[llvm] [SROA] Preserve inherited alignment when rewriting allocas (PR #222967)

[Yonah Goldberg via llvm-commits]

================
@@ -4,8 +4,8 @@
 ; RUN: opt < %s -mtriple nvptx64 -mcpu=sm_60 -mattr=ptx77 -passes=nvptx-lower-args -S | FileCheck %s --check-prefixes=COMMON,LOWER-ARGS,SM_60
 ; RUN: opt < %s -mtriple nvptx64 -mcpu=sm_70 -mattr=ptx77 -passes=nvptx-lower-args -S | FileCheck %s --check-prefixes=COMMON,LOWER-ARGS,SM_70
 ; RUN: opt < %s -mtriple nvptx64 -mcpu=sm_70 -mattr=ptx77 -passes=nvptx-copy-byval-args -S | FileCheck %s --check-prefixes=COMMON,COPY
-; RUN: llc < %s -mcpu=sm_60 -mattr=ptx77 | FileCheck %s --check-prefixes=PTX,PTX_60
-; RUN: llc < %s -mcpu=sm_70 -mattr=ptx77 | FileCheck %s --check-prefixes=PTX,PTX_70
+; RUN: opt < %s -mtriple=nvptx64 -mcpu=sm_60 -mattr=ptx77 -passes='nvptx-lower-args,sroa<preserve-cfg;aggregate-to-vector>,infer-alignment' | llc -mcpu=sm_60 -mattr=ptx77 | FileCheck %s --check-prefixes=PTX,PTX_60
----------------
YonahGoldberg wrote:

What is happening here? Please explain why you're updating these RUN lines.

https://github.com/llvm/llvm-project/pull/222967