[llvm] AMDGPU: VALU data fast-forwarding needs no s_delay (PR #205481)

[Jay Foad via llvm-commits]

================
@@ -2825,6 +2825,7 @@ define void @store_load_v3i32_unaligned(ptr addrspace(5) nocapture %arg) {
 ; GFX11-NEXT:    s_mov_b32 s2, 3
 ; GFX11-NEXT:    s_mov_b32 s1, 2
 ; GFX11-NEXT:    s_mov_b32 s0, 1
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
----------------
jayfoad wrote:

Not sure what is happening here, maybe also caused by implicit superreg use operands?

https://github.com/llvm/llvm-project/pull/205481