[PATCH] D133282: [TargetLowering][X86][AMDGPU] Teach expandMUL_LOHI to handle a mix of sign and zero extend.

[Craig Topper via Phabricator via llvm-commits]

craig.topper created this revision.
craig.topper added reviewers: RKSimon, efriedma, spatel.
Herald added subscribers: kosarev, StephenFan, kerbowa, pengfei, hiraditya, t-tye, tpr, dstuttard, yaxunl, jvesely, kzhuravl.
Herald added a project: All.
craig.topper requested review of this revision.
Herald added a subscriber: wdng.
Herald added a project: LLVM.

If one input of a MUL that needs to expanded is sign extended and
the other is zero extended, we can use an unsigned mul and apply
a correction. If the signed number is negative, we subtract the lower
bits of the zero extended input.

Fixes PR57549, but not in the same way.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D133282

Files:
  llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
  llvm/test/CodeGen/AMDGPU/mad_64_32.ll
  llvm/test/CodeGen/X86/extmul128.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D133282.457895.patch
Type: text/x-patch
Size: 4289 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20220905/fc373745/attachment.bin>