skia2/tools/skqp/README_GENERATING_MODELS.md

How SkQP Generates Render Test Models

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

0. Choose a commit to make the branch from

   COMMIT=origin/main
   

   Or use the script to find the best one:

   cd SKIA_SOURCE_DIRECTORY
   git fetch origin
   COMMIT=$(python tools/skqp/find_commit_with_best_gold_results.py \
            origin/main ^origin/skqp/dev)
   

1. Get the positively triaged results from Gold and generate models:

   cd SKIA_SOURCE_DIRECTORY
   git fetch origin
   git checkout "$COMMIT"
   python tools/skqp/cut_release.py HEAD~10 HEAD
   

   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, using this script:

   sh tools/skqp/branch_skqp_dev.sh
   

   Review and submit the change:

   git push origin HEAD:refs/for/skqp/dev
   bin/sysopen https://review.skia.org/$(bin/gerrit-number HEAD)
   

   (Optional) Make a SkQP APK.

   tools/skqp/docker_build_universal_apk.sh
   

   (Optional) Test the SkQP APK:

   tools/skqp/test_apk.sh (LOCATION)/skqp-universal-debug.apk
   

   (Once changes land) Upload the SkQP APK.

   tools/skqp/upload_apk HEAD (LOCATION)/skqp-universal-debug.apk
   

tools/skqp/cut_release.py

This tool will call make_skqp_model to generate the m{ax,in}.png files for each render test.

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 good.txt file and the model exists, 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.