[PATCH] D136267: [AMDGPU] Speedup GCNDownwardRPTracker::advanceBeforeNext

[Valery Pykhtin via Phabricator via llvm-commits]

vpykhtin added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/GCNRegPressure.cpp:360
+
+  auto const RegUses = collectVirtualRegUses(*LastTrackedMI, LIS, *MRI);
+  // each RegUnit is unique in RegUses
----------------
arsenm wrote:
> arsenm wrote:
> > Can you defer this until after skipDebugInstructionsForward?
> I guess that doesn't make sense. 
Well it can save some time at the end of the block


================
Comment at: llvm/lib/Target/AMDGPU/GCNRegPressure.cpp:387
+      if (It != LiveRegs.end() && It->second.none())
+        LiveRegs.erase(It);
     } else if (!LI.liveAt(SI)) {
----------------
arsenm wrote:
> vpykhtin wrote:
> > arsenm wrote:
> > > I don't understand why LiveRegs is a map in the first place
> > The map consists of pairs register:lanemask
> Yes, but tracking it this way doesn't really make sense to me. Does this really need to be erased?
I'm not sure: do you mean leaving zero lanemask instead of erasing?






Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D136267/new/

https://reviews.llvm.org/D136267