[PATCH] D18940: Loop vectorization with uniform load

[Ayal Zaks via llvm-commits]

Ayal added a comment.

This seems like the right cost to attribute to a uniform load.

In general, as in this testcase, uniform loads that are loop-invariant better be hoisted before the vectorizer, yielding potentially even lower costs. But that's an orthogonal issue.

It would be good to align the instructions emitted when vectorizing such a load (i.e., scalarizeInstruction's insertElements) with this cost attributed to it.


================
Comment at: ../test/Transforms/LoopVectorize/X86/uniform_load.ll:1
@@ +1,2 @@
+; RUN: opt -O2 -S -mcpu=core-avx2 < %s | FileCheck %s
+
----------------
Suffice to invoke -loop-vectorize rather than -O2?


Repository:
  rL LLVM

http://reviews.llvm.org/D18940