lld: pack Relocation from 32 to 24 bytes#38
Open
julesjacobs wants to merge 1 commit into
Open
Conversation
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Shrinks
struct lld::elf::Relocationfrom 32 to 24 bytes with no new input limits.exprandtypeare bit-packed into one 32-bit word (8 + 24 bits), and theSymbol *field becomes a 32-bit index (SymRef) into a single contiguous arena that now backs everySymbolUnionallocation;offsetandaddendkeep their full 64-bit width. ELF'sr_infoalready stores symbol references as 32-bit indices, so this caps nothing the format doesn't already cap.The relocation vector is among the largest live allocations during a link, so an 8-byte-per-record shrink shows up directly in peak memory: 13-15% lower peak RSS on relocation-dense executables, less in percentage terms on multi-GB outputs whose image dominates the peak. Byte-identical output, no wall-time regression.
Upstream lld still uses the 32-byte layout as of 21.1.8, so this is also a candidate to propose upstream (64-bit hosts benefit).