[PATCH] D60669: [APInt] Use __builtin_[su]mulll_overflow for [su]mul_ov if available

[Nikita Popov via Phabricator via llvm-commits]

nikic added a comment.

In D60669#1465834 <https://reviews.llvm.org/D60669#1465834>, @lebedev.ri wrote:

> Is this trying to address some performance problem? Any numbers?


I think even if it doesn't address a specific performance problem, this is a good change (well, once it is correct...), because you don't have to treat mul_ov as an expensive operation anymore. E.g. after this change I think we could drop the code in https://github.com/llvm-mirror/llvm/blob/2b38b84a12ff96d6862b7927f92cbdae77653deb/lib/Analysis/ValueTracking.cpp#L4001-L4008 and the big comment in the function without being concerned about a possible perf impact.


Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D60669/new/

https://reviews.llvm.org/D60669