[PATCH] D66085: [ARM] Add support for MVE vaddv

[Dave Green via Phabricator via llvm-commits]

dmgreen added a comment.

Thanks. I like the way this keeps the vecreduce_add as legal.



================
Comment at: llvm/lib/Target/ARM/ARMInstrMVE.td:553
+let Predicates = [HasMVEInt] in {
+  def : Pat<(i32 (vecreduce_add (v4i32 QPR:$src))), (i32 (MVE_VADDVu32no_acc $src))>;
+  def : Pat<(i32 (vecreduce_add (v8i16 QPR:$src))), (i32 (MVE_VADDVu16no_acc $src))>;
----------------
MQPR, over QPR.


================
Comment at: llvm/test/CodeGen/ARM/vaddv.ll:1
+; RUN: llc -mtriple=thumbv8.1m.main-arm-none-eabi -mattr=+mve.fp %s -o - | FileCheck %s
+
----------------
Can you move this test to test/Codegen/Thumb2/mve-vaddv.ll, to be with the others.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D66085/new/

https://reviews.llvm.org/D66085