[PATCH] D45201: Speeding up MIR-Canon using named-vregs.

[Puyan Lotfi via Phabricator via llvm-commits]

plotfi added a comment.

In https://reviews.llvm.org/D45201#1056209, @bogner wrote:

> Can we do better than just calling these namedVRegN? That is, is there a scheme where we could give these more meaningful names that would still be relatively stable between different input IRs?


Yes we can, but this is a first step. Also I want to compare the "namedVReg###" names with the old numbers as a sanity check too. I'm thinking of doing symbolic names in the future.


Repository:
  rL LLVM

https://reviews.llvm.org/D45201