[PATCH] D67383: Add new optimization pass of Tree-Height-Reduction

[Hal Finkel via Phabricator via llvm-commits]

hfinkel added inline comments.


================
Comment at: lib/Transforms/Scalar/TreeHeightReduction.cpp:379
+  const int InstLatency =
+      TTI->getInstructionCost(Inst, TargetTransformInfo::TCK_Latency);
+  setLatency(getLatency() + InstLatency);
----------------
This is the only call to TTI that I see, and I don't see it don't see it documented anywhere what objective function this transformation is effectively using. Can you please add some explanation of that? Also, I'd expect to see some check on the instruction throughput -- because it might not make sense to introduce more ILP than the target can use -- and I'd expect to see some register-pressure estimate.


Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D67383/new/

https://reviews.llvm.org/D67383