llvm-project

c5afcfe0 - [X86] combineINSERT_SUBVECTOR - fold insert_subvector(base,extract_subvector(broadcast)) -> blend shuffle(base,broadcast) (REAPPLIED) (#133724)

[X86] combineINSERT_SUBVECTOR - fold insert_subvector(base,extract_subvector(broadcast)) -> blend shuffle(base,broadcast) (REAPPLIED) (#133724) If the broadcast is already the full vector width, try to prefer a blend/vshuff64x2 over a vector insertion which is usually lower latency (and sometimes a lower uop count), and reduces changes in vector sizes that can interfere with further combines. Updated version of #133083 - which lead to infinite loops due to shuffle lowering recreating the INSERT_SUBVECTOR pattern, this variant creates the BLENDI/SHUF128 nodes directly.