[PATCH] D78989: [NFC] [PowerPC] Narrow fast-math flags in tests

Wed Apr 29 02:38:52 PDT 2020

qiucf marked 3 inline comments as done.
qiucf added inline comments.

================
Comment at: llvm/test/CodeGen/PowerPC/scalar-equal.ll:55
 entry:
-  %cmp = fcmp fast oeq double %a, %b
-  %cond = select fast i1 %cmp, double %c, double %d
+  %cmp = fcmp oeq double %a, %b
+  %cond = select i1 %cmp, double %c, double %d
----------------
steven.zhang wrote:
> The change here is questionable. If we have the fast flag set, we shouldn't see the different instruction sequence w/o --enable-unsafe-fp-math. Please double confirm with compiler implementation to see what happens.
That's not related to `unsafe-fp-math`. That's a bug about flag propagation. See D79074, thanks.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D78989/new/

https://reviews.llvm.org/D78989