[PATCH] D147243: [X86] MatchVectorAllZeroTest - add support for icmp(bitcast(icmp_ne(X,Y)),0) vector reduction patterns

[Alexander Kornienko via Phabricator via llvm-commits]

alexfh added a comment.

In D147243#4248129 <https://reviews.llvm.org/D147243#4248129>, @RKSimon wrote:

> I should have the fix done shortly - please don't start reverting anything....

Glad to hear that! In case you wonder, there's a C++ test case as well:

  $ cat q.ii
  struct V {
    unsigned : 28;
    unsigned p_ : 4;
    unsigned : 32;
    unsigned : 32;
    unsigned : 56;
    unsigned q_ : 4;
    unsigned : 2;
    unsigned r_ : 1;
    unsigned : 1;
    unsigned s : 27;
    unsigned t : 1;
    unsigned : 4;
    unsigned : 1;
    unsigned : 75;
    unsigned p() { return p_; }
    unsigned q() { return q_; }
    unsigned r() { return r_; }
  };
  
  bool x(V a, V b) {
    return a.p() != b.p() || a.s || a.q() != b.q() || a.r() != b.r() || a.t;
  }
  $ ./clang  -O2 --target=x86_64-unknown-linux-gnu -fsanitize=address -c q.ii -o bad.o
  Stack dump:
  0.      Program arguments: ./clang -O2 --target=x86_64-unknown-linux-gnu -fsanitize=address -c q.ii -o bad.o
  1.      <eof> parser at end of file
  2.      Code generation
  3.      Running pass 'Function Pass Manager' on module 'q.ii'.
  4.      Running pass 'X86 DAG->DAG Instruction Selection' on function '@_Z1x1VS_'
  ...

And here: https://gcc.godbolt.org/z/P51G8xrxG


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D147243/new/

https://reviews.llvm.org/D147243