[llvm] [AMDGPU] Enable unaligned scratch accesses (PR #110219)

[Matt Arsenault via llvm-commits]

================
@@ -2432,23 +2432,93 @@ define void @store_load_i64_unaligned(ptr addrspace(5) nocapture %arg) {
 ; GFX12-NEXT:    s_wait_samplecnt 0x0
 ; GFX12-NEXT:    s_wait_bvhcnt 0x0
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    v_mov_b32_e32 v1, 15
-; GFX12-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-NEXT:    v_dual_mov_b32 v1, 15 :: v_dual_mov_b32 v2, 0
 ; GFX12-NEXT:    s_wait_storecnt 0x0
-; GFX12-NEXT:    scratch_store_b64 v0, v[1:2], off scope:SCOPE_SYS
+; GFX12-NEXT:    scratch_store_b8 v0, v1, off scope:SCOPE_SYS
 ; GFX12-NEXT:    s_wait_storecnt 0x0
-; GFX12-NEXT:    scratch_load_b64 v[0:1], v0, off scope:SCOPE_SYS
+; GFX12-NEXT:    scratch_store_b8 v0, v2, off offset:1 scope:SCOPE_SYS
----------------
arsenm wrote:

I don't think there was any change here in gfx12 

https://github.com/llvm/llvm-project/pull/110219