu::subtract(const GpuMat& src, const Scalar& sc, GpuMat& dst)
|
||||
{
|
||||
typedef void (*caller_t)(const GpuMat& src, const Scalar& sc, GpuMat& dst);
|
||||
static const caller_t callers[] = {0, NppArithmScalar<1, nppiSubC_32f_C1R>::calc, NppArithmScalar<2, nppiSubC_32fc_C1R>::calc};
|
||||
|
||||
CV_Assert(src.type() == CV_32FC1 || src.type() == CV_32FC2);
|
||||
|
||||
callers[src.channels()](src, sc, dst);
|
||||
}
|
||||
|
||||
void cv::gpu::multiply(const GpuMat& src, const Scalar& sc, GpuMat& dst)
|
||||
{
|
||||
typedef void (*caller_t)(const GpuMat& src, const Scalar& sc, GpuMat& dst);
|
||||
static const caller_t callers[] = {0, NppArithmScalar<1, nppiMulC_32f_C1R>::calc, NppArithmScalar<2, nppiMulC_32fc_C1R>::calc};
|
||||
|
||||
CV_Assert(src.type() == CV_32FC1 || src.type() == CV_32FC2);
|
||||
|
||||
callers[src.channels()](src, sc, dst);
|
||||
}
|
||||
|
||||
void cv::gpu::divide(const GpuMat& src, const Scalar& sc, GpuMat& dst)
|
||||
{
|
||||
typedef void (*caller_t)(const GpuMat& src, const Scalar& sc, GpuMat& dst);
|
||||
static const caller_t callers[] = {0, NppArithmScalar<1, nppiDivC_32f_C1R>::calc, NppArithmScalar<2, nppiDivC_32fc_C1R>::calc};
|
||||
|
||||
CV_Assert(src.type() == CV_32FC1 || src.type() == CV_32FC2);
|
||||
|
||||
callers[src.channels()](src, sc, dst);
|
||||
}
|
||||
|
||||
////////////////////////////////////////////////////////////////////////
|
||||
// transpose
|
||||
|
||||
@@ -299,112 +118,6 @@ void cv::gpu::transpose(const GpuMat& src, GpuMat& dst)
|
||||
}
|
||||
}
|
||||
|
||||
////////////////////////////////////////////////////////////////////////
|
||||
// absdiff
|
||||
|
||||
void cv::gpu::absdiff(const GpuMat& src1, const GpuMat& src2, GpuMat& dst)
|
||||
{
|
||||
CV_DbgAssert(src1.size() == src2.size() && src1.type() == src2.type());
|
||||
|
||||
CV_Assert(src1.type() == CV_8UC1 || src1.type() == CV_8UC4 || src1.type() == CV_32SC1 || src1.type() == CV_32FC1);
|
||||
|
||||
dst.create( src1.size(), src1.type() );
|
||||
|
||||
NppiSize sz;
|
||||
sz.width = src1.cols;
|
||||
sz.height = src1.rows;
|
||||
|
||||
switch (src1.type())
|
||||
{
|
||||
case CV_8UC1:
|
||||
nppSafeCall( nppiAbsDiff_8u_C1R(src1.ptr<Npp8u>(), src1.step,
|
||||
src2.ptr<Npp8u>(), src2.step,
|
||||
dst.ptr<Npp8u>(), dst.step, sz) );
|
||||
break;
|
||||
case CV_8UC4:
|
||||
nppSafeCall( nppiAbsDiff_8u_C4R(src1.ptr<Npp8u>(), src1.step,
|
||||
src2.ptr<Npp8u>(), src2.step,
|
||||
dst.ptr<Npp8u>(), dst.step, sz) );
|
||||
break;
|
||||
case CV_32SC1:
|
||||
nppSafeCall( nppiAbsDiff_32s_C1R(src1.ptr<Npp32s>(), src1.step,
|
||||
src2.ptr<Npp32s>(), src2.step,
|
||||
dst.ptr<Npp32s>(), dst.step, sz) );
|
||||
break;
|
||||
case CV_32FC1:
|
||||
nppSafeCall( nppiAbsDiff_32f_C1R(src1.ptr<Npp32f>(), src1.step,
|
||||
src2.ptr<Npp32f>(), src2.step,
|
||||
dst.ptr<Npp32f>(), dst.step, sz) );
|
||||
break;
|
||||
default:
|
||||
CV_Assert(!"Unsupported source type");
|
||||
}
|
||||
}
|
||||
|
||||
void cv::gpu::absdiff(const GpuMat& src, const Scalar& s, GpuMat& dst)
|
||||
{
|
||||
CV_Assert(src.type() == CV_32FC1);
|
||||
|
||||
dst.create( src.size(), src.type() );
|
||||
|
||||
NppiSize sz;
|
||||
sz.width = src.cols;
|
||||
sz.height = src.rows;
|
||||
|
||||
nppSafeCall( nppiAbsDiffC_32f_C1R(src.ptr<Npp32f>(), src.step, dst.ptr<Npp32f>(), dst.step, sz, (Npp32f)s[0]) );
|
||||
}
|
||||
|
||||
////////////////////////////////////////////////////////////////////////
|
||||
// compare
|
||||
|
||||
namespace cv { namespace gpu { namespace mathfunc
|
||||
{
|
||||
void compare_ne_8uc4(const DevMem2D& src1, const DevMem2D& src2, const DevMem2D& dst);
|
||||
void compare_ne_32f(const DevMem2D& src1, const DevMem2D& src2, const DevMem2D& dst);
|
||||
}}}
|
||||
|
||||
void cv::gpu::compare(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, int cmpop)
|
||||
{
|
||||
CV_DbgAssert(src1.size() == src2.size() && src1.type() == src2.type());
|
||||
|
||||
CV_Assert(src1.type() == CV_8UC4 || src1.type() == CV_32FC1);
|
||||
|
||||
dst.create( src1.size(), CV_8UC1 );
|
||||
|
||||
static const NppCmpOp nppCmpOp[] = { NPP_CMP_EQ, NPP_CMP_GREATER, NPP_CMP_GREATER_EQ, NPP_CMP_LESS, NPP_CMP_LESS_EQ };
|
||||
|
||||
NppiSize sz;
|
||||
sz.width = src1.cols;
|
||||
sz.height = src1.rows;
|
||||
|
||||
if (src1.type() == CV_8UC4)
|
||||
{
|
||||
if (cmpop != CMP_NE)
|
||||
{
|
||||
nppSafeCall( nppiCompare_8u_C4R(src1.ptr<Npp8u>(), src1.step,
|
||||
src2.ptr<Npp8u>(), src2.step,
|
||||
dst.ptr<Npp8u>(), dst.step, sz, nppCmpOp[cmpop]) );
|
||||
}
|
||||
else
|
||||
{
|
||||
mathfunc::compare_ne_8uc4(src1, src2, dst);
|
||||
}
|
||||
}
|
||||
else
|
||||
{
|
||||
if (cmpop != CMP_NE)
|
||||
{
|
||||
nppSafeCall( nppiCompare_32f_C1R(src1.ptr<Npp32f>(), src1.step,
|
||||
src2.ptr<Npp32f>(), src2.step,
|
||||
dst.ptr<Npp8u>(), dst.step, sz, nppCmpOp[cmpop]) );
|
||||
}
|
||||
else
|
||||
{
|
||||
mathfunc::compare_ne_32f(src1, src2, dst);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
////////////////////////////////////////////////////////////////////////
|
||||
// meanStdDev
|
||||
|
||||
@@ -997,249 +710,6 @@ void cv::gpu::polarToCart(const GpuMat& magnitude, const GpuMat& angle, GpuMat&
|
||||
::polarToCart_caller(magnitude, angle, x, y, angleInDegrees, StreamAccessor::getStream(stream));
|
||||
}
|
||||
|
||||
//////////////////////////////////////////////////////////////////////////////
|
||||
// Per-element bit-wise logical matrix operations
|
||||
|
||||
namespace cv { namespace gpu { namespace mathfunc
|
||||
{
|
||||
void bitwise_not_caller(int rows, int cols, int elem_size1, int cn, const PtrStep src, PtrStep dst, cudaStream_t stream);
|
||||
|
||||
template <typename T>
|
||||
void bitwise_mask_not_caller(int rows, int cols, int cn, const PtrStep src, const PtrStep mask, PtrStep dst, cudaStream_t stream);
|
||||
|
||||
void bitwise_or_caller(int rows, int cols, int elem_size1, int cn, const PtrStep src1, const PtrStep src2, PtrStep dst, cudaStream_t stream);
|
||||
|
||||
template <typename T>
|
||||
void bitwise_mask_or_caller(int rows, int cols, int cn, const PtrStep src1, const PtrStep src2, const PtrStep mask, PtrStep dst, cudaStream_t stream);
|
||||
|
||||
void bitwise_and_caller(int rows, int cols, int elem_size1, int cn, const PtrStep src1, const PtrStep src2, PtrStep dst, cudaStream_t stream);
|
||||
|
||||
template <typename T>
|
||||
void bitwise_mask_and_caller(int rows, int cols, int cn, const PtrStep src1, const PtrStep src2, const PtrStep mask, PtrStep dst, cudaStream_t stream);
|
||||
|
||||
void bitwise_xor_caller(int rows, int cols, int elem_size1, int cn, const PtrStep src1, const PtrStep src2, PtrStep dst, cudaStream_t stream);
|
||||
|
||||
template <typename T>
|
||||
void bitwise_mask_xor_caller(int rows, int cols, int cn, const PtrStep src1, const PtrStep src2, const PtrStep mask, PtrStep dst, cudaStream_t stream);
|
||||
}}}
|
||||
|
||||
namespace
|
||||
{
|
||||
void bitwise_not_caller(const GpuMat& src, GpuMat& dst, cudaStream_t stream)
|
||||
{
|
||||
dst.create(src.size(), src.type());
|
||||
|
||||
cv::gpu::mathfunc::bitwise_not_caller(src.rows, src.cols, src.elemSize1(),
|
||||
dst.channels(), src, dst, stream);
|
||||
}
|
||||
|
||||
|
||||
void bitwise_not_caller(const GpuMat& src, GpuMat& dst, const GpuMat& mask, cudaStream_t stream)
|
||||
{
|
||||
using namespace cv::gpu;
|
||||
|
||||
typedef void (*Caller)(int, int, int, const PtrStep, const PtrStep, PtrStep, cudaStream_t);
|
||||
static Caller callers[] = {mathfunc::bitwise_mask_not_caller<unsigned char>, mathfunc::bitwise_mask_not_caller<unsigned char>,
|
||||
mathfunc::bitwise_mask_not_caller<unsigned short>, mathfunc::bitwise_mask_not_caller<unsigned short>,
|
||||
mathfunc::bitwise_mask_not_caller<unsigned int>, mathfunc::bitwise_mask_not_caller<unsigned int>,
|
||||
mathfunc::bitwise_mask_not_caller<unsigned int>};
|
||||
|
||||
CV_Assert(mask.type() == CV_8U && mask.size() == src.size());
|
||||
dst.create(src.size(), src.type());
|
||||
|
||||
Caller caller = callers[src.depth()];
|
||||
CV_Assert(caller);
|
||||
|
||||
int cn = src.depth() != CV_64F ? src.channels() : src.channels() * (sizeof(double) / sizeof(unsigned int));
|
||||
caller(src.rows, src.cols, cn, src, mask, dst, stream);
|
||||
}
|
||||
|
||||
|
||||
void bitwise_or_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());
|
||||
|
||||
cv::gpu::mathfunc::bitwise_or_caller(dst.rows, dst.cols, dst.elemSize1(),
|
||||
dst.channels(), src1, src2, dst, stream);
|
||||
}
|
||||
|
||||
|
||||
void bitwise_or_caller(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask, cudaStream_t stream)
|
||||
{
|
||||
using namespace cv::gpu;
|
||||
|
||||
typedef void (*Caller)(int, int, int, const PtrStep, const PtrStep, const PtrStep, PtrStep, cudaStream_t);
|
||||
static Caller callers[] = {mathfunc::bitwise_mask_or_caller<unsigned char>, mathfunc::bitwise_mask_or_caller<unsigned char>,
|
||||
mathfunc::bitwise_mask_or_caller<unsigned short>, mathfunc::bitwise_mask_or_caller<unsigned short>,
|
||||
mathfunc::bitwise_mask_or_caller<unsigned int>, mathfunc::bitwise_mask_or_caller<unsigned int>,
|
||||
mathfunc::bitwise_mask_or_caller<unsigned int>};
|
||||
|
||||
CV_Assert(src1.size() == src2.size() && src1.type() == src2.type());
|
||||
dst.create(src1.size(), src1.type());
|
||||
|
||||
Caller caller = callers[src1.depth()];
|
||||
CV_Assert(caller);
|
||||
|
||||
int cn = dst.depth() != CV_64F ? dst.channels() : dst.channels() * (sizeof(double) / sizeof(unsigned int));
|
||||
caller(dst.rows, dst.cols, cn, src1, src2, mask, dst, stream);
|
||||
}
|
||||
|
||||
|
||||
void bitwise_and_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());
|
||||
|
||||
cv::gpu::mathfunc::bitwise_and_caller(dst.rows, dst.cols, dst.elemSize1(),
|
||||
dst.channels(), src1, src2, dst, stream);
|
||||
}
|
||||
|
||||
|
||||
void bitwise_and_caller(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask, cudaStream_t stream)
|
||||
{
|
||||
using namespace cv::gpu;
|
||||
|
||||
typedef void (*Caller)(int, int, int, const PtrStep, const PtrStep, const PtrStep, PtrStep, cudaStream_t);
|
||||
static Caller callers[] = {mathfunc::bitwise_mask_and_caller<unsigned char>, mathfunc::bitwise_mask_and_caller<unsigned char>,
|
||||
mathfunc::bitwise_mask_and_caller<unsigned short>, mathfunc::bitwise_mask_and_caller<unsigned short>,
|
||||
mathfunc::bitwise_mask_and_caller<unsigned int>, mathfunc::bitwise_mask_and_caller<unsigned int>,
|
||||
mathfunc::bitwise_mask_and_caller<unsigned int>};
|
||||
|
||||
CV_Assert(src1.size() == src2.size() && src1.type() == src2.type());
|
||||
dst.create(src1.size(), src1.type());
|
||||
|
||||
Caller caller = callers[src1.depth()];
|
||||
CV_Assert(caller);
|
||||
|
||||
int cn = dst.depth() != CV_64F ? dst.channels() : dst.channels() * (sizeof(double) / sizeof(unsigned int));
|
||||
caller(dst.rows, dst.cols, cn, src1, src2, mask, dst, stream);
|
||||
}
|
||||
|
||||
|
||||
void bitwise_xor_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());
|
||||
|
||||
cv::gpu::mathfunc::bitwise_xor_caller(dst.rows, dst.cols, dst.elemSize1(),
|
||||
dst.channels(), src1, src2, dst, stream);
|
||||
}
|
||||
|
||||
|
||||
void bitwise_xor_caller(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask, cudaStream_t stream)
|
||||
{
|
||||
using namespace cv::gpu;
|
||||
|
||||
typedef void (*Caller)(int, int, int, const PtrStep, const PtrStep, const PtrStep, PtrStep, cudaStream_t);
|
||||
static Caller callers[] = {mathfunc::bitwise_mask_xor_caller<unsigned char>, mathfunc::bitwise_mask_xor_caller<unsigned char>,
|
||||
mathfunc::bitwise_mask_xor_caller<unsigned short>, mathfunc::bitwise_mask_xor_caller<unsigned short>,
|
||||
mathfunc::bitwise_mask_xor_caller<unsigned int>, mathfunc::bitwise_mask_xor_caller<unsigned int>,
|
||||
mathfunc::bitwise_mask_xor_caller<unsigned int>};
|
||||
|
||||
CV_Assert(src1.size() == src2.size() && src1.type() == src2.type());
|
||||
dst.create(src1.size(), src1.type());
|
||||
|
||||
Caller caller = callers[src1.depth()];
|
||||
CV_Assert(caller);
|
||||
|
||||
int cn = dst.depth() != CV_64F ? dst.channels() : dst.channels() * (sizeof(double) / sizeof(unsigned int));
|
||||
caller(dst.rows, dst.cols, cn, src1, src2, mask, dst, stream);
|
||||
}
|
||||
}
|
||||
|
||||
void cv::gpu::bitwise_not(const GpuMat& src, GpuMat& dst, const GpuMat& mask)
|
||||
{
|
||||
if (mask.empty())
|
||||
::bitwise_not_caller(src, dst, 0);
|
||||
else
|
||||
::bitwise_not_caller(src, dst, mask, 0);
|
||||
}
|
||||
|
||||
void cv::gpu::bitwise_not(const GpuMat& src, GpuMat& dst, const GpuMat& mask, const Stream& stream)
|
||||
{
|
||||
if (mask.empty())
|
||||
::bitwise_not_caller(src, dst, StreamAccessor::getStream(stream));
|
||||
else
|
||||
::bitwise_not_caller(src, dst, mask, StreamAccessor::getStream(stream));
|
||||
}
|
||||
|
||||
void cv::gpu::bitwise_or(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask)
|
||||
{
|
||||
if (mask.empty())
|
||||
::bitwise_or_caller(src1, src2, dst, 0);
|
||||
else
|
||||
::bitwise_or_caller(src1, src2, dst, mask, 0);
|
||||
}
|
||||
|
||||
void cv::gpu::bitwise_or(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask, const Stream& stream)
|
||||
{
|
||||
if (mask.empty())
|
||||
::bitwise_or_caller(src1, src2, dst, StreamAccessor::getStream(stream));
|
||||
else
|
||||
::bitwise_or_caller(src1, src2, dst, mask, StreamAccessor::getStream(stream));
|
||||
}
|
||||
|
||||
void cv::gpu::bitwise_and(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask)
|
||||
{
|
||||
if (mask.empty())
|
||||
::bitwise_and_caller(src1, src2, dst, 0);
|
||||
else
|
||||
::bitwise_and_caller(src1, src2, dst, mask, 0);
|
||||
}
|
||||
|
||||
void cv::gpu::bitwise_and(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask, const Stream& stream)
|
||||
{
|
||||
if (mask.empty())
|
||||
::bitwise_and_caller(src1, src2, dst, StreamAccessor::getStream(stream));
|
||||
else
|
||||
::bitwise_and_caller(src1, src2, dst, mask, StreamAccessor::getStream(stream));
|
||||
}
|
||||
|
||||
void cv::gpu::bitwise_xor(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask)
|
||||
{
|
||||
if (mask.empty())
|
||||
::bitwise_xor_caller(src1, src2, dst, 0);
|
||||
else
|
||||
::bitwise_xor_caller(src1, src2, dst, mask, 0);
|
||||
}
|
||||
|
||||
void cv::gpu::bitwise_xor(const GpuMat& src1, const GpuMat& src2, GpuMat& dst, const GpuMat& mask, const Stream& stream)
|
||||
{
|
||||
if (mask.empty())
|
||||
::bitwise_xor_caller(src1, src2, dst, StreamAccessor::getStream(stream));
|
||||
else
|
||||
::bitwise_xor_caller(src1, src2, dst, mask, StreamAccessor::getStream(stream));
|
||||
|
||||
}
|
||||
|
||||
cv::gpu::GpuMat cv::gpu::operator ~ (const GpuMat& src)
|
||||
{
|
||||
GpuMat dst;
|
||||
bitwise_not(src, dst);
|
||||
return dst;
|
||||
}
|
||||
|
||||
cv::gpu::GpuMat cv::gpu::operator | (const GpuMat& src1, const GpuMat& src2)
|
||||
{
|
||||
GpuMat dst;
|
||||
bitwise_or(src1, src2, dst);
|
||||
return dst;
|
||||
}
|
||||
|
||||
cv::gpu::GpuMat cv::gpu::operator & (const GpuMat& src1, const GpuMat& src2)
|
||||
{
|
||||
GpuMat dst;
|
||||
bitwise_and(src1, src2, dst);
|
||||
return dst;
|
||||
}
|
||||
|
||||
cv::gpu::GpuMat cv::gpu::operator ^ (const GpuMat& src1, const GpuMat& src2)
|
||||
{
|
||||
GpuMat dst;
|
||||
bitwise_xor(src1, src2, dst);
|
||||
return dst;
|
||||
}
|
||||
|
||||
//////////////////////////////////////////////////////////////////////////////
|
||||
// min/max
|
||||
|
||||
Reference in New Issue
Block a user