Type ref.null as exact - #7371
Merged
Merged
Conversation
`RefNull` expression now have type `(ref exact null bot)`, allowing them to be used wherever a nullable exact reference is expected. Update the fuzzer and fix a few bugs with exactness propagation this uncovers.
tlively
commented
Mar 13, 2025
| if (type.isRef()) { | ||
| auto heapType = type.getHeapType(); | ||
| if (auto it = typeIndices.find(heapType); it != typeIndices.end()) { | ||
| // TODO: Handle exactness. |
Member
Author
There was a problem hiding this comment.
The fuzzer does not seem to trip over this (yet), so I left it unfixed for now. Eventually I expect us to trip over this bug, at which point we will be able to reduce the reproducer to a test case and ensure we have test coverage.
tlively
commented
Mar 13, 2025
tlively
left a comment
tlively
left a comment
Member
Author
There was a problem hiding this comment.
Fuzzed for 80k iterations with no problems, so I think it's good enough to land in that regard.
kripken
approved these changes
Mar 13, 2025
| if (exactness == Exact) { | ||
| auto inexact = getRelevantTypes(Type(heapType, nullability, Inexact)); | ||
| ret.insert(ret.end(), inexact.begin(), inexact.end()); | ||
| // Do not consider exact references to supertypes. |
Member
Author
There was a problem hiding this comment.
Because (ref exact $super) is not a supertype of (ref exact $sub).
| // are better supported. | ||
| Type refType = ref->type; | ||
| if (refType.isExact()) { | ||
| refType = Type(refType.getHeapType(), refType.getNullability(), Inexact); |
Member
There was a problem hiding this comment.
Suggested change
| refType = Type(refType.getHeapType(), refType.getNullability(), Inexact); | |
| refType = refType.as(Inexact); |
Member
Author
There was a problem hiding this comment.
Will look for other instances of this, too.
| // exact references because the binary writer can always generalize the exact | ||
| // reference types away. | ||
| // | ||
| // if (wasm.features.hasGC() && oneIn(2)) { |
Member
There was a problem hiding this comment.
when we do this, perhaps less of a chance for exact refs? they are probably rarer in real-world code.
Member
Author
There was a problem hiding this comment.
Sure, will bump to one in 8.
| heapType = getSubType(heapType); | ||
| if (type.isExact()) { | ||
| // The only other possible heap type is bottom. | ||
| if (!heapType.isBottom() && oneIn(8)) { |
Member
There was a problem hiding this comment.
Even 8 seems very high to me, since emitting more nulls is risky in terms of traps. And this is getSubType which is called constantly (unlike other places we have constants like 8). How about
Suggested change
| if (!heapType.isBottom() && oneIn(8)) { | |
| if (!heapType.isBottom() && oneIn(20)) { |
| ;; CHECK-NEXT: (local.get $0) | ||
| ;; CHECK-NEXT: ) | ||
| ;; CHECK-NEXT: (ref.null none) | ||
| ;; CHECK-NEXT: ) |
Member
Author
There was a problem hiding this comment.
Now that ref.null is exact, we could remove the early return in OptimizeInstructions.cpp that prevented us from removing the cast and rematerializing the null value instead.
Merged
tlively
added a commit
that referenced
this pull request
Mar 26, 2025
We decided that Custom Descriptors should introduce exact heap types rather than exact reference types. Although these new features are very similar, the APIs we need to change for them are completely different. One option would have been to keep the existing exact reference type implementation while additionally implementing exact heap types, but there are not enough free bits in the type implementation to have both at once without increasing the alignment of HeapTypeInfo allocations. Portably increasing the alignment is annoying enough that it's easier to just eagerly remove exact reference types to free up the bit for use with exact heap types. Fully or partially revert the following PRs: - #7371 - #7365 - #7360 - #7357 - #7356 - #7355 - #7354 - #7353 - #7347 - #7342 - #7328 Keep the new `.with(...)` Type APIs and the relevant parts of the type relations gtest that were introduced as part of the reverted work.
tlively
added a commit
that referenced
this pull request
Mar 26, 2025
We decided that Custom Descriptors should introduce exact heap types rather than exact reference types. Although these new features are very similar, the APIs we need to change for them are completely different. One option would have been to keep the existing exact reference type implementation while additionally implementing exact heap types, but there are not enough free bits in the type implementation to have both at once without increasing the alignment of HeapTypeInfo allocations. Portably increasing the alignment is annoying enough that it's easier to just eagerly remove exact reference types to free up the bit for use with exact heap types. Fully or partially revert the following PRs: - #7371 - #7365 - #7360 - #7357 - #7356 - #7355 - #7354 - #7353 - #7347 - #7342 - #7328 Keep the new `.with(...)` Type APIs and the relevant parts of the type relations gtest that were introduced as part of the reverted work.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
RefNullexpression now have type(ref exact null bot), allowing themto be used wherever a nullable exact reference is expected. Update the
fuzzer and fix a few bugs with exactness propagation this uncovers.