[PATCH] D64325: RegUsageInfoCollector: Don't iterate all regs for every reg class

[Matt Arsenault via Phabricator via llvm-commits]

arsenm created this revision.
arsenm added a reviewer: qcolombet.
Herald added subscribers: tpr, nhaehnle, wdng, jvesely.

This is extremly slow on AMDGPU, which has a lot of physical register
and a lot of register classes. Visit registers with common regunits
instead. NFC, except for compile time improvement.


https://reviews.llvm.org/D64325

Files:
  lib/CodeGen/RegUsageInfoCollector.cpp
  test/CodeGen/AMDGPU/ipra-regmask.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D64325.208370.patch
Type: text/x-patch
Size: 5768 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20190708/95242480/attachment.bin>