5-10% regression on cc_perftests.calc_draw_props_time at 434322:434360 |
|||
Issue descriptiontouch_region_heavy and 10_10 regressed by 5-10% on the linux and android bots. e.g: https://chromeperf.appspot.com/report?sid=5c8dfe5df9bfbd391408adb714f362cb5a782a0e0801b896d90e2af06b144843 https://chromeperf.appspot.com/report?sid=72b63aa5d419c5621984c95285242f4853119a51e4849b51f32161a931c3be59 Bisect builds failed on the "Compare samples" step, but I locally bisected to r434349 (https://codereview.chromium.org/2517273002). The hash lookups added by that CL seem to be what's causing the regression.
,
Dec 16 2016
I'm confused at the state of this bug. The revert is the first thing on here but it's reverting a CL that points at this bug... the original CL seems to point here but its not shown in the bug comments. Can you clarify? This is marked P1 but has been idle some time.
,
Dec 16 2016
The current state of this bug is a fix landed but got reverted (bugdroid seems to have missed adding a comment for the original CL). I'm working on a better fix.
,
Dec 16 2016
Ok thanks!
,
Jan 5 2017
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/8cd27b3cd1d1fe047bcf2c169106ccac7261d4ed commit 8cd27b3cd1d1fe047bcf2c169106ccac7261d4ed Author: ajuma <ajuma@chromium.org> Date: Thu Jan 05 21:01:49 2017 cc: Remove map lookup from RenderSurfaceImpl::EffectTreeIndex This relands http://crrev.com/2540333002 with a fix for the crashes that caused that CL to get reverted. The previous CL set the effect tree index on surfaces during the surface update step (when EffectNode::render_surface gets updated). This is not sufficient, however, since the surface update step will be skipped when we can't draw, even though property trees can still change and surfaces can still get destroyed (because of their owning layer getting destroyed). This CL also updates the effect tree index on surfaces when new property trees are set (that is, at commit time and at activation time). BUG= 670074 CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_trusty_blink_rel Review-Url: https://codereview.chromium.org/2571263002 Cr-Commit-Position: refs/heads/master@{#441747} [modify] https://crrev.com/8cd27b3cd1d1fe047bcf2c169106ccac7261d4ed/cc/layers/render_surface_impl.cc [modify] https://crrev.com/8cd27b3cd1d1fe047bcf2c169106ccac7261d4ed/cc/layers/render_surface_impl.h [modify] https://crrev.com/8cd27b3cd1d1fe047bcf2c169106ccac7261d4ed/cc/layers/render_surface_unittest.cc [modify] https://crrev.com/8cd27b3cd1d1fe047bcf2c169106ccac7261d4ed/cc/trees/draw_property_utils.cc [modify] https://crrev.com/8cd27b3cd1d1fe047bcf2c169106ccac7261d4ed/cc/trees/layer_tree_host_in_process.cc [modify] https://crrev.com/8cd27b3cd1d1fe047bcf2c169106ccac7261d4ed/cc/trees/layer_tree_host_unittest.cc [modify] https://crrev.com/8cd27b3cd1d1fe047bcf2c169106ccac7261d4ed/cc/trees/layer_tree_impl.cc [modify] https://crrev.com/8cd27b3cd1d1fe047bcf2c169106ccac7261d4ed/cc/trees/property_tree.cc [modify] https://crrev.com/8cd27b3cd1d1fe047bcf2c169106ccac7261d4ed/cc/trees/property_tree.h [modify] https://crrev.com/8cd27b3cd1d1fe047bcf2c169106ccac7261d4ed/cc/trees/property_tree_builder.cc
,
Jan 5 2017
|
|||
►
Sign in to add a comment |
|||
Comment 1 by bugdroid1@chromium.org
, Dec 2 2016