[PATCH] D38611: Fix X86 regression on linpack

[Evgeny Stupachenko via Phabricator via llvm-commits]

evstupac added inline comments.


================
Comment at: lib/Target/X86/X86TargetTransformInfo.cpp:2538
+    // stores with folded address, treating each 2 such stores as 1
+    // additional instruction.
+    unsigned C1Insns = ((C1.FoldedStoreAddresses >> 1) + C1.Insns);
----------------
RKSimon wrote:
> This seems incredibly specific to a particular Intel architecture - you said you've tried to make a best average - based on what?
I have not seen significant difference on Atom CPUs and have seen gains on Cores. It is hard to cover all CPUs. We can adjust this (by limiting to specific CPU(s)) if get a regression.



================
Comment at: test/Transforms/LoopStrengthReduce/X86/folded_addresses.ll:8
+; Function Attrs: norecurse nounwind uwtable
+define void @foo(i32, float, float* noalias nocapture readonly, float* noalias nocapture) local_unnamed_addr #0 {
+; CHECK-LABEL: @foo(
----------------
RKSimon wrote:
> qcolombet wrote:
> > Could you use `opt -instnamer` on the input test?
> Is there any way that this can be reduced further?
Maybe we could get rid of vectorization, however now this is the test that got ~30% regression. It looks reasonable to track it. 


Repository:
  rL LLVM

https://reviews.llvm.org/D38611