[PATCH] D26855: New unsafe-fp-math implementation for X86 target

[Andrew V. Tischenko via Phabricator via llvm-commits]

avt77 added inline comments.


================
Comment at: lib/CodeGen/MachineTraceMetrics.cpp:521
 /// Invalidate traces through BadMBB.
+// TODO: this code should be refactored because it really increases compile time
 void
----------------
RKSimon wrote:
> Is this still true after D29627?
Yes, the profile shows that now (after D29627) this feature eats more time than any other

  58,19%  llc      llc                   [.] llvm::MachineTraceMetrics::Ensemble::invalidate
   3,44%  llc      llc                   [.] (anonymous namespace)::TwoAddressInstructionPass::scanUses
   3,19%  llc      llc                   [.] llvm::ScheduleDAGSDNodes::ClusterNeighboringLoads
   1,59%  llc      llc                   [.] llvm::SparseMultiSet<llvm::VReg2SUnit, llvm::VirtReg2IndexFunctor, unsigned char>::find
   1,08%  llc      llc                   [.] llvm::X86InstrInfo::areLoadsFromSameBasePtr



https://reviews.llvm.org/D26855