[llvm] [SLPVectorizer] Widen strided loads. (PR #153074)

[Mikhail Gudim via llvm-commits]

================
@@ -0,0 +1,483 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -mtriple=riscv64 -mattr=+m,+v,+unaligned-vector-mem \
+; RUN: -riscv-v-slp-prefer-alt-opc-vectorization=true \
+; RUN: -passes=slp-vectorizer -S < %s | FileCheck %s
+; Function Attrs: nounwind uwtable vscale_range(8,1024)
----------------
mgudim wrote:

this is the satd loop from x264. This test should probably get removed, but it's here for now just to show what code we generate with this patch.

https://github.com/llvm/llvm-project/pull/153074