[base] Experiment with PM_NOYIELD in Win's MessagePumpForUI
Jank analysis uncovered that Windows uses native ::PeekMessage calls as an opportunity to yield to other threads according to some heuristics (e.g. presumably when there's no input but perhaps a single WM_USER message posted later than another thread was readied). MessagePumpForUI doesn't intend to give this opportunity to the kernel when invoking ::PeekMessage however as it runs most tasks out-of-band. Hence, PM_NOYIELD should be used to tell ::PeekMessage it's not the only source of work for this thread. R=fdoray@chromium.org Bug: 1075960 Change-Id: Ic8808cab8c15aa690f7dfd8405741323285c107b Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2169728 Auto-Submit: Gabriel Charette <gab@chromium.org> Commit-Queue: François Doray <fdoray@chromium.org> Reviewed-by:François Doray <fdoray@chromium.org> Cr-Commit-Position: refs/heads/master@{#763349}
Showing
Please register or sign in to comment