[PATCH] D93132: [SVE][CodeGen] Vector + immediate addressing mode for masked gather/scatter

[Kerry McLaughlin via Phabricator via llvm-commits]

kmclaughlin added inline comments.


================
Comment at: llvm/test/CodeGen/AArch64/sve-masked-gather.ll:43
+
+define <vscale x 2 x half> @masked_gather_nxv2f16(<vscale x 2 x half*> %ptrs, i64 %offset, <vscale x 2 x i1> %mask) {
+; CHECK-LABEL: masked_gather_nxv2f16:
----------------
david-arm wrote:
> Don't know if it matters or not, but the scatter equivalents also have tests for bfloat?
Good find :) I've added more tests here after posting D93307, which lowers mgathers with bfloat types


================
Comment at: llvm/test/CodeGen/AArch64/sve-masked-scatter-legalise.ll:15
+; Code generate store of an illegal datatype via promotion.
+define void @masked_scatter_nxv2i16(<vscale x 2 x i16> %data, i16* %base, <vscale x 2 x i16*> %ptrs, <vscale x 2 x i1> %mask) {
+; CHECK-LABEL: masked_scatter_nxv2i16:
----------------
david-arm wrote:
> Do we need to worry about stuff like
> 
> masked_scatter_nxv2half
> masked_scatter_nxv2float?
These tests were duplicates of those added to `sve-masked-scatter.ll`, so I've removed them from this file - in the other file there are also tests for nxv2f16, nxv2bf16, nxv2f32 & nxv2f64


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D93132/new/

https://reviews.llvm.org/D93132