[PATCH] D139710: [AMDGPU] MachineScheduler: schedule execution metric added for the UnclusteredHighRPStage

[Valery Pykhtin via Phabricator via llvm-commits]

vpykhtin added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/GCNSchedStrategy.cpp:945
+  unsigned SumBubbles = 0;
+  DenseMap<MachineInstr *, unsigned> Model;
+  unsigned CurrCycle = 0;
----------------
Model -> ReadyCycles?


================
Comment at: llvm/lib/Target/AMDGPU/GCNSchedStrategy.cpp:960
+          unsigned DefIdx = Model[DefMI];
+          if (DefIdx + Latency > ReadyCycle)
+            ReadyCycle = DefIdx + Latency;
----------------
max


================
Comment at: llvm/lib/Target/AMDGPU/GCNSchedStrategy.cpp:965
+      SumBubbles += ReadyCycle - CurrCycle;
+      Model.insert(std::make_pair(&MI, ReadyCycle));
+      CurrCycle = ++ReadyCycle;
----------------
Model[&MI] = ReadyCycle;


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D139710/new/

https://reviews.llvm.org/D139710