Skip to content
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

[FIRRTL][LowerAnnotations] Fix non-probe type compat check. #6822

Merged
merged 2 commits into from
Mar 14, 2024

Conversation

dtzSiFive
Copy link
Contributor

Add test as "legacy wiring" as non-ref-type-port path is a hidden option.

Compatibility dest/source operands were backwards, introduced in #4656 .

Fixes #6819.

Add test as "legacy wiring" as non-ref-type-port path is a hidden option.

Compatibility dest/source operands were backwards,
introduced in llvm#4656 .

Fixes llvm#6819.
@dtzSiFive dtzSiFive added the FIRRTL Involving the `firrtl` dialect label Mar 13, 2024
Copy link
Member

@seldridge seldridge left a comment

Choose a reason for hiding this comment

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

Confused by the fix here. However, the test looks great.

!areTypesEquivalent(sourceFType, sinkFType)) {
!areTypesEquivalent(sinkFType, sourceFType)) {
Copy link
Member

Choose a reason for hiding this comment

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

Type equivalence is not commutative? This seems like an issue with the type equivalence check?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

LOL yeah it's not symmetric and therefore also not an equivalence relation, possibly for other reasons I'm forgetting. This is @mmaloney-sf's favorite part about it 🙃 😉 .

Copy link
Contributor Author

Choose a reason for hiding this comment

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

To be clear this is the FIRRTL spec's notion of "type equivalence", I don't think it's bugged (not for lack of being an equivalence anway) but at a very-quick glance it doesn't mention const there presently, although it's still not symmetric (Reset can connect to UInt, can't connect UInt to Reset).

Copy link
Member

Choose a reason for hiding this comment

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

Good point about resets which are obviously not commutative.

I think the problem here is more that the spec is presenting this as an "equivalence" when it's really "verification of connection source and destination". It's not inherently wrong that equality isn't commutative, it's just weird.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Right, shouldn't be calling its connectivity rules (or whatever) equivalence or equality probably as that's misleading if nothing else. I haven't checked but can't help but wonder if it was originally a proper equivalence and then drifted or something. Welp.

Anyway, mind approving the PR? 👍

Copy link
Member

Choose a reason for hiding this comment

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

I think it was before abstract reset. The spec was always written as something like "the source and destination must be equivalent types (see type equivalence section for a definition)." Notably, I don't think that "type equivalence" actually mattered except for connects. I.e., this was always that aforementioned connect verification. We should probably just roll it into the connect section.

test/Dialect/FIRRTL/legacy-wiring.mlir Outdated Show resolved Hide resolved
@dtzSiFive dtzSiFive merged commit b6f50d6 into llvm:main Mar 14, 2024
4 checks passed
@dtzSiFive dtzSiFive deleted the fix/issue-6819 branch March 14, 2024 14:49
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
FIRRTL Involving the `firrtl` dialect
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Wiring problem error with const source and non-const sink w/no-ref-type option
2 participants