• Xianzhu Wang's avatar
    [RasterInvalidation] Track explicit and implicit full composited layer invalidation · 44ab2704
    Xianzhu Wang authored
    1. For explicit full composited layer invalidation (Pre-SPv2 only) which
    is issued by GraphicsLayer::SetNeedsDisplay():
      - Use more descriptive reason kFullLayer instead of kFull;
      - Clear RasterInvalidator old states which are stale
    
    2. Track implicit full composited layer invalidation (Pre-SPv2 and
    SPv2) which happens when a layer is newly created or becomes non empty.
    We don't issue raster invaidation because the new tiles are invalid
    initially. However, it's a performance overhead when we unnecessarily
    re-create a composited layer e.g. just because the layer's id changed
    (crbug.com/842356), and we need to expose this overhead in layout test
    results.
    
    Bug: 842356
    Cq-Include-Trybots: luci.chromium.try:linux_layout_tests_slimming_paint_v2;master.tryserver.blink:linux_trusty_blink_rel
    Change-Id: I2ffcfa9dbc43668f57b0ea84795d3b0fbcb7b9c6
    Reviewed-on: https://chromium-review.googlesource.com/1112300Reviewed-by: default avatarChris Harrelson <chrishtr@chromium.org>
    Commit-Queue: Xianzhu Wang <wangxianzhu@chromium.org>
    Cr-Commit-Position: refs/heads/master@{#569855}
    44ab2704
scrolling-without-painting-expected.txt 1.66 KB