[PATCH] D23509: [LoopVectorize] Query TTI when deciding to splat IV

[Matthew Simpson via llvm-commits]

mssimpso added a comment.

In https://reviews.llvm.org/D23509#523061, @samparker wrote:

> Hi Matt,
>
> I can confirm that your approach works for me :)


Great! I hope you don't mind that a took a stab at implementing this approach over in https://reviews.llvm.org/D23889. We should now be able to scalarize instructions that LVL marks scalar after vectorization without increasing code size pre-instcombine. To do this we can check within the scalarization logic if an instruction is also uniform, and if so, only generate one scalar value per unroll iteration instead of VF values per iteration.


https://reviews.llvm.org/D23509