[PATCH] D113371: [X86] combinePMULH - recognise 'cheap' trunctions via PACKS/PACKUS as well as SEXT/ZEXT (WIP)

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon updated this revision to Diff 390720.
RKSimon added a comment.

Fix AVX2 regressions - we can use a similar trick to what we do for MADDWD - if we know that the upper bits of the extended input elements are zero, then the MULHU result for those (bitcasted) sub-elements will be zero as well - so we can perform the MULHU on the (bitcasted) inputs and then cheaply truncate/packus the results to vXi16.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D113371/new/

https://reviews.llvm.org/D113371

Files:
  llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
  llvm/lib/Target/X86/X86ISelLowering.cpp
  llvm/test/CodeGen/X86/pmulh.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D113371.390720.patch
Type: text/x-patch
Size: 21131 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20211130/51b75458/attachment-0001.bin>