different paddings in cvtColorTwoPlane() for biplane YUV420#19554
Conversation
|
@amirtu There is failed test in CI log: |
| { | ||
| CV_INSTRUMENT_REGION(); | ||
|
|
||
| const uchar* uv = src_data + src_step * static_cast<size_t>(dst_height); |
There was a problem hiding this comment.
Is the pointer arithmetics being done under dispatched branch of code by purpose?
0c861d8 to
e6f2066
Compare
|
jenkins cn please retry a build |
e6f2066 to
e84fb22
Compare
|
@alalek I fixed build issue for dispatch case and rebased the branch to current 3.4. Please take a look and merge the PR. |
| CALL_HAL(cvtTwoPlaneYUVtoBGREx, cv_hal_cvtTwoPlaneYUVtoBGREx, | ||
| y_data, y_step, uv_data, uv_step, dst_data, dst_step, dst_width, dst_height, dcn, swapBlue, uIdx); | ||
|
|
||
| CV_CPU_DISPATCH(cvtTwoPlaneYUVtoBGR, (y_data, y_step, uv_data, uv_step, dst_data, dst_step, dst_width, dst_height, dcn, swapBlue, uIdx), |
There was a problem hiding this comment.
We should have fallback on HAL code specialized for y_step == uv_step case.
Otherwise this patch introduces performance regressions (to see that, drop changes from the HAL implementations).
if (y_step == uv_step)
CALL_HAL(cvtTwoPlaneYUVtoBGR, cv_hal_cvtTwoPlaneYUVtoBGR, ...)
See discussion near removed CV_CheckEQ() for details (which is not resolved).
fixes #17036
Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
Patch to opencv_extra has the same branch name.