proposal for OCL_* performance tests for new T-API functions (Feature #3651)
| Status: | Open | Start date: | 2014-04-15 |
|---|---|---|---|
| Priority: | Normal | Due date: | |
| Assignee: |  Ilya Lavrenov |
% Done: | 0% |
| Category: | ocl | ||
| Target version: | 3.0 | ||
| Difficulty: | Medium | Pull request: |
Description
performance tests (OCL_*) for new T-API functions does not report which version (OCL or native) was run. Currently OCL* test could generate native performance number without any warning.
proposal:
1. Add some output for tests that will indicate which version (OCL or NATIVE) was really tested.
2. Implement some engineering mode that will FORCE OCL version execution. In this mode OCL version have to be run without any condition or failed. if OCL version failed then test has to report this fail but not to run native version instead.
This will allow to monitor OCL performance and compare it with native code.
This will allow to make real validation and performance testing for commits with modified OCL code.
Associated revisions
Merge pull request #3651 from mshabunin:shape-test
History
Updated by Evgeny Talanin almost 11 years ago
Andrey, could you comment on this?
- Assignee set to Andrey Pavlenko
Updated by Vadim Pisarevsky almost 11 years ago
reassigned to Ilya
- Difficulty set to Medium
- Assignee changed from Andrey Pavlenko to Ilya Lavrenov
- Target version set to 3.0
- Category set to ocl
Updated by Dmitry Retinskiy almost 11 years ago
- Status changed from New to Open
Updated by Maksim Shabunin over 9 years ago
Issue has been transferred to GitHub: https://github.com/Itseez/opencv/issues/4730