Merge pull request #3547 from ilya-lavrenov:ocl_setto

This commit is contained in:
Vadim Pisarevsky 2014-12-31 09:30:40 +00:00
commit 2f6db4dfac

View File

@ -771,25 +771,29 @@ UMat& UMat::setTo(InputArray _value, InputArray _mask)
{ {
bool haveMask = !_mask.empty(); bool haveMask = !_mask.empty();
#ifdef HAVE_OPENCL #ifdef HAVE_OPENCL
int tp = type(), cn = CV_MAT_CN(tp); int tp = type(), cn = CV_MAT_CN(tp), depth = CV_MAT_DEPTH(tp);
if( dims <= 2 && cn <= 4 && CV_MAT_DEPTH(tp) < CV_64F && ocl::useOpenCL() ) if( dims <= 2 && cn <= 4 && CV_MAT_DEPTH(tp) < CV_64F && ocl::useOpenCL() )
{ {
Mat value = _value.getMat(); Mat value = _value.getMat();
CV_Assert( checkScalar(value, type(), _value.kind(), _InputArray::UMAT) ); CV_Assert( checkScalar(value, type(), _value.kind(), _InputArray::UMAT) );
double buf[4] = { 0, 0, 0, 0 }; int kercn = haveMask || cn == 3 ? cn : std::max(cn, ocl::predictOptimalVectorWidth(*this)),
convertAndUnrollScalar(value, tp, (uchar *)buf, 1); kertp = CV_MAKE_TYPE(depth, kercn);
int scalarcn = cn == 3 ? 4 : cn, rowsPerWI = ocl::Device::getDefault().isIntel() ? 4 : 1; double buf[16] = { 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0 };
convertAndUnrollScalar(value, tp, (uchar *)buf, kercn / cn);
int scalarcn = kercn == 3 ? 4 : kercn, rowsPerWI = ocl::Device::getDefault().isIntel() ? 4 : 1;
String opts = format("-D dstT=%s -D rowsPerWI=%d -D dstST=%s -D dstT1=%s -D cn=%d", String opts = format("-D dstT=%s -D rowsPerWI=%d -D dstST=%s -D dstT1=%s -D cn=%d",
ocl::memopTypeToStr(tp), rowsPerWI, ocl::memopTypeToStr(kertp), rowsPerWI,
ocl::memopTypeToStr(CV_MAKETYPE(tp, scalarcn)), ocl::memopTypeToStr(CV_MAKETYPE(depth, scalarcn)),
ocl::memopTypeToStr(CV_MAT_DEPTH(tp)), cn); ocl::memopTypeToStr(depth), kercn);
ocl::Kernel setK(haveMask ? "setMask" : "set", ocl::core::copyset_oclsrc, opts); ocl::Kernel setK(haveMask ? "setMask" : "set", ocl::core::copyset_oclsrc, opts);
if( !setK.empty() ) if( !setK.empty() )
{ {
ocl::KernelArg scalararg(0, 0, 0, 0, buf, CV_ELEM_SIZE1(tp) * scalarcn); ocl::KernelArg scalararg(0, 0, 0, 0, buf, CV_ELEM_SIZE(depth) * scalarcn);
UMat mask; UMat mask;
if( haveMask ) if( haveMask )
@ -802,11 +806,11 @@ UMat& UMat::setTo(InputArray _value, InputArray _mask)
} }
else else
{ {
ocl::KernelArg dstarg = ocl::KernelArg::WriteOnly(*this); ocl::KernelArg dstarg = ocl::KernelArg::WriteOnly(*this, cn, kercn);
setK.args(dstarg, scalararg); setK.args(dstarg, scalararg);
} }
size_t globalsize[] = { cols, (rows + rowsPerWI - 1) / rowsPerWI }; size_t globalsize[] = { cols * cn / kercn, (rows + rowsPerWI - 1) / rowsPerWI };
if( setK.run(2, globalsize, NULL, false) ) if( setK.run(2, globalsize, NULL, false) )
{ {
CV_IMPL_ADD(CV_IMPL_OCL); CV_IMPL_ADD(CV_IMPL_OCL);