[llvm] [AMDGPU] Fold constants in v2bf16 canonicalization (PR #214384)

[Matt Arsenault via llvm-commits]

================
@@ -1283,9 +1303,10 @@ define <4 x bfloat> @v_test_canonicalize_reg_undef_reg_reg_v4bf16(bfloat %val0,
 ; FAKE16:       ; %bb.0:
 ; FAKE16-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; FAKE16-NEXT:    s_wait_kmcnt 0x0
-; FAKE16-NEXT:    v_perm_b32 v1, v2, v1, 0x5040100
 ; FAKE16-NEXT:    v_pk_mul_bf16 v0, 1.0, v0 op_sel_hi:[0,1]
-; FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; FAKE16-NEXT:    v_perm_b32 v1, v2, v1, 0x5040100
----------------
arsenm wrote:

All of these changes look worse 

https://github.com/llvm/llvm-project/pull/214384