[llvm] [AMDGPU] Remove s_delay_alu for VALU->SGPR->SALU (PR #127212)

[Ana Mihajlovic via llvm-commits]

================
@@ -2145,12 +2145,11 @@ define amdgpu_kernel void @add_i64_uniform(ptr addrspace(1) %out, ptr addrspace(
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1164-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
+; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mad_u64_u32 v[0:1], null, s4, v2, s[2:3]
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
----------------
mihajlovicana wrote:

I am not sure if this one waits for the va_sdst because it has a literal operand

https://github.com/llvm/llvm-project/pull/127212