[llvm] [LV]: Teach LV to recursively (de)interleave. (PR #89018)

[Paul Walker via llvm-commits]

================
@@ -1,5 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 2
-; RUN: opt < %s -passes=interleaved-access -mtriple=aarch64-linux-gnu -mattr=+sve -S | FileCheck %s
+; RUN: opt < %s -passes=loop-vectorize,interleaved-access -mtriple=aarch64-linux-gnu -mattr=+sve -S | FileCheck %s
----------------
paulwalker-arm wrote:

Structurally I don't think it makes sense for InterleavedAccess tests to also be testing LoopVectorize. I guess you want the comfort of proving the output from vectorisation will be converted to the target specific intrinsics so perhaps `llvm/test/Transforms/PhaseOrdering/AArch64/` is a better home?

We should also have a dedicated loop-vectorize only test in `llvm/tesst/LoopVectorize/AArch64/sve-interleaved-accesses.ll`.



https://github.com/llvm/llvm-project/pull/89018