Cranelift: shrink ABIArgSlot #8163
Open
+21
−18
Conversation
This file contains 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
ABIArgSlot is currently 16 bytes with 4 bytes of padding. Shrinking the offset field in its Stack variant from i64 to i32 therefore reduces the enum to 8 bytes. It also reduces its alignment to 4 bytes, which might reduce padding in containing structures. Some targets already limit stack frame sizes to 2GB or less, and in practice they must be much smaller than that, so an i32 is plenty. By making this type only one word in size, we can put two of them in ABIArgSlotVec for free, which guarantees that such SmallVecs will never spill to the heap. In principle we could use arrayvec or something that doesn't support spilling to the heap at all. There are many other places where we use i64 for stack frame offsets which would probably benefit from switching to i32, but I didn't want to change everything at once.
github-actions
bot
added
cranelift
bjorn3
reviewed
Mar 18, 2024
| @@ -320,7 +320,7 @@ impl ABIMachineSpec for AArch64MachineDeps { | |||
| Some((ty, slot_offset)) | |||
| }) | |||
| .map(|(ty, offset)| ABIArgSlot::Stack { | |||
| offset, | |||
| offset: i32::try_from(offset).unwrap(), | |||
There was a problem hiding this comment.
We don't currently check that the stack frame size is less than 2GB, so this unwrap can panic: #7916
There was a problem hiding this comment.
That's good to know! I'll hold off on merging this until I have a chance to think more about that or somebody else fixes it.
fitzgen
approved these changes
Mar 18, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
ABIArgSlot is currently 16 bytes with 4 bytes of padding. Shrinking the offset field in its Stack variant from i64 to i32 therefore reduces the enum to 8 bytes. It also reduces its alignment to 4 bytes, which might reduce padding in containing structures.
Some targets already limit stack frame sizes to 2GB or less, and in practice they must be much smaller than that, so an i32 is plenty.
By making this type only one word in size, we can put two of them in ABIArgSlotVec for free, which guarantees that such SmallVecs will never spill to the heap. In principle we could use arrayvec or something that doesn't support spilling to the heap at all.
There are many other places where we use i64 for stack frame offsets which would probably benefit from switching to i32, but I didn't want to change everything at once.