[PATCH] D56057: [X86] Individually simplify both operands of PMULDQ/PMULUDQ using the other entry point of SimplifyDemandedBits that allows the one use check of the root node to be suppressed.

[easyaspi314 (Devin) via Phabricator via llvm-commits]

easyaspi314 added a comment.

I am a little suspicious about these changes to the test results.

1. If these removals are all of redundant instructions that were added previously, I worry about how well people are checking the test result output. A large chunk of the code we are removing is from this: https://reviews.llvm.org/rL347181. As you can see, there are major changes to the test results, which, unless I am mistaken, passed without questioning whether these changes are a regression or an improvement. If nobody checks test results for regressions, why have tests at all?
2. I also worry that they might not return the correct result. I am especially concerned about the number of removed i64 vector multiplies.
3. Otherwise, if these all return the correct result, nice job.


Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D56057/new/

https://reviews.llvm.org/D56057