In macOS Big Sur, the Multiply and the Images sub-tests started to experience bi-modality in their scores especially on newer hardware.
This problem is due to a short warmup time. The initial score of a sub-test is crucial in deciding how much the complexity can grow. Because the system may be doing an expensive work only at the beginning of the sub-test (e.g. initializing a cache), drawing a simple object may take much longer than it usually does. MotionMark test harness will be mislead and believe that this drawing operation is always expensive and will not try to increase the complexity of the test much. This complexity is the score of the test.
Increasing the warm-up time will allow the system to finish its initialization before the test sampling starts. This should prevent the system initialization work from affecting the sub-test score.
Created attachment 416439 [details]
Committed r270959: <https://trac.webkit.org/changeset/270959>
All reviewed patches have been landed. Closing bug and clearing flags on attachment 416439 [details].