diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/mul-known-bits.i64.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/mul-known-bits.i64.ll
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/mul-known-bits.i64.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/mul-known-bits.i64.ll
@@ -74,9 +74,8 @@
 ; FUNC-NEXT:    global_load_dwordx2 v[0:1], v3, s[2:3]
 ; FUNC-NEXT:    s_waitcnt vmcnt(0)
 ; FUNC-NEXT:    v_mad_u64_u32 v[2:3], s0, v4, v0, 0
-; FUNC-NEXT:    v_mul_lo_u32 v1, v4, v1
-; FUNC-NEXT:    v_mul_lo_u32 v0, 0, v0
-; FUNC-NEXT:    v_add3_u32 v3, v3, v1, v0
+; FUNC-NEXT:    v_mul_lo_u32 v0, v4, v1
+; FUNC-NEXT:    v_add_nc_u32_e32 v3, v3, v0
 ; FUNC-NEXT:    v_mov_b32_e32 v0, 0
 ; FUNC-NEXT:    global_store_dwordx2 v0, v[2:3], s[4:5]
 ; FUNC-NEXT:    s_endpgm
@@ -105,8 +104,6 @@
 ; FUNC-NEXT:    global_load_dword v2, v0, s[2:3]
 ; FUNC-NEXT:    s_waitcnt vmcnt(0)
 ; FUNC-NEXT:    v_mad_u64_u32 v[0:1], s0, v1, v2, 0
-; FUNC-NEXT:    v_mul_lo_u32 v2, 0, v2
-; FUNC-NEXT:    v_add_nc_u32_e32 v1, v1, v2
 ; FUNC-NEXT:    v_mov_b32_e32 v2, 0
 ; FUNC-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
 ; FUNC-NEXT:    s_endpgm
@@ -136,9 +133,8 @@
 ; FUNC-NEXT:    global_load_dwordx2 v[0:1], v2, s[2:3]
 ; FUNC-NEXT:    s_waitcnt vmcnt(0)
 ; FUNC-NEXT:    v_mad_u64_u32 v[2:3], s0, v4, v0, 0
-; FUNC-NEXT:    v_mul_lo_u32 v1, v4, v1
-; FUNC-NEXT:    v_mul_lo_u32 v0, 0, v0
-; FUNC-NEXT:    v_add3_u32 v3, v3, v1, v0
+; FUNC-NEXT:    v_mul_lo_u32 v0, v4, v1
+; FUNC-NEXT:    v_add_nc_u32_e32 v3, v3, v0
 ; FUNC-NEXT:    v_mov_b32_e32 v0, 0
 ; FUNC-NEXT:    global_store_dwordx2 v0, v[2:3], s[4:5]
 ; FUNC-NEXT:    s_endpgm
@@ -158,21 +154,18 @@
 ; FUNC-LABEL: v_mul_i64_and_a_lo:
 ; FUNC:       ; %bb.0:
 ; FUNC-NEXT:    s_clause 0x1
-; FUNC-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
 ; FUNC-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; FUNC-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
 ; FUNC-NEXT:    v_lshlrev_b32_e32 v4, 3, v0
 ; FUNC-NEXT:    s_waitcnt lgkmcnt(0)
 ; FUNC-NEXT:    s_clause 0x1
-; FUNC-NEXT:    global_load_dwordx2 v[0:1], v4, s[2:3]
-; FUNC-NEXT:    global_load_dwordx2 v[2:3], v4, s[6:7]
+; FUNC-NEXT:    global_load_dwordx2 v[0:1], v4, s[6:7]
+; FUNC-NEXT:    global_load_dwordx2 v[2:3], v4, s[2:3]
 ; FUNC-NEXT:    s_waitcnt vmcnt(1)
-; FUNC-NEXT:    v_mad_u64_u32 v[4:5], s0, 0, v0, 0
-; FUNC-NEXT:    v_mul_lo_u32 v1, 0, v1
-; FUNC-NEXT:    s_waitcnt vmcnt(0)
-; FUNC-NEXT:    v_mul_lo_u32 v0, v3, v0
-; FUNC-NEXT:    v_add3_u32 v5, v5, v1, v0
 ; FUNC-NEXT:    v_mov_b32_e32 v0, 0
-; FUNC-NEXT:    global_store_dwordx2 v0, v[4:5], s[4:5]
+; FUNC-NEXT:    s_waitcnt vmcnt(0)
+; FUNC-NEXT:    v_mul_lo_u32 v1, v1, v2
+; FUNC-NEXT:    global_store_dwordx2 v0, v[0:1], s[4:5]
 ; FUNC-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep.a = getelementptr inbounds i64, i64 addrspace(1)* %aptr, i32 %tid
@@ -192,19 +185,18 @@
 ; FUNC-NEXT:    s_clause 0x1
 ; FUNC-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
 ; FUNC-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
-; FUNC-NEXT:    v_lshlrev_b32_e32 v4, 3, v0
+; FUNC-NEXT:    v_lshlrev_b32_e32 v3, 3, v0
+; FUNC-NEXT:    ; kill: killed $vgpr3
+; FUNC-NEXT:    ; kill: killed $sgpr6_sgpr7
 ; FUNC-NEXT:    s_waitcnt lgkmcnt(0)
 ; FUNC-NEXT:    s_clause 0x1
-; FUNC-NEXT:    global_load_dwordx2 v[0:1], v4, s[6:7]
-; FUNC-NEXT:    global_load_dwordx2 v[2:3], v4, s[2:3]
-; FUNC-NEXT:    s_waitcnt vmcnt(1)
-; FUNC-NEXT:    v_mad_u64_u32 v[4:5], s0, v0, 0, 0
+; FUNC-NEXT:    global_load_dwordx2 v[0:1], v3, s[6:7]
+; FUNC-NEXT:    global_load_dwordx2 v[1:2], v3, s[2:3]
+; FUNC-NEXT:    ; kill: killed $sgpr2_sgpr3
 ; FUNC-NEXT:    s_waitcnt vmcnt(0)
-; FUNC-NEXT:    v_mul_lo_u32 v0, v0, v3
-; FUNC-NEXT:    v_mul_lo_u32 v1, v1, 0
-; FUNC-NEXT:    v_add3_u32 v5, v5, v0, v1
+; FUNC-NEXT:    v_mul_lo_u32 v1, v0, v2
 ; FUNC-NEXT:    v_mov_b32_e32 v0, 0
-; FUNC-NEXT:    global_store_dwordx2 v0, v[4:5], s[4:5]
+; FUNC-NEXT:    global_store_dwordx2 v0, v[0:1], s[4:5]
 ; FUNC-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep.a = getelementptr inbounds i64, i64 addrspace(1)* %aptr, i32 %tid
@@ -282,35 +274,27 @@
 ; FUNC-NEXT:    s_clause 0x1
 ; FUNC-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
 ; FUNC-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
-; FUNC-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; FUNC-NEXT:    v_lshlrev_b32_e32 v4, 3, v0
 ; FUNC-NEXT:    s_waitcnt lgkmcnt(0)
 ; FUNC-NEXT:    s_clause 0x1
-; FUNC-NEXT:    global_load_dwordx2 v[2:3], v0, s[6:7]
-; FUNC-NEXT:    global_load_dwordx2 v[4:5], v0, s[2:3]
-; FUNC-NEXT:    ; implicit-def: $vgpr0_vgpr1
+; FUNC-NEXT:    global_load_dwordx2 v[2:3], v4, s[6:7]
+; FUNC-NEXT:    global_load_dwordx2 v[0:1], v4, s[2:3]
 ; FUNC-NEXT:    s_waitcnt vmcnt(1)
 ; FUNC-NEXT:    v_cmp_ge_u64_e32 vcc_lo, 0, v[2:3]
+; FUNC-NEXT:    s_waitcnt vmcnt(0)
+; FUNC-NEXT:    v_mul_lo_u32 v1, v2, v1
 ; FUNC-NEXT:    s_and_saveexec_b32 s0, vcc_lo
 ; FUNC-NEXT:    s_xor_b32 s0, exec_lo, s0
-; FUNC-NEXT:    s_cbranch_execz .LBB9_2
 ; FUNC-NEXT:  ; %bb.1: ; %else
-; FUNC-NEXT:    s_waitcnt vmcnt(0)
-; FUNC-NEXT:    v_mad_u64_u32 v[0:1], s1, v2, v4, 0
-; FUNC-NEXT:    v_mul_lo_u32 v2, v2, v5
-; FUNC-NEXT:    v_mul_lo_u32 v3, 0, v4
-; FUNC-NEXT:    ; implicit-def: $vgpr4_vgpr5
-; FUNC-NEXT:    v_add3_u32 v1, v1, v2, v3
-; FUNC-NEXT:    ; implicit-def: $vgpr2_vgpr3
-; FUNC-NEXT:  .LBB9_2: ; %Flow
+; FUNC-NEXT:    v_mad_u64_u32 v[2:3], s1, v2, v0, 0
+; FUNC-NEXT:    v_add_nc_u32_e32 v3, v3, v1
+; FUNC-NEXT:    v_mov_b32_e32 v0, v2
+; FUNC-NEXT:    v_mov_b32_e32 v1, v3
+; FUNC-NEXT:  ; %bb.2: ; %Flow
 ; FUNC-NEXT:    s_andn2_saveexec_b32 s0, s0
-; FUNC-NEXT:    s_cbranch_execz .LBB9_4
 ; FUNC-NEXT:  ; %bb.3: ; %if
-; FUNC-NEXT:    v_mad_u64_u32 v[0:1], s1, v2, 0, 0
-; FUNC-NEXT:    s_waitcnt vmcnt(0)
-; FUNC-NEXT:    v_mul_lo_u32 v2, v2, v5
-; FUNC-NEXT:    v_mul_lo_u32 v3, 0, 0
-; FUNC-NEXT:    v_add3_u32 v1, v1, v2, v3
-; FUNC-NEXT:  .LBB9_4: ; %endif
+; FUNC-NEXT:    v_mov_b32_e32 v0, 0
+; FUNC-NEXT:  ; %bb.4: ; %endif
 ; FUNC-NEXT:    s_or_b32 exec_lo, exec_lo, s0
 ; FUNC-NEXT:    v_mov_b32_e32 v2, 0
 ; FUNC-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]