[PATCH] D60318: [ExpandMemCmp][MergeICmps] Move passes out of CodeGen into opt pipeline.

Mon May 6 08:09:08 PDT 2019

courbet added inline comments.

================
Comment at: llvm/test/Transforms/ExpandMemCmp/PowerPC/memcmpIR.ll:68-70
+; CHECK-BE-NEXT:    [[TMP17]] = call i64 @llvm.bswap.i64(i64 [[TMP15]])
+; CHECK-BE-NEXT:    [[TMP18]] = call i64 @llvm.bswap.i64(i64 [[TMP16]])
+; CHECK-BE-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[TMP17]], [[TMP18]]
----------------
spatel wrote:
> Why are we generating bswap for a big-endian target?
Thanks for the catch. We're not, but contrary to `llc`, `opt` does not seem to get the data layout from the target, so I/m now explicitly specifying the data layout on the RUN line.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D60318/new/

https://reviews.llvm.org/D60318