[PATCH] D93030: [AArch64][SVE]Add cost model for masked gather and scatter for scalable vector.

[Sander de Smalen via Phabricator via llvm-commits]

sdesmalen added inline comments.


================
Comment at: llvm/include/llvm/Analysis/TargetTransformInfoImpl.h:355
 
+  Optional<unsigned> getMaxVScale() const { return 1; }
+
----------------
sdesmalen wrote:
> ctetreau wrote:
> > Is a concrete 1 here really the best default? I would think that for scalable architectures, the maximum vscale is almost certainly larger than 1.
> > 
> > Regardless, I would think that it should be none by default. It's an unknowable runtime constant after all.
> Agreed, None should be the default.
The idea is that not all targets may have an architectural maximum, and so the question 'getMaxVScale' may result in a valid answer, e.g. for SVE, the architectural maximum is 2048 bits, but for other targets that may not be the case.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D93030/new/

https://reviews.llvm.org/D93030