[PATCH] D55950: [ConstantFolding] Fold undef for integer intrinsics

[Nikita Popov via Phabricator via llvm-commits]

nikic created this revision.
nikic added reviewers: RKSimon, spatel.
Herald added a subscriber: llvm-commits.

This fixes https://bugs.llvm.org/show_bug.cgi?id=40110.

This implements handling of undef operands for integer intrinsics in ConstantFolding. In particular the bitcounting intrinsics (ctpop, cttz, ctlz), the with overflow intrinsics, the saturating math intrinsics and the funnel shift intrinsics.

The undef behavior follows what InstSimplify does for the general case of non-constant operands. For the bitcount intrinsics (where InstSimplify doesn't do undef handling -- there cannot be a combination of an undef + non-constant operand) I'm using a 0 result if the intrinsic is defined for zero and undef otherwise.


Repository:
  rL LLVM

https://reviews.llvm.org/D55950

Files:
  lib/Analysis/ConstantFolding.cpp
  test/Analysis/ConstantFolding/bitcount.ll
  test/Analysis/ConstantFolding/funnel-shift.ll
  test/Analysis/ConstantFolding/saturating-add-sub.ll
  test/Transforms/ConstProp/overflow-ops.ll
  test/Transforms/InstCombine/saturating-add-sub-vector.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D55950.179129.patch
Type: text/x-patch
Size: 49761 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20181220/b090661f/attachment-0001.bin>