[AMDGPU] Fix VOPD dependency checks during combine

[rampitec]

Summary

Check superreg/subreg defs of an instruction when checking for
dependencies. This may cause some regressions, but better be
safe than sorry. Changed tests are affected because of the
implicit-defs of the superregs.

[foad]

LGTM.

Changed tests are affected because of the imlict-defs of the superregs.

Ugh, yet another reason to get rid of those implicit defs...

[Joe_Nash]

LGTM

llvm/lib/Target/AMDGPU/GCNVOPDUtils.cpp

for (auto MII = MachineBasicBlock::const_iterator(&FirstMI);
          MII != FirstMI.getParent()->instr_end(); ++MII) {
       if (&*MII == &SecondMI)
         return true;
     }
     return false;
   }() && "Expected FirstMI to precede SecondMI");
   // Cannot pair dependent instructions
   for (const auto &Use : SecondMI.uses())
-    if (Use.isReg() && FirstMI.modifiesRegister(Use.getReg()))
+    if (Use.isReg() && FirstMI.modifiesRegister(Use.getReg(), TRI))
       return false;
 
   auto getVRegIdx = [&](unsigned OpcodeIdx, unsigned OperandIdx) {
     const MachineInstr &MI = (OpcodeIdx == VOPD::X) ? FirstMI : SecondMI;
     const MachineOperand &Operand = MI.getOperand(OperandIdx);
     if (Operand.isReg() && TRI->isVectorRegister(MRI, Operand.getReg()))
       return Operand.getReg();
     return Register();

llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.i16.ll

 ; GFX11-NEXT:    v_and_b32_e32 v2, 1, v1
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v4, 1, v1
 ; GFX11-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
 ; GFX11-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 1, v4
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_lshlrev_b32 v2, 4, v2
 ; GFX11-NEXT:    v_cndmask_b32_e32 v5, s0, v1, vcc_lo
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_3) | instid1(VALU_DEP_4)
 ; GFX11-NEXT:    v_lshlrev_b32_e64 v3, v2, 0xffff
 ; GFX11-NEXT:    v_lshlrev_b32_e32 v2, v2, v0
-; GFX11-NEXT:    v_not_b32_e32 v3, v3
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_4)
-; GFX11-NEXT:    v_and_or_b32 v5, v5, v3, v2
 ; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-NEXT:    v_cmp_eq_u32_e64 s0, 0, v4
+; GFX11-NEXT:    v_not_b32_e32 v3, v3
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-NEXT:    v_and_or_b32 v5, v5, v3, v2
 ; GFX11-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-NEXT:    v_mov_b32_e32 v3, 0
-; GFX11-NEXT:    v_cndmask_b32_e32 v1, v1, v5, vcc_lo
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_4)
 ; GFX11-NEXT:    v_cndmask_b32_e64 v0, v0, v5, s0
+; GFX11-NEXT:    v_cndmask_b32_e32 v1, v1, v5, vcc_lo
 ; GFX11-NEXT:    global_store_b64 v[2:3], v[0:1], off
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %vec = load <4 x i16>, ptr addrspace(4) %ptr
   %insert = insertelement <4 x i16> %vec, i16 %val, i32 %idx
   store <4 x i16> %insert, ptr addrspace(1) null
   ret void
 }

 ; GFX11-NEXT:    v_cmp_eq_u32_e64 s0, 0, v5
 ; GFX11-NEXT:    v_not_b32_e32 v3, v4
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_cndmask_b32_e32 v4, v0, v1, vcc_lo
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
 ; GFX11-NEXT:    v_and_or_b32 v4, v4, v3, v2
 ; GFX11-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-NEXT:    v_mov_b32_e32 v3, 0
-; GFX11-NEXT:    v_cndmask_b32_e64 v0, v0, v4, s0
 ; GFX11-NEXT:    v_cndmask_b32_e32 v1, v1, v4, vcc_lo
+; GFX11-NEXT:    v_cndmask_b32_e64 v0, v0, v4, s0
 ; GFX11-NEXT:    global_store_b64 v[2:3], v[0:1], off
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %vec = load <4 x i16>, ptr addrspace(1) %ptr
   %insert = insertelement <4 x i16> %vec, i16 %val, i32 %idx
   store <4 x i16> %insert, ptr addrspace(1) null
   ret void
 }

 ; GFX10-NEXT:    global_store_dwordx4 v[11:12], v[7:10], off
 ; GFX10-NEXT:    s_endpgm
 ;
 ; GFX11-LABEL: insertelement_v_v16i16_v_s:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_clause 0x1
 ; GFX11-NEXT:    global_load_b128 v[3:6], v[0:1], off
 ; GFX11-NEXT:    global_load_b128 v[7:10], v[0:1], off offset:16
-; GFX11-NEXT:    v_and_b32_e32 v0, 0xffff, v2
+; GFX11-NEXT:    v_dual_mov_b32 v11, 16 :: v_dual_and_b32 v0, 0xffff, v2
 ; GFX11-NEXT:    s_and_b32 s0, s2, 1
 ; GFX11-NEXT:    s_lshr_b32 m0, s2, 1
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, 4
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX11-NEXT:    v_dual_mov_b32 v11, 16 :: v_dual_lshlrev_b32 v0, s0, v0
-; GFX11-NEXT:    s_lshl_b32 s0, 0xffff, s0
 ; GFX11-NEXT:    v_mov_b32_e32 v12, 0
+; GFX11-NEXT:    v_lshlrev_b32_e32 v0, s0, v0
+; GFX11-NEXT:    s_lshl_b32 s0, 0xffff, s0
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
 ; GFX11-NEXT:    s_not_b32 s0, s0
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_movrels_b32_e32 v1, v3
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
 ; GFX11-NEXT:    v_and_or_b32 v2, v1, s0, v0
 ; GFX11-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_3)
 ; GFX11-NEXT:    v_movreld_b32_e32 v3, v2
 ; GFX11-NEXT:    s_clause 0x1
 ; GFX11-NEXT:    global_store_b128 v[0:1], v[3:6], off
 ; GFX11-NEXT:    global_store_b128 v[11:12], v[7:10], off
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %vec = load <16 x i16>, ptr addrspace(1) %ptr
   %insert = insertelement <16 x i16> %vec, i16 %val, i32 %idx

llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.i8.ll

 ; GFX11-NEXT:    v_and_b32_e32 v2, 3, v1
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v4, 2, v1
 ; GFX11-NEXT:    v_and_b32_e32 v0, 0xff, v0
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
 ; GFX11-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 1, v4
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_lshlrev_b32 v2, 3, v2
 ; GFX11-NEXT:    v_cndmask_b32_e32 v5, s0, v1, vcc_lo
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_3) | instid1(VALU_DEP_4)
 ; GFX11-NEXT:    v_lshlrev_b32_e64 v3, v2, 0xff
 ; GFX11-NEXT:    v_lshlrev_b32_e32 v2, v2, v0
-; GFX11-NEXT:    v_not_b32_e32 v3, v3
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_4)
-; GFX11-NEXT:    v_and_or_b32 v5, v5, v3, v2
 ; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-NEXT:    v_cmp_eq_u32_e64 s0, 0, v4
+; GFX11-NEXT:    v_not_b32_e32 v3, v3
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-NEXT:    v_and_or_b32 v5, v5, v3, v2
 ; GFX11-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-NEXT:    v_mov_b32_e32 v3, 0
-; GFX11-NEXT:    v_cndmask_b32_e32 v1, v1, v5, vcc_lo
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_4)
 ; GFX11-NEXT:    v_cndmask_b32_e64 v0, v0, v5, s0
+; GFX11-NEXT:    v_cndmask_b32_e32 v1, v1, v5, vcc_lo
 ; GFX11-NEXT:    global_store_b64 v[2:3], v[0:1], off
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %vec = load <8 x i8>, ptr addrspace(4) %ptr
   %insert = insertelement <8 x i8> %vec, i8 %val, i32 %idx
   store <8 x i8> %insert, ptr addrspace(1) null
   ret void
 }

 ; GFX11-NEXT:    v_cmp_eq_u32_e64 s0, 0, v5
 ; GFX11-NEXT:    v_not_b32_e32 v3, v4
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_cndmask_b32_e32 v4, v0, v1, vcc_lo
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
 ; GFX11-NEXT:    v_and_or_b32 v4, v4, v3, v2
 ; GFX11-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-NEXT:    v_mov_b32_e32 v3, 0
-; GFX11-NEXT:    v_cndmask_b32_e64 v0, v0, v4, s0
 ; GFX11-NEXT:    v_cndmask_b32_e32 v1, v1, v4, vcc_lo
+; GFX11-NEXT:    v_cndmask_b32_e64 v0, v0, v4, s0
 ; GFX11-NEXT:    global_store_b64 v[2:3], v[0:1], off
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %vec = load <8 x i8>, ptr addrspace(1) %ptr
   %insert = insertelement <8 x i8> %vec, i8 %val, i32 %idx
   store <8 x i8> %insert, ptr addrspace(1) null
   ret void
 }

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.load.2d.ll

 ;
 ; GFX11-LABEL: load_2d_v4f32_xyzw_tfe:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    v_mov_b32_e32 v7, 0
 ; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v6, v1
 ; GFX11-NEXT:    s_mov_b32 s0, s2
 ; GFX11-NEXT:    s_mov_b32 s1, s3
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT:    v_mov_b32_e32 v8, v7
 ; GFX11-NEXT:    v_mov_b32_e32 v9, v7
-; GFX11-NEXT:    v_mov_b32_e32 v11, v7
 ; GFX11-NEXT:    v_mov_b32_e32 v10, v7
-; GFX11-NEXT:    v_mov_b32_e32 v8, v7
+; GFX11-NEXT:    v_mov_b32_e32 v11, v7
 ; GFX11-NEXT:    s_mov_b32 s2, s4
 ; GFX11-NEXT:    s_mov_b32 s3, s5
 ; GFX11-NEXT:    s_mov_b32 s4, s6
 ; GFX11-NEXT:    s_mov_b32 s5, s7
 ; GFX11-NEXT:    s_mov_b32 s6, s8
 ; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    v_mov_b32_e32 v0, v7
+; GFX11-NEXT:    v_dual_mov_b32 v0, v7 :: v_dual_mov_b32 v1, v8
 ; GFX11-NEXT:    v_dual_mov_b32 v2, v9 :: v_dual_mov_b32 v3, v10
-; GFX11-NEXT:    v_dual_mov_b32 v1, v8 :: v_dual_mov_b32 v4, v11
+; GFX11-NEXT:    v_mov_b32_e32 v4, v11
 ; GFX11-NEXT:    image_load v[0:4], v[5:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm tfe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v7, v4, s[10:11]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
   %v = call { <4 x float>, i32 } @llvm.amdgcn.image.load.2d.sl_v4f32i32s.i32(i32 15, i32 %s, i32 %t, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue { <4 x float>, i32 } %v, 0
   %v.err = extractvalue { <4 x float>, i32 } %v, 1

 ;
 ; GFX11-LABEL: load_2d_v4f32_xyzw_tfe_lwe:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    v_mov_b32_e32 v7, 0
 ; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v6, v1
 ; GFX11-NEXT:    s_mov_b32 s0, s2
 ; GFX11-NEXT:    s_mov_b32 s1, s3
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT:    v_mov_b32_e32 v8, v7
 ; GFX11-NEXT:    v_mov_b32_e32 v9, v7
-; GFX11-NEXT:    v_mov_b32_e32 v11, v7
 ; GFX11-NEXT:    v_mov_b32_e32 v10, v7
-; GFX11-NEXT:    v_mov_b32_e32 v8, v7
+; GFX11-NEXT:    v_mov_b32_e32 v11, v7
 ; GFX11-NEXT:    s_mov_b32 s2, s4
 ; GFX11-NEXT:    s_mov_b32 s3, s5
 ; GFX11-NEXT:    s_mov_b32 s4, s6
 ; GFX11-NEXT:    s_mov_b32 s5, s7
 ; GFX11-NEXT:    s_mov_b32 s6, s8
 ; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    v_mov_b32_e32 v0, v7
+; GFX11-NEXT:    v_dual_mov_b32 v0, v7 :: v_dual_mov_b32 v1, v8
 ; GFX11-NEXT:    v_dual_mov_b32 v2, v9 :: v_dual_mov_b32 v3, v10
-; GFX11-NEXT:    v_dual_mov_b32 v1, v8 :: v_dual_mov_b32 v4, v11
+; GFX11-NEXT:    v_mov_b32_e32 v4, v11
 ; GFX11-NEXT:    image_load v[0:4], v[5:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm tfe lwe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v7, v4, s[10:11]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
   %v = call { <4 x float>, i32 } @llvm.amdgcn.image.load.2d.sl_v4f32i32s.i32(i32 15, i32 %s, i32 %t, <8 x i32> %rsrc, i32 3, i32 0)
   %v.vec = extractvalue { <4 x float>, i32 } %v, 0
   %v.err = extractvalue { <4 x float>, i32 } %v, 1
   store i32 %v.err, ptr addrspace(1) %out, align 4
   ret <4 x float> %v.vec
 }
 
 declare <4 x float> @llvm.amdgcn.image.load.2d.v4f32.i32(i32 immarg, i32, i32, <8 x i32>, i32 immarg, i32 immarg) #0
 declare { <4 x float>, i32 } @llvm.amdgcn.image.load.2d.sl_v4f32i32s.i32(i32 immarg, i32, i32, <8 x i32>, i32 immarg, i32 immarg) #0
 
 attributes #0 = { nounwind readonly }

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.load.2darraymsaa.ll

 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: load_2darraymsaa_v4f32_xyzw_tfe:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    v_mov_b32_e32 v9, 0
 ; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v6, v1
 ; GFX11-NEXT:    v_dual_mov_b32 v7, v2 :: v_dual_mov_b32 v8, v3
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_3)
+; GFX11-NEXT:    v_mov_b32_e32 v10, v9
 ; GFX11-NEXT:    v_mov_b32_e32 v11, v9
-; GFX11-NEXT:    v_mov_b32_e32 v13, v9
 ; GFX11-NEXT:    v_mov_b32_e32 v12, v9
-; GFX11-NEXT:    v_mov_b32_e32 v10, v9
+; GFX11-NEXT:    v_mov_b32_e32 v13, v9
 ; GFX11-NEXT:    s_mov_b32 s0, s2
 ; GFX11-NEXT:    s_mov_b32 s1, s3
 ; GFX11-NEXT:    s_mov_b32 s2, s4
 ; GFX11-NEXT:    s_mov_b32 s3, s5
 ; GFX11-NEXT:    s_mov_b32 s4, s6
 ; GFX11-NEXT:    s_mov_b32 s5, s7
 ; GFX11-NEXT:    s_mov_b32 s6, s8
 ; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    v_mov_b32_e32 v0, v9
+; GFX11-NEXT:    v_dual_mov_b32 v0, v9 :: v_dual_mov_b32 v1, v10
 ; GFX11-NEXT:    v_dual_mov_b32 v2, v11 :: v_dual_mov_b32 v3, v12
-; GFX11-NEXT:    v_dual_mov_b32 v1, v10 :: v_dual_mov_b32 v4, v13
+; GFX11-NEXT:    v_mov_b32_e32 v4, v13
 ; GFX11-NEXT:    image_load v[0:4], v[5:8], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm tfe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v9, v4, s[10:11]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
   %v = call { <4 x float>, i32 } @llvm.amdgcn.image.load.2darraymsaa.sl_v4f32i32s.i32(i32 15, i32 %s, i32 %t, i32 %slice, i32 %fragid, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue { <4 x float>, i32 } %v, 0
   %v.err = extractvalue { <4 x float>, i32 } %v, 1

 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: load_2darraymsaa_v4f32_xyzw_tfe_lwe:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    v_mov_b32_e32 v9, 0
 ; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v6, v1
 ; GFX11-NEXT:    v_dual_mov_b32 v7, v2 :: v_dual_mov_b32 v8, v3
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_3)
+; GFX11-NEXT:    v_mov_b32_e32 v10, v9
 ; GFX11-NEXT:    v_mov_b32_e32 v11, v9
-; GFX11-NEXT:    v_mov_b32_e32 v13, v9
 ; GFX11-NEXT:    v_mov_b32_e32 v12, v9
-; GFX11-NEXT:    v_mov_b32_e32 v10, v9
+; GFX11-NEXT:    v_mov_b32_e32 v13, v9
 ; GFX11-NEXT:    s_mov_b32 s0, s2
 ; GFX11-NEXT:    s_mov_b32 s1, s3
 ; GFX11-NEXT:    s_mov_b32 s2, s4
 ; GFX11-NEXT:    s_mov_b32 s3, s5
 ; GFX11-NEXT:    s_mov_b32 s4, s6
 ; GFX11-NEXT:    s_mov_b32 s5, s7
 ; GFX11-NEXT:    s_mov_b32 s6, s8
 ; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    v_mov_b32_e32 v0, v9
+; GFX11-NEXT:    v_dual_mov_b32 v0, v9 :: v_dual_mov_b32 v1, v10
 ; GFX11-NEXT:    v_dual_mov_b32 v2, v11 :: v_dual_mov_b32 v3, v12
-; GFX11-NEXT:    v_dual_mov_b32 v1, v10 :: v_dual_mov_b32 v4, v13
+; GFX11-NEXT:    v_mov_b32_e32 v4, v13
 ; GFX11-NEXT:    image_load v[0:4], v[5:8], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm tfe lwe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v9, v4, s[10:11]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
   %v = call { <4 x float>, i32 } @llvm.amdgcn.image.load.2darraymsaa.sl_v4f32i32s.i32(i32 15, i32 %s, i32 %t, i32 %slice, i32 %fragid, <8 x i32> %rsrc, i32 3, i32 0)
   %v.vec = extractvalue { <4 x float>, i32 } %v, 0
   %v.err = extractvalue { <4 x float>, i32 } %v, 1
   store i32 %v.err, ptr addrspace(1) %out, align 4
   ret <4 x float> %v.vec
 }
 
 declare <4 x float> @llvm.amdgcn.image.load.2darraymsaa.v4f32.i32(i32 immarg, i32, i32, i32, i32, <8 x i32>, i32 immarg, i32 immarg) #0
 declare { <4 x float>, i32 } @llvm.amdgcn.image.load.2darraymsaa.sl_v4f32i32s.i32(i32 immarg, i32, i32, i32, i32, <8 x i32>, i32 immarg, i32 immarg) #0
 
 attributes #0 = { nounwind readonly }

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.load.3d.a16.ll

 ; GFX11-LABEL: load_3d_v4f32_xyzw_tfe:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    v_dual_mov_b32 v6, v2 :: v_dual_mov_b32 v7, 0
 ; GFX11-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX11-NEXT:    s_mov_b32 s0, s2
 ; GFX11-NEXT:    s_mov_b32 s1, s3
 ; GFX11-NEXT:    s_mov_b32 s2, s4
 ; GFX11-NEXT:    v_mov_b32_e32 v9, v7
-; GFX11-NEXT:    v_mov_b32_e32 v11, v7
-; GFX11-NEXT:    v_mov_b32_e32 v10, v7
 ; GFX11-NEXT:    v_mov_b32_e32 v8, v7
+; GFX11-NEXT:    v_mov_b32_e32 v10, v7
+; GFX11-NEXT:    v_mov_b32_e32 v11, v7
 ; GFX11-NEXT:    v_lshl_or_b32 v5, v1, 16, v0
 ; GFX11-NEXT:    s_mov_b32 s3, s5
 ; GFX11-NEXT:    s_mov_b32 s4, s6
 ; GFX11-NEXT:    s_mov_b32 s5, s7
 ; GFX11-NEXT:    s_mov_b32 s6, s8
 ; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    v_dual_mov_b32 v0, v7 :: v_dual_mov_b32 v3, v10
-; GFX11-NEXT:    v_dual_mov_b32 v1, v8 :: v_dual_mov_b32 v2, v9
+; GFX11-NEXT:    v_dual_mov_b32 v0, v7 :: v_dual_mov_b32 v1, v8
+; GFX11-NEXT:    v_dual_mov_b32 v2, v9 :: v_dual_mov_b32 v3, v10
 ; GFX11-NEXT:    v_mov_b32_e32 v4, v11
 ; GFX11-NEXT:    image_load v[0:4], v[5:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16 tfe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v7, v4, s[10:11]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
   %v = call { <4 x float>, i32 } @llvm.amdgcn.image.load.3d.sl_v4f32i32s.i16(i32 15, i16 %s, i16 %t, i16 %r, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue { <4 x float>, i32 } %v, 0

 ; GFX11-LABEL: load_3d_v4f32_xyzw_tfe_lwe:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    v_dual_mov_b32 v6, v2 :: v_dual_mov_b32 v7, 0
 ; GFX11-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX11-NEXT:    s_mov_b32 s0, s2
 ; GFX11-NEXT:    s_mov_b32 s1, s3
 ; GFX11-NEXT:    s_mov_b32 s2, s4
 ; GFX11-NEXT:    v_mov_b32_e32 v9, v7
-; GFX11-NEXT:    v_mov_b32_e32 v11, v7
-; GFX11-NEXT:    v_mov_b32_e32 v10, v7
 ; GFX11-NEXT:    v_mov_b32_e32 v8, v7
+; GFX11-NEXT:    v_mov_b32_e32 v10, v7
+; GFX11-NEXT:    v_mov_b32_e32 v11, v7
 ; GFX11-NEXT:    v_lshl_or_b32 v5, v1, 16, v0
 ; GFX11-NEXT:    s_mov_b32 s3, s5
 ; GFX11-NEXT:    s_mov_b32 s4, s6
 ; GFX11-NEXT:    s_mov_b32 s5, s7
 ; GFX11-NEXT:    s_mov_b32 s6, s8
 ; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    v_dual_mov_b32 v0, v7 :: v_dual_mov_b32 v3, v10
-; GFX11-NEXT:    v_dual_mov_b32 v1, v8 :: v_dual_mov_b32 v2, v9
+; GFX11-NEXT:    v_dual_mov_b32 v0, v7 :: v_dual_mov_b32 v1, v8
+; GFX11-NEXT:    v_dual_mov_b32 v2, v9 :: v_dual_mov_b32 v3, v10
 ; GFX11-NEXT:    v_mov_b32_e32 v4, v11
 ; GFX11-NEXT:    image_load v[0:4], v[5:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16 tfe lwe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v7, v4, s[10:11]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
   %v = call { <4 x float>, i32 } @llvm.amdgcn.image.load.3d.sl_v4f32i32s.i16(i32 15, i16 %s, i16 %t, i16 %r, <8 x i32> %rsrc, i32 3, i32 0)
   %v.vec = extractvalue { <4 x float>, i32 } %v, 0

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.load.3d.ll

 ;
 ; GFX11-LABEL: load_3d_v4f32_xyzw_tfe:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    v_dual_mov_b32 v7, v2 :: v_dual_mov_b32 v8, 0
 ; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v6, v1
 ; GFX11-NEXT:    s_mov_b32 s0, s2
 ; GFX11-NEXT:    s_mov_b32 s1, s3
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT:    v_mov_b32_e32 v9, v8
+; GFX11-NEXT:    v_mov_b32_e32 v10, v8
 ; GFX11-NEXT:    v_mov_b32_e32 v11, v8
 ; GFX11-NEXT:    v_mov_b32_e32 v12, v8
-; GFX11-NEXT:    v_mov_b32_e32 v10, v8
-; GFX11-NEXT:    v_mov_b32_e32 v9, v8
 ; GFX11-NEXT:    s_mov_b32 s2, s4
 ; GFX11-NEXT:    s_mov_b32 s3, s5
 ; GFX11-NEXT:    s_mov_b32 s4, s6
 ; GFX11-NEXT:    s_mov_b32 s5, s7
 ; GFX11-NEXT:    s_mov_b32 s6, s8
 ; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    v_mov_b32_e32 v0, v8
-; GFX11-NEXT:    v_dual_mov_b32 v2, v10 :: v_dual_mov_b32 v1, v9
-; GFX11-NEXT:    v_dual_mov_b32 v3, v11 :: v_dual_mov_b32 v4, v12
+; GFX11-NEXT:    v_dual_mov_b32 v0, v8 :: v_dual_mov_b32 v1, v9
+; GFX11-NEXT:    v_dual_mov_b32 v2, v10 :: v_dual_mov_b32 v3, v11
+; GFX11-NEXT:    v_mov_b32_e32 v4, v12
 ; GFX11-NEXT:    image_load v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm tfe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v8, v4, s[10:11]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
   %v = call { <4 x float>, i32 } @llvm.amdgcn.image.load.3d.sl_v4f32i32s.i32(i32 15, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue { <4 x float>, i32 } %v, 0
   %v.err = extractvalue { <4 x float>, i32 } %v, 1

 ;
 ; GFX11-LABEL: load_3d_v4f32_xyzw_tfe_lwe:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    v_dual_mov_b32 v7, v2 :: v_dual_mov_b32 v8, 0
 ; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v6, v1
 ; GFX11-NEXT:    s_mov_b32 s0, s2
 ; GFX11-NEXT:    s_mov_b32 s1, s3
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT:    v_mov_b32_e32 v9, v8
+; GFX11-NEXT:    v_mov_b32_e32 v10, v8
 ; GFX11-NEXT:    v_mov_b32_e32 v11, v8
 ; GFX11-NEXT:    v_mov_b32_e32 v12, v8
-; GFX11-NEXT:    v_mov_b32_e32 v10, v8
-; GFX11-NEXT:    v_mov_b32_e32 v9, v8
 ; GFX11-NEXT:    s_mov_b32 s2, s4
 ; GFX11-NEXT:    s_mov_b32 s3, s5
 ; GFX11-NEXT:    s_mov_b32 s4, s6
 ; GFX11-NEXT:    s_mov_b32 s5, s7
 ; GFX11-NEXT:    s_mov_b32 s6, s8
 ; GFX11-NEXT:    s_mov_b32 s7, s9
-; GFX11-NEXT:    v_mov_b32_e32 v0, v8
-; GFX11-NEXT:    v_dual_mov_b32 v2, v10 :: v_dual_mov_b32 v1, v9
-; GFX11-NEXT:    v_dual_mov_b32 v3, v11 :: v_dual_mov_b32 v4, v12
+; GFX11-NEXT:    v_dual_mov_b32 v0, v8 :: v_dual_mov_b32 v1, v9
+; GFX11-NEXT:    v_dual_mov_b32 v2, v10 :: v_dual_mov_b32 v3, v11
+; GFX11-NEXT:    v_mov_b32_e32 v4, v12
 ; GFX11-NEXT:    image_load v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm tfe lwe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v8, v4, s[10:11]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
   %v = call { <4 x float>, i32 } @llvm.amdgcn.image.load.3d.sl_v4f32i32s.i32(i32 15, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 3, i32 0)
   %v.vec = extractvalue { <4 x float>, i32 } %v, 0
   %v.err = extractvalue { <4 x float>, i32 } %v, 1
   store i32 %v.err, ptr addrspace(1) %out, align 4
   ret <4 x float> %v.vec
 }
 
 declare <4 x float> @llvm.amdgcn.image.load.3d.v4f32.i32(i32 immarg, i32, i32, i32, <8 x i32>, i32 immarg, i32 immarg) #0
 declare { <4 x float>, i32 } @llvm.amdgcn.image.load.3d.sl_v4f32i32s.i32(i32 immarg, i32, i32, i32, <8 x i32>, i32 immarg, i32 immarg) #0
 
 attributes #0 = { nounwind readonly }

llvm/test/CodeGen/AMDGPU/GlobalISel/mul-known-bits.i64.ll

 ; GFX11-NEXT:    s_mov_b32 s0, exec_lo
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_mul_lo_u32 v1, v2, v1
 ; GFX11-NEXT:    v_cmpx_ge_u64_e32 0, v[2:3]
 ; GFX11-NEXT:    s_xor_b32 s0, exec_lo, s0
 ; GFX11-NEXT:  ; %bb.1: ; %else
 ; GFX11-NEXT:    v_mad_u64_u32 v[2:3], null, v2, v0, 0
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT:    v_dual_mov_b32 v0, v2 :: v_dual_add_nc_u32 v3, v3, v1
-; GFX11-NEXT:    v_mov_b32_e32 v1, v3
+; GFX11-NEXT:    v_add_nc_u32_e32 v3, v3, v1
+; GFX11-NEXT:    v_dual_mov_b32 v0, v2 :: v_dual_mov_b32 v1, v3
 ; GFX11-NEXT:  ; %bb.2: ; %Flow
 ; GFX11-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX11-NEXT:  ; %bb.3: ; %if
 ; GFX11-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX11-NEXT:  ; %bb.4: ; %endif
 ; GFX11-NEXT:    s_or_b32 exec_lo, exec_lo, s0
 ; GFX11-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[4:5]

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.dim.ll

 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    global_store_dword v6, v4, s[8:9]
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: load_1d_tfe:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v6, 0
+; GFX11-NEXT:    v_mov_b32_e32 v7, v6
+; GFX11-NEXT:    v_mov_b32_e32 v8, v6
 ; GFX11-NEXT:    v_mov_b32_e32 v9, v6
 ; GFX11-NEXT:    v_mov_b32_e32 v10, v6
-; GFX11-NEXT:    v_mov_b32_e32 v8, v6
-; GFX11-NEXT:    v_mov_b32_e32 v7, v6
-; GFX11-NEXT:    v_mov_b32_e32 v0, v6
-; GFX11-NEXT:    v_dual_mov_b32 v2, v8 :: v_dual_mov_b32 v1, v7
-; GFX11-NEXT:    v_dual_mov_b32 v3, v9 :: v_dual_mov_b32 v4, v10
+; GFX11-NEXT:    v_dual_mov_b32 v0, v6 :: v_dual_mov_b32 v1, v7
+; GFX11-NEXT:    v_dual_mov_b32 v2, v8 :: v_dual_mov_b32 v3, v9
+; GFX11-NEXT:    v_mov_b32_e32 v4, v10
 ; GFX11-NEXT:    image_load v[0:4], v5, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm tfe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v6, v4, s[8:9]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0

 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    global_store_dword v6, v4, s[8:9]
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: load_1d_lwe:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v6, 0
+; GFX11-NEXT:    v_mov_b32_e32 v7, v6
+; GFX11-NEXT:    v_mov_b32_e32 v8, v6
 ; GFX11-NEXT:    v_mov_b32_e32 v9, v6
 ; GFX11-NEXT:    v_mov_b32_e32 v10, v6
-; GFX11-NEXT:    v_mov_b32_e32 v8, v6
-; GFX11-NEXT:    v_mov_b32_e32 v7, v6
-; GFX11-NEXT:    v_mov_b32_e32 v0, v6
-; GFX11-NEXT:    v_dual_mov_b32 v2, v8 :: v_dual_mov_b32 v1, v7
-; GFX11-NEXT:    v_dual_mov_b32 v3, v9 :: v_dual_mov_b32 v4, v10
+; GFX11-NEXT:    v_dual_mov_b32 v0, v6 :: v_dual_mov_b32 v1, v7
+; GFX11-NEXT:    v_dual_mov_b32 v2, v8 :: v_dual_mov_b32 v3, v9
+; GFX11-NEXT:    v_mov_b32_e32 v4, v10
 ; GFX11-NEXT:    image_load v[0:4], v5, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm lwe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v6, v4, s[8:9]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
 main_body:
   %v = call {<4 x float>, i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 2, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0

 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    global_store_dword v7, v4, s[8:9]
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: load_2d_tfe:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    v_dual_mov_b32 v7, 0 :: v_dual_mov_b32 v6, v1
-; GFX11-NEXT:    v_mov_b32_e32 v5, v0
+; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v8, v7
 ; GFX11-NEXT:    v_mov_b32_e32 v9, v7
-; GFX11-NEXT:    v_mov_b32_e32 v11, v7
 ; GFX11-NEXT:    v_mov_b32_e32 v10, v7
-; GFX11-NEXT:    v_mov_b32_e32 v8, v7
-; GFX11-NEXT:    v_mov_b32_e32 v0, v7
+; GFX11-NEXT:    v_mov_b32_e32 v11, v7
+; GFX11-NEXT:    v_dual_mov_b32 v0, v7 :: v_dual_mov_b32 v1, v8
 ; GFX11-NEXT:    v_dual_mov_b32 v2, v9 :: v_dual_mov_b32 v3, v10
-; GFX11-NEXT:    v_dual_mov_b32 v1, v8 :: v_dual_mov_b32 v4, v11
+; GFX11-NEXT:    v_mov_b32_e32 v4, v11
 ; GFX11-NEXT:    image_load v[0:4], v[5:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm tfe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v7, v4, s[8:9]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.2d.v4f32i32.i32(i32 15, i32 %s, i32 %t, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0

 ; GFX10-NEXT:    global_store_dword v8, v4, s[8:9]
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: load_3d_tfe_lwe:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v8, 0
 ; GFX11-NEXT:    v_dual_mov_b32 v7, v2 :: v_dual_mov_b32 v6, v1
+; GFX11-NEXT:    v_mov_b32_e32 v9, v8
+; GFX11-NEXT:    v_mov_b32_e32 v10, v8
 ; GFX11-NEXT:    v_mov_b32_e32 v11, v8
 ; GFX11-NEXT:    v_mov_b32_e32 v12, v8
-; GFX11-NEXT:    v_mov_b32_e32 v10, v8
-; GFX11-NEXT:    v_mov_b32_e32 v9, v8
-; GFX11-NEXT:    v_mov_b32_e32 v0, v8
-; GFX11-NEXT:    v_dual_mov_b32 v2, v10 :: v_dual_mov_b32 v1, v9
-; GFX11-NEXT:    v_dual_mov_b32 v3, v11 :: v_dual_mov_b32 v4, v12
+; GFX11-NEXT:    v_dual_mov_b32 v0, v8 :: v_dual_mov_b32 v1, v9
+; GFX11-NEXT:    v_dual_mov_b32 v2, v10 :: v_dual_mov_b32 v3, v11
+; GFX11-NEXT:    v_mov_b32_e32 v4, v12
 ; GFX11-NEXT:    image_load v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm tfe lwe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v8, v4, s[8:9]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.3d.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 3, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0

 ; GFX10-NEXT:    global_store_dword v8, v4, s[8:9]
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: load_cube_lwe:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v8, 0
 ; GFX11-NEXT:    v_dual_mov_b32 v7, v2 :: v_dual_mov_b32 v6, v1
+; GFX11-NEXT:    v_mov_b32_e32 v9, v8
+; GFX11-NEXT:    v_mov_b32_e32 v10, v8
 ; GFX11-NEXT:    v_mov_b32_e32 v11, v8
 ; GFX11-NEXT:    v_mov_b32_e32 v12, v8
-; GFX11-NEXT:    v_mov_b32_e32 v10, v8
-; GFX11-NEXT:    v_mov_b32_e32 v9, v8
-; GFX11-NEXT:    v_mov_b32_e32 v0, v8
-; GFX11-NEXT:    v_dual_mov_b32 v2, v10 :: v_dual_mov_b32 v1, v9
-; GFX11-NEXT:    v_dual_mov_b32 v3, v11 :: v_dual_mov_b32 v4, v12
+; GFX11-NEXT:    v_dual_mov_b32 v0, v8 :: v_dual_mov_b32 v1, v9
+; GFX11-NEXT:    v_dual_mov_b32 v2, v10 :: v_dual_mov_b32 v3, v11
+; GFX11-NEXT:    v_mov_b32_e32 v4, v12
 ; GFX11-NEXT:    image_load v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm lwe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v8, v4, s[8:9]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.cube.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 2, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0

 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    global_store_dword v7, v4, s[8:9]
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: load_1darray_tfe:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    v_dual_mov_b32 v7, 0 :: v_dual_mov_b32 v6, v1
-; GFX11-NEXT:    v_mov_b32_e32 v5, v0
+; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v8, v7
 ; GFX11-NEXT:    v_mov_b32_e32 v9, v7
-; GFX11-NEXT:    v_mov_b32_e32 v11, v7
 ; GFX11-NEXT:    v_mov_b32_e32 v10, v7
-; GFX11-NEXT:    v_mov_b32_e32 v8, v7
-; GFX11-NEXT:    v_mov_b32_e32 v0, v7
+; GFX11-NEXT:    v_mov_b32_e32 v11, v7
+; GFX11-NEXT:    v_dual_mov_b32 v0, v7 :: v_dual_mov_b32 v1, v8
 ; GFX11-NEXT:    v_dual_mov_b32 v2, v9 :: v_dual_mov_b32 v3, v10
-; GFX11-NEXT:    v_dual_mov_b32 v1, v8 :: v_dual_mov_b32 v4, v11
+; GFX11-NEXT:    v_mov_b32_e32 v4, v11
 ; GFX11-NEXT:    image_load v[0:4], v[5:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm tfe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v7, v4, s[8:9]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1darray.v4f32i32.i32(i32 15, i32 %s, i32 %slice, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0

 ; GFX10-NEXT:    global_store_dword v8, v4, s[8:9]
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: load_2darray_lwe:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v8, 0
 ; GFX11-NEXT:    v_dual_mov_b32 v7, v2 :: v_dual_mov_b32 v6, v1
+; GFX11-NEXT:    v_mov_b32_e32 v9, v8
+; GFX11-NEXT:    v_mov_b32_e32 v10, v8
 ; GFX11-NEXT:    v_mov_b32_e32 v11, v8
 ; GFX11-NEXT:    v_mov_b32_e32 v12, v8
-; GFX11-NEXT:    v_mov_b32_e32 v10, v8
-; GFX11-NEXT:    v_mov_b32_e32 v9, v8
-; GFX11-NEXT:    v_mov_b32_e32 v0, v8
-; GFX11-NEXT:    v_dual_mov_b32 v2, v10 :: v_dual_mov_b32 v1, v9
-; GFX11-NEXT:    v_dual_mov_b32 v3, v11 :: v_dual_mov_b32 v4, v12
+; GFX11-NEXT:    v_dual_mov_b32 v0, v8 :: v_dual_mov_b32 v1, v9
+; GFX11-NEXT:    v_dual_mov_b32 v2, v10 :: v_dual_mov_b32 v3, v11
+; GFX11-NEXT:    v_mov_b32_e32 v4, v12
 ; GFX11-NEXT:    image_load v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm lwe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v8, v4, s[8:9]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.2darray.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 2, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0

 ; GFX10-NEXT:    global_store_dword v8, v4, s[8:9]
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: load_2dmsaa_both:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v8, 0
 ; GFX11-NEXT:    v_dual_mov_b32 v7, v2 :: v_dual_mov_b32 v6, v1
+; GFX11-NEXT:    v_mov_b32_e32 v9, v8
+; GFX11-NEXT:    v_mov_b32_e32 v10, v8
 ; GFX11-NEXT:    v_mov_b32_e32 v11, v8
 ; GFX11-NEXT:    v_mov_b32_e32 v12, v8
-; GFX11-NEXT:    v_mov_b32_e32 v10, v8
-; GFX11-NEXT:    v_mov_b32_e32 v9, v8
-; GFX11-NEXT:    v_mov_b32_e32 v0, v8
-; GFX11-NEXT:    v_dual_mov_b32 v2, v10 :: v_dual_mov_b32 v1, v9
-; GFX11-NEXT:    v_dual_mov_b32 v3, v11 :: v_dual_mov_b32 v4, v12
+; GFX11-NEXT:    v_dual_mov_b32 v0, v8 :: v_dual_mov_b32 v1, v9
+; GFX11-NEXT:    v_dual_mov_b32 v2, v10 :: v_dual_mov_b32 v3, v11
+; GFX11-NEXT:    v_mov_b32_e32 v4, v12
 ; GFX11-NEXT:    image_load v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA unorm tfe lwe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v8, v4, s[8:9]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.2dmsaa.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %fragid, <8 x i32> %rsrc, i32 3, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0

 ; GFX10-NEXT:    global_store_dword v9, v4, s[8:9]
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: load_2darraymsaa_tfe:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    v_dual_mov_b32 v9, 0 :: v_dual_mov_b32 v8, v3
 ; GFX11-NEXT:    v_dual_mov_b32 v7, v2 :: v_dual_mov_b32 v6, v1
-; GFX11-NEXT:    v_mov_b32_e32 v5, v0
+; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v10, v9
 ; GFX11-NEXT:    v_mov_b32_e32 v11, v9
-; GFX11-NEXT:    v_mov_b32_e32 v13, v9
 ; GFX11-NEXT:    v_mov_b32_e32 v12, v9
-; GFX11-NEXT:    v_mov_b32_e32 v10, v9
-; GFX11-NEXT:    v_mov_b32_e32 v0, v9
+; GFX11-NEXT:    v_mov_b32_e32 v13, v9
+; GFX11-NEXT:    v_dual_mov_b32 v0, v9 :: v_dual_mov_b32 v1, v10
 ; GFX11-NEXT:    v_dual_mov_b32 v2, v11 :: v_dual_mov_b32 v3, v12
-; GFX11-NEXT:    v_dual_mov_b32 v1, v10 :: v_dual_mov_b32 v4, v13
+; GFX11-NEXT:    v_mov_b32_e32 v4, v13
 ; GFX11-NEXT:    image_load v[0:4], v[5:8], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm tfe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v9, v4, s[8:9]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.2darraymsaa.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %slice, i32 %fragid, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0

 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    global_store_dword v7, v4, s[8:9]
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: load_mip_1d_lwe:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    v_dual_mov_b32 v7, 0 :: v_dual_mov_b32 v6, v1
-; GFX11-NEXT:    v_mov_b32_e32 v5, v0
+; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v8, v7
 ; GFX11-NEXT:    v_mov_b32_e32 v9, v7
-; GFX11-NEXT:    v_mov_b32_e32 v11, v7
 ; GFX11-NEXT:    v_mov_b32_e32 v10, v7
-; GFX11-NEXT:    v_mov_b32_e32 v8, v7
-; GFX11-NEXT:    v_mov_b32_e32 v0, v7
+; GFX11-NEXT:    v_mov_b32_e32 v11, v7
+; GFX11-NEXT:    v_dual_mov_b32 v0, v7 :: v_dual_mov_b32 v1, v8
 ; GFX11-NEXT:    v_dual_mov_b32 v2, v9 :: v_dual_mov_b32 v3, v10
-; GFX11-NEXT:    v_dual_mov_b32 v1, v8 :: v_dual_mov_b32 v4, v11
+; GFX11-NEXT:    v_mov_b32_e32 v4, v11
 ; GFX11-NEXT:    image_load_mip v[0:4], v[5:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm lwe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v7, v4, s[8:9]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.mip.1d.v4f32i32.i32(i32 15, i32 %s, i32 %mip, <8 x i32> %rsrc, i32 2, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0

 ; GFX10-NEXT:    global_store_dword v8, v4, s[8:9]
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: load_mip_2d_tfe:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v8, 0
 ; GFX11-NEXT:    v_dual_mov_b32 v7, v2 :: v_dual_mov_b32 v6, v1
+; GFX11-NEXT:    v_mov_b32_e32 v9, v8
+; GFX11-NEXT:    v_mov_b32_e32 v10, v8
 ; GFX11-NEXT:    v_mov_b32_e32 v11, v8
 ; GFX11-NEXT:    v_mov_b32_e32 v12, v8
-; GFX11-NEXT:    v_mov_b32_e32 v10, v8
-; GFX11-NEXT:    v_mov_b32_e32 v9, v8
-; GFX11-NEXT:    v_mov_b32_e32 v0, v8
-; GFX11-NEXT:    v_dual_mov_b32 v2, v10 :: v_dual_mov_b32 v1, v9
-; GFX11-NEXT:    v_dual_mov_b32 v3, v11 :: v_dual_mov_b32 v4, v12
+; GFX11-NEXT:    v_dual_mov_b32 v0, v8 :: v_dual_mov_b32 v1, v9
+; GFX11-NEXT:    v_dual_mov_b32 v2, v10 :: v_dual_mov_b32 v3, v11
+; GFX11-NEXT:    v_mov_b32_e32 v4, v12
 ; GFX11-NEXT:    image_load_mip v[0:4], v[5:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm tfe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v8, v4, s[8:9]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.mip.2d.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %mip, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0

 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    global_store_dword v5, v3, s[8:9]
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: load_1d_tfe_V4_dmask3:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    v_dual_mov_b32 v4, v0 :: v_dual_mov_b32 v5, 0
+; GFX11-NEXT:    v_mov_b32_e32 v6, v5
 ; GFX11-NEXT:    v_mov_b32_e32 v7, v5
 ; GFX11-NEXT:    v_mov_b32_e32 v8, v5
-; GFX11-NEXT:    v_mov_b32_e32 v6, v5
-; GFX11-NEXT:    v_dual_mov_b32 v0, v5 :: v_dual_mov_b32 v3, v8
-; GFX11-NEXT:    v_dual_mov_b32 v1, v6 :: v_dual_mov_b32 v2, v7
+; GFX11-NEXT:    v_dual_mov_b32 v0, v5 :: v_dual_mov_b32 v1, v6
+; GFX11-NEXT:    v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
 ; GFX11-NEXT:    image_load v[0:3], v4, s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_1D unorm tfe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v5, v3, s[8:9]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 7, i32 %s, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0

 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    global_store_dword v4, v2, s[8:9]
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: load_1d_tfe_V4_dmask2:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    v_dual_mov_b32 v3, v0 :: v_dual_mov_b32 v4, 0
-; GFX11-NEXT:    v_mov_b32_e32 v6, v4
 ; GFX11-NEXT:    v_mov_b32_e32 v5, v4
-; GFX11-NEXT:    v_mov_b32_e32 v0, v4
-; GFX11-NEXT:    v_dual_mov_b32 v2, v6 :: v_dual_mov_b32 v1, v5
+; GFX11-NEXT:    v_mov_b32_e32 v6, v4
+; GFX11-NEXT:    v_dual_mov_b32 v0, v4 :: v_dual_mov_b32 v1, v5
+; GFX11-NEXT:    v_mov_b32_e32 v2, v6
 ; GFX11-NEXT:    image_load v[0:2], v3, s[0:7] dmask:0x6 dim:SQ_RSRC_IMG_1D unorm tfe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v4, v2, s[8:9]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 6, i32 %s, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.dim.ll

 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: sample_1d_tfe:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    s_mov_b32 s14, exec_lo
 ; GFX11-NEXT:    s_wqm_b32 exec_lo, exec_lo
 ; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v6, 0
+; GFX11-NEXT:    v_mov_b32_e32 v7, v6
+; GFX11-NEXT:    v_mov_b32_e32 v8, v6
 ; GFX11-NEXT:    v_mov_b32_e32 v9, v6
 ; GFX11-NEXT:    v_mov_b32_e32 v10, v6
-; GFX11-NEXT:    v_mov_b32_e32 v8, v6
-; GFX11-NEXT:    v_mov_b32_e32 v7, v6
-; GFX11-NEXT:    v_mov_b32_e32 v0, v6
-; GFX11-NEXT:    v_dual_mov_b32 v2, v8 :: v_dual_mov_b32 v1, v7
-; GFX11-NEXT:    v_dual_mov_b32 v3, v9 :: v_dual_mov_b32 v4, v10
+; GFX11-NEXT:    v_dual_mov_b32 v0, v6 :: v_dual_mov_b32 v1, v7
+; GFX11-NEXT:    v_dual_mov_b32 v2, v8 :: v_dual_mov_b32 v3, v9
+; GFX11-NEXT:    v_mov_b32_e32 v4, v10
 ; GFX11-NEXT:    s_and_b32 exec_lo, exec_lo, s14
 ; GFX11-NEXT:    image_sample v[0:4], v5, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D tfe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v6, v4, s[12:13]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.sample.1d.v4f32i32.f32(i32 15, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 1, i32 0)

 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: sample_1d_lwe:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    s_mov_b32 s14, exec_lo
 ; GFX11-NEXT:    s_wqm_b32 exec_lo, exec_lo
 ; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v6, 0
+; GFX11-NEXT:    v_mov_b32_e32 v7, v6
+; GFX11-NEXT:    v_mov_b32_e32 v8, v6
 ; GFX11-NEXT:    v_mov_b32_e32 v9, v6
 ; GFX11-NEXT:    v_mov_b32_e32 v10, v6
-; GFX11-NEXT:    v_mov_b32_e32 v8, v6
-; GFX11-NEXT:    v_mov_b32_e32 v7, v6
-; GFX11-NEXT:    v_mov_b32_e32 v0, v6
-; GFX11-NEXT:    v_dual_mov_b32 v2, v8 :: v_dual_mov_b32 v1, v7
-; GFX11-NEXT:    v_dual_mov_b32 v3, v9 :: v_dual_mov_b32 v4, v10
+; GFX11-NEXT:    v_dual_mov_b32 v0, v6 :: v_dual_mov_b32 v1, v7
+; GFX11-NEXT:    v_dual_mov_b32 v2, v8 :: v_dual_mov_b32 v3, v9
+; GFX11-NEXT:    v_mov_b32_e32 v4, v10
 ; GFX11-NEXT:    s_and_b32 exec_lo, exec_lo, s14
 ; GFX11-NEXT:    image_sample v[0:4], v5, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D lwe
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v6, v4, s[12:13]
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    ; return to shader part epilog
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.sample.1d.v4f32i32.f32(i32 15, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 2, i32 0)

llvm/test/CodeGen/AMDGPU/vopd-combine.mir

--- |
   define void @vopd_constants_inlinable() { ret void }
   define void @vopd_constants_same() { ret void }
   define void @vopd_mov_fmaak_constants_same() { ret void }
   define void @vopd_debug() { ret void }
   define void @vopd_schedule_unconstrained() { ret void }
   define void @vopd_schedule_unconstrained_2() { ret void }
   define void @vopd_mov_fixup() { ret void }
   define void @vopd_mov_fixup_fail() { ret void }
+  define void @vopd_no_combine_dependent_subreg() { ret void }
 ...
 
 ---
 name:            vopd_schedule
 tracksRegLiveness: true
 body:             |
   bb.0:
 
bb.0:
     ; SCHED: $vgpr0 = V_MOV_B32_e32 target-flags(amdgpu-abs32-lo) @lds, implicit $exec
     ; SCHED-NEXT: $vgpr1 = V_MOV_B32_e32 target-flags(amdgpu-abs32-lo) @lds + 4, implicit $exec
     ; PAIR-LABEL: name: vopd_mov_fixup_fail
     ; PAIR: $vgpr0 = V_MOV_B32_e32 target-flags(amdgpu-abs32-lo) @lds, implicit $exec
     ; PAIR-NEXT: $vgpr1 = V_MOV_B32_e32 target-flags(amdgpu-abs32-lo) @lds + 4, implicit $exec
     $vgpr0 = V_MOV_B32_e32 target-flags(amdgpu-abs32-lo) @lds, implicit $exec
     $vgpr1 = V_MOV_B32_e32 target-flags(amdgpu-abs32-lo) @lds + 4, implicit $exec
 ...
+
+---
+name:            vopd_no_combine_dependent_subreg
+tracksRegLiveness: true
+body:             |
+  bb.0:
+
+    ; SCHED-LABEL: name: vopd_no_combine_dependent_subreg
+    ; SCHED: $vgpr0 = IMPLICIT_DEF
+    ; SCHED-NEXT: $vgpr2 = V_MOV_B32_e32 0, implicit-def $vgpr2_vgpr3, implicit $exec
+    ; SCHED-NEXT: $vgpr5 = V_ADD_F32_e32 killed $vgpr0, killed $vgpr3, implicit $mode, implicit $exec
+    ; PAIR-LABEL: name: vopd_no_combine_dependent_subreg
+    ; PAIR: $vgpr0 = IMPLICIT_DEF
+    ; PAIR-NEXT: $vgpr2 = V_MOV_B32_e32 0, implicit-def $vgpr2_vgpr3, implicit $exec
+    ; PAIR-NEXT: $vgpr5 = V_ADD_F32_e32 killed $vgpr0, killed $vgpr3, implicit $mode, implicit $exec
+    $vgpr0 = IMPLICIT_DEF
+    $vgpr2 = V_MOV_B32_e32 0, implicit-def $vgpr2_vgpr3, implicit $exec
+    $vgpr5 = V_ADD_F32_e32 $vgpr0, $vgpr3, implicit $mode, implicit $exec
+...