Skip to content

Conversation

@c-rhodes
Copy link
Collaborator

@c-rhodes c-rhodes commented Dec 2, 2025

For many of these tests the inputs are very similiar with slight divergences, ideally we'd have common sources where we can to avoid subtle differences.

Changes:

  • For basic tests there's a relatively large diff between V1 and all other cores because of 24f0901 ([MCA] Adding missing instructions in AArch64 Neoverse V1 tests #128892) which makes it more complete. I've been thru the entire diff and 99% of the time V1 makes more sense, except for a couple of small changes (might post a separate patch for). Therefore I decided it's best to take V1-basic-instructions.s as the common source.
  • Split out FEAT_RCPC_IMMO tests from basic since N1 doesnt have this feature.
  • Split out FEAT_MTE tests. V2/V3 also have this feature but were missing tests, so I've added them.

The NEON and SVE tests are also quite substantial and could be made common, but there's also subtle differences that take time to go thru.

…s (NFC)

For many of these tests the inputs are very similiar with slight
divergences, we should work to make common sources where we can to avoid
subtle differences.

Changes:

- For basic tests there's a relatively large diff between V1 and all
  other cores because of 24f0901 (llvm#128892) which makes it more complete.
  I've been thru the entire diff and 99% of the time V1 makes more
  sense, except for a couple of small changes (might post a separate
  patch for). Therefore I decided it's best to take
  V1-basic-instructions.s as the common source.
- Split out FEAT_RCPC_IMMO tests from basic since N1 doesnt have this feature.
- Split out FEAT_MTE tests. V2/V3 also have this feature but were
  missing tests, so I've added them.

The NEON and SVE tests are also quite substantial and could be made
common, but there's also subtle differences that take time to go thru.
Copy link
Contributor

@Asher8118 Asher8118 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

 I think the change makes sense and I agree that a common source for testing is a good direction.
 

 I've been thru the entire diff and 99% of the time V1 makes more sense, except for a couple of small changes (might post a separate patch for).

One question, in the cases where we do have even slight divergence, what is the plan? Add those instructions specifically as a separate test to avoid duplication?

The NEON and SVE tests are also quite substantial and could be made common, but there's also subtle differences that take time to go thru

I think as long as exceptions are handled as well, that would still be an improvement to the amount of duplication we currently have in tests.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants