ImageDecodeBench packeted runs should resume decoding where it left off |
||
Issue descriptionImageDecodeBench has a mode where it fills the buffer with a given packet size. It then decodes as much as it can for each virtual packet received. I believe the point of this mode is to test a more real-world decoding scenario. However, ImageDecodeBench restarts animated images for each packet. For example, if we could decode the first 3 frames with packet 1, upon packet 2 coming in we decode those first 3 frames again before decoding frame 4. This deviates from the real-world decoding scenario. So ImageDecodeBench should resume where it left off for packeted timing mode.
,
May 22 2017
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/40bdf8dbdc4089b110e76ea1d063a9b92f615301 commit 40bdf8dbdc4089b110e76ea1d063a9b92f615301 Author: cblume <cblume@chromium.org> Date: Mon May 22 19:07:21 2017 Resume ImageDecodeBench packeted runs where it left off ImageDecodeBench has a mode where it fills the buffer with data of a given packet size. It then decodes as much as it can for each data packet received. I believe the point of this mode is to test a more real-world decoding scenario (using network-sized packets). However, ImageDecodeBench fetches already-decoded frames from animated images when a new packet arrives, rather than continuing to decode the next frame. For example, if we decoded the first 3 frames with packet 1, upon packet 2 coming in, we fetch those first 3 frames again before decoding frame 4. This does not cause an expensive decode, but still deviates from the real-world decoding scenario. In this change, make ImageDecodeBench resume where it left off when in packet timing mode. BUG= 721260 Review-Url: https://codereview.chromium.org/2880533002 Cr-Commit-Position: refs/heads/master@{#473645} [modify] https://crrev.com/40bdf8dbdc4089b110e76ea1d063a9b92f615301/third_party/WebKit/Source/platform/testing/ImageDecodeBench.cpp
,
Mar 8 2018
|
||
►
Sign in to add a comment |
||
Comment 1 by cblume@chromium.org
, May 14 2017