Commit b4cd3a68b27af91620c3a519bb8ff302e3bc1e16
2024-06-07T13:53:01
Perf tests: trials 4->3, samples 6->3. Currently we're running each test for 6*4*10 seconds, which is a LOT. This CL reduces to 3*3*10, which is still a lot, but less so :) I think the reasoning for getting more samples was to reduce variability, but 1) thermals need to be taken into account and 2) we shouldn't be spending that much time per test. 9 data points results in 1 outlier dropped on each side: """ Truncation: Removing the 1 highest and lowest times from wall_times. """ Bug: b/345514719 Change-Id: I15b31c9d4ff94965e1bf7932571f97cd392f2006 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/5606164 Reviewed-by: Cody Northrop <cnorthrop@google.com>