You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Mar 17, 2026. It is now read-only.
Here's a concise summary of the major changes in this LLVM IR diff:
Replacement of manual alignment checks with llvm.assume alignment metadata: Across multiple files (ChecksumFile.ll, Cr2Decoder.ll, FileReader.ll, RawImage.ll, rawspeed-identify.ll, ZendAccelerator.ll, vm.ll), sequences computing pointer alignment via ptrtoint + and + icmp are replaced with direct llvm.assume(i1 true) [ "align"(ptr, i64 N) ]. This simplifies IR and enables better optimization by conveying alignment assumptions more directly to the optimizer.
Refinement of no-alias scope usage in XXH3 hash functions: In input-sections.cc.X86_64.ll, new llvm.experimental.noalias.scope.decl intrinsics are added, and existing !noalias and !alias.scope metadata on loads/stores in XXH3 accumulation loops are updated to reference newly declared scopes (e.g., !555, !563, !574). This improves alias analysis precision for vectorized memory operations.
Fixes to phi node predecessor lists: Several phi instructions in rawspeed-identify.ll and ChecksumFile.ll had incorrect predecessor blocks listed (e.g., %475 → %472, %91 → %88, %95 → %92). These were corrected to match the actual control-flow predecessors, ensuring IR validity and correct SSA semantics.
Cleanup of redundant or obsolete llvm.assume calls and TBAA metadata: In ZendAccelerator.ll and vm.ll, redundant alignment checks (e.g., ptrtoint/and/icmp) are removed entirely, replaced solely by llvm.assume with alignment metadata. Additionally, some outdated or unused llvm.assume calls (e.g., in zend_shared_alloc.ll) and stale !tbaa references are removed or updated.
Consistent attribute and metadata index updates across TBB-related code: In input-sections.cc.X86_64.ll, numerous function attributes, loop metadata IDs (e.g., !572 → !573, !612 → !613), and TBAA node references (e.g., !602 → !603, !631 → !632) are incremented to reflect structural changes in the module’s metadata hierarchy, maintaining consistency after prior transformations.
These changes collectively improve IR clarity, strengthen optimization hints (especially around alignment and aliasing), and fix correctness issues in control-flow and metadata, primarily targeting performance-critical hashing and memory management code.
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 freeto subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Labels
None yet
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.
Link: llvm/llvm-project#177597
Requested by: @nikic