skia2/tools/skqp/README_GENERATING_MODELS.md

How SkQP Generates Render Test Models

We will, at regular intervals, generate new models from the master branch of Skia. Here is how that process works:

1. Get the positively triaged results from Gold:

   Go to Skia Gold's search and search for results that are

   • Positive
   • config=gles OR config=vk
   • Span as many commits in the past as possible.

   Then go to "Actions" → "Export" and save the resulting meta.json file.

2. From a checkout of Skia's master branch, execute:

   origin	https://skia.googlesource.com/skia.git
   git checkout origin/master
   tools/skqp/cut_release META_JSON_FILE
   

   This will create the following files:

   platform_tools/android/apps/skqp/src/main/assets/files.checksum
   platform_tools/android/apps/skqp/src/main/assets/skqp/rendertests.txt
   platform_tools/android/apps/skqp/src/main/assets/skqp/unittests.txt
   

   These three files can be commited to Skia to create a new commit. Make origin/skqp/dev a parent of this commit (without merging it in), and push this new commit to origin/skqp/dev:

   git merge -s ours origin/skqp/dev -m "Cut SkQP $(date +%Y-%m-%d)"
   git add \
     platform_tools/android/apps/skqp/src/main/assets/files.checksum \
     platform_tools/android/apps/skqp/src/main/assets/skqp/rendertests.txt \
     platform_tools/android/apps/skqp/src/main/assets/skqp/unittests.txt
   git commit --amend --reuse-message=HEAD
   git push origin HEAD:refs/for/skqp/dev
   

   Review and submit the change:

   NUM=$(experimental/tools/gerrit-change-id-to-number HEAD)
   tools/skqp/sysopen.py https://review.skia.org/$NUM
   

tools/skqp/cut_release

This tool will call make_gmkb.go to generate the m{ax,in}.png files for each render test. Additionaly, a models.txt file enumerates all of the models.

Then it calls jitter_gms to see which render tests pass the jitter test. jitter_gms respects the bad_gms.txt file by ignoring the render tests enumerated in that file. Tests which pass the jitter test are enumerated in the file good.txt, those that fail in the bad.txt file.

Next, the skqp/rendertests.txt file is created. This file lists the render tests that will be executed by SkQP. These are the union of the tests enumerated in the good.txt and bad.txt files. If the render test is found in the models.txt file and the good.txt file, its per-test threshold is set to 0 (a later CL can manually change this, if needed). Otherwise, the threshold is set to -1; this indicated that the rendertest will be executed (to verify that the driver will not crash), but the output will not be compared against the model. Unnecessary models will be removed.

Next, all of the files that represent the models are uploaded to cloud storage. A single checksum hash is kept in the files.checksum file. This is enough to re-download those files later, but we don't have to fill the git repository with a lot of binary data.

Finally, a list of the current gpu unit tests is created and stored in skqp/unittests.txt.