cmd/compile/internal/ssa: allow memcombine on proved aligned addresses #77300
+518
−4
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Outdated suggestions cannot be applied.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.
Currently, memcombine is disabled for architectures where unaligned
access might be slow or unsupported (unalignedOK = false), such as riscv64.
This leads to significant performance degradation for common operations
like encoding/binary.Uint64, which expand into long sequences of
byte loads, shifts, and ORs.
This change relaxes the global unalignedOK constraint in memcombine.
Instead of a total bailout, it now allows merging loads and stores
if the combined address can be statically proven to be naturally aligned.
By analyzing OpAddr, OpLocalAddr, and constant offsets through OpOffPtr,
the compiler can safely emit a single wide load (e.g., LD on RISC-V)
even when the architecture generally forbids unaligned access.
This provides the performance benefits of load merging for well-aligned
data structures without risking expensive kernel traps on unaligned-unfriendly
hardware.
Updates #736940