mirror/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2025-06-12 20:42:53 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

ocl: validate arguments in KernelArgs constructor

Browse Source 
- don't use undefined flag=0. It should be CONSTANT instead.
- don't allow 'UMat* m=NULL' argument (except LOCAL/CONSTANT flags).
  This case is not handled well to provide NULL __global pointers.
  It is better to use '-D' macro defines instead (at least for performance)
This commit is contained in:
Alexander Alekhin 2017-02-14 14:58:52 +03:00
parent 47ae5f14f5
commit 4c7aa8645a
4 changed files with 12 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
modules/core/src/arithm.cpp
View File

@ -138,7 +138,7 @@ static bool ocl_binary_op(InputArray _src1, InputArray _src2, OutputArray _dst,
convertAndUnrollScalar(src2sc, srctype, (uchar*)buf, 1);
}
ocl::KernelArg scalararg = ocl::KernelArg(0, 0, 0, 0, buf, esz);
ocl::KernelArg scalararg = ocl::KernelArg(ocl::KernelArg::CONSTANT, 0, 0, 0, buf, esz);
if( !haveMask )
k.args(src1arg, dstarg, scalararg);
@ -550,7 +550,7 @@ static bool ocl_arithm_op(InputArray _src1, InputArray _src2, OutputArray _dst,
if( !src2sc.empty() )
convertAndUnrollScalar(src2sc, wtype, (uchar*)buf, 1);
ocl::KernelArg scalararg = ocl::KernelArg(0, 0, 0, 0, buf, esz);
ocl::KernelArg scalararg = ocl::KernelArg(ocl::KernelArg::CONSTANT, 0, 0, 0, buf, esz);
if( !haveMask )
{
@ -558,7 +558,7 @@ static bool ocl_arithm_op(InputArray _src1, InputArray _src2, OutputArray _dst,
k.args(src1arg, dstarg, scalararg);
else if(n == 1)
k.args(src1arg, dstarg, scalararg,
ocl::KernelArg(0, 0, 0, 0, usrdata_p, usrdata_esz));
ocl::KernelArg(ocl::KernelArg::CONSTANT, 0, 0, 0, usrdata_p, usrdata_esz));
else
CV_Error(Error::StsNotImplemented, "unsupported number of extra parameters");
}
@ -576,12 +576,12 @@ static bool ocl_arithm_op(InputArray _src1, InputArray _src2, OutputArray _dst,
k.args(src1arg, src2arg, dstarg);
else if (n == 1)
k.args(src1arg, src2arg, dstarg,
ocl::KernelArg(0, 0, 0, 0, usrdata_p, usrdata_esz));
ocl::KernelArg(ocl::KernelArg::CONSTANT, 0, 0, 0, usrdata_p, usrdata_esz));
else if (n == 3)
k.args(src1arg, src2arg, dstarg,
ocl::KernelArg(0, 0, 0, 0, usrdata_p, usrdata_esz),
ocl::KernelArg(0, 0, 0, 0, usrdata_p + usrdata_esz, usrdata_esz),
ocl::KernelArg(0, 0, 0, 0, usrdata_p + usrdata_esz*2, usrdata_esz));
ocl::KernelArg(ocl::KernelArg::CONSTANT, 0, 0, 0, usrdata_p, usrdata_esz),
ocl::KernelArg(ocl::KernelArg::CONSTANT, 0, 0, 0, usrdata_p + usrdata_esz, usrdata_esz),
ocl::KernelArg(ocl::KernelArg::CONSTANT, 0, 0, 0, usrdata_p + usrdata_esz*2, usrdata_esz));
else
CV_Error(Error::StsNotImplemented, "unsupported number of extra parameters");
}
@ -1204,7 +1204,7 @@ static bool ocl_compare(InputArray _src1, InputArray _src2, OutputArray _dst, in
convertAndUnrollScalar(Mat(1, 1, CV_32S, &ival), depth1, (uchar *)buf, kercn);
}
ocl::KernelArg scalararg = ocl::KernelArg(0, 0, 0, 0, buf, esz);
ocl::KernelArg scalararg = ocl::KernelArg(ocl::KernelArg::CONSTANT, 0, 0, 0, buf, esz);
k.args(ocl::KernelArg::ReadOnlyNoSize(src1, cn, kercn),
ocl::KernelArg::WriteOnly(dst, cn, kercn), scalararg);

1
modules/core/src/ocl.cpp
View File

@ -3155,6 +3155,7 @@ KernelArg::KernelArg()
KernelArg::KernelArg(int _flags, UMat* _m, int _wscale, int _iwscale, const void* _obj, size_t _sz)
: flags(_flags), m(_m), obj(_obj), sz(_sz), wscale(_wscale), iwscale(_iwscale)
{
CV_Assert(_flags == LOCAL || _flags == CONSTANT || _m != NULL);
}
KernelArg KernelArg::Constant(const Mat& m)

2
modules/core/src/umatrix.cpp
View File

@ -986,7 +986,7 @@ UMat& UMat::setTo(InputArray _value, InputArray _mask)
ocl::Kernel setK(haveMask ? "setMask" : "set", ocl::core::copyset_oclsrc, opts);
if( !setK.empty() )
{
ocl::KernelArg scalararg(0, 0, 0, 0, buf, CV_ELEM_SIZE(d) * scalarcn);
ocl::KernelArg scalararg(ocl::KernelArg::CONSTANT, 0, 0, 0, buf, CV_ELEM_SIZE(d) * scalarcn);
UMat mask;
if( haveMask )

4
modules/imgproc/src/imgwarp.cpp
View File

@ -5814,7 +5814,7 @@ static bool ocl_warpTransform_cols4(InputArray _src, OutputArray _dst, InputArra
matM.convertTo(M0, CV_32F);
k.args(ocl::KernelArg::ReadOnly(src), ocl::KernelArg::WriteOnly(dst), ocl::KernelArg::PtrReadOnly(M0),
ocl::KernelArg(0, 0, 0, 0, borderBuf, CV_ELEM_SIZE(sctype)));
ocl::KernelArg(ocl::KernelArg::CONSTANT, 0, 0, 0, borderBuf, CV_ELEM_SIZE(sctype)));
size_t globalThreads[2];
globalThreads[0] = (size_t)(dst.cols / 4);
@ -5913,7 +5913,7 @@ static bool ocl_warpTransform(InputArray _src, OutputArray _dst, InputArray _M0,
matM.convertTo(M0, doubleSupport ? CV_64F : CV_32F);
k.args(ocl::KernelArg::ReadOnly(src), ocl::KernelArg::WriteOnly(dst), ocl::KernelArg::PtrReadOnly(M0),
ocl::KernelArg(0, 0, 0, 0, borderBuf, CV_ELEM_SIZE(sctype)));
ocl::KernelArg(ocl::KernelArg::CONSTANT, 0, 0, 0, borderBuf, CV_ELEM_SIZE(sctype)));
size_t globalThreads[2] = { (size_t)dst.cols, ((size_t)dst.rows + rowsPerWI - 1) / rowsPerWI };
return k.run(2, globalThreads, NULL, false);