potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-29 15:23:05 +04:00
Code Packages Releases Wiki Activity

Replaced most of the usages of parallel_for with that of parallel_for_.

Browse Source 
This should allow many algorithms to take advantage of more parallelization
technologies.
This commit is contained in:
Roman Donchenko
2013-05-30 18:44:33 +04:00
parent 37091b086c
commit 29b13ec1de
24 changed files with 232 additions and 352 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/imgproc/src/color.cpp
+42 -54
View File
@@ -2755,7 +2755,7 @@ const int ITUR_BT_601_CGV = -385875;
const int ITUR_BT_601_CBV = -74448;
template<int bIdx, int uIdx>
struct YUV420sp2RGB888Invoker
struct YUV420sp2RGB888Invoker : ParallelLoopBody
{
Mat* dst;
const uchar* my1, *muv;
@@ -2764,10 +2764,10 @@ struct YUV420sp2RGB888Invoker
YUV420sp2RGB888Invoker(Mat* _dst, int _stride, const uchar* _y1, const uchar* _uv)
: dst(_dst), my1(_y1), muv(_uv), width(_dst->cols), stride(_stride) {}
void operator()(const BlockedRange& range) const
void operator()(const Range& range) const
{
int rangeBegin = range.begin() * 2;
int rangeEnd = range.end() * 2;
int rangeBegin = range.start * 2;
int rangeEnd = range.end * 2;
//R = 1.164(Y - 16) + 1.596(V - 128)
//G = 1.164(Y - 16) - 0.813(V - 128) - 0.391(U - 128)
@@ -2824,7 +2824,7 @@ struct YUV420sp2RGB888Invoker
};
template<int bIdx, int uIdx>
struct YUV420sp2RGBA8888Invoker
struct YUV420sp2RGBA8888Invoker : ParallelLoopBody
{
Mat* dst;
const uchar* my1, *muv;
@@ -2833,10 +2833,10 @@ struct YUV420sp2RGBA8888Invoker
YUV420sp2RGBA8888Invoker(Mat* _dst, int _stride, const uchar* _y1, const uchar* _uv)
: dst(_dst), my1(_y1), muv(_uv), width(_dst->cols), stride(_stride) {}
void operator()(const BlockedRange& range) const
void operator()(const Range& range) const
{
int rangeBegin = range.begin() * 2;
int rangeEnd = range.end() * 2;
int rangeBegin = range.start * 2;
int rangeEnd = range.end * 2;
//R = 1.164(Y - 16) + 1.596(V - 128)
//G = 1.164(Y - 16) - 0.813(V - 128) - 0.391(U - 128)
@@ -2897,7 +2897,7 @@ struct YUV420sp2RGBA8888Invoker
};
template<int bIdx>
struct YUV420p2RGB888Invoker
struct YUV420p2RGB888Invoker : ParallelLoopBody
{
Mat* dst;
const uchar* my1, *mu, *mv;
@@ -2907,19 +2907,19 @@ struct YUV420p2RGB888Invoker
YUV420p2RGB888Invoker(Mat* _dst, int _stride, const uchar* _y1, const uchar* _u, const uchar* _v, int _ustepIdx, int _vstepIdx)
: dst(_dst), my1(_y1), mu(_u), mv(_v), width(_dst->cols), stride(_stride), ustepIdx(_ustepIdx), vstepIdx(_vstepIdx) {}
void operator()(const BlockedRange& range) const
void operator()(const Range& range) const
{
const int rangeBegin = range.begin() * 2;
const int rangeEnd = range.end() * 2;
const int rangeBegin = range.start * 2;
const int rangeEnd = range.end * 2;
size_t uvsteps[2] = {width/2, stride - width/2};
int usIdx = ustepIdx, vsIdx = vstepIdx;
const uchar* y1 = my1 + rangeBegin * stride;
const uchar* u1 = mu + (range.begin() / 2) * stride;
const uchar* v1 = mv + (range.begin() / 2) * stride;
const uchar* u1 = mu + (range.start / 2) * stride;
const uchar* v1 = mv + (range.start / 2) * stride;
if(range.begin() % 2 == 1)
if(range.start % 2 == 1)
{
u1 += uvsteps[(usIdx++) & 1];
v1 += uvsteps[(vsIdx++) & 1];
@@ -2965,7 +2965,7 @@ struct YUV420p2RGB888Invoker
};
template<int bIdx>
struct YUV420p2RGBA8888Invoker
struct YUV420p2RGBA8888Invoker : ParallelLoopBody
{
Mat* dst;
const uchar* my1, *mu, *mv;
@@ -2975,19 +2975,19 @@ struct YUV420p2RGBA8888Invoker
YUV420p2RGBA8888Invoker(Mat* _dst, int _stride, const uchar* _y1, const uchar* _u, const uchar* _v, int _ustepIdx, int _vstepIdx)
: dst(_dst), my1(_y1), mu(_u), mv(_v), width(_dst->cols), stride(_stride), ustepIdx(_ustepIdx), vstepIdx(_vstepIdx) {}
void operator()(const BlockedRange& range) const
void operator()(const Range& range) const
{
int rangeBegin = range.begin() * 2;
int rangeEnd = range.end() * 2;
int rangeBegin = range.start * 2;
int rangeEnd = range.end * 2;
size_t uvsteps[2] = {width/2, stride - width/2};
int usIdx = ustepIdx, vsIdx = vstepIdx;
const uchar* y1 = my1 + rangeBegin * stride;
const uchar* u1 = mu + (range.begin() / 2) * stride;
const uchar* v1 = mv + (range.begin() / 2) * stride;
const uchar* u1 = mu + (range.start / 2) * stride;
const uchar* v1 = mv + (range.start / 2) * stride;
if(range.begin() % 2 == 1)
if(range.start % 2 == 1)
{
u1 += uvsteps[(usIdx++) & 1];
v1 += uvsteps[(vsIdx++) & 1];
@@ -3042,48 +3042,40 @@ template<int bIdx, int uIdx>
inline void cvtYUV420sp2RGB(Mat& _dst, int _stride, const uchar* _y1, const uchar* _uv)
{
YUV420sp2RGB888Invoker<bIdx, uIdx> converter(&_dst, _stride, _y1, _uv);
#ifdef HAVE_TBB
if (_dst.total() >= MIN_SIZE_FOR_PARALLEL_YUV420_CONVERSION)
parallel_for(BlockedRange(0, _dst.rows/2), converter);
parallel_for_(Range(0, _dst.rows/2), converter);
else
#endif
converter(BlockedRange(0, _dst.rows/2));
converter(Range(0, _dst.rows/2));
}
template<int bIdx, int uIdx>
inline void cvtYUV420sp2RGBA(Mat& _dst, int _stride, const uchar* _y1, const uchar* _uv)
{
YUV420sp2RGBA8888Invoker<bIdx, uIdx> converter(&_dst, _stride, _y1, _uv);
#ifdef HAVE_TBB
if (_dst.total() >= MIN_SIZE_FOR_PARALLEL_YUV420_CONVERSION)
parallel_for(BlockedRange(0, _dst.rows/2), converter);
parallel_for_(Range(0, _dst.rows/2), converter);
else
#endif
converter(BlockedRange(0, _dst.rows/2));
converter(Range(0, _dst.rows/2));
}
template<int bIdx>
inline void cvtYUV420p2RGB(Mat& _dst, int _stride, const uchar* _y1, const uchar* _u, const uchar* _v, int ustepIdx, int vstepIdx)
{
YUV420p2RGB888Invoker<bIdx> converter(&_dst, _stride, _y1, _u, _v, ustepIdx, vstepIdx);
#ifdef HAVE_TBB
if (_dst.total() >= MIN_SIZE_FOR_PARALLEL_YUV420_CONVERSION)
parallel_for(BlockedRange(0, _dst.rows/2), converter);
parallel_for_(Range(0, _dst.rows/2), converter);
else
#endif
converter(BlockedRange(0, _dst.rows/2));
converter(Range(0, _dst.rows/2));
}
template<int bIdx>
inline void cvtYUV420p2RGBA(Mat& _dst, int _stride, const uchar* _y1, const uchar* _u, const uchar* _v, int ustepIdx, int vstepIdx)
{
YUV420p2RGBA8888Invoker<bIdx> converter(&_dst, _stride, _y1, _u, _v, ustepIdx, vstepIdx);
#ifdef HAVE_TBB
if (_dst.total() >= MIN_SIZE_FOR_PARALLEL_YUV420_CONVERSION)
parallel_for(BlockedRange(0, _dst.rows/2), converter);
parallel_for_(Range(0, _dst.rows/2), converter);
else
#endif
converter(BlockedRange(0, _dst.rows/2));
converter(Range(0, _dst.rows/2));
}
///////////////////////////////////// RGB -> YUV420p /////////////////////////////////////
@@ -3167,7 +3159,7 @@ static void cvtRGBtoYUV420p(const Mat& src, Mat& dst)
///////////////////////////////////// YUV422 -> RGB /////////////////////////////////////
template<int bIdx, int uIdx, int yIdx>
struct YUV422toRGB888Invoker
struct YUV422toRGB888Invoker : ParallelLoopBody
{
Mat* dst;
const uchar* src;
@@ -3176,10 +3168,10 @@ struct YUV422toRGB888Invoker
YUV422toRGB888Invoker(Mat* _dst, int _stride, const uchar* _yuv)
: dst(_dst), src(_yuv), width(_dst->cols), stride(_stride) {}
void operator()(const BlockedRange& range) const
void operator()(const Range& range) const
{
int rangeBegin = range.begin();
int rangeEnd = range.end();
int rangeBegin = range.start;
int rangeEnd = range.end;
const int uidx = 1 - yIdx + uIdx * 2;
const int vidx = (2 + uidx) % 4;
@@ -3213,7 +3205,7 @@ struct YUV422toRGB888Invoker
};
template<int bIdx, int uIdx, int yIdx>
struct YUV422toRGBA8888Invoker
struct YUV422toRGBA8888Invoker : ParallelLoopBody
{
Mat* dst;
const uchar* src;
@@ -3222,10 +3214,10 @@ struct YUV422toRGBA8888Invoker
YUV422toRGBA8888Invoker(Mat* _dst, int _stride, const uchar* _yuv)
: dst(_dst), src(_yuv), width(_dst->cols), stride(_stride) {}
void operator()(const BlockedRange& range) const
void operator()(const Range& range) const
{
int rangeBegin = range.begin();
int rangeEnd = range.end();
int rangeBegin = range.start;
int rangeEnd = range.end;
const int uidx = 1 - yIdx + uIdx * 2;
const int vidx = (2 + uidx) % 4;
@@ -3266,24 +3258,20 @@ template<int bIdx, int uIdx, int yIdx>
inline void cvtYUV422toRGB(Mat& _dst, int _stride, const uchar* _yuv)
{
YUV422toRGB888Invoker<bIdx, uIdx, yIdx> converter(&_dst, _stride, _yuv);
#ifdef HAVE_TBB
if (_dst.total() >= MIN_SIZE_FOR_PARALLEL_YUV422_CONVERSION)
parallel_for(BlockedRange(0, _dst.rows), converter);
parallel_for_(Range(0, _dst.rows), converter);
else
#endif
converter(BlockedRange(0, _dst.rows));
converter(Range(0, _dst.rows));
}
template<int bIdx, int uIdx, int yIdx>
inline void cvtYUV422toRGBA(Mat& _dst, int _stride, const uchar* _yuv)
{
YUV422toRGBA8888Invoker<bIdx, uIdx, yIdx> converter(&_dst, _stride, _yuv);
#ifdef HAVE_TBB
if (_dst.total() >= MIN_SIZE_FOR_PARALLEL_YUV422_CONVERSION)
parallel_for(BlockedRange(0, _dst.rows), converter);
parallel_for_(Range(0, _dst.rows), converter);
else
#endif
converter(BlockedRange(0, _dst.rows));
converter(Range(0, _dst.rows));
}
/////////////////////////// RGBA <-> mRGBA (alpha premultiplied) //////////////