[llvm] [DAGCombine] Propagate truncate to operands (PR #98666)

[Matt Arsenault via llvm-commits]

================
@@ -182,8 +182,12 @@ define amdgpu_kernel void @v_xor_i1(ptr addrspace(1) %out, ptr addrspace(1) %in0
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    s_mov_b32 s8, s4
 ; SI-NEXT:    s_mov_b32 s9, s5
-; SI-NEXT:    v_xor_b32_e32 v0, v0, v1
 ; SI-NEXT:    v_and_b32_e32 v0, 1, v0
+; SI-NEXT:    v_and_b32_e32 v1, 1, v1
+; SI-NEXT:    v_cmp_eq_u32_e32 vcc, 1, v0
+; SI-NEXT:    v_cmp_eq_u32_e64 s[0:1], 1, v1
----------------
arsenm wrote:

Another i1 regression 

https://github.com/llvm/llvm-project/pull/98666