[llvm] [AMDGPU] Filter candidates of LiveRegOptimizer for profitable cases (PR #124624)

[Jeffrey Byrnes via llvm-commits]

================
@@ -632,18 +2353,24 @@ define amdgpu_kernel void @v8i8_multi_block(ptr addrspace(1) %src1, ptr addrspac
 ; GFX906-LABEL: v8i8_multi_block:
 ; GFX906:       ; %bb.0: ; %entry
 ; GFX906-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
-; GFX906-NEXT:    v_lshlrev_b32_e32 v6, 3, v0
+; GFX906-NEXT:    v_lshlrev_b32_e32 v12, 3, v0
 ; GFX906-NEXT:    v_mov_b32_e32 v5, 0
 ; GFX906-NEXT:    v_cmp_gt_u32_e32 vcc, 15, v0
 ; GFX906-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX906-NEXT:    global_load_dwordx2 v[3:4], v6, s[8:9]
+; GFX906-NEXT:    global_load_dwordx2 v[3:4], v12, s[8:9]
 ; GFX906-NEXT:    s_waitcnt vmcnt(0)
+; GFX906-NEXT:    v_lshrrev_b32_e32 v6, 24, v4
+; GFX906-NEXT:    v_lshrrev_b32_e32 v7, 16, v4
+; GFX906-NEXT:    v_lshrrev_b32_e32 v8, 8, v4
+; GFX906-NEXT:    v_lshrrev_b32_e32 v9, 24, v3
+; GFX906-NEXT:    v_lshrrev_b32_e32 v10, 16, v3
+; GFX906-NEXT:    v_lshrrev_b32_e32 v11, 8, v3
----------------
jrbyrnes wrote:

Ditto

https://github.com/llvm/llvm-project/pull/124624