[PATCH] D142472: [TableGen] Speed up computeUberSets. NFC.

[Jay Foad via Phabricator via llvm-commits]

foad added a comment.

For AMDGPUGenRegisterInfo.in this speeds up computeUberSets from 11 ms to 0.4 ms. It is still dwarfed by other parts of CodeGenRegisters which take several seconds.



================
Comment at: llvm/utils/TableGen/CodeGenRegisters.cpp:1707
     USet->Regs.push_back(&Reg);
-    sortAndUniqueRegisters(USet->Regs);
     RegSets[i++] = USet;
----------------
This is very expensive and completely pointless, since the registers added to each uber set are unique and sorted by construction.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D142472/new/

https://reviews.llvm.org/D142472