[PATCH] D95273: AMDGPU: Reduce the number of expensive calls in SIFormMemoryClause

[Changpeng Fang via Phabricator via llvm-commits]

cfang added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/SIFormMemoryClauses.cpp:340
 
-      if (!processRegUses(MI, Defs, Uses, RPT))
+      auto LiveRegs = RPT.getLiveRegs();
+      RegUse Defs, Uses;
----------------
rampitec wrote:
> LiveRegSet LiveRegs = ...
> Make sure you are actually copying it, not taking a reference.
How to clone a copy? I can not find a suitable RPT function for that purpose.


================
Comment at: llvm/lib/Target/AMDGPU/SIFormMemoryClauses.cpp:363
+      // Restore the state after processing the bundle.
+      RPT.reset(MI, &LiveRegs);
+
----------------
rampitec wrote:
> Move it down after "Ind->insertMachineInstrInMaps(*B);" and reset on B, not on MI. Technically MI's iterator is not valid for RPT purposes after bundling.
Shouldn't  we also restore  before the continue when there is just one instruction?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D95273/new/

https://reviews.llvm.org/D95273