You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Previously, a single 't_context_t' instance was used for all statistical
tests, limiting the ability to assess multiple cropping thresholds
independently.
This change allows separate t-tests for different percentiles of
execution times, ensuring more robust statistical analysis.
To achieve this, the following changes were made:
- Replaced the single 't_context_t' instance with an array 'ctxs[]' to
track multiple contexts.
- Introduced 'DUDECT_TESTS' to define the total number of tests,
including the baseline.
- Updated 'update_statistics()' to push values into the appropriate
context based on percentile filtering.
- Implemented 'max_test()' to determine the test context with the
maximum t-value.
- Adjusted memory allocation and cleanup for multiple contexts.
This enhances the ability to detect timing differences at various
thresholds, improving test reliability.
Change-Id: Icfad85854765316dece6d1303c5e13ad1e8ab8e9
0 commit comments