提交 6fa41c5a 编写于 作者: I Ilya Lavrenov

some experiments

上级 afa62c41
......@@ -3313,6 +3313,56 @@ static bool ocl_filter2D( InputArray _src, OutputArray _dst, int ddepth,
return kernel.run(2, globalsize, localsize, true);
}
template <typename T>
static std::string kerToStr(const Mat & k)
{
int width = k.cols - 1, depth = k.depth();
const T * const data = reinterpret_cast<const T *>(k.data);
std::ostringstream stream;
stream.precision(10);
if (depth <= CV_8S)
{
for (int i = 0; i < width; ++i)
stream << (int)data[i] << ",";
stream << (int)data[width];
}
else if (depth == CV_32F)
{
for (int i = 0; i < width; ++i)
stream << data[i] << "f,";
stream << data[width] << "f";
}
else
{
for (int i = 0; i < width; ++i)
stream << data[i] << ",";
}
return stream.str();
}
static String kernelToStr(InputArray _kernel, int ddepth = -1)
{
Mat kernel = _kernel.getMat().reshape(1, 1);
int depth = kernel.depth();
if (ddepth < 0)
ddepth = depth;
if (ddepth != depth)
kernel.convertTo(kernel, ddepth);
typedef std::string (*func_t)(const Mat &);
static const func_t funcs[] = { kerToStr<uchar>, kerToStr<char>, kerToStr<ushort>,kerToStr<short>,
kerToStr<int>, kerToStr<float>, kerToStr<double>, 0 };
const func_t func = funcs[depth];
CV_Assert(func != 0);
return cv::format(" -D COEFF=%s", func(kernel).c_str());
}
static bool ocl_sepRowFilter2D( UMat &src, UMat &buf, Mat &kernelX, int anchor, int borderType, bool sync)
{
int type = src.type();
......@@ -3378,6 +3428,7 @@ static bool ocl_sepRowFilter2D( UMat &src, UMat &buf, Mat &kernelX, int anchor,
btype,
extra_extrapolation ? "EXTRA_EXTRAPOLATION" : "NO_EXTRA_EXTRAPOLATION",
isIsolatedBorder ? "BORDER_ISOLATED" : "NO_BORDER_ISOLATED");
build_options += kernelToStr(kernelX, CV_32F);
Size srcWholeSize; Point srcOffset;
src.locateROI(srcWholeSize, srcOffset);
......@@ -3390,7 +3441,8 @@ static bool ocl_sepRowFilter2D( UMat &src, UMat &buf, Mat &kernelX, int anchor,
strKernel << "_D" << sdepth;
ocl::Kernel kernelRow;
if (!kernelRow.create(strKernel.str().c_str(), cv::ocl::imgproc::filterSepRow_oclsrc, build_options))
if (!kernelRow.create(strKernel.str().c_str(), cv::ocl::imgproc::filterSepRow_oclsrc,
build_options))
return false;
int idxArg = 0;
......@@ -3409,7 +3461,7 @@ static bool ocl_sepRowFilter2D( UMat &src, UMat &buf, Mat &kernelX, int anchor,
idxArg = kernelRow.set(idxArg, buf.cols);
idxArg = kernelRow.set(idxArg, buf.rows);
idxArg = kernelRow.set(idxArg, radiusY);
idxArg = kernelRow.set(idxArg, ocl::KernelArg::PtrReadOnly(kernelX.getUMat(ACCESS_READ)));
// idxArg = kernelRow.set(idxArg, ocl::KernelArg::PtrReadOnly(kernelX.getUMat(ACCESS_READ)));
return kernelRow.run(2, globalsize, localsize, sync);
}
......@@ -3479,6 +3531,8 @@ static bool ocl_sepColFilter2D(UMat &buf, UMat &dst, Mat &kernelY, int anchor, b
}
}
build_options += kernelToStr(kernelY, CV_32F);
ocl::Kernel kernelCol;
if (!kernelCol.create("col_filter", cv::ocl::imgproc::filterSepCol_oclsrc, build_options))
return false;
......@@ -3494,7 +3548,7 @@ static bool ocl_sepColFilter2D(UMat &buf, UMat &dst, Mat &kernelY, int anchor, b
idxArg = kernelCol.set(idxArg, (int)(dst.step / dst.elemSize()));
idxArg = kernelCol.set(idxArg, dst.cols);
idxArg = kernelCol.set(idxArg, dst.rows);
idxArg = kernelCol.set(idxArg, ocl::KernelArg::PtrReadOnly(kernelY.getUMat(ACCESS_READ)));
// idxArg = kernelCol.set(idxArg, ocl::KernelArg::PtrReadOnly(kernelY.getUMat(ACCESS_READ)));
return kernelCol.run(2, globalsize, localsize, sync);
}
......@@ -3508,7 +3562,7 @@ static bool ocl_sepFilter2D( InputArray _src, OutputArray _dst, int ddepth,
int type = _src.type();
if ( !( (CV_8UC1 == type || CV_8UC4 == type || CV_32FC1 == type || CV_32FC4 == type) &&
(ddepth == CV_32F || ddepth == CV_8U) ) )
(ddepth == CV_32F || ddepth == CV_8U || ddepth < 0) ) )
return false;
int cn = CV_MAT_CN(type);
......
......@@ -60,6 +60,7 @@ Niko
The info above maybe obsolete.
***********************************************************************************/
__constant float mat_kernel[] = { COEFF };
__kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void col_filter
(__global const GENTYPE_SRC * restrict src,
......@@ -70,8 +71,7 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void col_filter
const int dst_offset_in_pixel,
const int dst_step_in_pixel,
const int dst_cols,
const int dst_rows,
__constant float * mat_kernel)
const int dst_rows)
{
int x = get_global_id(0);
int y = get_global_id(1);
......
......@@ -144,6 +144,8 @@ Niko
The info above maybe obsolete.
***********************************************************************************/
__constant float mat_kernel[] = { COEFF };
__kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_C1_D0
(__global uchar * restrict src,
int src_step_in_pixel,
......@@ -153,8 +155,7 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
__global float * dst,
int dst_step_in_pixel,
int dst_cols, int dst_rows,
int radiusy,
__constant float * mat_kernel)
int radiusy)
{
int x = get_global_id(0)<<2;
int y = get_global_id(1);
......@@ -297,8 +298,7 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
__global float4 * dst,
int dst_step_in_pixel,
int dst_cols, int dst_rows,
int radiusy,
__constant float * mat_kernel)
int radiusy)
{
int x = get_global_id(0);
int y = get_global_id(1);
......@@ -391,8 +391,7 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
__global float * dst,
int dst_step_in_pixel,
int dst_cols, int dst_rows,
int radiusy,
__constant float * mat_kernel)
int radiusy)
{
int x = get_global_id(0);
int y = get_global_id(1);
......@@ -484,8 +483,7 @@ __kernel __attribute__((reqd_work_group_size(LSIZE0,LSIZE1,1))) void row_filter_
__global float4 * dst,
int dst_step_in_pixel,
int dst_cols, int dst_rows,
int radiusy,
__constant float * mat_kernel)
int radiusy)
{
int x = get_global_id(0);
int y = get_global_id(1);
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册