[PATCH] D106333: [AArch64][SVE] Handle svbool_t VLST <-> VLAT/GNUT conversion

Tue Jul 20 03:12:20 PDT 2021

junparser added a comment.

In D106333#2889859 <https://reviews.llvm.org/D106333#2889859>, @paulwalker-arm wrote:

> In D106333#2889168 <https://reviews.llvm.org/D106333#2889168>, @junparser wrote:
>
>> @efriedma with this patch,  all of conversion between VLST and VLAT should have same vector size(getElementType() * getElementCount()). The regression in D105097 <https://reviews.llvm.org/D105097> will be fixed by using bitcast + vector.insert/extract directly
>
> I hope I've not got the wrong end of the stick here but the above is our intention.  As in, Arm is looking at replacing the "via memory predicate casting" with a method that uses vector_of_i8s vector insert/extract with the necessary bitcasting.  Before doing this we just had to fix up a bunch of failing INSERT_SUBVECTOR cases when it comes to the illegal types this idiom introduces.

Good to know this.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D106333/new/

https://reviews.llvm.org/D106333