[llvm] [AArch64][SDAG] Enable +sme2p2/+sve2p2 lowering for MSTORE (PR #217609)

[Benjamin Maxwell via llvm-commits]

================
@@ -1,17 +1,51 @@
-; RUN: llc -mtriple=aarch64 -mattr=+sve2p2 < %s
-
-;; These masked.compressstore operations could be natively supported with +sve2p2
-;; (or by promoting to 32/64 bit elements + a truncstore), but currently are not
-;; supported.
-
-; XFAIL: *
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+; RUN: llc -mtriple=aarch64 -mattr=+sve2p2 < %s | FileCheck %s --check-prefixes=CHECK,CHECK-BASE --allow-unused-prefixes
+; RUN: llc -mtriple=aarch64 -aarch64-sve-vector-bits-min=256 -mattr=+sve2p2 < %s | FileCheck %s --check-prefixes=CHECK,CHECK-VL256 --allow-unused-prefixes
----------------
MacDue wrote:

Could you add runlines for SME with/without streaming mode:

E.g.

```
; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve,+sme2p2 < %s | FileCheck %s --check-prefixes=CHECK,CHECK-SVE2p2
; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sme2p2 -force-streaming < %s | FileCheck %s --check-prefixes=CHECK,CHECK-SME2p2-STREAMING
; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sme2p2,+sve2p2 -force-streaming-compatible < %s | FileCheck %s --check-prefixes=CHECK,CHECK-STREAMING-COMPAT
```

I think all these cases should be able to use the new `compact` instruction.


https://github.com/llvm/llvm-project/pull/217609