Bug: skia:9588
Change-Id: I8dc915f9cedd40b94f37f3d1ae89bb7d4767814f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/252186
Commit-Queue: Eric Boren <borenet@google.com>
Reviewed-by: Kevin Lubick <kjlubick@google.com>
The lack of 'rmfile' steps is OK because all of the temporary paths
are created within swarming's sandbox; swarming will clean everything
up after the recipe has quit.
R=borenet@google.com
Change-Id: Ibb8dc612ec8eb98fe966c9caa90ba59db48dbfba
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/239797
Auto-Submit: Robbie Iannucci <iannucci@google.com>
Commit-Queue: Eric Boren <borenet@google.com>
Reviewed-by: Eric Boren <borenet@google.com>
Adds the following new metrics:
* aborted_frames: These are the number of frames which have "main_frame_aborted" terminations in the chosen 25 samples.
* drawn_frame_max_us / drawn_frame_min_us / drawn_frame_avg_us: These metrics are similar to the existing "frame_*_us" metrics but are calculated using only drawn frames (frames terminated with missed_frame or submitted_frame).
Bug: skia:9237
Change-Id: Ib6ed29f7ae6a1f7b6660f231aad528fe7bb6f203
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/229419
Commit-Queue: Ravi Mistry <rmistry@google.com>
Reviewed-by: Florin Malita <fmalita@chromium.org>
Added new "backend" flag to lottie-web-perf.js and created a new lottie-web-canvas-perf.html file.
Bug: skia:9237
Change-Id: Ibca2f121666daf25fbbccac48fed80eb12ccedf7
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/229420
Commit-Queue: Ravi Mistry <rmistry@google.com>
Reviewed-by: Florin Malita <fmalita@chromium.org>
* Also use DISPLAY=:0 for non-headless chrome via puppeeteer.
* Output gpu-gl-renderer, gpu-driver, gpu-gl-vendor from the trace file. Will be useful for debugging.
NoTry: true
Bug: skia:9237
Change-Id: Ieb70dbe540aeca89e88dbaeace3cdba1b50eb7ef
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/227156
Reviewed-by: Florin Malita <fmalita@chromium.org>
Commit-Queue: Ravi Mistry <rmistry@google.com>
This enables us to use the same termination checks as for lottie-web.
NoTry: true
Bug: skia:9237
Change-Id: I9d5d0c450196c3800020e12b45868c656cdcc46a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226858
Commit-Queue: Ravi Mistry <rmistry@google.com>
Reviewed-by: Joe Gregorio <jcgregorio@google.com>
Doing this will avoid the unreliable frames at the start (setup related) and at the end (teardown related). Also:
* Increase timeouts in lottie-web-perf.js and skottie-wasm-perf.js since we are looping twice now.
* Fix bug where ret code was 0 if puppeteer test failed.
NoTry: true
Bug: skia:9237
Change-Id: I0933ecb87113b40c162712076c38fb7379b352d4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/226836
Reviewed-by: Joe Gregorio <jcgregorio@google.com>
Reviewed-by: Florin Malita <fmalita@chromium.org>
Commit-Queue: Ravi Mistry <rmistry@google.com>
Uses the following algorithm for Skottie+WASM:
* Start timer when you see BeginImplFrameToSendBeginMainFrame
* Stop timer for main_frame_aborted
* Abandon frame_id if you see replaced_by_new_reporter_at_same_stage, did_not_produce_frame or missed_frame
Getting consistently 25 frames for all lottie-samples with the above algorithm.
Uses the following algorithm for Lottie-Web:
* Start timer when you see BeginImplFrameToSendBeginMainFrame
* Stop timer for missed_frame or submitted_frame
* Abandon frame_id if you see replaced_by_new_reporter_at_same_stage, did_not_produce_frame or main_frame_aborted
Getting 25 sometimes but also getting other numbers like 20, 27, etc. This is probably because there are more moving parts
and it is hard to predict when Chrome is going to render frames and how many. We might not get exactly 25 here like for
Skottie+WASM, but we will use whatever the tracer gives us (we can always revisit this later).
NoTry: true
Bug: skia:9237
Change-Id: Ia4d5aaf6ac95399b80574d1579358abda4e4e2ea
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/225898
Commit-Queue: Ravi Mistry <rmistry@google.com>
Reviewed-by: Joe Gregorio <jcgregorio@google.com>
Might as well do the same in lottie-web-perf.
Also mark as infra failures to distinguish from real failures.
NoTry: true
Bug: skia:9207
Change-Id: Id1f3cc5719094d0bf0179ea1a84529c6f376e356
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/223917
Reviewed-by: Ravi Mistry <rmistry@google.com>
Commit-Queue: Ravi Mistry <rmistry@google.com>
* Added support for LottieWeb bots in perf_skottiewasm_lottieweb.py
* Adds a LOTTIE_WEB_BLACKLIST for lottie files that crash lottie-web as described in https://bugs.chromium.org/p/skia/issues/detail?id=9187#c4
NoTry: true
Bug: skia:9187
Change-Id: I8dbf485e7162e027cb4c4db2bf8f8e9c3ec4d966
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/222157
Reviewed-by: Joe Gregorio <jcgregorio@google.com>
Commit-Queue: Ravi Mistry <rmistry@google.com>
The lottieweb perf recipe will be very similar to the skottiewasm perf
recipe, so decided to eventually use the same one for both.
Bug: skia:9187
Change-Id: Id9da91c9cf79df7c01c61c7920f944bec974e9f8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/222156
Reviewed-by: Joe Gregorio <jcgregorio@google.com>
Commit-Queue: Ravi Mistry <rmistry@google.com>
