[PATCH] D93129: [LV] Do not use vector type to compute cost of scalar address comp.

[Florian Hahn via Phabricator via llvm-commits]

fhahn added a comment.

In D93129#2509597 <https://reviews.llvm.org/D93129#2509597>, @pengfei wrote:

> I don't have much experience on performance tuning. Adding more people who contributed to the affected tests.
>
>> I checked on both ARM64 and X86 (skylake), and the vectorized versions are roughly 2x faster than the scalar loops.
>
> How did you check the performance? Can we evaluate the affected tests `inv_val_store_to_inv_address_conditional` and `masked_strided2` using the same method?

I built used a C wrapper which called the IR function and linked them together. I can try the same for the others ,but it's a bit time consuming.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D93129/new/

https://reviews.llvm.org/D93129