[llvm] [LoadStoreVectorizer] Propagate alignment through contiguous chain (PR #145733)

[Drew Kersnar via llvm-commits]

================
@@ -0,0 +1,450 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -passes=load-store-vectorizer -S < %s | FileCheck %s
+
+; The IR has the first float3 labeled with align 16, and that 16 should
+; be propagated such that the second set of 4 values
+; can also be vectorized together.
----------------
dakersnar wrote:

> We should definitely not infer better alignments in InstCombine. We specifically split out the InferAlignments pass out of InstCombine to reduce compile-time from repeatedly inferring alignment.

I agree. There is technically a solution that involves expanding the aggregate unpacking logic to traverse multiple layers at once, which would theoretically solve this problem for "free", but I think there are other tradeoffs that come with that.

> I'm not sure we'll be losing much in practice. I'd generally expect that after CSE, cases where alignment can be increased in this manner will have the form of constant offset GEPs from a common base.

This is a reasonable argument. Let me investigate this possibility and get back to you all with my findings.

https://github.com/llvm/llvm-project/pull/145733