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

[Jay Foad via llvm-commits]

================
@@ -196,8 +196,9 @@ define float @v_fdiv_f32(float %a, float %b) #1 {
 ; GFX11-FLUSH-NEXT:    v_fmac_f32_e32 v5, v6, v3
 ; GFX11-FLUSH-NEXT:    v_fma_f32 v2, -v2, v5, v4
 ; GFX11-FLUSH-NEXT:    s_denorm_mode 0
-; GFX11-FLUSH-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FLUSH-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
----------------
jayfoad wrote:

New delay due to implicit read/write of MODE register.

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