[PATCH] D85725: [Transforms][SROA] Skip uses of allocas where the type is scalable

[Cullen Rhodes via Phabricator via llvm-commits]

c-rhodes marked an inline comment as done.
c-rhodes added inline comments.


================
Comment at: llvm/test/Transforms/SROA/scalable-vectors.ll:54
+  %type.addr = alloca <16 x i32>
+  %1 = bitcast <16 x i32>* %type to <vscale x 4 x i32>*
+  store <vscale x 4 x i32> %type.coerce, <vscale x 4 x i32>* %1
----------------
c-rhodes wrote:
> david-arm wrote:
> > I think this is undefined behaviour for anything other than 512 fixed length vectors. Is it possible to amend the opt line to specify the fixed length vector size?
> > I think this is undefined behaviour for anything other than 512 fixed length vectors. Is it possible to amend the opt line to specify the fixed length vector size?
> 
> Good point, I can add the `-aarch64-sve-vector-bits-min=` flag.
>> I think this is undefined behaviour for anything other than 512 fixed length vectors. Is it possible to amend the opt line to specify the fixed length vector size?

> Good point, I can add the -aarch64-sve-vector-bits-min= flag.

I've added new check lines for these tests that specify `-aarch64-sve-vector-bits-min=512`. I also tried specifying it in an attribute group but I wasn't sure it was having any effect since it didn't complain when I put garbage in there.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D85725/new/

https://reviews.llvm.org/D85725