[SOL] Correctly copy 16-byte aligned memory #97
Merged
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
LucasSte
changed the title
dmakarov
approved these changes
Jun 20, 2024
LucasSte
merged commit 7d055ed
into
anza-xyz:solana-rustc/17.0-2023-12-14
13 checks passed
LucasSte
added a commit
to LucasSte/llvm-project
that referenced
this pull request
Jun 29, 2024
* Fix issue with copying 16-byte aligned memory
LucasSte
added a commit
that referenced
this pull request
Aug 19, 2024
* Fix issue with copying 16-byte aligned memory
LucasSte
added a commit
to LucasSte/llvm-project
that referenced
this pull request
Feb 12, 2025
* Fix issue with copying 16-byte aligned memory
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Outdated suggestions cannot be applied.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.
Problem
When compiling a Rust struct marked with
#[repr(C, align(16)], we force a 16-byte alignment for it and all the necessary copies. The compiler wasn't handling this case correctly and would panic if we try to callmemcpywith 16-byte aligned memory.Changes
The actual problem lies in
SBFSelectionDAGInfo.cpp, which was not estimating the number of loads and stores correctly. It was using the alignment, not the number of bytes per operation, for the estimation. This has been fixed.Although not used in production, the backend has an option
-sbf-expand-memcpy-in-orderto expand loads and stores with increasing offsets. The function that does so wasn't handling 16-byte aligned memory and was emitting inefficient code (more loads and stores than necessary). This problem has also been fixed.There are tests for these two fixes included in this PR.