-
Notifications
You must be signed in to change notification settings - Fork 809
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Propose A Better Scheme for Id Allocation Specifically For The Testing Purpose #3512
Comments
I think the concern about spurious diffs is a valid one, and it's been an issue for me as well when trying to review changes that add or remove some instructions at the beginning of a SPIR-V file and thus result in big diffs to test results where the actual change is really small and everything else is just changes in |
To clarify, I'm proposing this scheme only for testing. Those modules shouldn't be generated in real world for end users. Perhaps this could also be done by separately maintained tools instead of directly in glslang.
|
It's great that we maintain a large repository of shaders in this project to ensure that glslang is composed with high quality code. However, the exact-match model in test result and the counting approach of id allocation strategy are making it difficult to reason about the test result change. Even a really simple change in glslang could lead to an id shift and all relevant test results may be edited entirely.
Could we add a new scheme of id allocation for testing only so that those ids are somewhat more stable, and the test result changes are easier to reason about? For example, we may let "id = opcode * 10000 + offset" so at least ids are stable across different opcodes?
The text was updated successfully, but these errors were encountered: