[Sink] Allow sinking of invariant loads across critical edges

[critson]

Summary

Invariant loads can always be sunk.

[foad]

Makes sense to me.

Could do a similar thing in InstCombine's TryToSinkInstruction.

[nikic]

Please add an IR test (in llvm/test/Transforms/Sink/).

[arsenm]

llvm/lib/Transforms/Scalar/Sink.cpp

83

Probably should go through the AA query for pointsToConstantMemory instead of directly checking the metadata

[arsenm]

llvm/lib/Transforms/Scalar/Sink.cpp

83

I guess that's a bit different since it's on the source operand, so a future possible improvements

[critson]

• Add test llvm/test/Transforms/Sink/invariant-load.ll

llvm/lib/Transforms/Scalar/Sink.cpp

static bool IsAcceptableTarget(Instruction *Inst, BasicBlock *SuccToSinkTo,
 
   // If the block has multiple predecessors, this would introduce computation
   // on different code paths.  We could split the critical edge, but for now we
   // just punt.
   // FIXME: Split critical edges if not backedges.
   if (SuccToSinkTo->getUniquePredecessor() != Inst->getParent()) {
     // We cannot sink a load across a critical edge - there may be stores in
     // other code paths.
-    if (Inst->mayReadFromMemory())
+    if (Inst->mayReadFromMemory() &&
+        !Inst->hasMetadata(LLVMContext::MD_invariant_load))
       return false;
 
     // We don't want to sink across a critical edge if we don't dominate the
     // successor. We could be introducing calculations to new code paths.
     if (!DT.dominates(Inst->getParent(), SuccToSinkTo))
       return false;
 
     // Don't sink instructions into a loop.

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.div.fmas.ll

 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0xa
 ; GFX7-NEXT:    v_lshlrev_b32_e32 v1, 2, v0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX7-NEXT:    s_mov_b32 s6, 0
 ; GFX7-NEXT:    s_mov_b32 s7, 0xf000
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    buffer_load_dwordx3 v[1:3], v[1:2], s[4:7], 0 addr64
-; GFX7-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
 ; GFX7-NEXT:    s_mov_b32 s6, 0
 ; GFX7-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX7-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX7-NEXT:    s_cbranch_execz .LBB13_2
 ; GFX7-NEXT:  ; %bb.1: ; %bb
-; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x14
+; GFX7-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x14
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7-NEXT:    s_load_dword s0, s[0:1], 0x0
+; GFX7-NEXT:    s_load_dword s4, s[4:5], 0x0
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7-NEXT:    s_cmp_lg_u32 s0, 0
+; GFX7-NEXT:    s_cmp_lg_u32 s4, 0
 ; GFX7-NEXT:    s_cselect_b32 s6, 1, 0
 ; GFX7-NEXT:  .LBB13_2: ; %exit
 ; GFX7-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX7-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
 ; GFX7-NEXT:    s_and_b32 s0, 1, s6
 ; GFX7-NEXT:    v_cmp_ne_u32_e64 vcc, 0, s0
 ; GFX7-NEXT:    s_mov_b32 s6, -1
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    s_nop 1
 ; GFX7-NEXT:    v_div_fmas_f32 v0, v1, v2, v3
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    buffer_store_dword v0, off, s[4:7], 0 offset:8
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: test_div_fmas_f32_i1_phi_vcc:
 ; GFX8:       ; %bb.0: ; %entry
 ; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x28
 ; GFX8-NEXT:    v_lshlrev_b32_e32 v3, 2, v0
-; GFX8-NEXT:    s_mov_b32 s6, 0
+; GFX8-NEXT:    s_mov_b32 s4, 0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s3
 ; GFX8-NEXT:    v_add_u32_e32 v1, vcc, v1, v3
 ; GFX8-NEXT:    v_addc_u32_e32 v2, vcc, 0, v2, vcc
 ; GFX8-NEXT:    flat_load_dwordx3 v[1:3], v[1:2]
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB13_2
 ; GFX8-NEXT:  ; %bb.1: ; %bb
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x50
+; GFX8-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x50
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s0, s[0:1], 0x0
+; GFX8-NEXT:    s_load_dword s4, s[4:5], 0x0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_cmp_lg_u32 s0, 0
-; GFX8-NEXT:    s_cselect_b32 s6, 1, 0
+; GFX8-NEXT:    s_cmp_lg_u32 s4, 0
+; GFX8-NEXT:    s_cselect_b32 s4, 1, 0
 ; GFX8-NEXT:  .LBB13_2: ; %exit
-; GFX8-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_add_u32 s0, s2, 8
-; GFX8-NEXT:    s_addc_u32 s1, s3, 0
-; GFX8-NEXT:    s_and_b32 s2, 1, s6
+; GFX8-NEXT:    s_add_u32 s0, s0, 8
+; GFX8-NEXT:    s_addc_u32 s1, s1, 0
+; GFX8-NEXT:    s_and_b32 s2, 1, s4
 ; GFX8-NEXT:    v_cmp_ne_u32_e64 vcc, 0, s2
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
 ; GFX8-NEXT:    s_nop 2
 ; GFX8-NEXT:    v_div_fmas_f32 v2, v1, v2, v3
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    flat_store_dword v[0:1], v2
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX10_W32-LABEL: test_div_fmas_f32_i1_phi_vcc:
 ; GFX10_W32:       ; %bb.0: ; %entry
 ; GFX10_W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x28
 ; GFX10_W32-NEXT:    v_lshlrev_b32_e32 v1, 2, v0
 ; GFX10_W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX10_W32-NEXT:    s_mov_b32 s5, 0
 ; GFX10_W32-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10_W32-NEXT:    global_load_dwordx3 v[1:3], v1, s[2:3]
 ; GFX10_W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10_W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
-; GFX10_W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX10_W32-NEXT:    s_mov_b32 s2, 0
+; GFX10_W32-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX10_W32-NEXT:    s_cbranch_execz .LBB13_2
 ; GFX10_W32-NEXT:  ; %bb.1: ; %bb
-; GFX10_W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x50
+; GFX10_W32-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x50
 ; GFX10_W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10_W32-NEXT:    s_load_dword s0, s[0:1], 0x0
+; GFX10_W32-NEXT:    s_load_dword s2, s[4:5], 0x0
 ; GFX10_W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10_W32-NEXT:    s_cmp_lg_u32 s0, 0
-; GFX10_W32-NEXT:    s_cselect_b32 s5, 1, 0
+; GFX10_W32-NEXT:    s_cmp_lg_u32 s2, 0
+; GFX10_W32-NEXT:    s_cselect_b32 s2, 1, 0
 ; GFX10_W32-NEXT:  .LBB13_2: ; %exit
-; GFX10_W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
-; GFX10_W32-NEXT:    s_and_b32 s0, 1, s5
-; GFX10_W32-NEXT:    v_cmp_ne_u32_e64 vcc_lo, 0, s0
+; GFX10_W32-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX10_W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x0
+; GFX10_W32-NEXT:    s_and_b32 s2, 1, s2
+; GFX10_W32-NEXT:    v_cmp_ne_u32_e64 vcc_lo, 0, s2
 ; GFX10_W32-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10_W32-NEXT:    v_div_fmas_f32 v0, v1, v2, v3
 ; GFX10_W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10_W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10_W32-NEXT:    global_store_dword v1, v0, s[2:3] offset:8
+; GFX10_W32-NEXT:    global_store_dword v1, v0, s[0:1] offset:8
 ; GFX10_W32-NEXT:    s_endpgm
 ;
 ; GFX10_W64-LABEL: test_div_fmas_f32_i1_phi_vcc:
 ; GFX10_W64:       ; %bb.0: ; %entry
 ; GFX10_W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x28
 ; GFX10_W64-NEXT:    v_lshlrev_b32_e32 v1, 2, v0
 ; GFX10_W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX10_W64-NEXT:    s_mov_b32 s6, 0
+; GFX10_W64-NEXT:    s_mov_b32 s4, 0
 ; GFX10_W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10_W64-NEXT:    global_load_dwordx3 v[1:3], v1, s[2:3]
 ; GFX10_W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10_W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
-; GFX10_W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX10_W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10_W64-NEXT:    s_cbranch_execz .LBB13_2
 ; GFX10_W64-NEXT:  ; %bb.1: ; %bb
-; GFX10_W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x50
+; GFX10_W64-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x50
 ; GFX10_W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10_W64-NEXT:    s_load_dword s0, s[0:1], 0x0
+; GFX10_W64-NEXT:    s_load_dword s4, s[4:5], 0x0
 ; GFX10_W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10_W64-NEXT:    s_cmp_lg_u32 s0, 0
-; GFX10_W64-NEXT:    s_cselect_b32 s6, 1, 0
+; GFX10_W64-NEXT:    s_cmp_lg_u32 s4, 0
+; GFX10_W64-NEXT:    s_cselect_b32 s4, 1, 0
 ; GFX10_W64-NEXT:  .LBB13_2: ; %exit
-; GFX10_W64-NEXT:    s_or_b64 exec, exec, s[4:5]
-; GFX10_W64-NEXT:    s_and_b32 s0, 1, s6
-; GFX10_W64-NEXT:    v_cmp_ne_u32_e64 vcc, 0, s0
+; GFX10_W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10_W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x0
+; GFX10_W64-NEXT:    s_and_b32 s2, 1, s4
+; GFX10_W64-NEXT:    v_cmp_ne_u32_e64 vcc, 0, s2
 ; GFX10_W64-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10_W64-NEXT:    v_div_fmas_f32 v0, v1, v2, v3
 ; GFX10_W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10_W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10_W64-NEXT:    global_store_dword v1, v0, s[2:3] offset:8
+; GFX10_W64-NEXT:    global_store_dword v1, v0, s[0:1] offset:8
 ; GFX10_W64-NEXT:    s_endpgm
 ;
 ; GFX11_W32-LABEL: test_div_fmas_f32_i1_phi_vcc:
 ; GFX11_W32:       ; %bb.0: ; %entry
 ; GFX11_W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x28
 ; GFX11_W32-NEXT:    v_lshlrev_b32_e32 v1, 2, v0
-; GFX11_W32-NEXT:    s_mov_b32 s5, 0
-; GFX11_W32-NEXT:    s_mov_b32 s4, exec_lo
 ; GFX11_W32-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11_W32-NEXT:    global_load_b96 v[1:3], v1, s[2:3]
-; GFX11_W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
+; GFX11_W32-NEXT:    s_mov_b32 s2, 0
+; GFX11_W32-NEXT:    s_mov_b32 s3, exec_lo
 ; GFX11_W32-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11_W32-NEXT:    s_cbranch_execz .LBB13_2
 ; GFX11_W32-NEXT:  ; %bb.1: ; %bb
-; GFX11_W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x50
+; GFX11_W32-NEXT:    s_load_b64 s[4:5], s[0:1], 0x50
 ; GFX11_W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11_W32-NEXT:    s_load_b32 s0, s[0:1], 0x0
+; GFX11_W32-NEXT:    s_load_b32 s2, s[4:5], 0x0
 ; GFX11_W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11_W32-NEXT:    s_cmp_lg_u32 s0, 0
-; GFX11_W32-NEXT:    s_cselect_b32 s5, 1, 0
+; GFX11_W32-NEXT:    s_cmp_lg_u32 s2, 0
+; GFX11_W32-NEXT:    s_cselect_b32 s2, 1, 0
 ; GFX11_W32-NEXT:  .LBB13_2: ; %exit
-; GFX11_W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
-; GFX11_W32-NEXT:    s_and_b32 s0, 1, s5
-; GFX11_W32-NEXT:    v_cmp_ne_u32_e64 vcc_lo, 0, s0
+; GFX11_W32-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX11_W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x0
+; GFX11_W32-NEXT:    s_and_b32 s2, 1, s2
+; GFX11_W32-NEXT:    v_cmp_ne_u32_e64 vcc_lo, 0, s2
 ; GFX11_W32-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11_W32-NEXT:    v_div_fmas_f32 v0, v1, v2, v3
 ; GFX11_W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11_W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11_W32-NEXT:    global_store_b32 v1, v0, s[2:3] offset:8
+; GFX11_W32-NEXT:    global_store_b32 v1, v0, s[0:1] offset:8
 ; GFX11_W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11_W32-NEXT:    s_endpgm
 ;
 ; GFX11_W64-LABEL: test_div_fmas_f32_i1_phi_vcc:
 ; GFX11_W64:       ; %bb.0: ; %entry
 ; GFX11_W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x28
 ; GFX11_W64-NEXT:    v_lshlrev_b32_e32 v1, 2, v0
-; GFX11_W64-NEXT:    s_mov_b32 s6, 0
-; GFX11_W64-NEXT:    s_mov_b64 s[4:5], exec
+; GFX11_W64-NEXT:    s_mov_b32 s4, 0
 ; GFX11_W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11_W64-NEXT:    global_load_b96 v[1:3], v1, s[2:3]
-; GFX11_W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
+; GFX11_W64-NEXT:    s_mov_b64 s[2:3], exec
 ; GFX11_W64-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11_W64-NEXT:    s_cbranch_execz .LBB13_2
 ; GFX11_W64-NEXT:  ; %bb.1: ; %bb
-; GFX11_W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x50
+; GFX11_W64-NEXT:    s_load_b64 s[4:5], s[0:1], 0x50
 ; GFX11_W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11_W64-NEXT:    s_load_b32 s0, s[0:1], 0x0
+; GFX11_W64-NEXT:    s_load_b32 s4, s[4:5], 0x0
 ; GFX11_W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11_W64-NEXT:    s_cmp_lg_u32 s0, 0
-; GFX11_W64-NEXT:    s_cselect_b32 s6, 1, 0
+; GFX11_W64-NEXT:    s_cmp_lg_u32 s4, 0
+; GFX11_W64-NEXT:    s_cselect_b32 s4, 1, 0
 ; GFX11_W64-NEXT:  .LBB13_2: ; %exit
-; GFX11_W64-NEXT:    s_or_b64 exec, exec, s[4:5]
-; GFX11_W64-NEXT:    s_and_b32 s0, 1, s6
-; GFX11_W64-NEXT:    v_cmp_ne_u32_e64 vcc, 0, s0
+; GFX11_W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11_W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x0
+; GFX11_W64-NEXT:    s_and_b32 s2, 1, s4
+; GFX11_W64-NEXT:    v_cmp_ne_u32_e64 vcc, 0, s2
 ; GFX11_W64-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11_W64-NEXT:    v_div_fmas_f32 v0, v1, v2, v3
 ; GFX11_W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11_W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11_W64-NEXT:    global_store_b32 v1, v0, s[2:3] offset:8
+; GFX11_W64-NEXT:    global_store_b32 v1, v0, s[0:1] offset:8
 ; GFX11_W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11_W64-NEXT:    s_endpgm
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep.a = getelementptr float, float addrspace(1)* %in, i32 %tid
   %gep.b = getelementptr float, float addrspace(1)* %gep.a, i32 1
   %gep.c = getelementptr float, float addrspace(1)* %gep.a, i32 2
 

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.end.cf.i32.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -global-isel -mtriple=amdgcn--amdhsa -mcpu=gfx1010 -mattr=+wavefrontsize32,-wavefrontsize64 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX10 %s
 ; RUN: llc -global-isel -mtriple=amdgcn--amdhsa -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 -mattr=+wavefrontsize32,-wavefrontsize64 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX11 %s
 
 define amdgpu_kernel void @test_wave32(i32 %arg0, [8 x i32], i32 %saved) {
 ; GFX10-LABEL: test_wave32:
 ; GFX10:       ; %bb.0: ; %entry
-; GFX10-NEXT:    s_clause 0x1
-; GFX10-NEXT:    s_load_dword s1, s[4:5], 0x0
-; GFX10-NEXT:    s_load_dword s0, s[4:5], 0x24
+; GFX10-NEXT:    s_load_dword s0, s[4:5], 0x0
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    s_cmp_lg_u32 s1, 0
+; GFX10-NEXT:    s_cmp_lg_u32 s0, 0
 ; GFX10-NEXT:    s_cbranch_scc1 .LBB0_2
 ; GFX10-NEXT:  ; %bb.1: ; %mid
 ; GFX10-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-NEXT:    global_store_dword v[0:1], v0, off
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:  .LBB0_2: ; %bb
+; GFX10-NEXT:    s_load_dword s0, s[4:5], 0x24
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX10-NEXT:    s_or_b32 exec_lo, exec_lo, s0
 ; GFX10-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-NEXT:    global_store_dword v[0:1], v0, off
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    s_endpgm
 ;
 ; GFX11-LABEL: test_wave32:
 ; GFX11:       ; %bb.0: ; %entry
-; GFX11-NEXT:    s_clause 0x1
 ; GFX11-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX11-NEXT:    s_load_b32 s0, s[0:1], 0x24
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    s_cmp_lg_u32 s2, 0
 ; GFX11-NEXT:    s_cbranch_scc1 .LBB0_2
 ; GFX11-NEXT:  ; %bb.1: ; %mid
 ; GFX11-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX11-NEXT:    global_store_b32 v[0:1], v0, off dlc
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:  .LBB0_2: ; %bb
+; GFX11-NEXT:    s_load_b32 s0, s[0:1], 0x24
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    s_or_b32 exec_lo, exec_lo, s0
 ; GFX11-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX11-NEXT:    global_store_b32 v[0:1], v0, off dlc
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 entry:
   %cond = icmp eq i32 %arg0, 0

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.end.cf.i64.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -global-isel -mtriple=amdgcn--amdhsa -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
 
 define amdgpu_kernel void @test_wave64(i32 %arg0, i64 %saved) {
 ; GCN-LABEL: test_wave64:
 ; GCN:       ; %bb.0: ; %entry
-; GCN-NEXT:    s_load_dword s2, s[4:5], 0x0
-; GCN-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x8
+; GCN-NEXT:    s_load_dword s0, s[4:5], 0x0
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NEXT:    s_cmp_lg_u32 s2, 0
+; GCN-NEXT:    s_cmp_lg_u32 s0, 0
 ; GCN-NEXT:    s_cbranch_scc1 .LBB0_2
 ; GCN-NEXT:  ; %bb.1: ; %mid
 ; GCN-NEXT:    v_mov_b32_e32 v0, 0
 ; GCN-NEXT:    global_store_dword v[0:1], v0, off
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:  .LBB0_2: ; %bb
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x8
+; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; GCN-NEXT:    v_mov_b32_e32 v0, 0
 ; GCN-NEXT:    global_store_dword v[0:1], v0, off
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_endpgm
 entry:
   %cond = icmp eq i32 %arg0, 0
   br i1 %cond, label %mid, label %bb

llvm/test/CodeGen/AMDGPU/GlobalISel/non-entry-alloca.ll

 ; GCN-NEXT:    s_cmp_lg_u32 s6, 0
 ; GCN-NEXT:    s_cbranch_scc1 .LBB0_3
 ; GCN-NEXT:  ; %bb.1: ; %bb.0
 ; GCN-NEXT:    s_load_dword s6, s[4:5], 0xc
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_cmp_lg_u32 s6, 0
 ; GCN-NEXT:    s_cbranch_scc1 .LBB0_3
 ; GCN-NEXT:  ; %bb.2: ; %bb.1
-; GCN-NEXT:    s_load_dword s6, s[4:5], 0x10
-; GCN-NEXT:    s_add_u32 s7, s32, 0x1000
+; GCN-NEXT:    s_load_dword s7, s[4:5], 0x10
+; GCN-NEXT:    s_add_u32 s6, s32, 0x1000
 ; GCN-NEXT:    v_mov_b32_e32 v1, 0
-; GCN-NEXT:    v_mov_b32_e32 v2, s7
+; GCN-NEXT:    v_mov_b32_e32 v2, s6
 ; GCN-NEXT:    v_mov_b32_e32 v3, 1
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NEXT:    s_lshl_b32 s6, s6, 2
-; GCN-NEXT:    s_add_u32 s6, s7, s6
+; GCN-NEXT:    s_lshl_b32 s7, s7, 2
+; GCN-NEXT:    s_add_u32 s6, s6, s7
 ; GCN-NEXT:    buffer_store_dword v1, v2, s[0:3], 0 offen
 ; GCN-NEXT:    buffer_store_dword v3, v2, s[0:3], 0 offen offset:4
 ; GCN-NEXT:    v_mov_b32_e32 v2, s6
 ; GCN-NEXT:    buffer_load_dword v2, v2, s[0:3], 0 offen
 ; GCN-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    v_add_u32_e32 v0, v2, v0
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)

llvm/test/CodeGen/AMDGPU/atomic_optimizations_buffer.ll

 declare i32 @llvm.amdgcn.struct.buffer.atomic.add(i32, <4 x i32>, i32, i32, i32, i32 immarg)
 declare i32 @llvm.amdgcn.raw.buffer.atomic.sub(i32, <4 x i32>, i32, i32, i32 immarg)
 
 ; Show what the atomic optimization pass will do for raw buffers.
 
 define amdgpu_kernel void @add_i32_constant(i32 addrspace(1)* %out, <4 x i32> %inout) {
 ; GFX6-LABEL: add_i32_constant:
 ; GFX6:       ; %bb.0: ; %entry
-; GFX6-NEXT:    s_mov_b64 s[2:3], exec
-; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
-; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
+; GFX6-NEXT:    s_mov_b64 s[4:5], exec
+; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s4, 0
+; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s5, v0
 ; GFX6-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX6-NEXT:    ; implicit-def: $vgpr1
-; GFX6-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX6-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX6-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX6-NEXT:  ; %bb.1:
 ; GFX6-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0xd
-; GFX6-NEXT:    s_bcnt1_i32_b64 s0, s[2:3]
-; GFX6-NEXT:    s_mul_i32 s0, s0, 5
-; GFX6-NEXT:    v_mov_b32_e32 v1, s0
+; GFX6-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX6-NEXT:    s_mul_i32 s4, s4, 5
+; GFX6-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
 ; GFX6-NEXT:  .LBB0_2:
-; GFX6-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX6-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
-; GFX6-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX6-NEXT:    s_mov_b32 s7, 0xf000
-; GFX6-NEXT:    v_mad_u32_u24 v0, v0, 5, s0
-; GFX6-NEXT:    s_mov_b32 s6, -1
+; GFX6-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX6-NEXT:    v_mad_u32_u24 v0, v0, 5, s4
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: add_i32_constant:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[6:7], exec
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX8-NEXT:    s_mov_b64 s[4:5], exec
+; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr1
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX8-NEXT:  ; %bb.1:
 ; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX8-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
-; GFX8-NEXT:    s_mul_i32 s0, s0, 5
-; GFX8-NEXT:    v_mov_b32_e32 v1, s0
+; GFX8-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX8-NEXT:    s_mul_i32 s4, s4, 5
+; GFX8-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
 ; GFX8-NEXT:  .LBB0_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX8-NEXT:    v_mad_u32_u24 v2, v0, 5, s0
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX8-NEXT:    v_mad_u32_u24 v2, v0, 5, s2
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    flat_store_dword v[0:1], v2
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: add_i32_constant:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[6:7], exec
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX9-NEXT:    s_mov_b64 s[4:5], exec
+; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr1
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX9-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
-; GFX9-NEXT:    s_mul_i32 s0, s0, 5
-; GFX9-NEXT:    v_mov_b32_e32 v1, s0
+; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX9-NEXT:    s_mul_i32 s4, s4, 5
+; GFX9-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
 ; GFX9-NEXT:  .LBB0_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX9-NEXT:    v_mad_u32_u24 v0, v0, 5, s0
-; GFX9-NEXT:    v_mov_b32_e32 v1, 0
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX9-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: add_i32_constant:
 ; GFX10W64:       ; %bb.0: ; %entry
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_mov_b64 s[6:7], exec
+; GFX10W64-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr1
-; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX10W64-NEXT:  ; %bb.1:
 ; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W64-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
-; GFX10W64-NEXT:    s_mul_i32 s0, s0, 5
-; GFX10W64-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX10W64-NEXT:    s_mul_i32 s4, s4, 5
+; GFX10W64-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10W64-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
 ; GFX10W64-NEXT:  .LBB0_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W64-NEXT:    v_mad_u32_u24 v0, v0, 5, s0
+; GFX10W64-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W64-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: add_i32_constant:
 ; GFX10W32:       ; %bb.0: ; %entry
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_mov_b32 s5, exec_lo
+; GFX10W32-NEXT:    s_mov_b32 s3, exec_lo
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr1
-; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s5, 0
+; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX10W32-NEXT:  ; %bb.1:
-; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    s_bcnt1_i32_b32 s0, s5
-; GFX10W32-NEXT:    s_mul_i32 s0, s0, 5
-; GFX10W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10W32-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x34
+; GFX10W32-NEXT:    s_bcnt1_i32_b32 s3, s3
+; GFX10W32-NEXT:    s_mul_i32 s3, s3, 5
+; GFX10W32-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
+; GFX10W32-NEXT:    buffer_atomic_add v1, off, s[4:7], 0 glc
 ; GFX10W32-NEXT:  .LBB0_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W32-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W32-NEXT:    v_mad_u32_u24 v0, v0, 5, s0
+; GFX10W32-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W32-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: add_i32_constant:
 ; GFX11W64:       ; %bb.0: ; %entry
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_mov_b64 s[6:7], exec
 ; GFX11W64-NEXT:    s_mov_b64 s[4:5], exec
-; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX11W64-NEXT:    s_mov_b64 s[2:3], exec
+; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX11W64-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX11W64-NEXT:  ; %bb.1:
 ; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W64-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
+; GFX11W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX11W64-NEXT:    s_mul_i32 s0, s0, 5
-; GFX11W64-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11W64-NEXT:    s_mul_i32 s4, s4, 5
+; GFX11W64-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11W64-NEXT:    buffer_atomic_add_u32 v1, off, s[8:11], 0 glc
 ; GFX11W64-NEXT:  .LBB0_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W64-NEXT:    v_mad_u32_u24 v0, v0, 5, s0
+; GFX11W64-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11W64-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: add_i32_constant:
 ; GFX11W32:       ; %bb.0: ; %entry
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_mov_b32 s5, exec_lo
-; GFX11W32-NEXT:    s_mov_b32 s4, exec_lo
-; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s5, 0
+; GFX11W32-NEXT:    s_mov_b32 s3, exec_lo
+; GFX11W32-NEXT:    s_mov_b32 s2, exec_lo
+; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX11W32-NEXT:  ; %bb.1:
-; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    s_bcnt1_i32_b32 s0, s5
+; GFX11W32-NEXT:    s_load_b128 s[4:7], s[0:1], 0x34
+; GFX11W32-NEXT:    s_bcnt1_i32_b32 s3, s3
 ; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX11W32-NEXT:    s_mul_i32 s0, s0, 5
-; GFX11W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11W32-NEXT:    s_mul_i32 s3, s3, 5
+; GFX11W32-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    buffer_atomic_add_u32 v1, off, s[8:11], 0 glc
+; GFX11W32-NEXT:    buffer_atomic_add_u32 v1, off, s[4:7], 0 glc
 ; GFX11W32-NEXT:  .LBB0_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W32-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W32-NEXT:    v_mad_u32_u24 v0, v0, 5, s0
+; GFX11W32-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11W32-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %old = call i32 @llvm.amdgcn.raw.buffer.atomic.add(i32 5, <4 x i32> %inout, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @add_i32_uniform(i32 addrspace(1)* %out, <4 x i32> %inout, i32 %additive) {
 ; GFX6-LABEL: add_i32_uniform:
 ; GFX6:       ; %bb.0: ; %entry
-; GFX6-NEXT:    s_mov_b64 s[2:3], exec
-; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
-; GFX6-NEXT:    s_load_dword s8, s[0:1], 0x11
-; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
+; GFX6-NEXT:    s_mov_b64 s[4:5], exec
+; GFX6-NEXT:    s_load_dword s6, s[0:1], 0x11
+; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s4, 0
+; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s5, v0
 ; GFX6-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX6-NEXT:    ; implicit-def: $vgpr1
-; GFX6-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX6-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX6-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX6-NEXT:  ; %bb.1:
-; GFX6-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0xd
-; GFX6-NEXT:    s_bcnt1_i32_b64 s0, s[2:3]
+; GFX6-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0xd
+; GFX6-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    s_mul_i32 s0, s8, s0
-; GFX6-NEXT:    v_mov_b32_e32 v1, s0
-; GFX6-NEXT:    buffer_atomic_add v1, off, s[12:15], 0 glc
+; GFX6-NEXT:    s_mul_i32 s4, s6, s4
+; GFX6-NEXT:    v_mov_b32_e32 v1, s4
+; GFX6-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
 ; GFX6-NEXT:  .LBB1_2:
-; GFX6-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX6-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
-; GFX6-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX6-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    v_mul_lo_u32 v0, s8, v0
-; GFX6-NEXT:    s_mov_b32 s7, 0xf000
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, s0, v0
-; GFX6-NEXT:    s_mov_b32 s6, -1
-; GFX6-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX6-NEXT:    v_mul_lo_u32 v0, s6, v0
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, s4, v0
+; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: add_i32_uniform:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_load_dword s8, s[0:1], 0x44
+; GFX8-NEXT:    s_load_dword s6, s[0:1], 0x44
 ; GFX8-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr1
-; GFX8-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0x34
-; GFX8-NEXT:    s_bcnt1_i32_b64 s0, s[4:5]
+; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX8-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_mul_i32 s0, s8, s0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s0
-; GFX8-NEXT:    buffer_atomic_add v1, off, s[12:15], 0 glc
+; GFX8-NEXT:    s_mul_i32 s4, s6, s4
+; GFX8-NEXT:    v_mov_b32_e32 v1, s4
+; GFX8-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
 ; GFX8-NEXT:  .LBB1_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mul_lo_u32 v0, s8, v0
+; GFX8-NEXT:    v_mul_lo_u32 v0, s6, v0
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX8-NEXT:    v_add_u32_e32 v2, vcc, s0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX8-NEXT:    v_add_u32_e32 v2, vcc, s2, v0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    flat_store_dword v[0:1], v2
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: add_i32_uniform:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_load_dword s8, s[0:1], 0x44
+; GFX9-NEXT:    s_load_dword s6, s[0:1], 0x44
 ; GFX9-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr1
-; GFX9-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0x34
-; GFX9-NEXT:    s_bcnt1_i32_b64 s0, s[4:5]
+; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_mul_i32 s0, s8, s0
-; GFX9-NEXT:    v_mov_b32_e32 v1, s0
-; GFX9-NEXT:    buffer_atomic_add v1, off, s[12:15], 0 glc
+; GFX9-NEXT:    s_mul_i32 s4, s6, s4
+; GFX9-NEXT:    v_mov_b32_e32 v1, s4
+; GFX9-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
 ; GFX9-NEXT:  .LBB1_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mul_lo_u32 v0, s8, v0
+; GFX9-NEXT:    v_mul_lo_u32 v0, s6, v0
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX9-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-NEXT:    v_add_u32_e32 v0, s0, v0
-; GFX9-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_add_u32_e32 v0, s2, v0
+; GFX9-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: add_i32_uniform:
 ; GFX10W64:       ; %bb.0: ; %entry
-; GFX10W64-NEXT:    s_clause 0x1
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_load_dword s8, s[0:1], 0x44
+; GFX10W64-NEXT:    s_load_dword s6, s[0:1], 0x44
 ; GFX10W64-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr1
 ; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX10W64-NEXT:  ; %bb.1:
-; GFX10W64-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0x34
-; GFX10W64-NEXT:    s_bcnt1_i32_b64 s0, s[4:5]
+; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX10W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    s_mul_i32 s0, s8, s0
-; GFX10W64-NEXT:    v_mov_b32_e32 v1, s0
-; GFX10W64-NEXT:    buffer_atomic_add v1, off, s[12:15], 0 glc
+; GFX10W64-NEXT:    s_mul_i32 s4, s6, s4
+; GFX10W64-NEXT:    v_mov_b32_e32 v1, s4
+; GFX10W64-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
 ; GFX10W64-NEXT:  .LBB1_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], s8, v0, s[0:1]
+; GFX10W64-NEXT:    v_mad_u64_u32 v[0:1], s[2:3], s6, v0, s[2:3]
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W64-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W64-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: add_i32_uniform:
 ; GFX10W32:       ; %bb.0: ; %entry
-; GFX10W32-NEXT:    s_clause 0x1
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_load_dword s4, s[0:1], 0x44
-; GFX10W32-NEXT:    s_mov_b32 s6, exec_lo
+; GFX10W32-NEXT:    s_load_dword s2, s[0:1], 0x44
+; GFX10W32-NEXT:    s_mov_b32 s4, exec_lo
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr1
-; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s5, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX10W32-NEXT:  ; %bb.1:
 ; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    s_bcnt1_i32_b32 s0, s6
+; GFX10W32-NEXT:    s_bcnt1_i32_b32 s4, s4
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    s_mul_i32 s0, s4, s0
-; GFX10W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10W32-NEXT:    s_mul_i32 s4, s2, s4
+; GFX10W32-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX10W32-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
 ; GFX10W32-NEXT:  .LBB1_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s5
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W32-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    v_mad_u64_u32 v[0:1], s0, s4, v0, s[0:1]
+; GFX10W32-NEXT:    v_mad_u64_u32 v[0:1], s2, s2, v0, s[4:5]
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W32-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W32-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: add_i32_uniform:
 ; GFX11W64:       ; %bb.0: ; %entry
-; GFX11W64-NEXT:    s_clause 0x1
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_load_b32 s8, s[0:1], 0x44
+; GFX11W64-NEXT:    s_load_b32 s6, s[0:1], 0x44
 ; GFX11W64-NEXT:    s_mov_b64 s[4:5], exec
-; GFX11W64-NEXT:    s_mov_b64 s[6:7], exec
+; GFX11W64-NEXT:    s_mov_b64 s[2:3], exec
 ; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX11W64-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX11W64-NEXT:  ; %bb.1:
-; GFX11W64-NEXT:    s_load_b128 s[12:15], s[0:1], 0x34
-; GFX11W64-NEXT:    s_bcnt1_i32_b64 s0, s[4:5]
+; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
+; GFX11W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    s_mul_i32 s0, s8, s0
+; GFX11W64-NEXT:    s_mul_i32 s4, s6, s4
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11W64-NEXT:    v_mov_b32_e32 v1, s0
-; GFX11W64-NEXT:    buffer_atomic_add_u32 v1, off, s[12:15], 0 glc
+; GFX11W64-NEXT:    v_mov_b32_e32 v1, s4
+; GFX11W64-NEXT:    buffer_atomic_add_u32 v1, off, s[8:11], 0 glc
 ; GFX11W64-NEXT:  .LBB1_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11W64-NEXT:    v_mad_u64_u32 v[1:2], null, s8, v0, s[0:1]
+; GFX11W64-NEXT:    v_mad_u64_u32 v[1:2], null, s6, v0, s[2:3]
 ; GFX11W64-NEXT:    v_mov_b32_e32 v0, 0
-; GFX11W64-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11W64-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: add_i32_uniform:
 ; GFX11W32:       ; %bb.0: ; %entry
-; GFX11W32-NEXT:    s_clause 0x1
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_load_b32 s4, s[0:1], 0x44
-; GFX11W32-NEXT:    s_mov_b32 s6, exec_lo
-; GFX11W32-NEXT:    s_mov_b32 s5, exec_lo
-; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX11W32-NEXT:    s_load_b32 s2, s[0:1], 0x44
+; GFX11W32-NEXT:    s_mov_b32 s4, exec_lo
+; GFX11W32-NEXT:    s_mov_b32 s3, exec_lo
+; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX11W32-NEXT:  ; %bb.1:
 ; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    s_bcnt1_i32_b32 s0, s6
+; GFX11W32-NEXT:    s_bcnt1_i32_b32 s4, s4
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    s_mul_i32 s0, s4, s0
+; GFX11W32-NEXT:    s_mul_i32 s4, s2, s4
 ; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11W32-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX11W32-NEXT:    buffer_atomic_add_u32 v1, off, s[8:11], 0 glc
 ; GFX11W32-NEXT:  .LBB1_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s5
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W32-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11W32-NEXT:    v_mad_u64_u32 v[1:2], null, s4, v0, s[0:1]
+; GFX11W32-NEXT:    v_mad_u64_u32 v[1:2], null, s2, v0, s[4:5]
 ; GFX11W32-NEXT:    v_mov_b32_e32 v0, 0
-; GFX11W32-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11W32-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %old = call i32 @llvm.amdgcn.raw.buffer.atomic.add(i32 %additive, <4 x i32> %inout, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @add_i32_varying_vdata(i32 addrspace(1)* %out, <4 x i32> %inout) {
 ; GFX6-LABEL: add_i32_varying_vdata:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0xd
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_add v0, off, s[4:7], 0 glc
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: add_i32_varying_vdata:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
-; GFX8-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX8-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX8-NEXT:    s_not_b64 exec, exec
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX8-NEXT:    s_not_b64 exec, exec
-; GFX8-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX8-NEXT:    v_readlane_b32 s6, v2, 63
+; GFX8-NEXT:    v_readlane_b32 s4, v2, 63
 ; GFX8-NEXT:    s_nop 0
 ; GFX8-NEXT:    v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf
-; GFX8-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
 ; GFX8-NEXT:    ; implicit-def: $vgpr0
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX8-NEXT:  ; %bb.1:
 ; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX8-NEXT:    v_mov_b32_e32 v0, s6
+; GFX8-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    buffer_atomic_add v0, off, s[8:11], 0 glc
 ; GFX8-NEXT:  .LBB2_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, v1
+; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v4, s3
-; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s2
+; GFX8-NEXT:    v_mov_b32_e32 v4, s1
+; GFX8-NEXT:    v_mov_b32_e32 v3, s0
 ; GFX8-NEXT:    flat_store_dword v[3:4], v0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: add_i32_varying_vdata:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX9-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX9-NEXT:    s_not_b64 exec, exec
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    s_not_b64 exec, exec
-; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-NEXT:    v_readlane_b32 s6, v2, 63
+; GFX9-NEXT:    v_readlane_b32 s4, v2, 63
 ; GFX9-NEXT:    s_nop 0
 ; GFX9-NEXT:    v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf
-; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
 ; GFX9-NEXT:    ; implicit-def: $vgpr0
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX9-NEXT:    v_mov_b32_e32 v0, s6
+; GFX9-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    buffer_atomic_add v0, off, s[8:11], 0 glc
 ; GFX9-NEXT:  .LBB2_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v1
-; GFX9-NEXT:    v_add_u32_e32 v0, s0, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, 0
+; GFX9-NEXT:    v_add_u32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    global_store_dword v3, v0, s[2:3]
+; GFX9-NEXT:    global_store_dword v3, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: add_i32_varying_vdata:
 ; GFX10W64:       ; %bb.0: ; %entry
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX10W64-NEXT:    s_not_b64 exec, exec
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10W64-NEXT:    s_not_b64 exec, exec
 ; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX10W64-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX10W64-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 15
 ; GFX10W64-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX10W64-NEXT:    v_readlane_b32 s5, v1, 31
+; GFX10W64-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_readlane_b32 s7, v1, 31
-; GFX10W64-NEXT:    v_writelane_b32 v3, s6, 16
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
 ; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 63
-; GFX10W64-NEXT:    v_readlane_b32 s8, v1, 47
-; GFX10W64-NEXT:    v_writelane_b32 v3, s7, 32
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 63
+; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 47
+; GFX10W64-NEXT:    v_writelane_b32 v3, s5, 32
+; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_writelane_b32 v3, s8, 48
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX10W64-NEXT:    v_writelane_b32 v3, s6, 48
+; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX10W64-NEXT:  ; %bb.1:
 ; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W64-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10W64-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10W64-NEXT:    buffer_atomic_add v0, off, s[8:11], 0 glc
 ; GFX10W64-NEXT:  .LBB2_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX10W64-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX10W64-NEXT:    v_mov_b32_e32 v4, 0
-; GFX10W64-NEXT:    v_add_nc_u32_e32 v0, s0, v0
+; GFX10W64-NEXT:    v_add_nc_u32_e32 v0, s2, v0
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    global_store_dword v4, v0, s[2:3]
+; GFX10W64-NEXT:    global_store_dword v4, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: add_i32_varying_vdata:
 ; GFX10W32:       ; %bb.0: ; %entry
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX10W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX10W32-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX10W32-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX10W32-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_or_saveexec_b32 s4, -1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX10W32-NEXT:    v_mov_b32_e32 v3, 0
-; GFX10W32-NEXT:    v_readlane_b32 s6, v1, 31
+; GFX10W32-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX10W32-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX10W32-NEXT:    v_readlane_b32 s5, v1, 15
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10W32-NEXT:    v_readlane_b32 s3, v1, 15
+; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX10W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX10W32-NEXT:    v_writelane_b32 v3, s5, 16
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10W32-NEXT:    s_or_saveexec_b32 s2, -1
+; GFX10W32-NEXT:    v_writelane_b32 v3, s3, 16
+; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX10W32-NEXT:  ; %bb.1:
-; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    v_mov_b32_e32 v0, s6
-; GFX10W32-NEXT:    s_mov_b32 s5, s6
+; GFX10W32-NEXT:    s_mov_b32 s3, s4
+; GFX10W32-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x34
+; GFX10W32-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    buffer_atomic_add v0, off, s[8:11], 0 glc
+; GFX10W32-NEXT:    buffer_atomic_add v0, off, s[4:7], 0 glc
 ; GFX10W32-NEXT:  .LBB2_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10W32-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX10W32-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX10W32-NEXT:    v_mov_b32_e32 v4, 0
-; GFX10W32-NEXT:    v_add_nc_u32_e32 v0, s0, v0
+; GFX10W32-NEXT:    v_add_nc_u32_e32 v0, s2, v0
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    global_store_dword v4, v0, s[2:3]
+; GFX10W32-NEXT:    global_store_dword v4, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: add_i32_varying_vdata:
 ; GFX11W64:       ; %bb.0: ; %entry
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX11W64-NEXT:    s_not_b64 exec, exec
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W64-NEXT:    s_not_b64 exec, exec

 ; GFX11W64-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX11W64-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W64-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX11W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 15
+; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
 ; GFX11W64-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX11W64-NEXT:    v_readlane_b32 s5, v1, 31
+; GFX11W64-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_readlane_b32 s7, v1, 31
-; GFX11W64-NEXT:    v_writelane_b32 v3, s6, 16
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 63
-; GFX11W64-NEXT:    v_readlane_b32 s8, v1, 47
-; GFX11W64-NEXT:    v_writelane_b32 v3, s7, 32
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX11W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 63
+; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 47
+; GFX11W64-NEXT:    v_writelane_b32 v3, s5, 32
+; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
 ; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_writelane_b32 v3, s8, 48
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX11W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX11W64-NEXT:    v_writelane_b32 v3, s6, 48
+; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX11W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr0
-; GFX11W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX11W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX11W64-NEXT:  ; %bb.1:
 ; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W64-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11W64-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11W64-NEXT:    buffer_atomic_add_u32 v0, off, s[8:11], 0 glc
 ; GFX11W64-NEXT:  .LBB2_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX11W64-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX11W64-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W64-NEXT:    v_add_nc_u32_e32 v0, s0, v0
+; GFX11W64-NEXT:    v_add_nc_u32_e32 v0, s2, v0
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    global_store_b32 v4, v0, s[2:3]
+; GFX11W64-NEXT:    global_store_b32 v4, v0, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: add_i32_varying_vdata:
 ; GFX11W32:       ; %bb.0: ; %entry
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX11W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX11W32-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX11W32-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX11W32-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX11W32-NEXT:    v_mov_b32_e32 v3, 0
-; GFX11W32-NEXT:    v_readlane_b32 s6, v1, 31
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
+; GFX11W32-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX11W32-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX11W32-NEXT:    v_readlane_b32 s5, v1, 15
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX11W32-NEXT:    v_readlane_b32 s3, v1, 15
+; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
+; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX11W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11W32-NEXT:    v_writelane_b32 v3, s5, 16
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX11W32-NEXT:    s_or_saveexec_b32 s2, -1
+; GFX11W32-NEXT:    v_writelane_b32 v3, s3, 16
+; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr0
-; GFX11W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX11W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX11W32-NEXT:  ; %bb.1:
-; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    v_mov_b32_e32 v0, s6
-; GFX11W32-NEXT:    s_mov_b32 s5, s6
+; GFX11W32-NEXT:    s_mov_b32 s3, s4
+; GFX11W32-NEXT:    s_load_b128 s[4:7], s[0:1], 0x34
+; GFX11W32-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    buffer_atomic_add_u32 v0, off, s[8:11], 0 glc
+; GFX11W32-NEXT:    buffer_atomic_add_u32 v0, off, s[4:7], 0 glc
 ; GFX11W32-NEXT:  .LBB2_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX11W32-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX11W32-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX11W32-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W32-NEXT:    v_add_nc_u32_e32 v0, s0, v0
+; GFX11W32-NEXT:    v_add_nc_u32_e32 v0, s2, v0
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    global_store_b32 v4, v0, s[2:3]
+; GFX11W32-NEXT:    global_store_b32 v4, v0, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
   %old = call i32 @llvm.amdgcn.raw.buffer.atomic.add(i32 %lane, <4 x i32> %inout, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }

 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: struct_add_i32_varying_vdata:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
-; GFX8-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX8-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX8-NEXT:    s_not_b64 exec, exec
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX8-NEXT:    s_not_b64 exec, exec
-; GFX8-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX8-NEXT:    v_readlane_b32 s6, v2, 63
+; GFX8-NEXT:    v_readlane_b32 s4, v2, 63
 ; GFX8-NEXT:    s_nop 0
 ; GFX8-NEXT:    v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf
-; GFX8-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
 ; GFX8-NEXT:    ; implicit-def: $vgpr0
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB3_2
 ; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dword s7, s[0:1], 0x44
+; GFX8-NEXT:    s_load_dword s5, s[0:1], 0x44
 ; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX8-NEXT:    v_mov_b32_e32 v0, s6
+; GFX8-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v3, s7
+; GFX8-NEXT:    v_mov_b32_e32 v3, s5
 ; GFX8-NEXT:    buffer_atomic_add v0, v3, s[8:11], 0 idxen glc
 ; GFX8-NEXT:  .LBB3_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, v1
+; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v4, s3
-; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s2
+; GFX8-NEXT:    v_mov_b32_e32 v4, s1
+; GFX8-NEXT:    v_mov_b32_e32 v3, s0
 ; GFX8-NEXT:    flat_store_dword v[3:4], v0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: struct_add_i32_varying_vdata:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX9-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX9-NEXT:    s_not_b64 exec, exec
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    s_not_b64 exec, exec
-; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-NEXT:    v_readlane_b32 s6, v2, 63
+; GFX9-NEXT:    v_readlane_b32 s4, v2, 63
 ; GFX9-NEXT:    s_nop 0
 ; GFX9-NEXT:    v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf
-; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
 ; GFX9-NEXT:    ; implicit-def: $vgpr0
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB3_2
 ; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dword s7, s[0:1], 0x44
+; GFX9-NEXT:    s_load_dword s5, s[0:1], 0x44
 ; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX9-NEXT:    v_mov_b32_e32 v0, s6
+; GFX9-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v3, s7
+; GFX9-NEXT:    v_mov_b32_e32 v3, s5
 ; GFX9-NEXT:    buffer_atomic_add v0, v3, s[8:11], 0 idxen glc
 ; GFX9-NEXT:  .LBB3_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v1
-; GFX9-NEXT:    v_add_u32_e32 v0, s0, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, 0
+; GFX9-NEXT:    v_add_u32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    global_store_dword v3, v0, s[2:3]
+; GFX9-NEXT:    global_store_dword v3, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: struct_add_i32_varying_vdata:
 ; GFX10W64:       ; %bb.0: ; %entry
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX10W64-NEXT:    s_not_b64 exec, exec
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10W64-NEXT:    s_not_b64 exec, exec
 ; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX10W64-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX10W64-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 15
 ; GFX10W64-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX10W64-NEXT:    v_readlane_b32 s5, v1, 31
+; GFX10W64-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_readlane_b32 s7, v1, 31
-; GFX10W64-NEXT:    v_writelane_b32 v3, s6, 16
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
 ; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 63
-; GFX10W64-NEXT:    v_readlane_b32 s8, v1, 47
-; GFX10W64-NEXT:    v_writelane_b32 v3, s7, 32
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 63
+; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 47
+; GFX10W64-NEXT:    v_writelane_b32 v3, s5, 32
+; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_writelane_b32 v3, s8, 48
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX10W64-NEXT:    v_writelane_b32 v3, s6, 48
+; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB3_2
 ; GFX10W64-NEXT:  ; %bb.1:
 ; GFX10W64-NEXT:    s_clause 0x1
-; GFX10W64-NEXT:    s_load_dword s7, s[0:1], 0x44
+; GFX10W64-NEXT:    s_load_dword s5, s[0:1], 0x44
 ; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W64-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10W64-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    v_mov_b32_e32 v4, s7
+; GFX10W64-NEXT:    v_mov_b32_e32 v4, s5
 ; GFX10W64-NEXT:    buffer_atomic_add v0, v4, s[8:11], 0 idxen glc
 ; GFX10W64-NEXT:  .LBB3_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX10W64-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX10W64-NEXT:    v_mov_b32_e32 v4, 0
-; GFX10W64-NEXT:    v_add_nc_u32_e32 v0, s0, v0
+; GFX10W64-NEXT:    v_add_nc_u32_e32 v0, s2, v0
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    global_store_dword v4, v0, s[2:3]
+; GFX10W64-NEXT:    global_store_dword v4, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: struct_add_i32_varying_vdata:
 ; GFX10W32:       ; %bb.0: ; %entry
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX10W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX10W32-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX10W32-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX10W32-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_or_saveexec_b32 s4, -1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX10W32-NEXT:    v_mov_b32_e32 v3, 0
-; GFX10W32-NEXT:    v_readlane_b32 s6, v1, 31
+; GFX10W32-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX10W32-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX10W32-NEXT:    v_readlane_b32 s5, v1, 15
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10W32-NEXT:    v_readlane_b32 s3, v1, 15
+; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX10W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX10W32-NEXT:    v_writelane_b32 v3, s5, 16
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10W32-NEXT:    s_or_saveexec_b32 s2, -1
+; GFX10W32-NEXT:    v_writelane_b32 v3, s3, 16
+; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB3_2
 ; GFX10W32-NEXT:  ; %bb.1:
-; GFX10W32-NEXT:    s_mov_b32 s5, s6
+; GFX10W32-NEXT:    s_mov_b32 s3, s4
 ; GFX10W32-NEXT:    s_clause 0x1
-; GFX10W32-NEXT:    s_load_dword s6, s[0:1], 0x44
-; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    v_mov_b32_e32 v0, s5
+; GFX10W32-NEXT:    s_load_dword s8, s[0:1], 0x44
+; GFX10W32-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x34
+; GFX10W32-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    v_mov_b32_e32 v4, s6
-; GFX10W32-NEXT:    buffer_atomic_add v0, v4, s[8:11], 0 idxen glc
+; GFX10W32-NEXT:    v_mov_b32_e32 v4, s8
+; GFX10W32-NEXT:    buffer_atomic_add v0, v4, s[4:7], 0 idxen glc
 ; GFX10W32-NEXT:  .LBB3_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10W32-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX10W32-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX10W32-NEXT:    v_mov_b32_e32 v4, 0
-; GFX10W32-NEXT:    v_add_nc_u32_e32 v0, s0, v0
+; GFX10W32-NEXT:    v_add_nc_u32_e32 v0, s2, v0
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    global_store_dword v4, v0, s[2:3]
+; GFX10W32-NEXT:    global_store_dword v4, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: struct_add_i32_varying_vdata:
 ; GFX11W64:       ; %bb.0: ; %entry
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX11W64-NEXT:    s_not_b64 exec, exec
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W64-NEXT:    s_not_b64 exec, exec

 ; GFX11W64-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX11W64-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W64-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX11W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 15
+; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
 ; GFX11W64-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX11W64-NEXT:    v_readlane_b32 s5, v1, 31
+; GFX11W64-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_readlane_b32 s7, v1, 31
-; GFX11W64-NEXT:    v_writelane_b32 v3, s6, 16
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 63
-; GFX11W64-NEXT:    v_readlane_b32 s8, v1, 47
-; GFX11W64-NEXT:    v_writelane_b32 v3, s7, 32
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX11W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 63
+; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 47
+; GFX11W64-NEXT:    v_writelane_b32 v3, s5, 32
+; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
 ; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_writelane_b32 v3, s8, 48
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX11W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX11W64-NEXT:    v_writelane_b32 v3, s6, 48
+; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX11W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr0
-; GFX11W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX11W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB3_2
 ; GFX11W64-NEXT:  ; %bb.1:
 ; GFX11W64-NEXT:    s_clause 0x1
-; GFX11W64-NEXT:    s_load_b32 s7, s[0:1], 0x44
+; GFX11W64-NEXT:    s_load_b32 s5, s[0:1], 0x44
 ; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W64-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11W64-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    v_mov_b32_e32 v4, s7
+; GFX11W64-NEXT:    v_mov_b32_e32 v4, s5
 ; GFX11W64-NEXT:    buffer_atomic_add_u32 v0, v4, s[8:11], 0 idxen glc
 ; GFX11W64-NEXT:  .LBB3_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX11W64-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX11W64-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W64-NEXT:    v_add_nc_u32_e32 v0, s0, v0
+; GFX11W64-NEXT:    v_add_nc_u32_e32 v0, s2, v0
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    global_store_b32 v4, v0, s[2:3]
+; GFX11W64-NEXT:    global_store_b32 v4, v0, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: struct_add_i32_varying_vdata:
 ; GFX11W32:       ; %bb.0: ; %entry
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX11W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX11W32-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX11W32-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX11W32-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX11W32-NEXT:    v_mov_b32_e32 v3, 0
-; GFX11W32-NEXT:    v_readlane_b32 s6, v1, 31
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
+; GFX11W32-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX11W32-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX11W32-NEXT:    v_readlane_b32 s5, v1, 15
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX11W32-NEXT:    v_readlane_b32 s3, v1, 15
+; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
+; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX11W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11W32-NEXT:    v_writelane_b32 v3, s5, 16
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX11W32-NEXT:    s_or_saveexec_b32 s2, -1
+; GFX11W32-NEXT:    v_writelane_b32 v3, s3, 16
+; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr0
-; GFX11W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX11W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB3_2
 ; GFX11W32-NEXT:  ; %bb.1:
-; GFX11W32-NEXT:    s_mov_b32 s5, s6
+; GFX11W32-NEXT:    s_mov_b32 s3, s4
 ; GFX11W32-NEXT:    s_clause 0x1
-; GFX11W32-NEXT:    s_load_b32 s6, s[0:1], 0x44
-; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    v_mov_b32_e32 v0, s5
+; GFX11W32-NEXT:    s_load_b32 s8, s[0:1], 0x44
+; GFX11W32-NEXT:    s_load_b128 s[4:7], s[0:1], 0x34
+; GFX11W32-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    v_mov_b32_e32 v4, s6
-; GFX11W32-NEXT:    buffer_atomic_add_u32 v0, v4, s[8:11], 0 idxen glc
+; GFX11W32-NEXT:    v_mov_b32_e32 v4, s8
+; GFX11W32-NEXT:    buffer_atomic_add_u32 v0, v4, s[4:7], 0 idxen glc
 ; GFX11W32-NEXT:  .LBB3_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX11W32-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX11W32-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX11W32-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W32-NEXT:    v_add_nc_u32_e32 v0, s0, v0
+; GFX11W32-NEXT:    v_add_nc_u32_e32 v0, s2, v0
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    global_store_b32 v4, v0, s[2:3]
+; GFX11W32-NEXT:    global_store_b32 v4, v0, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
   %old = call i32 @llvm.amdgcn.struct.buffer.atomic.add(i32 %lane, <4 x i32> %inout, i32 %vindex, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
entry:
   %old = call i32 @llvm.amdgcn.raw.buffer.atomic.add(i32 1, <4 x i32> %inout, i32 %lane, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @sub_i32_constant(i32 addrspace(1)* %out, <4 x i32> %inout) {
 ; GFX6-LABEL: sub_i32_constant:
 ; GFX6:       ; %bb.0: ; %entry
-; GFX6-NEXT:    s_mov_b64 s[2:3], exec
-; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
-; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
+; GFX6-NEXT:    s_mov_b64 s[4:5], exec
+; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s4, 0
+; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s5, v0
 ; GFX6-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX6-NEXT:    ; implicit-def: $vgpr1
-; GFX6-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX6-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX6-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX6-NEXT:  ; %bb.1:
 ; GFX6-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0xd
-; GFX6-NEXT:    s_bcnt1_i32_b64 s0, s[2:3]
-; GFX6-NEXT:    s_mul_i32 s0, s0, 5
-; GFX6-NEXT:    v_mov_b32_e32 v1, s0
+; GFX6-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX6-NEXT:    s_mul_i32 s4, s4, 5
+; GFX6-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
 ; GFX6-NEXT:  .LBB5_2:
-; GFX6-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX6-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
-; GFX6-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX6-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX6-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
-; GFX6-NEXT:    s_mov_b32 s7, 0xf000
-; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, s0, v0
-; GFX6-NEXT:    s_mov_b32 s6, -1
+; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, s4, v0
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: sub_i32_constant:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[6:7], exec
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX8-NEXT:    s_mov_b64 s[4:5], exec
+; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr1
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX8-NEXT:  ; %bb.1:
 ; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX8-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
-; GFX8-NEXT:    s_mul_i32 s0, s0, 5
-; GFX8-NEXT:    v_mov_b32_e32 v1, s0
+; GFX8-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX8-NEXT:    s_mul_i32 s4, s4, 5
+; GFX8-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
 ; GFX8-NEXT:  .LBB5_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX8-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
-; GFX8-NEXT:    v_sub_u32_e32 v2, vcc, s0, v0
+; GFX8-NEXT:    v_sub_u32_e32 v2, vcc, s2, v0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    flat_store_dword v[0:1], v2
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sub_i32_constant:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[6:7], exec
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX9-NEXT:    s_mov_b64 s[4:5], exec
+; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr1
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX9-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
-; GFX9-NEXT:    s_mul_i32 s0, s0, 5
-; GFX9-NEXT:    v_mov_b32_e32 v1, s0
+; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX9-NEXT:    s_mul_i32 s4, s4, 5
+; GFX9-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
 ; GFX9-NEXT:  .LBB5_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX9-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
-; GFX9-NEXT:    v_sub_u32_e32 v0, s0, v0
-; GFX9-NEXT:    v_mov_b32_e32 v1, 0
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_sub_u32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX9-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: sub_i32_constant:
 ; GFX10W64:       ; %bb.0: ; %entry
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_mov_b64 s[6:7], exec
+; GFX10W64-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr1
-; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX10W64-NEXT:  ; %bb.1:
 ; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W64-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
-; GFX10W64-NEXT:    s_mul_i32 s0, s0, 5
-; GFX10W64-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX10W64-NEXT:    s_mul_i32 s4, s4, 5
+; GFX10W64-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10W64-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
 ; GFX10W64-NEXT:  .LBB5_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W64-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W64-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX10W64-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W64-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: sub_i32_constant:
 ; GFX10W32:       ; %bb.0: ; %entry
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_mov_b32 s5, exec_lo
+; GFX10W32-NEXT:    s_mov_b32 s3, exec_lo
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr1
-; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s5, 0
+; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX10W32-NEXT:  ; %bb.1:
-; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    s_bcnt1_i32_b32 s0, s5
-; GFX10W32-NEXT:    s_mul_i32 s0, s0, 5
-; GFX10W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10W32-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x34
+; GFX10W32-NEXT:    s_bcnt1_i32_b32 s3, s3
+; GFX10W32-NEXT:    s_mul_i32 s3, s3, 5
+; GFX10W32-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
+; GFX10W32-NEXT:    buffer_atomic_sub v1, off, s[4:7], 0 glc
 ; GFX10W32-NEXT:  .LBB5_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W32-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W32-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W32-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX10W32-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W32-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: sub_i32_constant:
 ; GFX11W64:       ; %bb.0: ; %entry
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_mov_b64 s[6:7], exec
 ; GFX11W64-NEXT:    s_mov_b64 s[4:5], exec
-; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX11W64-NEXT:    s_mov_b64 s[2:3], exec
+; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX11W64-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX11W64-NEXT:  ; %bb.1:
 ; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W64-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
+; GFX11W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX11W64-NEXT:    s_mul_i32 s0, s0, 5
-; GFX11W64-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11W64-NEXT:    s_mul_i32 s4, s4, 5
+; GFX11W64-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11W64-NEXT:    buffer_atomic_sub_u32 v1, off, s[8:11], 0 glc
 ; GFX11W64-NEXT:  .LBB5_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W64-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W64-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX11W64-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11W64-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: sub_i32_constant:
 ; GFX11W32:       ; %bb.0: ; %entry
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_mov_b32 s5, exec_lo
-; GFX11W32-NEXT:    s_mov_b32 s4, exec_lo
-; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s5, 0
+; GFX11W32-NEXT:    s_mov_b32 s3, exec_lo
+; GFX11W32-NEXT:    s_mov_b32 s2, exec_lo
+; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX11W32-NEXT:  ; %bb.1:
-; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    s_bcnt1_i32_b32 s0, s5
+; GFX11W32-NEXT:    s_load_b128 s[4:7], s[0:1], 0x34
+; GFX11W32-NEXT:    s_bcnt1_i32_b32 s3, s3
 ; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX11W32-NEXT:    s_mul_i32 s0, s0, 5
-; GFX11W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11W32-NEXT:    s_mul_i32 s3, s3, 5
+; GFX11W32-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    buffer_atomic_sub_u32 v1, off, s[8:11], 0 glc
+; GFX11W32-NEXT:    buffer_atomic_sub_u32 v1, off, s[4:7], 0 glc
 ; GFX11W32-NEXT:  .LBB5_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W32-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W32-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W32-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX11W32-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11W32-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %old = call i32 @llvm.amdgcn.raw.buffer.atomic.sub(i32 5, <4 x i32> %inout, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @sub_i32_uniform(i32 addrspace(1)* %out, <4 x i32> %inout, i32 %subitive) {
 ; GFX6-LABEL: sub_i32_uniform:
 ; GFX6:       ; %bb.0: ; %entry
-; GFX6-NEXT:    s_mov_b64 s[2:3], exec
-; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
-; GFX6-NEXT:    s_load_dword s8, s[0:1], 0x11
-; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
+; GFX6-NEXT:    s_mov_b64 s[4:5], exec
+; GFX6-NEXT:    s_load_dword s6, s[0:1], 0x11
+; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s4, 0
+; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s5, v0
 ; GFX6-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX6-NEXT:    ; implicit-def: $vgpr1
-; GFX6-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX6-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX6-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX6-NEXT:  ; %bb.1:
-; GFX6-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0xd
-; GFX6-NEXT:    s_bcnt1_i32_b64 s0, s[2:3]
+; GFX6-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0xd
+; GFX6-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    s_mul_i32 s0, s8, s0
-; GFX6-NEXT:    v_mov_b32_e32 v1, s0
-; GFX6-NEXT:    buffer_atomic_sub v1, off, s[12:15], 0 glc
+; GFX6-NEXT:    s_mul_i32 s4, s6, s4
+; GFX6-NEXT:    v_mov_b32_e32 v1, s4
+; GFX6-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
 ; GFX6-NEXT:  .LBB6_2:
-; GFX6-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX6-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
-; GFX6-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX6-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    v_mul_lo_u32 v0, s8, v0
-; GFX6-NEXT:    s_mov_b32 s7, 0xf000
-; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, s0, v0
-; GFX6-NEXT:    s_mov_b32 s6, -1
-; GFX6-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX6-NEXT:    v_mul_lo_u32 v0, s6, v0
+; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, s4, v0
+; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: sub_i32_uniform:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_load_dword s8, s[0:1], 0x44
+; GFX8-NEXT:    s_load_dword s6, s[0:1], 0x44
 ; GFX8-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr1
-; GFX8-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0x34
-; GFX8-NEXT:    s_bcnt1_i32_b64 s0, s[4:5]
+; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX8-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_mul_i32 s0, s8, s0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s0
-; GFX8-NEXT:    buffer_atomic_sub v1, off, s[12:15], 0 glc
+; GFX8-NEXT:    s_mul_i32 s4, s6, s4
+; GFX8-NEXT:    v_mov_b32_e32 v1, s4
+; GFX8-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
 ; GFX8-NEXT:  .LBB6_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mul_lo_u32 v0, s8, v0
+; GFX8-NEXT:    v_mul_lo_u32 v0, s6, v0
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX8-NEXT:    v_sub_u32_e32 v2, vcc, s0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX8-NEXT:    v_sub_u32_e32 v2, vcc, s2, v0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    flat_store_dword v[0:1], v2
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sub_i32_uniform:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_load_dword s8, s[0:1], 0x44
+; GFX9-NEXT:    s_load_dword s6, s[0:1], 0x44
 ; GFX9-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr1
-; GFX9-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0x34
-; GFX9-NEXT:    s_bcnt1_i32_b64 s0, s[4:5]
+; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_mul_i32 s0, s8, s0
-; GFX9-NEXT:    v_mov_b32_e32 v1, s0
-; GFX9-NEXT:    buffer_atomic_sub v1, off, s[12:15], 0 glc
+; GFX9-NEXT:    s_mul_i32 s4, s6, s4
+; GFX9-NEXT:    v_mov_b32_e32 v1, s4
+; GFX9-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
 ; GFX9-NEXT:  .LBB6_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mul_lo_u32 v0, s8, v0
+; GFX9-NEXT:    v_mul_lo_u32 v0, s6, v0
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX9-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-NEXT:    v_sub_u32_e32 v0, s0, v0
-; GFX9-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_sub_u32_e32 v0, s2, v0
+; GFX9-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: sub_i32_uniform:
 ; GFX10W64:       ; %bb.0: ; %entry
-; GFX10W64-NEXT:    s_clause 0x1
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_load_dword s8, s[0:1], 0x44
+; GFX10W64-NEXT:    s_load_dword s6, s[0:1], 0x44
 ; GFX10W64-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr1
 ; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX10W64-NEXT:  ; %bb.1:
-; GFX10W64-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0x34
-; GFX10W64-NEXT:    s_bcnt1_i32_b64 s0, s[4:5]
+; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX10W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    s_mul_i32 s0, s8, s0
-; GFX10W64-NEXT:    v_mov_b32_e32 v1, s0
-; GFX10W64-NEXT:    buffer_atomic_sub v1, off, s[12:15], 0 glc
+; GFX10W64-NEXT:    s_mul_i32 s4, s6, s4
+; GFX10W64-NEXT:    v_mov_b32_e32 v1, s4
+; GFX10W64-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
 ; GFX10W64-NEXT:  .LBB6_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    v_mul_lo_u32 v0, s8, v0
+; GFX10W64-NEXT:    v_mul_lo_u32 v0, s6, v0
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W64-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
-; GFX10W64-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W64-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
+; GFX10W64-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: sub_i32_uniform:
 ; GFX10W32:       ; %bb.0: ; %entry
-; GFX10W32-NEXT:    s_clause 0x1
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_load_dword s4, s[0:1], 0x44
-; GFX10W32-NEXT:    s_mov_b32 s6, exec_lo
+; GFX10W32-NEXT:    s_load_dword s2, s[0:1], 0x44
+; GFX10W32-NEXT:    s_mov_b32 s4, exec_lo
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr1
-; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s5, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX10W32-NEXT:  ; %bb.1:
 ; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    s_bcnt1_i32_b32 s0, s6
+; GFX10W32-NEXT:    s_bcnt1_i32_b32 s4, s4
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    s_mul_i32 s0, s4, s0
-; GFX10W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10W32-NEXT:    s_mul_i32 s4, s2, s4
+; GFX10W32-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX10W32-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
 ; GFX10W32-NEXT:  .LBB6_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s5
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    v_mul_lo_u32 v0, s4, v0
+; GFX10W32-NEXT:    v_mul_lo_u32 v0, s2, v0
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W32-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W32-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
-; GFX10W32-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W32-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
+; GFX10W32-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: sub_i32_uniform:
 ; GFX11W64:       ; %bb.0: ; %entry
-; GFX11W64-NEXT:    s_clause 0x1
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_load_b32 s8, s[0:1], 0x44
+; GFX11W64-NEXT:    s_load_b32 s6, s[0:1], 0x44
 ; GFX11W64-NEXT:    s_mov_b64 s[4:5], exec
-; GFX11W64-NEXT:    s_mov_b64 s[6:7], exec
+; GFX11W64-NEXT:    s_mov_b64 s[2:3], exec
 ; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX11W64-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX11W64-NEXT:  ; %bb.1:
-; GFX11W64-NEXT:    s_load_b128 s[12:15], s[0:1], 0x34
-; GFX11W64-NEXT:    s_bcnt1_i32_b64 s0, s[4:5]
+; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
+; GFX11W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    s_mul_i32 s0, s8, s0
+; GFX11W64-NEXT:    s_mul_i32 s4, s6, s4
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11W64-NEXT:    v_mov_b32_e32 v1, s0
-; GFX11W64-NEXT:    buffer_atomic_sub_u32 v1, off, s[12:15], 0 glc
+; GFX11W64-NEXT:    v_mov_b32_e32 v1, s4
+; GFX11W64-NEXT:    buffer_atomic_sub_u32 v1, off, s[8:11], 0 glc
 ; GFX11W64-NEXT:  .LBB6_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    v_mul_lo_u32 v0, s8, v0
+; GFX11W64-NEXT:    v_mul_lo_u32 v0, s6, v0
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W64-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
-; GFX11W64-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11W64-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
+; GFX11W64-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: sub_i32_uniform:
 ; GFX11W32:       ; %bb.0: ; %entry
-; GFX11W32-NEXT:    s_clause 0x1
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_load_b32 s4, s[0:1], 0x44
-; GFX11W32-NEXT:    s_mov_b32 s6, exec_lo
-; GFX11W32-NEXT:    s_mov_b32 s5, exec_lo
-; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX11W32-NEXT:    s_load_b32 s2, s[0:1], 0x44
+; GFX11W32-NEXT:    s_mov_b32 s4, exec_lo
+; GFX11W32-NEXT:    s_mov_b32 s3, exec_lo
+; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX11W32-NEXT:  ; %bb.1:
 ; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    s_bcnt1_i32_b32 s0, s6
+; GFX11W32-NEXT:    s_bcnt1_i32_b32 s4, s4
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    s_mul_i32 s0, s4, s0
+; GFX11W32-NEXT:    s_mul_i32 s4, s2, s4
 ; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11W32-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX11W32-NEXT:    buffer_atomic_sub_u32 v1, off, s[8:11], 0 glc
 ; GFX11W32-NEXT:  .LBB6_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s5
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    v_mul_lo_u32 v0, s4, v0
+; GFX11W32-NEXT:    v_mul_lo_u32 v0, s2, v0
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W32-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W32-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
-; GFX11W32-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11W32-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
+; GFX11W32-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %old = call i32 @llvm.amdgcn.raw.buffer.atomic.sub(i32 %subitive, <4 x i32> %inout, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @sub_i32_varying_vdata(i32 addrspace(1)* %out, <4 x i32> %inout) {
 ; GFX6-LABEL: sub_i32_varying_vdata:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0xd
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_sub v0, off, s[4:7], 0 glc
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: sub_i32_varying_vdata:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
-; GFX8-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX8-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX8-NEXT:    s_not_b64 exec, exec
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX8-NEXT:    s_not_b64 exec, exec
-; GFX8-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX8-NEXT:    v_readlane_b32 s6, v2, 63
+; GFX8-NEXT:    v_readlane_b32 s4, v2, 63
 ; GFX8-NEXT:    s_nop 0
 ; GFX8-NEXT:    v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf
-; GFX8-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
 ; GFX8-NEXT:    ; implicit-def: $vgpr0
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX8-NEXT:  ; %bb.1:
 ; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX8-NEXT:    v_mov_b32_e32 v0, s6
+; GFX8-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    buffer_atomic_sub v0, off, s[8:11], 0 glc
 ; GFX8-NEXT:  .LBB7_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, v1
+; GFX8-NEXT:    v_sub_u32_e32 v0, vcc, s2, v0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v4, s3
-; GFX8-NEXT:    v_sub_u32_e32 v0, vcc, s0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s2
+; GFX8-NEXT:    v_mov_b32_e32 v4, s1
+; GFX8-NEXT:    v_mov_b32_e32 v3, s0
 ; GFX8-NEXT:    flat_store_dword v[3:4], v0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sub_i32_varying_vdata:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX9-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX9-NEXT:    s_not_b64 exec, exec
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    s_not_b64 exec, exec
-; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-NEXT:    v_readlane_b32 s6, v2, 63
+; GFX9-NEXT:    v_readlane_b32 s4, v2, 63
 ; GFX9-NEXT:    s_nop 0
 ; GFX9-NEXT:    v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf
-; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
 ; GFX9-NEXT:    ; implicit-def: $vgpr0
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX9-NEXT:    v_mov_b32_e32 v0, s6
+; GFX9-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    buffer_atomic_sub v0, off, s[8:11], 0 glc
 ; GFX9-NEXT:  .LBB7_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v1
-; GFX9-NEXT:    v_sub_u32_e32 v0, s0, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, 0
+; GFX9-NEXT:    v_sub_u32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    global_store_dword v3, v0, s[2:3]
+; GFX9-NEXT:    global_store_dword v3, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: sub_i32_varying_vdata:
 ; GFX10W64:       ; %bb.0: ; %entry
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX10W64-NEXT:    s_not_b64 exec, exec
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10W64-NEXT:    s_not_b64 exec, exec
 ; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX10W64-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX10W64-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 15
 ; GFX10W64-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX10W64-NEXT:    v_readlane_b32 s5, v1, 31
+; GFX10W64-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_readlane_b32 s7, v1, 31
-; GFX10W64-NEXT:    v_writelane_b32 v3, s6, 16
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
 ; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 63
-; GFX10W64-NEXT:    v_readlane_b32 s8, v1, 47
-; GFX10W64-NEXT:    v_writelane_b32 v3, s7, 32
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 63
+; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 47
+; GFX10W64-NEXT:    v_writelane_b32 v3, s5, 32
+; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_writelane_b32 v3, s8, 48
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX10W64-NEXT:    v_writelane_b32 v3, s6, 48
+; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX10W64-NEXT:  ; %bb.1:
 ; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W64-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10W64-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10W64-NEXT:    buffer_atomic_sub v0, off, s[8:11], 0 glc
 ; GFX10W64-NEXT:  .LBB7_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX10W64-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX10W64-NEXT:    v_mov_b32_e32 v4, 0
-; GFX10W64-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX10W64-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    global_store_dword v4, v0, s[2:3]
+; GFX10W64-NEXT:    global_store_dword v4, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: sub_i32_varying_vdata:
 ; GFX10W32:       ; %bb.0: ; %entry
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX10W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX10W32-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX10W32-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX10W32-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_or_saveexec_b32 s4, -1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX10W32-NEXT:    v_mov_b32_e32 v3, 0
-; GFX10W32-NEXT:    v_readlane_b32 s6, v1, 31
+; GFX10W32-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX10W32-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX10W32-NEXT:    v_readlane_b32 s5, v1, 15
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10W32-NEXT:    v_readlane_b32 s3, v1, 15
+; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX10W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX10W32-NEXT:    v_writelane_b32 v3, s5, 16
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10W32-NEXT:    s_or_saveexec_b32 s2, -1
+; GFX10W32-NEXT:    v_writelane_b32 v3, s3, 16
+; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX10W32-NEXT:  ; %bb.1:
-; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    v_mov_b32_e32 v0, s6
-; GFX10W32-NEXT:    s_mov_b32 s5, s6
+; GFX10W32-NEXT:    s_mov_b32 s3, s4
+; GFX10W32-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x34
+; GFX10W32-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    buffer_atomic_sub v0, off, s[8:11], 0 glc
+; GFX10W32-NEXT:    buffer_atomic_sub v0, off, s[4:7], 0 glc
 ; GFX10W32-NEXT:  .LBB7_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10W32-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX10W32-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX10W32-NEXT:    v_mov_b32_e32 v4, 0
-; GFX10W32-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX10W32-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    global_store_dword v4, v0, s[2:3]
+; GFX10W32-NEXT:    global_store_dword v4, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: sub_i32_varying_vdata:
 ; GFX11W64:       ; %bb.0: ; %entry
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX11W64-NEXT:    s_not_b64 exec, exec
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W64-NEXT:    s_not_b64 exec, exec

 ; GFX11W64-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX11W64-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W64-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX11W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 15
+; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
 ; GFX11W64-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX11W64-NEXT:    v_readlane_b32 s5, v1, 31
+; GFX11W64-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_readlane_b32 s7, v1, 31
-; GFX11W64-NEXT:    v_writelane_b32 v3, s6, 16
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 63
-; GFX11W64-NEXT:    v_readlane_b32 s8, v1, 47
-; GFX11W64-NEXT:    v_writelane_b32 v3, s7, 32
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX11W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 63
+; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 47
+; GFX11W64-NEXT:    v_writelane_b32 v3, s5, 32
+; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
 ; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_writelane_b32 v3, s8, 48
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX11W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX11W64-NEXT:    v_writelane_b32 v3, s6, 48
+; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX11W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr0
-; GFX11W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX11W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX11W64-NEXT:  ; %bb.1:
 ; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W64-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11W64-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11W64-NEXT:    buffer_atomic_sub_u32 v0, off, s[8:11], 0 glc
 ; GFX11W64-NEXT:  .LBB7_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX11W64-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX11W64-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W64-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX11W64-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    global_store_b32 v4, v0, s[2:3]
+; GFX11W64-NEXT:    global_store_b32 v4, v0, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: sub_i32_varying_vdata:
 ; GFX11W32:       ; %bb.0: ; %entry
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX11W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX11W32-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX11W32-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX11W32-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX11W32-NEXT:    v_mov_b32_e32 v3, 0
-; GFX11W32-NEXT:    v_readlane_b32 s6, v1, 31
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
+; GFX11W32-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX11W32-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX11W32-NEXT:    v_readlane_b32 s5, v1, 15
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX11W32-NEXT:    v_readlane_b32 s3, v1, 15
+; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
+; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX11W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11W32-NEXT:    v_writelane_b32 v3, s5, 16
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX11W32-NEXT:    s_or_saveexec_b32 s2, -1
+; GFX11W32-NEXT:    v_writelane_b32 v3, s3, 16
+; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr0
-; GFX11W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX11W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX11W32-NEXT:  ; %bb.1:
-; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    v_mov_b32_e32 v0, s6
-; GFX11W32-NEXT:    s_mov_b32 s5, s6
+; GFX11W32-NEXT:    s_mov_b32 s3, s4
+; GFX11W32-NEXT:    s_load_b128 s[4:7], s[0:1], 0x34
+; GFX11W32-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    buffer_atomic_sub_u32 v0, off, s[8:11], 0 glc
+; GFX11W32-NEXT:    buffer_atomic_sub_u32 v0, off, s[4:7], 0 glc
 ; GFX11W32-NEXT:  .LBB7_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX11W32-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX11W32-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX11W32-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W32-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX11W32-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    global_store_b32 v4, v0, s[2:3]
+; GFX11W32-NEXT:    global_store_b32 v4, v0, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
   %old = call i32 @llvm.amdgcn.raw.buffer.atomic.sub(i32 %lane, <4 x i32> %inout, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }

llvm/test/CodeGen/AMDGPU/atomic_optimizations_local_pointer.ll

 
 ; Show what the atomic optimization pass will do for local pointers.
 
 define amdgpu_kernel void @add_i32_constant(i32 addrspace(1)* %out) {
 ;
 ;
 ; GFX7LESS-LABEL: add_i32_constant:
 ; GFX7LESS:       ; %bb.0: ; %entry
-; GFX7LESS-NEXT:    s_mov_b64 s[2:3], exec
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
+; GFX7LESS-NEXT:    s_mov_b64 s[4:5], exec
+; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s4, 0
+; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s5, v0
 ; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX7LESS-NEXT:    ; implicit-def: $vgpr1
-; GFX7LESS-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX7LESS-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX7LESS-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX7LESS-NEXT:  ; %bb.1:
-; GFX7LESS-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX7LESS-NEXT:    s_mul_i32 s2, s2, 5
+; GFX7LESS-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX7LESS-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v1, 0
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s2
+; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX7LESS-NEXT:    s_mov_b32 m0, -1
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    ds_add_rtn_u32 v1, v1, v2
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:  .LBB0_2:
-; GFX7LESS-NEXT:    s_or_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX7LESS-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
 ; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX7LESS-NEXT:    s_mov_b32 s2, -1
+; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX7LESS-NEXT:    v_mad_u32_u24 v0, v0, 5, s4
 ; GFX7LESS-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: add_i32_constant:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX8-NEXT:    s_mov_b64 s[4:5], exec
+; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr1
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX8-NEXT:    s_mul_i32 s2, s2, 5
+; GFX8-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX8-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s2
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_add_rtn_u32 v1, v1, v2
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB0_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s2, v1
-; GFX8-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX8-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX8-NEXT:    s_mov_b32 s2, -1
-; GFX8-NEXT:    s_nop 1
+; GFX8-NEXT:    v_mad_u32_u24 v0, v0, 5, s4
 ; GFX8-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: add_i32_constant:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX9-NEXT:    s_mov_b64 s[4:5], exec
+; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr1
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX9-NEXT:    s_mul_i32 s2, s2, 5
+; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX9-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_add_rtn_u32 v1, v1, v2
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB0_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s2, v1
-; GFX9-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
+; GFX9-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1
-; GFX9-NEXT:    s_nop 1
+; GFX9-NEXT:    v_mad_u32_u24 v0, v0, 5, s4
 ; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: add_i32_constant:
 ; GFX1064:       ; %bb.0: ; %entry
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
+; GFX1064-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX1064-NEXT:    ; implicit-def: $vgpr1
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX1064-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX1064-NEXT:  ; %bb.1:
-; GFX1064-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
+; GFX1064-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
-; GFX1064-NEXT:    s_mul_i32 s2, s2, 5
-; GFX1064-NEXT:    v_mov_b32_e32 v2, s2
+; GFX1064-NEXT:    s_mul_i32 s4, s4, 5
+; GFX1064-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_add_rtn_u32 v1, v1, v2
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB0_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1064-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX1064-NEXT:    s_mov_b32 s2, -1
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: add_i32_constant:
 ; GFX1032:       ; %bb.0: ; %entry
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    s_mov_b32 s3, exec_lo
 ; GFX1032-NEXT:    ; implicit-def: $vgpr1
 ; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1032-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    s_bcnt1_i32_b32 s3, s3
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1032-NEXT:    s_mul_i32 s3, s3, 5
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, s3
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_add_rtn_u32 v1, v1, v2
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB0_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1032-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;
 ; GFX1164-LABEL: add_i32_constant:
 ; GFX1164:       ; %bb.0: ; %entry
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
 ; GFX1164-NEXT:    s_mov_b64 s[4:5], exec
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
+; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
+; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX1164-NEXT:    ; implicit-def: $vgpr1
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX1164-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX1164-NEXT:  ; %bb.1:
-; GFX1164-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
+; GFX1164-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX1164-NEXT:    v_mov_b32_e32 v1, 0
-; GFX1164-NEXT:    s_mul_i32 s2, s2, 5
+; GFX1164-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    v_mov_b32_e32 v2, s2
+; GFX1164-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_add_rtn_u32 v1, v1, v2
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB0_2:
-; GFX1164-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1164-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX1164-NEXT:    s_mov_b32 s2, -1
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: add_i32_constant:
 ; GFX1132:       ; %bb.0: ; %entry
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
 ; GFX1132-NEXT:    s_mov_b32 s2, exec_lo
 ; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
 ; GFX1132-NEXT:    ; implicit-def: $vgpr1
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX1132-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX1132-NEXT:  ; %bb.1:
 ; GFX1132-NEXT:    s_bcnt1_i32_b32 s3, s3
 ; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
 ; GFX1132-NEXT:    s_mul_i32 s3, s3, 5
 ; GFX1132-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s3
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_add_rtn_u32 v1, v1, v2
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB0_2:
 ; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %old = atomicrmw add i32 addrspace(3)* @local_var32, i32 5 acq_rel
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @add_i32_uniform(i32 addrspace(1)* %out, i32 %additive) {
 ;
 ;
 ; GFX7LESS-LABEL: add_i32_uniform:
 ; GFX7LESS:       ; %bb.0: ; %entry
-; GFX7LESS-NEXT:    s_mov_b64 s[2:3], exec
-; GFX7LESS-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
+; GFX7LESS-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX7LESS-NEXT:    s_load_dword s6, s[0:1], 0xb
-; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
+; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s4, 0
+; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s5, v0
 ; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX7LESS-NEXT:    ; implicit-def: $vgpr1
-; GFX7LESS-NEXT:    s_and_saveexec_b64 s[0:1], vcc
+; GFX7LESS-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX7LESS-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX7LESS-NEXT:  ; %bb.1:
-; GFX7LESS-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
+; GFX7LESS-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_mul_i32 s2, s6, s2
+; GFX7LESS-NEXT:    s_mul_i32 s4, s6, s4
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v1, 0
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s2
+; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX7LESS-NEXT:    s_mov_b32 m0, -1
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    ds_add_rtn_u32 v1, v1, v2
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:  .LBB1_2:
-; GFX7LESS-NEXT:    s_or_b64 exec, exec, s[0:1]
+; GFX7LESS-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX7LESS-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX7LESS-NEXT:    v_mul_lo_u32 v0, s6, v0
-; GFX7LESS-NEXT:    s_mov_b32 s7, 0xf000
-; GFX7LESS-NEXT:    v_add_i32_e32 v0, vcc, s0, v0
-; GFX7LESS-NEXT:    s_mov_b32 s6, -1
-; GFX7LESS-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX7LESS-NEXT:    v_add_i32_e32 v0, vcc, s4, v0
+; GFX7LESS-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: add_i32_uniform:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
 ; GFX8-NEXT:    s_load_dword s6, s[0:1], 0x2c
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX8-NEXT:    s_mov_b64 s[4:5], exec
+; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr1
-; GFX8-NEXT:    s_and_saveexec_b64 s[0:1], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
+; GFX8-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_mul_i32 s2, s6, s2
+; GFX8-NEXT:    s_mul_i32 s4, s6, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s2
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_add_rtn_u32 v1, v1, v2
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB1_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[0:1]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    v_mul_lo_u32 v0, s6, v0
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX8-NEXT:    s_mov_b32 s7, 0xf000
-; GFX8-NEXT:    s_mov_b32 s6, -1
-; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v0
-; GFX8-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX8-NEXT:    s_mov_b32 s3, 0xf000
+; GFX8-NEXT:    s_mov_b32 s2, -1
+; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s4, v0
+; GFX8-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: add_i32_uniform:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
 ; GFX9-NEXT:    s_load_dword s6, s[0:1], 0x2c
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX9-NEXT:    s_mov_b64 s[4:5], exec
+; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr1
-; GFX9-NEXT:    s_and_saveexec_b64 s[0:1], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
+; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_mul_i32 s2, s6, s2
+; GFX9-NEXT:    s_mul_i32 s4, s6, s4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_add_rtn_u32 v1, v1, v2
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB1_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[0:1]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_mul_lo_u32 v0, s6, v0
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX9-NEXT:    s_mov_b32 s7, 0xf000
-; GFX9-NEXT:    s_mov_b32 s6, -1
-; GFX9-NEXT:    v_add_u32_e32 v0, s0, v0
-; GFX9-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX9-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX9-NEXT:    s_mov_b32 s3, 0xf000
+; GFX9-NEXT:    s_mov_b32 s2, -1
+; GFX9-NEXT:    v_add_u32_e32 v0, s4, v0
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: add_i32_uniform:
 ; GFX1064:       ; %bb.0: ; %entry
-; GFX1064-NEXT:    s_clause 0x1
-; GFX1064-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
 ; GFX1064-NEXT:    s_load_dword s6, s[0:1], 0x2c
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
+; GFX1064-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX1064-NEXT:    ; implicit-def: $vgpr1
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[0:1], vcc
+; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX1064-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX1064-NEXT:  ; %bb.1:
-; GFX1064-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
+; GFX1064-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_mul_i32 s2, s6, s2
-; GFX1064-NEXT:    v_mov_b32_e32 v2, s2
+; GFX1064-NEXT:    s_mul_i32 s4, s6, s4
+; GFX1064-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_add_rtn_u32 v1, v1, v2
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB1_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX1064-NEXT:    s_or_b64 exec, exec, s[0:1]
-; GFX1064-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX1064-NEXT:    s_mov_b32 s7, 0x31016000
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], s6, v0, s[0:1]
-; GFX1064-NEXT:    s_mov_b32 s6, -1
-; GFX1064-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX1064-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1064-NEXT:    v_mad_u64_u32 v[0:1], s[2:3], s6, v0, s[2:3]
+; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
+; GFX1064-NEXT:    s_mov_b32 s2, -1
+; GFX1064-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: add_i32_uniform:
 ; GFX1032:       ; %bb.0: ; %entry
-; GFX1032-NEXT:    s_clause 0x1
-; GFX1032-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
 ; GFX1032-NEXT:    s_load_dword s2, s[0:1], 0x2c
-; GFX1032-NEXT:    s_mov_b32 s3, exec_lo
+; GFX1032-NEXT:    s_mov_b32 s4, exec_lo
 ; GFX1032-NEXT:    ; implicit-def: $vgpr1
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
+; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s0, vcc_lo
+; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX1032-NEXT:  ; %bb.1:
-; GFX1032-NEXT:    s_bcnt1_i32_b32 s1, s3
+; GFX1032-NEXT:    s_bcnt1_i32_b32 s4, s4
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_mul_i32 s1, s2, s1
-; GFX1032-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1032-NEXT:    s_mul_i32 s4, s2, s4
+; GFX1032-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_add_rtn_u32 v1, v1, v2
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB1_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s0
-; GFX1032-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX1032-NEXT:    s_mov_b32 s7, 0x31016000
-; GFX1032-NEXT:    s_mov_b32 s6, -1
+; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX1032-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mad_u64_u32 v[0:1], s0, s2, v0, s[0:1]
-; GFX1032-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX1032-NEXT:    v_mad_u64_u32 v[0:1], s2, s2, v0, s[4:5]
+; GFX1032-NEXT:    s_mov_b32 s2, -1
+; GFX1032-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;
 ; GFX1164-LABEL: add_i32_uniform:
 ; GFX1164:       ; %bb.0: ; %entry
-; GFX1164-NEXT:    s_clause 0x1
-; GFX1164-NEXT:    s_load_b64 s[4:5], s[0:1], 0x24
 ; GFX1164-NEXT:    s_load_b32 s6, s[0:1], 0x2c
+; GFX1164-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_mov_b64 s[0:1], exec
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
+; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX1164-NEXT:    ; implicit-def: $vgpr1
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX1164-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX1164-NEXT:  ; %bb.1:
-; GFX1164-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
+; GFX1164-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX1164-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_mul_i32 s2, s6, s2
+; GFX1164-NEXT:    s_mul_i32 s4, s6, s4
 ; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    v_mov_b32_e32 v2, s2
+; GFX1164-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_add_rtn_u32 v1, v1, v2
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB1_2:
-; GFX1164-NEXT:    s_or_b64 exec, exec, s[0:1]
-; GFX1164-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX1164-NEXT:    s_mov_b32 s7, 0x31016000
+; GFX1164-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX1164-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_mad_u64_u32 v[1:2], null, s6, v0, s[0:1]
-; GFX1164-NEXT:    s_mov_b32 s6, -1
-; GFX1164-NEXT:    buffer_store_b32 v1, off, s[4:7], 0
+; GFX1164-NEXT:    v_mad_u64_u32 v[1:2], null, s6, v0, s[2:3]
+; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
+; GFX1164-NEXT:    s_mov_b32 s2, -1
+; GFX1164-NEXT:    buffer_store_b32 v1, off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: add_i32_uniform:
 ; GFX1132:       ; %bb.0: ; %entry
-; GFX1132-NEXT:    s_clause 0x1
-; GFX1132-NEXT:    s_load_b64 s[4:5], s[0:1], 0x24
-; GFX1132-NEXT:    s_load_b32 s0, s[0:1], 0x2c
-; GFX1132-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1132-NEXT:    s_mov_b32 s1, exec_lo
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
+; GFX1132-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX1132-NEXT:    s_mov_b32 s4, exec_lo
+; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
+; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX1132-NEXT:    ; implicit-def: $vgpr1
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX1132-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX1132-NEXT:  ; %bb.1:
-; GFX1132-NEXT:    s_bcnt1_i32_b32 s2, s2
+; GFX1132-NEXT:    s_bcnt1_i32_b32 s4, s4
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_mul_i32 s2, s0, s2
+; GFX1132-NEXT:    s_mul_i32 s4, s2, s4
 ; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s2
+; GFX1132-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s4
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_add_rtn_u32 v1, v1, v2
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB1_2:
-; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s1
-; GFX1132-NEXT:    v_readfirstlane_b32 s2, v1
-; GFX1132-NEXT:    s_mov_b32 s7, 0x31016000
-; GFX1132-NEXT:    s_mov_b32 s6, -1
+; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX1132-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_mad_u64_u32 v[1:2], null, s0, v0, s[2:3]
-; GFX1132-NEXT:    buffer_store_b32 v1, off, s[4:7], 0
+; GFX1132-NEXT:    v_mad_u64_u32 v[1:2], null, s2, v0, s[4:5]
+; GFX1132-NEXT:    s_mov_b32 s2, -1
+; GFX1132-NEXT:    buffer_store_b32 v1, off, s[0:3], 0
 ; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %old = atomicrmw add i32 addrspace(3)* @local_var32, i32 %additive acq_rel
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 

 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: add_i32_varying:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX8-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX8-NEXT:    s_not_b64 exec, exec
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0

 ; GFX8-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_add_rtn_u32 v0, v0, v3
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB2_2:
 ; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, v1
-; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
 ; GFX8-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX8-NEXT:    s_mov_b32 s2, -1
-; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s4, v0
 ; GFX8-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: add_i32_varying:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX9-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX9-NEXT:    s_not_b64 exec, exec
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0

 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_add_rtn_u32 v0, v0, v3
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB2_2:
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v1
-; GFX9-NEXT:    v_add_u32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1
-; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_add_u32_e32 v0, s4, v0
 ; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: add_i32_varying:
 ; GFX1064:       ; %bb.0: ; %entry
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1064-NEXT:    s_not_b64 exec, exec
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1064-NEXT:    s_not_b64 exec, exec
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1064-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1064-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1064-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX1064-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX1064-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1064-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX1064-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1064-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX1064-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX1064-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1064-NEXT:    v_readlane_b32 s5, v1, 31
 ; GFX1064-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX1064-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1064-NEXT:    v_readlane_b32 s7, v1, 63
 ; GFX1064-NEXT:    v_readlane_b32 s6, v1, 47
 ; GFX1064-NEXT:    v_writelane_b32 v3, s5, 32

 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_add_rtn_u32 v0, v0, v4
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB2_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1064-NEXT:    v_add_nc_u32_e32 v0, s3, v0
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: add_i32_varying:
 ; GFX1032:       ; %bb.0: ; %entry
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1032-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1032-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1032-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1032-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1032-NEXT:    v_readlane_b32 s3, v1, 15
 ; GFX1032-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1032-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_writelane_b32 v3, s3, 16
 ; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    ; implicit-def: $vgpr0
 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1032-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_add_rtn_u32 v0, v0, v4
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB2_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1032-NEXT:    v_add_nc_u32_e32 v0, s3, v0
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;

 ; GFX1164-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1164-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX1164-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1164-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
 ; GFX1164-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX1164-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX1164-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1164-NEXT:    v_readlane_b32 s5, v1, 31
 ; GFX1164-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX1164-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1164-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1164-NEXT:    v_readlane_b32 s7, v1, 63
 ; GFX1164-NEXT:    v_readlane_b32 s6, v1, 47

 ; GFX1164-NEXT:    s_mov_b32 s3, s7
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_add_rtn_u32 v0, v0, v4
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB2_2:
 ; GFX1164-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1164-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1164-NEXT:    v_add_nc_u32_e32 v0, s3, v0
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: add_i32_varying:
 ; GFX1132:       ; %bb.0: ; %entry
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1132-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1132-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1132-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1132-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1132-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1132-NEXT:    v_readlane_b32 s3, v1, 15
 ; GFX1132-NEXT:    v_readlane_b32 s4, v1, 31
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
 ; GFX1132-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
+; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1132-NEXT:    v_writelane_b32 v3, s3, 16
 ; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_2)
 ; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    ; implicit-def: $vgpr0
 ; GFX1132-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1132-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX1132-NEXT:  ; %bb.1:
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1132-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_add_rtn_u32 v0, v0, v4
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB2_2:
 ; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_add_nc_u32_e32 v0, s3, v0
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)

 }
 
 define amdgpu_kernel void @add_i64_constant(i64 addrspace(1)* %out) {
 ;
 ;
 ; GFX7LESS-LABEL: add_i64_constant:
 ; GFX7LESS:       ; %bb.0: ; %entry
 ; GFX7LESS-NEXT:    s_mov_b64 s[4:5], exec
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
 ; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s4, 0
 ; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v2, s5, v0
 ; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v2
 ; GFX7LESS-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX7LESS-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX7LESS-NEXT:    s_cbranch_execz .LBB4_2
 ; GFX7LESS-NEXT:  ; %bb.1:
 ; GFX7LESS-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX7LESS-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7LESS-NEXT:    s_mov_b32 m0, -1
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    ds_add_rtn_u64 v[0:1], v1, v[0:1]
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:  .LBB4_2:
 ; GFX7LESS-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX7LESS-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX7LESS-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX7LESS-NEXT:    v_readfirstlane_b32 s5, v1
 ; GFX7LESS-NEXT:    v_mul_hi_u32_u24_e32 v1, 5, v2
 ; GFX7LESS-NEXT:    v_mul_u32_u24_e32 v0, 5, v2
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s4
-; GFX7LESS-NEXT:    v_add_i32_e32 v0, vcc, s2, v0
+; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s5
+; GFX7LESS-NEXT:    v_add_i32_e32 v0, vcc, s4, v0
 ; GFX7LESS-NEXT:    v_addc_u32_e32 v1, vcc, v2, v1, vcc
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: add_i64_constant:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v2, s5, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v2
 ; GFX8-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB4_2
 ; GFX8-NEXT:  ; %bb.1:
 ; GFX8-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX8-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_add_rtn_u64 v[0:1], v1, v[0:1]
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB4_2:
 ; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX8-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX8-NEXT:    v_mad_u64_u32 v[0:1], s[2:3], v2, 5, v[0:1]
 ; GFX8-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX8-NEXT:    s_mov_b32 s2, -1
-; GFX8-NEXT:    s_nop 2
+; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: add_i64_constant:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v2, s5, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v2
 ; GFX9-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB4_2
 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX9-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX9-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_add_rtn_u64 v[0:1], v1, v[0:1]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB4_2:
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX9-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX9-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX9-NEXT:    v_mad_u64_u32 v[0:1], s[2:3], v2, 5, v[0:1]
 ; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1
-; GFX9-NEXT:    s_nop 2
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: add_i64_constant:
 ; GFX1064:       ; %bb.0: ; %entry
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v2, s5, v0
 ; GFX1064-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v2
 ; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX1064-NEXT:    s_cbranch_execz .LBB4_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1064-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_add_rtn_u64 v[0:1], v1, v[0:1]
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB4_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1064-NEXT:    v_mad_u64_u32 v[0:1], s[2:3], v2, 5, s[2:3]
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1064-NEXT:    s_mov_b32 s2, -1
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: add_i64_constant:
 ; GFX1032:       ; %bb.0: ; %entry
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    s_mov_b32 s3, exec_lo
 ; GFX1032-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v2, s3, 0
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v2
 ; GFX1032-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB4_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    s_bcnt1_i32_b32 s3, s3
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1032-NEXT:    s_mul_i32 s3, s3, 5
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_add_rtn_u64 v[0:1], v1, v[0:1]
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB4_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1032-NEXT:    v_mad_u64_u32 v[0:1], s2, v2, 5, s[2:3]
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;
 ; GFX1164-LABEL: add_i64_constant:
 ; GFX1164:       ; %bb.0: ; %entry
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
 ; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v2, s5, v0
 ; GFX1164-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v2
 ; GFX1164-NEXT:    s_cbranch_execz .LBB4_2
 ; GFX1164-NEXT:  ; %bb.1:
 ; GFX1164-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX1164-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1164-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1164-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_add_rtn_u64 v[0:1], v1, v[0:1]
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB4_2:
 ; GFX1164-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1164-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mad_u64_u32 v[0:1], null, v2, 5, s[2:3]
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_mov_b32 s2, -1
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: add_i64_constant:
 ; GFX1132:       ; %bb.0: ; %entry
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
 ; GFX1132-NEXT:    s_mov_b32 s2, exec_lo
 ; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v2, s3, 0
 ; GFX1132-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v2
 ; GFX1132-NEXT:    s_cbranch_execz .LBB4_2
 ; GFX1132-NEXT:  ; %bb.1:
 ; GFX1132-NEXT:    s_bcnt1_i32_b32 s3, s3
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1132-NEXT:    s_mul_i32 s3, s3, 5
 ; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_add_rtn_u64 v[0:1], v1, v[0:1]
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB4_2:
 ; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1132-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_mad_u64_u32 v[0:1], null, v2, 5, s[2:3]
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
entry:
   ret void
 }
 
 define amdgpu_kernel void @sub_i32_constant(i32 addrspace(1)* %out) {
 ;
 ;
 ; GFX7LESS-LABEL: sub_i32_constant:
 ; GFX7LESS:       ; %bb.0: ; %entry
-; GFX7LESS-NEXT:    s_mov_b64 s[2:3], exec
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
+; GFX7LESS-NEXT:    s_mov_b64 s[4:5], exec
+; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s4, 0
+; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s5, v0
 ; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX7LESS-NEXT:    ; implicit-def: $vgpr1
-; GFX7LESS-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX7LESS-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX7LESS-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX7LESS-NEXT:  ; %bb.1:
-; GFX7LESS-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX7LESS-NEXT:    s_mul_i32 s2, s2, 5
+; GFX7LESS-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX7LESS-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v1, 0
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s2
+; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX7LESS-NEXT:    s_mov_b32 m0, -1
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    ds_sub_rtn_u32 v1, v1, v2
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:  .LBB7_2:
-; GFX7LESS-NEXT:    s_or_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_readfirstlane_b32 s2, v1
-; GFX7LESS-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
+; GFX7LESS-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
 ; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    v_sub_i32_e32 v0, vcc, s2, v0
 ; GFX7LESS-NEXT:    s_mov_b32 s2, -1
+; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX7LESS-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX7LESS-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
+; GFX7LESS-NEXT:    v_sub_i32_e32 v0, vcc, s4, v0
 ; GFX7LESS-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: sub_i32_constant:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX8-NEXT:    s_mov_b64 s[4:5], exec
+; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr1
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX8-NEXT:    s_mul_i32 s2, s2, 5
+; GFX8-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX8-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s2
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_sub_rtn_u32 v1, v1, v2
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB7_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX8-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
-; GFX8-NEXT:    v_sub_u32_e32 v0, vcc, s2, v0
 ; GFX8-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX8-NEXT:    s_mov_b32 s2, -1
-; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_sub_u32_e32 v0, vcc, s4, v0
 ; GFX8-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sub_i32_constant:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX9-NEXT:    s_mov_b64 s[4:5], exec
+; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr1
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX9-NEXT:    s_mul_i32 s2, s2, 5
+; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX9-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_sub_rtn_u32 v1, v1, v2
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB7_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX9-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX9-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
-; GFX9-NEXT:    v_sub_u32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1
-; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_sub_u32_e32 v0, s4, v0
 ; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: sub_i32_constant:
 ; GFX1064:       ; %bb.0: ; %entry
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
+; GFX1064-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX1064-NEXT:    ; implicit-def: $vgpr1
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX1064-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX1064-NEXT:  ; %bb.1:
-; GFX1064-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
+; GFX1064-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
-; GFX1064-NEXT:    s_mul_i32 s2, s2, 5
-; GFX1064-NEXT:    v_mov_b32_e32 v2, s2
+; GFX1064-NEXT:    s_mul_i32 s4, s4, 5
+; GFX1064-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_sub_rtn_u32 v1, v1, v2
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB7_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX1064-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1064-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX1064-NEXT:    s_mov_b32 s2, -1
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: sub_i32_constant:
 ; GFX1032:       ; %bb.0: ; %entry
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    s_mov_b32 s3, exec_lo
 ; GFX1032-NEXT:    ; implicit-def: $vgpr1
 ; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1032-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    s_bcnt1_i32_b32 s3, s3
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1032-NEXT:    s_mul_i32 s3, s3, 5
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, s3
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_sub_rtn_u32 v1, v1, v2
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB7_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX1032-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1032-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;
 ; GFX1164-LABEL: sub_i32_constant:
 ; GFX1164:       ; %bb.0: ; %entry
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
 ; GFX1164-NEXT:    s_mov_b64 s[4:5], exec
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
+; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
+; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX1164-NEXT:    ; implicit-def: $vgpr1
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX1164-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX1164-NEXT:  ; %bb.1:
-; GFX1164-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
+; GFX1164-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX1164-NEXT:    v_mov_b32_e32 v1, 0
-; GFX1164-NEXT:    s_mul_i32 s2, s2, 5
+; GFX1164-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    v_mov_b32_e32 v2, s2
+; GFX1164-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_sub_rtn_u32 v1, v1, v2
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB7_2:
-; GFX1164-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1164-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX1164-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1164-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX1164-NEXT:    s_mov_b32 s2, -1
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: sub_i32_constant:
 ; GFX1132:       ; %bb.0: ; %entry
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
 ; GFX1132-NEXT:    s_mov_b32 s2, exec_lo
 ; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
 ; GFX1132-NEXT:    ; implicit-def: $vgpr1
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX1132-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX1132-NEXT:  ; %bb.1:
 ; GFX1132-NEXT:    s_bcnt1_i32_b32 s3, s3
 ; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
 ; GFX1132-NEXT:    s_mul_i32 s3, s3, 5
 ; GFX1132-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s3
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_sub_rtn_u32 v1, v1, v2
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB7_2:
 ; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX1132-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %old = atomicrmw sub i32 addrspace(3)* @local_var32, i32 5 acq_rel
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @sub_i32_uniform(i32 addrspace(1)* %out, i32 %subitive) {
 ;
 ;
 ; GFX7LESS-LABEL: sub_i32_uniform:
 ; GFX7LESS:       ; %bb.0: ; %entry
-; GFX7LESS-NEXT:    s_mov_b64 s[2:3], exec
-; GFX7LESS-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
+; GFX7LESS-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX7LESS-NEXT:    s_load_dword s6, s[0:1], 0xb
-; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
+; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s4, 0
+; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s5, v0
 ; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX7LESS-NEXT:    ; implicit-def: $vgpr1
-; GFX7LESS-NEXT:    s_and_saveexec_b64 s[0:1], vcc
+; GFX7LESS-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX7LESS-NEXT:    s_cbranch_execz .LBB8_2
 ; GFX7LESS-NEXT:  ; %bb.1:
-; GFX7LESS-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
+; GFX7LESS-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_mul_i32 s2, s6, s2
+; GFX7LESS-NEXT:    s_mul_i32 s4, s6, s4
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v1, 0
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s2
+; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX7LESS-NEXT:    s_mov_b32 m0, -1
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    ds_sub_rtn_u32 v1, v1, v2
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:  .LBB8_2:
-; GFX7LESS-NEXT:    s_or_b64 exec, exec, s[0:1]
+; GFX7LESS-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX7LESS-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX7LESS-NEXT:    v_mul_lo_u32 v0, s6, v0
-; GFX7LESS-NEXT:    s_mov_b32 s7, 0xf000
-; GFX7LESS-NEXT:    v_sub_i32_e32 v0, vcc, s0, v0
-; GFX7LESS-NEXT:    s_mov_b32 s6, -1
-; GFX7LESS-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX7LESS-NEXT:    v_sub_i32_e32 v0, vcc, s4, v0
+; GFX7LESS-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: sub_i32_uniform:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
 ; GFX8-NEXT:    s_load_dword s6, s[0:1], 0x2c
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX8-NEXT:    s_mov_b64 s[4:5], exec
+; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr1
-; GFX8-NEXT:    s_and_saveexec_b64 s[0:1], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB8_2
 ; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
+; GFX8-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_mul_i32 s2, s6, s2
+; GFX8-NEXT:    s_mul_i32 s4, s6, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
-; GFX8-NEXT:    v_mov_b32_e32 v2, s2
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_sub_rtn_u32 v1, v1, v2
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB8_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[0:1]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    v_mul_lo_u32 v0, s6, v0
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX8-NEXT:    s_mov_b32 s7, 0xf000
-; GFX8-NEXT:    s_mov_b32 s6, -1
-; GFX8-NEXT:    v_sub_u32_e32 v0, vcc, s0, v0
-; GFX8-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX8-NEXT:    s_mov_b32 s3, 0xf000
+; GFX8-NEXT:    s_mov_b32 s2, -1
+; GFX8-NEXT:    v_sub_u32_e32 v0, vcc, s4, v0
+; GFX8-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sub_i32_uniform:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
 ; GFX9-NEXT:    s_load_dword s6, s[0:1], 0x2c
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX9-NEXT:    s_mov_b64 s[4:5], exec
+; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr1
-; GFX9-NEXT:    s_and_saveexec_b64 s[0:1], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB8_2
 ; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
+; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_mul_i32 s2, s6, s2
+; GFX9-NEXT:    s_mul_i32 s4, s6, s4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_sub_rtn_u32 v1, v1, v2
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB8_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[0:1]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_mul_lo_u32 v0, s6, v0
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX9-NEXT:    s_mov_b32 s7, 0xf000
-; GFX9-NEXT:    s_mov_b32 s6, -1
-; GFX9-NEXT:    v_sub_u32_e32 v0, s0, v0
-; GFX9-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX9-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX9-NEXT:    s_mov_b32 s3, 0xf000
+; GFX9-NEXT:    s_mov_b32 s2, -1
+; GFX9-NEXT:    v_sub_u32_e32 v0, s4, v0
+; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: sub_i32_uniform:
 ; GFX1064:       ; %bb.0: ; %entry
-; GFX1064-NEXT:    s_clause 0x1
-; GFX1064-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
 ; GFX1064-NEXT:    s_load_dword s6, s[0:1], 0x2c
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
+; GFX1064-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX1064-NEXT:    ; implicit-def: $vgpr1
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[0:1], vcc
+; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX1064-NEXT:    s_cbranch_execz .LBB8_2
 ; GFX1064-NEXT:  ; %bb.1:
-; GFX1064-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
+; GFX1064-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_mul_i32 s2, s6, s2
-; GFX1064-NEXT:    v_mov_b32_e32 v2, s2
+; GFX1064-NEXT:    s_mul_i32 s4, s6, s4
+; GFX1064-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_sub_rtn_u32 v1, v1, v2
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB8_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX1064-NEXT:    s_or_b64 exec, exec, s[0:1]
+; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    v_mul_lo_u32 v0, s6, v0
-; GFX1064-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX1064-NEXT:    s_mov_b32 s7, 0x31016000
-; GFX1064-NEXT:    s_mov_b32 s6, -1
-; GFX1064-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
-; GFX1064-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX1064-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
+; GFX1064-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
+; GFX1064-NEXT:    s_mov_b32 s2, -1
+; GFX1064-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: sub_i32_uniform:
 ; GFX1032:       ; %bb.0: ; %entry
-; GFX1032-NEXT:    s_clause 0x1
-; GFX1032-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
 ; GFX1032-NEXT:    s_load_dword s2, s[0:1], 0x2c
-; GFX1032-NEXT:    s_mov_b32 s3, exec_lo
+; GFX1032-NEXT:    s_mov_b32 s4, exec_lo
 ; GFX1032-NEXT:    ; implicit-def: $vgpr1
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
+; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s0, vcc_lo
+; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB8_2
 ; GFX1032-NEXT:  ; %bb.1:
-; GFX1032-NEXT:    s_bcnt1_i32_b32 s1, s3
+; GFX1032-NEXT:    s_bcnt1_i32_b32 s4, s4
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_mul_i32 s1, s2, s1
-; GFX1032-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1032-NEXT:    s_mul_i32 s4, s2, s4
+; GFX1032-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_sub_rtn_u32 v1, v1, v2
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB8_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s0
+; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    v_mul_lo_u32 v0, s2, v0
-; GFX1032-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX1032-NEXT:    s_mov_b32 s7, 0x31016000
-; GFX1032-NEXT:    s_mov_b32 s6, -1
-; GFX1032-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
-; GFX1032-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX1032-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
+; GFX1032-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
+; GFX1032-NEXT:    s_mov_b32 s2, -1
+; GFX1032-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;
 ; GFX1164-LABEL: sub_i32_uniform:
 ; GFX1164:       ; %bb.0: ; %entry
-; GFX1164-NEXT:    s_clause 0x1
-; GFX1164-NEXT:    s_load_b64 s[4:5], s[0:1], 0x24
 ; GFX1164-NEXT:    s_load_b32 s6, s[0:1], 0x2c
+; GFX1164-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_mov_b64 s[0:1], exec
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
+; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX1164-NEXT:    ; implicit-def: $vgpr1
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX1164-NEXT:    s_cbranch_execz .LBB8_2
 ; GFX1164-NEXT:  ; %bb.1:
-; GFX1164-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
+; GFX1164-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX1164-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_mul_i32 s2, s6, s2
+; GFX1164-NEXT:    s_mul_i32 s4, s6, s4
 ; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    v_mov_b32_e32 v2, s2
+; GFX1164-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_sub_rtn_u32 v1, v1, v2
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB8_2:
-; GFX1164-NEXT:    s_or_b64 exec, exec, s[0:1]
+; GFX1164-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    v_mul_lo_u32 v0, s6, v0
-; GFX1164-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX1164-NEXT:    s_mov_b32 s7, 0x31016000
-; GFX1164-NEXT:    s_mov_b32 s6, -1
+; GFX1164-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
-; GFX1164-NEXT:    buffer_store_b32 v0, off, s[4:7], 0
+; GFX1164-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
+; GFX1164-NEXT:    s_mov_b32 s2, -1
+; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: sub_i32_uniform:
 ; GFX1132:       ; %bb.0: ; %entry
-; GFX1132-NEXT:    s_clause 0x1
-; GFX1132-NEXT:    s_load_b64 s[4:5], s[0:1], 0x24
-; GFX1132-NEXT:    s_load_b32 s0, s[0:1], 0x2c
-; GFX1132-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1132-NEXT:    s_mov_b32 s1, exec_lo
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
+; GFX1132-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX1132-NEXT:    s_mov_b32 s4, exec_lo
+; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
+; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX1132-NEXT:    ; implicit-def: $vgpr1
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX1132-NEXT:    s_cbranch_execz .LBB8_2
 ; GFX1132-NEXT:  ; %bb.1:
-; GFX1132-NEXT:    s_bcnt1_i32_b32 s2, s2
+; GFX1132-NEXT:    s_bcnt1_i32_b32 s4, s4
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_mul_i32 s2, s0, s2
+; GFX1132-NEXT:    s_mul_i32 s4, s2, s4
 ; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s2
+; GFX1132-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s4
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_sub_rtn_u32 v1, v1, v2
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB8_2:
-; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s1
+; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_mul_lo_u32 v0, s0, v0
-; GFX1132-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX1132-NEXT:    s_mov_b32 s7, 0x31016000
-; GFX1132-NEXT:    s_mov_b32 s6, -1
+; GFX1132-NEXT:    v_mul_lo_u32 v0, s2, v0
+; GFX1132-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
-; GFX1132-NEXT:    buffer_store_b32 v0, off, s[4:7], 0
+; GFX1132-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
+; GFX1132-NEXT:    s_mov_b32 s2, -1
+; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %old = atomicrmw sub i32 addrspace(3)* @local_var32, i32 %subitive acq_rel
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 

 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: sub_i32_varying:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX8-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX8-NEXT:    s_not_b64 exec, exec
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0

 ; GFX8-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_sub_rtn_u32 v0, v0, v3
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB9_2:
 ; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, v1
-; GFX8-NEXT:    v_sub_u32_e32 v0, vcc, s2, v0
 ; GFX8-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX8-NEXT:    s_mov_b32 s2, -1
-; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_sub_u32_e32 v0, vcc, s4, v0
 ; GFX8-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sub_i32_varying:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX9-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX9-NEXT:    s_not_b64 exec, exec
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0

 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_sub_rtn_u32 v0, v0, v3
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB9_2:
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v1
-; GFX9-NEXT:    v_sub_u32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1
-; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_sub_u32_e32 v0, s4, v0
 ; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: sub_i32_varying:
 ; GFX1064:       ; %bb.0: ; %entry
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1064-NEXT:    s_not_b64 exec, exec
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1064-NEXT:    s_not_b64 exec, exec
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1064-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1064-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1064-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX1064-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX1064-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1064-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX1064-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1064-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX1064-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX1064-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1064-NEXT:    v_readlane_b32 s5, v1, 31
 ; GFX1064-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX1064-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1064-NEXT:    v_readlane_b32 s7, v1, 63
 ; GFX1064-NEXT:    v_readlane_b32 s6, v1, 47
 ; GFX1064-NEXT:    v_writelane_b32 v3, s5, 32

 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_sub_rtn_u32 v0, v0, v4
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB9_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1064-NEXT:    v_sub_nc_u32_e32 v0, s3, v0
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: sub_i32_varying:
 ; GFX1032:       ; %bb.0: ; %entry
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1032-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1032-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1032-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1032-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1032-NEXT:    v_readlane_b32 s3, v1, 15
 ; GFX1032-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1032-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_writelane_b32 v3, s3, 16
 ; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    ; implicit-def: $vgpr0
 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB9_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1032-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_sub_rtn_u32 v0, v0, v4
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB9_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1032-NEXT:    v_sub_nc_u32_e32 v0, s3, v0
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;

 ; GFX1164-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1164-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX1164-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1164-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
 ; GFX1164-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX1164-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX1164-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1164-NEXT:    v_readlane_b32 s5, v1, 31
 ; GFX1164-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX1164-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1164-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1164-NEXT:    v_readlane_b32 s7, v1, 63
 ; GFX1164-NEXT:    v_readlane_b32 s6, v1, 47

 ; GFX1164-NEXT:    s_mov_b32 s3, s7
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_sub_rtn_u32 v0, v0, v4
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB9_2:
 ; GFX1164-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1164-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1164-NEXT:    v_sub_nc_u32_e32 v0, s3, v0
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: sub_i32_varying:
 ; GFX1132:       ; %bb.0: ; %entry
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1132-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1132-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1132-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1132-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1132-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1132-NEXT:    v_readlane_b32 s3, v1, 15
 ; GFX1132-NEXT:    v_readlane_b32 s4, v1, 31
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
 ; GFX1132-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
+; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1132-NEXT:    v_writelane_b32 v3, s3, 16
 ; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_2)
 ; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    ; implicit-def: $vgpr0
 ; GFX1132-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1132-NEXT:    s_cbranch_execz .LBB9_2
 ; GFX1132-NEXT:  ; %bb.1:
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1132-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_sub_rtn_u32 v0, v0, v4
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB9_2:
 ; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_sub_nc_u32_e32 v0, s3, v0
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)

 }
 
 define amdgpu_kernel void @sub_i64_constant(i64 addrspace(1)* %out) {
 ;
 ;
 ; GFX7LESS-LABEL: sub_i64_constant:
 ; GFX7LESS:       ; %bb.0: ; %entry
 ; GFX7LESS-NEXT:    s_mov_b64 s[4:5], exec
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
 ; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s4, 0
 ; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v2, s5, v0
 ; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v2
 ; GFX7LESS-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX7LESS-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX7LESS-NEXT:    s_cbranch_execz .LBB11_2
 ; GFX7LESS-NEXT:  ; %bb.1:
 ; GFX7LESS-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX7LESS-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7LESS-NEXT:    s_mov_b32 m0, -1
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    ds_sub_rtn_u64 v[0:1], v1, v[0:1]
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:  .LBB11_2:
 ; GFX7LESS-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX7LESS-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX7LESS-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX7LESS-NEXT:    v_readfirstlane_b32 s5, v1
 ; GFX7LESS-NEXT:    v_mul_hi_u32_u24_e32 v1, 5, v2
 ; GFX7LESS-NEXT:    v_mul_u32_u24_e32 v0, 5, v2
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s4
-; GFX7LESS-NEXT:    v_sub_i32_e32 v0, vcc, s2, v0
+; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s5
+; GFX7LESS-NEXT:    v_sub_i32_e32 v0, vcc, s4, v0
 ; GFX7LESS-NEXT:    v_subb_u32_e32 v1, vcc, v2, v1, vcc
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: sub_i64_constant:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v2, s5, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v2
 ; GFX8-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB11_2
 ; GFX8-NEXT:  ; %bb.1:
 ; GFX8-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX8-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_sub_rtn_u64 v[0:1], v1, v[0:1]
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB11_2:
 ; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX8-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s5, v1
 ; GFX8-NEXT:    v_mul_u32_u24_e32 v0, 5, v2
 ; GFX8-NEXT:    v_mul_hi_u32_u24_e32 v1, 5, v2
-; GFX8-NEXT:    v_mov_b32_e32 v2, s3
-; GFX8-NEXT:    v_sub_u32_e32 v0, vcc, s2, v0
-; GFX8-NEXT:    v_subb_u32_e32 v1, vcc, v2, v1, vcc
+; GFX8-NEXT:    v_mov_b32_e32 v2, s5
+; GFX8-NEXT:    v_sub_u32_e32 v0, vcc, s4, v0
 ; GFX8-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX8-NEXT:    s_mov_b32 s2, -1
+; GFX8-NEXT:    v_subb_u32_e32 v1, vcc, v2, v1, vcc
 ; GFX8-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sub_i64_constant:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v2, s5, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v2
 ; GFX9-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB11_2
 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX9-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX9-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_sub_rtn_u64 v[0:1], v1, v[0:1]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB11_2:
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX9-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX9-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s5, v1
 ; GFX9-NEXT:    v_mul_u32_u24_e32 v0, 5, v2
 ; GFX9-NEXT:    v_mul_hi_u32_u24_e32 v1, 5, v2
-; GFX9-NEXT:    v_mov_b32_e32 v2, s3
-; GFX9-NEXT:    v_sub_co_u32_e32 v0, vcc, s2, v0
-; GFX9-NEXT:    v_subb_co_u32_e32 v1, vcc, v2, v1, vcc
+; GFX9-NEXT:    v_mov_b32_e32 v2, s5
+; GFX9-NEXT:    v_sub_co_u32_e32 v0, vcc, s4, v0
 ; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1
+; GFX9-NEXT:    v_subb_co_u32_e32 v1, vcc, v2, v1, vcc
 ; GFX9-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: sub_i64_constant:
 ; GFX1064:       ; %bb.0: ; %entry
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v2, s5, v0
 ; GFX1064-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v2
 ; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX1064-NEXT:    s_cbranch_execz .LBB11_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1064-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_sub_rtn_u64 v[0:1], v1, v[0:1]
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB11_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1064-NEXT:    v_mul_u32_u24_e32 v0, 5, v2
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1064-NEXT:    v_mul_hi_u32_u24_e32 v1, 5, v2
 ; GFX1064-NEXT:    v_sub_co_u32 v0, vcc, s2, v0
 ; GFX1064-NEXT:    v_sub_co_ci_u32_e32 v1, vcc, s3, v1, vcc
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1064-NEXT:    s_mov_b32 s2, -1
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: sub_i64_constant:
 ; GFX1032:       ; %bb.0: ; %entry
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    s_mov_b32 s3, exec_lo
 ; GFX1032-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v2, s3, 0
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v2
 ; GFX1032-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB11_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    s_bcnt1_i32_b32 s3, s3
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1032-NEXT:    s_mul_i32 s3, s3, 5
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_sub_rtn_u64 v[0:1], v1, v[0:1]
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB11_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1032-NEXT:    v_mul_u32_u24_e32 v0, 5, v2
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1032-NEXT:    v_mul_hi_u32_u24_e32 v1, 5, v2
 ; GFX1032-NEXT:    v_sub_co_u32 v0, vcc_lo, s2, v0
 ; GFX1032-NEXT:    v_sub_co_ci_u32_e32 v1, vcc_lo, s3, v1, vcc_lo
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;
 ; GFX1164-LABEL: sub_i64_constant:
 ; GFX1164:       ; %bb.0: ; %entry
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
 ; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v2, s5, v0
 ; GFX1164-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v2
 ; GFX1164-NEXT:    s_cbranch_execz .LBB11_2
 ; GFX1164-NEXT:  ; %bb.1:
 ; GFX1164-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX1164-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1164-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1164-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_sub_rtn_u64 v[0:1], v1, v[0:1]
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB11_2:
 ; GFX1164-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1164-NEXT:    v_mul_u32_u24_e32 v0, 5, v2
 ; GFX1164-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1164-NEXT:    v_mul_hi_u32_u24_e32 v1, 5, v2
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
 ; GFX1164-NEXT:    v_sub_co_u32 v0, vcc, s2, v0
 ; GFX1164-NEXT:    v_sub_co_ci_u32_e32 v1, vcc, s3, v1, vcc
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_mov_b32 s2, -1
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: sub_i64_constant:
 ; GFX1132:       ; %bb.0: ; %entry
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
 ; GFX1132-NEXT:    s_mov_b32 s2, exec_lo
 ; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v2, s3, 0
 ; GFX1132-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v2
 ; GFX1132-NEXT:    s_cbranch_execz .LBB11_2
 ; GFX1132-NEXT:  ; %bb.1:
 ; GFX1132-NEXT:    s_bcnt1_i32_b32 s3, s3
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1132-NEXT:    s_mul_i32 s3, s3, 5
 ; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_sub_rtn_u64 v[0:1], v1, v[0:1]
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB11_2:
 ; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1132-NEXT:    v_mul_u32_u24_e32 v0, 5, v2
 ; GFX1132-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1132-NEXT:    v_mul_hi_u32_u24_e32 v1, 5, v2
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_sub_co_u32 v0, vcc_lo, s2, v0
 ; GFX1132-NEXT:    v_sub_co_ci_u32_e32 v1, vcc_lo, s3, v1, vcc_lo
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000

 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: and_i32_varying:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_mov_b32_e32 v1, -1
 ; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX8-NEXT:    s_not_b64 exec, exec
 ; GFX8-NEXT:    v_mov_b32_e32 v2, -1

 ; GFX8-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_and_rtn_b32 v0, v0, v3
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB14_2:
 ; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, v1
-; GFX8-NEXT:    v_and_b32_e32 v0, s2, v0
 ; GFX8-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX8-NEXT:    s_mov_b32 s2, -1
-; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_and_b32_e32 v0, s4, v0
 ; GFX8-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: and_i32_varying:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_mov_b32_e32 v1, -1
 ; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX9-NEXT:    s_not_b64 exec, exec
 ; GFX9-NEXT:    v_mov_b32_e32 v2, -1

 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_and_rtn_b32 v0, v0, v3
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB14_2:
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v1
-; GFX9-NEXT:    v_and_b32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1
-; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_and_b32_e32 v0, s4, v0
 ; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: and_i32_varying:
 ; GFX1064:       ; %bb.0: ; %entry
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1064-NEXT:    s_not_b64 exec, exec
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, -1
 ; GFX1064-NEXT:    s_not_b64 exec, exec
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1064-NEXT:    v_and_b32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1064-NEXT:    v_mov_b32_e32 v3, -1
 ; GFX1064-NEXT:    v_and_b32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf
 ; GFX1064-NEXT:    v_and_b32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf
 ; GFX1064-NEXT:    v_and_b32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1064-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX1064-NEXT:    v_and_b32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX1064-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1064-NEXT:    v_and_b32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX1064-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1064-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX1064-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX1064-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1064-NEXT:    v_readlane_b32 s5, v1, 31
 ; GFX1064-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX1064-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1064-NEXT:    v_readlane_b32 s7, v1, 63
 ; GFX1064-NEXT:    v_readlane_b32 s6, v1, 47
 ; GFX1064-NEXT:    v_writelane_b32 v3, s5, 32

 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_and_rtn_b32 v0, v0, v4
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB14_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1064-NEXT:    v_and_b32_e32 v0, s3, v0
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: and_i32_varying:
 ; GFX1032:       ; %bb.0: ; %entry
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, -1
 ; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_and_b32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX1032-NEXT:    v_mov_b32_e32 v3, -1
 ; GFX1032-NEXT:    v_and_b32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    v_and_b32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    v_and_b32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1032-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_and_b32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX1032-NEXT:    v_mov_b32_e32 v3, -1
 ; GFX1032-NEXT:    v_readlane_b32 s3, v1, 15
 ; GFX1032-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1032-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_writelane_b32 v3, s3, 16
 ; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    ; implicit-def: $vgpr0
 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB14_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1032-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_and_rtn_b32 v0, v0, v4
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB14_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1032-NEXT:    v_and_b32_e32 v0, s3, v0
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;

 ; GFX1164-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1164-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_and_b32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX1164-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1164-NEXT:    v_and_b32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
 ; GFX1164-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX1164-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX1164-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1164-NEXT:    v_readlane_b32 s5, v1, 31
 ; GFX1164-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX1164-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1164-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1164-NEXT:    v_readlane_b32 s7, v1, 63
 ; GFX1164-NEXT:    v_readlane_b32 s6, v1, 47

 ; GFX1164-NEXT:    s_mov_b32 s3, s7
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_and_rtn_b32 v0, v0, v4
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB14_2:
 ; GFX1164-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1164-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1164-NEXT:    v_and_b32_e32 v0, s3, v0
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: and_i32_varying:
 ; GFX1132:       ; %bb.0: ; %entry
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1132-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, -1
 ; GFX1132-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_and_b32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX1132-NEXT:    v_mov_b32_e32 v3, -1
 ; GFX1132-NEXT:    v_and_b32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_and_b32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    v_and_b32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1132-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_and_b32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX1132-NEXT:    v_mov_b32_e32 v3, -1
 ; GFX1132-NEXT:    v_readlane_b32 s3, v1, 15
 ; GFX1132-NEXT:    v_readlane_b32 s4, v1, 31
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
 ; GFX1132-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
+; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1132-NEXT:    v_writelane_b32 v3, s3, 16
 ; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_2)
 ; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    ; implicit-def: $vgpr0
 ; GFX1132-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1132-NEXT:    s_cbranch_execz .LBB14_2
 ; GFX1132-NEXT:  ; %bb.1:
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1132-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_and_rtn_b32 v0, v0, v4
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB14_2:
 ; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_and_b32_e32 v0, s3, v0
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)

 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: or_i32_varying:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX8-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX8-NEXT:    s_not_b64 exec, exec
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0

 ; GFX8-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_or_rtn_b32 v0, v0, v3
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB15_2:
 ; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, v1
-; GFX8-NEXT:    v_or_b32_e32 v0, s2, v0
 ; GFX8-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX8-NEXT:    s_mov_b32 s2, -1
-; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_or_b32_e32 v0, s4, v0
 ; GFX8-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: or_i32_varying:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX9-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX9-NEXT:    s_not_b64 exec, exec
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0

 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_or_rtn_b32 v0, v0, v3
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB15_2:
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v1
-; GFX9-NEXT:    v_or_b32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1
-; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_or_b32_e32 v0, s4, v0
 ; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: or_i32_varying:
 ; GFX1064:       ; %bb.0: ; %entry
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1064-NEXT:    s_not_b64 exec, exec
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1064-NEXT:    s_not_b64 exec, exec
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1064-NEXT:    v_or_b32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1064-NEXT:    v_or_b32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_or_b32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_or_b32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1064-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX1064-NEXT:    v_or_b32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX1064-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1064-NEXT:    v_or_b32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX1064-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1064-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX1064-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX1064-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1064-NEXT:    v_readlane_b32 s5, v1, 31
 ; GFX1064-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX1064-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1064-NEXT:    v_readlane_b32 s7, v1, 63
 ; GFX1064-NEXT:    v_readlane_b32 s6, v1, 47
 ; GFX1064-NEXT:    v_writelane_b32 v3, s5, 32

 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_or_rtn_b32 v0, v0, v4
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB15_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1064-NEXT:    v_or_b32_e32 v0, s3, v0
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: or_i32_varying:
 ; GFX1032:       ; %bb.0: ; %entry
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_or_b32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1032-NEXT:    v_or_b32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1032-NEXT:    v_or_b32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1032-NEXT:    v_or_b32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1032-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_or_b32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1032-NEXT:    v_readlane_b32 s3, v1, 15
 ; GFX1032-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1032-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_writelane_b32 v3, s3, 16
 ; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    ; implicit-def: $vgpr0
 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB15_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1032-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_or_rtn_b32 v0, v0, v4
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB15_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1032-NEXT:    v_or_b32_e32 v0, s3, v0
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;

 ; GFX1164-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1164-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_or_b32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX1164-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1164-NEXT:    v_or_b32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
 ; GFX1164-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX1164-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX1164-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1164-NEXT:    v_readlane_b32 s5, v1, 31
 ; GFX1164-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX1164-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1164-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1164-NEXT:    v_readlane_b32 s7, v1, 63
 ; GFX1164-NEXT:    v_readlane_b32 s6, v1, 47

 ; GFX1164-NEXT:    s_mov_b32 s3, s7
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_or_rtn_b32 v0, v0, v4
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB15_2:
 ; GFX1164-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1164-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1164-NEXT:    v_or_b32_e32 v0, s3, v0
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: or_i32_varying:
 ; GFX1132:       ; %bb.0: ; %entry
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1132-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1132-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_or_b32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1132-NEXT:    v_or_b32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_or_b32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1132-NEXT:    v_or_b32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1132-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_or_b32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1132-NEXT:    v_readlane_b32 s3, v1, 15
 ; GFX1132-NEXT:    v_readlane_b32 s4, v1, 31
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
 ; GFX1132-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
+; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1132-NEXT:    v_writelane_b32 v3, s3, 16
 ; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_2)
 ; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    ; implicit-def: $vgpr0
 ; GFX1132-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1132-NEXT:    s_cbranch_execz .LBB15_2
 ; GFX1132-NEXT:  ; %bb.1:
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1132-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_or_rtn_b32 v0, v0, v4
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB15_2:
 ; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_or_b32_e32 v0, s3, v0
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)

 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: xor_i32_varying:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX8-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX8-NEXT:    s_not_b64 exec, exec
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0

 ; GFX8-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_xor_rtn_b32 v0, v0, v3
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB16_2:
 ; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, v1
-; GFX8-NEXT:    v_xor_b32_e32 v0, s2, v0
 ; GFX8-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX8-NEXT:    s_mov_b32 s2, -1
-; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_xor_b32_e32 v0, s4, v0
 ; GFX8-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: xor_i32_varying:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX9-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX9-NEXT:    s_not_b64 exec, exec
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0

 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_xor_rtn_b32 v0, v0, v3
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB16_2:
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v1
-; GFX9-NEXT:    v_xor_b32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1
-; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_xor_b32_e32 v0, s4, v0
 ; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: xor_i32_varying:
 ; GFX1064:       ; %bb.0: ; %entry
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1064-NEXT:    s_not_b64 exec, exec
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1064-NEXT:    s_not_b64 exec, exec
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1064-NEXT:    v_xor_b32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1064-NEXT:    v_xor_b32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_xor_b32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_xor_b32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1064-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX1064-NEXT:    v_xor_b32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX1064-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1064-NEXT:    v_xor_b32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX1064-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1064-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX1064-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX1064-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1064-NEXT:    v_readlane_b32 s5, v1, 31
 ; GFX1064-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX1064-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1064-NEXT:    v_readlane_b32 s7, v1, 63
 ; GFX1064-NEXT:    v_readlane_b32 s6, v1, 47
 ; GFX1064-NEXT:    v_writelane_b32 v3, s5, 32

 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_xor_rtn_b32 v0, v0, v4
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB16_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1064-NEXT:    v_xor_b32_e32 v0, s3, v0
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: xor_i32_varying:
 ; GFX1032:       ; %bb.0: ; %entry
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_xor_b32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1032-NEXT:    v_xor_b32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1032-NEXT:    v_xor_b32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1032-NEXT:    v_xor_b32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1032-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_xor_b32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1032-NEXT:    v_readlane_b32 s3, v1, 15
 ; GFX1032-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1032-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_writelane_b32 v3, s3, 16
 ; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    ; implicit-def: $vgpr0
 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB16_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1032-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_xor_rtn_b32 v0, v0, v4
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB16_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1032-NEXT:    v_xor_b32_e32 v0, s3, v0
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;

 ; GFX1164-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1164-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_xor_b32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX1164-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1164-NEXT:    v_xor_b32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
 ; GFX1164-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX1164-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX1164-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1164-NEXT:    v_readlane_b32 s5, v1, 31
 ; GFX1164-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX1164-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1164-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1164-NEXT:    v_readlane_b32 s7, v1, 63
 ; GFX1164-NEXT:    v_readlane_b32 s6, v1, 47

 ; GFX1164-NEXT:    s_mov_b32 s3, s7
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_xor_rtn_b32 v0, v0, v4
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB16_2:
 ; GFX1164-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1164-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1164-NEXT:    v_xor_b32_e32 v0, s3, v0
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: xor_i32_varying:
 ; GFX1132:       ; %bb.0: ; %entry
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1132-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1132-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_xor_b32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1132-NEXT:    v_xor_b32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_xor_b32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1132-NEXT:    v_xor_b32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1132-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_xor_b32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1132-NEXT:    v_readlane_b32 s3, v1, 15
 ; GFX1132-NEXT:    v_readlane_b32 s4, v1, 31
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
 ; GFX1132-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
+; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1132-NEXT:    v_writelane_b32 v3, s3, 16
 ; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_2)
 ; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    ; implicit-def: $vgpr0
 ; GFX1132-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1132-NEXT:    s_cbranch_execz .LBB16_2
 ; GFX1132-NEXT:  ; %bb.1:
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1132-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_xor_rtn_b32 v0, v0, v4
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB16_2:
 ; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_xor_b32_e32 v0, s3, v0
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)

 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: max_i32_varying:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_bfrev_b32_e32 v1, 1
 ; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX8-NEXT:    s_not_b64 exec, exec
 ; GFX8-NEXT:    v_bfrev_b32_e32 v2, 1

 ; GFX8-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_max_rtn_i32 v0, v0, v3
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB17_2:
 ; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, v1
-; GFX8-NEXT:    v_max_i32_e32 v0, s2, v0
 ; GFX8-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX8-NEXT:    s_mov_b32 s2, -1
-; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_max_i32_e32 v0, s4, v0
 ; GFX8-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: max_i32_varying:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_bfrev_b32_e32 v1, 1
 ; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX9-NEXT:    s_not_b64 exec, exec
 ; GFX9-NEXT:    v_bfrev_b32_e32 v2, 1

 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_max_rtn_i32 v0, v0, v3
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB17_2:
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v1
-; GFX9-NEXT:    v_max_i32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1
-; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_max_i32_e32 v0, s4, v0
 ; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: max_i32_varying:
 ; GFX1064:       ; %bb.0: ; %entry
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1064-NEXT:    s_not_b64 exec, exec
 ; GFX1064-NEXT:    v_bfrev_b32_e32 v1, 1
 ; GFX1064-NEXT:    s_not_b64 exec, exec
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1064-NEXT:    v_max_i32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1064-NEXT:    v_bfrev_b32_e32 v3, 1
 ; GFX1064-NEXT:    v_max_i32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf
 ; GFX1064-NEXT:    v_max_i32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf
 ; GFX1064-NEXT:    v_max_i32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1064-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX1064-NEXT:    v_max_i32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX1064-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1064-NEXT:    v_max_i32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX1064-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1064-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX1064-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX1064-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1064-NEXT:    v_readlane_b32 s5, v1, 31
 ; GFX1064-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX1064-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1064-NEXT:    v_readlane_b32 s7, v1, 63
 ; GFX1064-NEXT:    v_readlane_b32 s6, v1, 47
 ; GFX1064-NEXT:    v_writelane_b32 v3, s5, 32

 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_max_rtn_i32 v0, v0, v4
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB17_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1064-NEXT:    v_max_i32_e32 v0, s3, v0
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: max_i32_varying:
 ; GFX1032:       ; %bb.0: ; %entry
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1032-NEXT:    v_bfrev_b32_e32 v1, 1
 ; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_max_i32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX1032-NEXT:    v_bfrev_b32_e32 v3, 1
 ; GFX1032-NEXT:    v_max_i32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    v_max_i32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    v_max_i32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1032-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_max_i32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX1032-NEXT:    v_bfrev_b32_e32 v3, 1
 ; GFX1032-NEXT:    v_readlane_b32 s3, v1, 15
 ; GFX1032-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1032-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_writelane_b32 v3, s3, 16
 ; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    ; implicit-def: $vgpr0
 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB17_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1032-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_max_rtn_i32 v0, v0, v4
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB17_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1032-NEXT:    v_max_i32_e32 v0, s3, v0
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;

 ; GFX1164-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1164-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_max_i32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX1164-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1164-NEXT:    v_max_i32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
 ; GFX1164-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX1164-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX1164-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1164-NEXT:    v_readlane_b32 s5, v1, 31
 ; GFX1164-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX1164-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1164-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1164-NEXT:    v_readlane_b32 s7, v1, 63
 ; GFX1164-NEXT:    v_readlane_b32 s6, v1, 47

 ; GFX1164-NEXT:    s_mov_b32 s3, s7
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_max_rtn_i32 v0, v0, v4
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB17_2:
 ; GFX1164-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1164-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1164-NEXT:    v_max_i32_e32 v0, s3, v0
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: max_i32_varying:
 ; GFX1132:       ; %bb.0: ; %entry
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1132-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1132-NEXT:    v_bfrev_b32_e32 v1, 1
 ; GFX1132-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_max_i32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX1132-NEXT:    v_bfrev_b32_e32 v3, 1
 ; GFX1132-NEXT:    v_max_i32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_max_i32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    v_max_i32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1132-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_max_i32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX1132-NEXT:    v_bfrev_b32_e32 v3, 1
 ; GFX1132-NEXT:    v_readlane_b32 s3, v1, 15
 ; GFX1132-NEXT:    v_readlane_b32 s4, v1, 31
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
 ; GFX1132-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
+; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1132-NEXT:    v_writelane_b32 v3, s3, 16
 ; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_2)
 ; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    ; implicit-def: $vgpr0
 ; GFX1132-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1132-NEXT:    s_cbranch_execz .LBB17_2
 ; GFX1132-NEXT:  ; %bb.1:
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1132-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_max_rtn_i32 v0, v0, v4
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB17_2:
 ; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_max_i32_e32 v0, s3, v0
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
   %old = atomicrmw max i32 addrspace(3)* @local_var32, i32 %lane acq_rel
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @max_i64_constant(i64 addrspace(1)* %out) {
 ;
 ;
 ; GFX7LESS-LABEL: max_i64_constant:
 ; GFX7LESS:       ; %bb.0: ; %entry
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
 ; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
 ; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
 ; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX7LESS-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX7LESS-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX7LESS-NEXT:    s_cbranch_execz .LBB18_2
 ; GFX7LESS-NEXT:  ; %bb.1:
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX7LESS-NEXT:    s_mov_b32 m0, -1
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    ds_max_rtn_i64 v[0:1], v2, v[0:1]
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:  .LBB18_2:
 ; GFX7LESS-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX7LESS-NEXT:    v_readfirstlane_b32 s5, v1
 ; GFX7LESS-NEXT:    v_bfrev_b32_e32 v1, 1
 ; GFX7LESS-NEXT:    v_cndmask_b32_e64 v0, 5, 0, vcc
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7LESS-NEXT:    v_cndmask_b32_e32 v1, 0, v1, vcc
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s5
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX7LESS-NEXT:    v_cmp_gt_i64_e32 vcc, s[4:5], v[0:1]
 ; GFX7LESS-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc
-; GFX7LESS-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
+; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s4
+; GFX7LESS-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
 ; GFX7LESS-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: max_i64_constant:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB18_2
 ; GFX8-NEXT:  ; %bb.1:
 ; GFX8-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_max_rtn_i64 v[0:1], v2, v[0:1]
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB18_2:
 ; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX8-NEXT:    v_bfrev_b32_e32 v0, 1
-; GFX8-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX8-NEXT:    v_readfirstlane_b32 s5, v1
 ; GFX8-NEXT:    v_cndmask_b32_e32 v1, 0, v0, vcc
 ; GFX8-NEXT:    v_cndmask_b32_e64 v0, 5, 0, vcc
-; GFX8-NEXT:    v_cmp_gt_i64_e32 vcc, s[2:3], v[0:1]
-; GFX8-NEXT:    v_mov_b32_e32 v2, s3
+; GFX8-NEXT:    v_cmp_gt_i64_e32 vcc, s[4:5], v[0:1]
+; GFX8-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX8-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc
-; GFX8-NEXT:    v_mov_b32_e32 v2, s2
-; GFX8-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX8-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX8-NEXT:    s_mov_b32 s2, -1
+; GFX8-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
+; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: max_i64_constant:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB18_2
 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_max_rtn_i64 v[0:1], v2, v[0:1]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB18_2:
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX9-NEXT:    v_bfrev_b32_e32 v0, 1
-; GFX9-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    v_readfirstlane_b32 s5, v1
 ; GFX9-NEXT:    v_cndmask_b32_e32 v1, 0, v0, vcc
 ; GFX9-NEXT:    v_cndmask_b32_e64 v0, 5, 0, vcc
-; GFX9-NEXT:    v_cmp_gt_i64_e32 vcc, s[2:3], v[0:1]
-; GFX9-NEXT:    v_mov_b32_e32 v2, s3
+; GFX9-NEXT:    v_cmp_gt_i64_e32 vcc, s[4:5], v[0:1]
+; GFX9-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX9-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc
-; GFX9-NEXT:    v_mov_b32_e32 v2, s2
-; GFX9-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
+; GFX9-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1
+; GFX9-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: max_i64_constant:
 ; GFX1064:       ; %bb.0: ; %entry
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
 ; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX1064-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX1064-NEXT:    s_cbranch_execz .LBB18_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_max_rtn_i64 v[0:1], v2, v[0:1]
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB18_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1064-NEXT:    v_cndmask_b32_e64 v1, 0, 0x80000000, vcc
 ; GFX1064-NEXT:    v_cndmask_b32_e64 v0, 5, 0, vcc
 ; GFX1064-NEXT:    v_cmp_gt_i64_e32 vcc, s[2:3], v[0:1]
 ; GFX1064-NEXT:    v_cndmask_b32_e64 v1, v1, s3, vcc
 ; GFX1064-NEXT:    v_cndmask_b32_e64 v0, v0, s2, vcc
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1064-NEXT:    s_mov_b32 s2, -1
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: max_i64_constant:
 ; GFX1032:       ; %bb.0: ; %entry
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1032-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1032-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB18_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_max_rtn_i64 v[0:1], v2, v[0:1]
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB18_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1032-NEXT:    v_cndmask_b32_e64 v1, 0, 0x80000000, vcc_lo
 ; GFX1032-NEXT:    v_cndmask_b32_e64 v0, 5, 0, vcc_lo
 ; GFX1032-NEXT:    v_cmp_gt_i64_e32 vcc_lo, s[2:3], v[0:1]
 ; GFX1032-NEXT:    v_cndmask_b32_e64 v1, v1, s3, vcc_lo
 ; GFX1032-NEXT:    v_cndmask_b32_e64 v0, v0, s2, vcc_lo
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;
 ; GFX1164-LABEL: max_i64_constant:
 ; GFX1164:       ; %bb.0: ; %entry
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
 ; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX1164-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1164-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX1164-NEXT:    s_cbranch_execz .LBB18_2
 ; GFX1164-NEXT:  ; %bb.1:
 ; GFX1164-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1164-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1164-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_max_rtn_i64 v[0:1], v2, v[0:1]
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB18_2:
 ; GFX1164-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1164-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1164-NEXT:    v_cndmask_b32_e64 v1, 0, 0x80000000, vcc
 ; GFX1164-NEXT:    v_cndmask_b32_e64 v0, 5, 0, vcc
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1164-NEXT:    v_cmp_gt_i64_e32 vcc, s[2:3], v[0:1]
 ; GFX1164-NEXT:    v_cndmask_b32_e64 v1, v1, s3, vcc
 ; GFX1164-NEXT:    v_cndmask_b32_e64 v0, v0, s2, vcc
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_mov_b32 s2, -1
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: max_i64_constant:
 ; GFX1132:       ; %bb.0: ; %entry
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1132-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1132-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX1132-NEXT:    s_cbranch_execz .LBB18_2
 ; GFX1132-NEXT:  ; %bb.1:
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1132-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_max_rtn_i64 v[0:1], v2, v[0:1]
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB18_2:
 ; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1132-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1132-NEXT:    v_cndmask_b32_e64 v1, 0, 0x80000000, vcc_lo
 ; GFX1132-NEXT:    v_cndmask_b32_e64 v0, 5, 0, vcc_lo
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_cmp_gt_i64_e32 vcc_lo, s[2:3], v[0:1]
 ; GFX1132-NEXT:    v_cndmask_b32_e64 v1, v1, s3, vcc_lo
 ; GFX1132-NEXT:    v_cndmask_b32_e64 v0, v0, s2, vcc_lo

 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: min_i32_varying:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_bfrev_b32_e32 v1, -2
 ; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX8-NEXT:    s_not_b64 exec, exec
 ; GFX8-NEXT:    v_bfrev_b32_e32 v2, -2

 ; GFX8-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_min_rtn_i32 v0, v0, v3
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB19_2:
 ; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, v1
-; GFX8-NEXT:    v_min_i32_e32 v0, s2, v0
 ; GFX8-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX8-NEXT:    s_mov_b32 s2, -1
-; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_min_i32_e32 v0, s4, v0
 ; GFX8-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: min_i32_varying:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_bfrev_b32_e32 v1, -2
 ; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX9-NEXT:    s_not_b64 exec, exec
 ; GFX9-NEXT:    v_bfrev_b32_e32 v2, -2

 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_min_rtn_i32 v0, v0, v3
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB19_2:
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v1
-; GFX9-NEXT:    v_min_i32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1
-; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_min_i32_e32 v0, s4, v0
 ; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: min_i32_varying:
 ; GFX1064:       ; %bb.0: ; %entry
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1064-NEXT:    s_not_b64 exec, exec
 ; GFX1064-NEXT:    v_bfrev_b32_e32 v1, -2
 ; GFX1064-NEXT:    s_not_b64 exec, exec
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1064-NEXT:    v_min_i32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1064-NEXT:    v_bfrev_b32_e32 v3, -2
 ; GFX1064-NEXT:    v_min_i32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf
 ; GFX1064-NEXT:    v_min_i32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf
 ; GFX1064-NEXT:    v_min_i32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1064-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX1064-NEXT:    v_min_i32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX1064-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1064-NEXT:    v_min_i32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX1064-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1064-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX1064-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX1064-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1064-NEXT:    v_readlane_b32 s5, v1, 31
 ; GFX1064-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX1064-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1064-NEXT:    v_readlane_b32 s7, v1, 63
 ; GFX1064-NEXT:    v_readlane_b32 s6, v1, 47
 ; GFX1064-NEXT:    v_writelane_b32 v3, s5, 32

 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_min_rtn_i32 v0, v0, v4
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB19_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1064-NEXT:    v_min_i32_e32 v0, s3, v0
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: min_i32_varying:
 ; GFX1032:       ; %bb.0: ; %entry
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1032-NEXT:    v_bfrev_b32_e32 v1, -2
 ; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_min_i32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX1032-NEXT:    v_bfrev_b32_e32 v3, -2
 ; GFX1032-NEXT:    v_min_i32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    v_min_i32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    v_min_i32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1032-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_min_i32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX1032-NEXT:    v_bfrev_b32_e32 v3, -2
 ; GFX1032-NEXT:    v_readlane_b32 s3, v1, 15
 ; GFX1032-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1032-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_writelane_b32 v3, s3, 16
 ; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    ; implicit-def: $vgpr0
 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB19_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1032-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_min_rtn_i32 v0, v0, v4
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB19_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1032-NEXT:    v_min_i32_e32 v0, s3, v0
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;

 ; GFX1164-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1164-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_min_i32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX1164-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1164-NEXT:    v_min_i32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
 ; GFX1164-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX1164-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX1164-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1164-NEXT:    v_readlane_b32 s5, v1, 31
 ; GFX1164-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX1164-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1164-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1164-NEXT:    v_readlane_b32 s7, v1, 63
 ; GFX1164-NEXT:    v_readlane_b32 s6, v1, 47

 ; GFX1164-NEXT:    s_mov_b32 s3, s7
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_min_rtn_i32 v0, v0, v4
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB19_2:
 ; GFX1164-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1164-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1164-NEXT:    v_min_i32_e32 v0, s3, v0
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: min_i32_varying:
 ; GFX1132:       ; %bb.0: ; %entry
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1132-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1132-NEXT:    v_bfrev_b32_e32 v1, -2
 ; GFX1132-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_min_i32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX1132-NEXT:    v_bfrev_b32_e32 v3, -2
 ; GFX1132-NEXT:    v_min_i32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_min_i32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    v_min_i32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1132-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_min_i32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX1132-NEXT:    v_bfrev_b32_e32 v3, -2
 ; GFX1132-NEXT:    v_readlane_b32 s3, v1, 15
 ; GFX1132-NEXT:    v_readlane_b32 s4, v1, 31
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
 ; GFX1132-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
+; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1132-NEXT:    v_writelane_b32 v3, s3, 16
 ; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_2)
 ; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    ; implicit-def: $vgpr0
 ; GFX1132-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1132-NEXT:    s_cbranch_execz .LBB19_2
 ; GFX1132-NEXT:  ; %bb.1:
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1132-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_min_rtn_i32 v0, v0, v4
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB19_2:
 ; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_min_i32_e32 v0, s3, v0
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
   %old = atomicrmw min i32 addrspace(3)* @local_var32, i32 %lane acq_rel
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @min_i64_constant(i64 addrspace(1)* %out) {
 ;
 ;
 ; GFX7LESS-LABEL: min_i64_constant:
 ; GFX7LESS:       ; %bb.0: ; %entry
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
 ; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
 ; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
 ; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX7LESS-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX7LESS-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX7LESS-NEXT:    s_cbranch_execz .LBB20_2
 ; GFX7LESS-NEXT:  ; %bb.1:
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX7LESS-NEXT:    s_mov_b32 m0, -1
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    ds_min_rtn_i64 v[0:1], v2, v[0:1]
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:  .LBB20_2:
 ; GFX7LESS-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX7LESS-NEXT:    v_readfirstlane_b32 s5, v1
 ; GFX7LESS-NEXT:    v_bfrev_b32_e32 v1, -2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    v_cndmask_b32_e64 v0, 5, -1, vcc
 ; GFX7LESS-NEXT:    v_cndmask_b32_e32 v1, 0, v1, vcc
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s5
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX7LESS-NEXT:    v_cmp_lt_i64_e32 vcc, s[4:5], v[0:1]
 ; GFX7LESS-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc
-; GFX7LESS-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s4
+; GFX7LESS-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
 ; GFX7LESS-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: min_i64_constant:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB20_2
 ; GFX8-NEXT:  ; %bb.1:
 ; GFX8-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_min_rtn_i64 v[0:1], v2, v[0:1]
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB20_2:
 ; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX8-NEXT:    v_bfrev_b32_e32 v0, -2
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    v_readfirstlane_b32 s5, v1
 ; GFX8-NEXT:    v_cndmask_b32_e32 v1, 0, v0, vcc
 ; GFX8-NEXT:    v_cndmask_b32_e64 v0, 5, -1, vcc
 ; GFX8-NEXT:    v_cmp_lt_i64_e32 vcc, s[4:5], v[0:1]
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX8-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s4
+; GFX8-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX8-NEXT:    s_mov_b32 s2, -1
 ; GFX8-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
-; GFX8-NEXT:    s_mov_b32 s3, 0xf000
+; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: min_i64_constant:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB20_2
 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_min_rtn_i64 v[0:1], v2, v[0:1]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB20_2:
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX9-NEXT:    v_bfrev_b32_e32 v0, -2
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    v_readfirstlane_b32 s5, v1
 ; GFX9-NEXT:    v_cndmask_b32_e32 v1, 0, v0, vcc
 ; GFX9-NEXT:    v_cndmask_b32_e64 v0, 5, -1, vcc
 ; GFX9-NEXT:    v_cmp_lt_i64_e32 vcc, s[4:5], v[0:1]
 ; GFX9-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX9-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc
 ; GFX9-NEXT:    v_mov_b32_e32 v2, s4
+; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1
 ; GFX9-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
-; GFX9-NEXT:    s_mov_b32 s3, 0xf000
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: min_i64_constant:
 ; GFX1064:       ; %bb.0: ; %entry
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
 ; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX1064-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX1064-NEXT:    s_cbranch_execz .LBB20_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_min_rtn_i64 v[0:1], v2, v[0:1]
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB20_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1064-NEXT:    v_cndmask_b32_e64 v1, 0, 0x7fffffff, vcc
 ; GFX1064-NEXT:    v_cndmask_b32_e64 v0, 5, -1, vcc
 ; GFX1064-NEXT:    v_cmp_lt_i64_e32 vcc, s[2:3], v[0:1]
 ; GFX1064-NEXT:    v_cndmask_b32_e64 v1, v1, s3, vcc
 ; GFX1064-NEXT:    v_cndmask_b32_e64 v0, v0, s2, vcc
-; GFX1064-NEXT:    s_mov_b32 s2, -1
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
+; GFX1064-NEXT:    s_mov_b32 s2, -1
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: min_i64_constant:
 ; GFX1032:       ; %bb.0: ; %entry
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1032-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1032-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB20_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_min_rtn_i64 v[0:1], v2, v[0:1]
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB20_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1032-NEXT:    v_cndmask_b32_e64 v1, 0, 0x7fffffff, vcc_lo
 ; GFX1032-NEXT:    v_cndmask_b32_e64 v0, 5, -1, vcc_lo
 ; GFX1032-NEXT:    v_cmp_lt_i64_e32 vcc_lo, s[2:3], v[0:1]
 ; GFX1032-NEXT:    v_cndmask_b32_e64 v1, v1, s3, vcc_lo
 ; GFX1032-NEXT:    v_cndmask_b32_e64 v0, v0, s2, vcc_lo
-; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
+; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;
 ; GFX1164-LABEL: min_i64_constant:
 ; GFX1164:       ; %bb.0: ; %entry
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
 ; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX1164-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1164-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX1164-NEXT:    s_cbranch_execz .LBB20_2
 ; GFX1164-NEXT:  ; %bb.1:
 ; GFX1164-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1164-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1164-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_min_rtn_i64 v[0:1], v2, v[0:1]
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB20_2:
 ; GFX1164-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1164-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1164-NEXT:    v_cndmask_b32_e64 v1, 0, 0x7fffffff, vcc
 ; GFX1164-NEXT:    v_cndmask_b32_e64 v0, 5, -1, vcc
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1164-NEXT:    v_cmp_lt_i64_e32 vcc, s[2:3], v[0:1]
 ; GFX1164-NEXT:    v_cndmask_b32_e64 v1, v1, s3, vcc
 ; GFX1164-NEXT:    v_cndmask_b32_e64 v0, v0, s2, vcc
-; GFX1164-NEXT:    s_mov_b32 s2, -1
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
+; GFX1164-NEXT:    s_mov_b32 s2, -1
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: min_i64_constant:
 ; GFX1132:       ; %bb.0: ; %entry
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1132-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1132-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX1132-NEXT:    s_cbranch_execz .LBB20_2
 ; GFX1132-NEXT:  ; %bb.1:
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1132-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_min_rtn_i64 v[0:1], v2, v[0:1]
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB20_2:
 ; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1132-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1132-NEXT:    v_cndmask_b32_e64 v1, 0, 0x7fffffff, vcc_lo
 ; GFX1132-NEXT:    v_cndmask_b32_e64 v0, 5, -1, vcc_lo
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_cmp_lt_i64_e32 vcc_lo, s[2:3], v[0:1]
 ; GFX1132-NEXT:    v_cndmask_b32_e64 v1, v1, s3, vcc_lo
 ; GFX1132-NEXT:    v_cndmask_b32_e64 v0, v0, s2, vcc_lo
-; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
+; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
 ; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %old = atomicrmw min i64 addrspace(3)* @local_var64, i64 5 acq_rel
   store i64 %old, i64 addrspace(1)* %out
   ret void

 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: umax_i32_varying:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX8-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX8-NEXT:    s_not_b64 exec, exec
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0

 ; GFX8-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_max_rtn_u32 v0, v0, v3
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB21_2:
 ; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, v1
-; GFX8-NEXT:    v_max_u32_e32 v0, s2, v0
 ; GFX8-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX8-NEXT:    s_mov_b32 s2, -1
-; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_max_u32_e32 v0, s4, v0
 ; GFX8-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: umax_i32_varying:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX9-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX9-NEXT:    s_not_b64 exec, exec
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0

 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_max_rtn_u32 v0, v0, v3
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB21_2:
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v1
-; GFX9-NEXT:    v_max_u32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1
-; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_max_u32_e32 v0, s4, v0
 ; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: umax_i32_varying:
 ; GFX1064:       ; %bb.0: ; %entry
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1064-NEXT:    s_not_b64 exec, exec
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1064-NEXT:    s_not_b64 exec, exec
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1064-NEXT:    v_max_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1064-NEXT:    v_max_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_max_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_max_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1064-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX1064-NEXT:    v_max_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX1064-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1064-NEXT:    v_max_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX1064-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1064-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX1064-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX1064-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1064-NEXT:    v_readlane_b32 s5, v1, 31
 ; GFX1064-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX1064-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1064-NEXT:    v_readlane_b32 s7, v1, 63
 ; GFX1064-NEXT:    v_readlane_b32 s6, v1, 47
 ; GFX1064-NEXT:    v_writelane_b32 v3, s5, 32

 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_max_rtn_u32 v0, v0, v4
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB21_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1064-NEXT:    v_max_u32_e32 v0, s3, v0
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: umax_i32_varying:
 ; GFX1032:       ; %bb.0: ; %entry
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_max_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1032-NEXT:    v_max_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1032-NEXT:    v_max_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1032-NEXT:    v_max_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1032-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_max_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1032-NEXT:    v_readlane_b32 s3, v1, 15
 ; GFX1032-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1032-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_writelane_b32 v3, s3, 16
 ; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    ; implicit-def: $vgpr0
 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB21_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1032-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_max_rtn_u32 v0, v0, v4
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB21_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1032-NEXT:    v_max_u32_e32 v0, s3, v0
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;

 ; GFX1164-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1164-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_max_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX1164-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1164-NEXT:    v_max_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
 ; GFX1164-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX1164-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX1164-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1164-NEXT:    v_readlane_b32 s5, v1, 31
 ; GFX1164-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX1164-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1164-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1164-NEXT:    v_readlane_b32 s7, v1, 63
 ; GFX1164-NEXT:    v_readlane_b32 s6, v1, 47

 ; GFX1164-NEXT:    s_mov_b32 s3, s7
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_max_rtn_u32 v0, v0, v4
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB21_2:
 ; GFX1164-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1164-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1164-NEXT:    v_max_u32_e32 v0, s3, v0
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: umax_i32_varying:
 ; GFX1132:       ; %bb.0: ; %entry
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1132-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1132-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_max_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1132-NEXT:    v_max_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_max_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1132-NEXT:    v_max_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1132-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_max_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX1132-NEXT:    v_readlane_b32 s3, v1, 15
 ; GFX1132-NEXT:    v_readlane_b32 s4, v1, 31
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
 ; GFX1132-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
+; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1132-NEXT:    v_writelane_b32 v3, s3, 16
 ; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_2)
 ; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    ; implicit-def: $vgpr0
 ; GFX1132-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1132-NEXT:    s_cbranch_execz .LBB21_2
 ; GFX1132-NEXT:  ; %bb.1:
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1132-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_max_rtn_u32 v0, v0, v4
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB21_2:
 ; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_max_u32_e32 v0, s3, v0
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
   %old = atomicrmw umax i32 addrspace(3)* @local_var32, i32 %lane acq_rel
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @umax_i64_constant(i64 addrspace(1)* %out) {
 ;
 ;
 ; GFX7LESS-LABEL: umax_i64_constant:
 ; GFX7LESS:       ; %bb.0: ; %entry
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
 ; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
 ; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
 ; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX7LESS-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX7LESS-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX7LESS-NEXT:    s_cbranch_execz .LBB22_2
 ; GFX7LESS-NEXT:  ; %bb.1:
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX7LESS-NEXT:    s_mov_b32 m0, -1
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    ds_max_rtn_u64 v[0:1], v2, v[0:1]
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:  .LBB22_2:
 ; GFX7LESS-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX7LESS-NEXT:    v_readfirstlane_b32 s5, v1
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX7LESS-NEXT:    v_cndmask_b32_e64 v0, 5, 0, vcc
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX7LESS-NEXT:    v_cmp_gt_u64_e32 vcc, s[4:5], v[0:1]
 ; GFX7LESS-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7LESS-NEXT:    v_cndmask_b32_e32 v1, 0, v1, vcc
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: umax_i64_constant:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB22_2
 ; GFX8-NEXT:  ; %bb.1:
 ; GFX8-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_max_rtn_u64 v[0:1], v2, v[0:1]
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB22_2:
 ; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX8-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s5, v1
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX8-NEXT:    v_cndmask_b32_e64 v0, 5, 0, vcc
-; GFX8-NEXT:    v_cmp_gt_u64_e32 vcc, s[2:3], v[0:1]
-; GFX8-NEXT:    v_mov_b32_e32 v2, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, s3
-; GFX8-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
-; GFX8-NEXT:    v_cndmask_b32_e32 v1, 0, v1, vcc
+; GFX8-NEXT:    v_cmp_gt_u64_e32 vcc, s[4:5], v[0:1]
+; GFX8-NEXT:    v_mov_b32_e32 v2, s4
+; GFX8-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX8-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX8-NEXT:    s_mov_b32 s2, -1
+; GFX8-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
+; GFX8-NEXT:    v_cndmask_b32_e32 v1, 0, v1, vcc
 ; GFX8-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: umax_i64_constant:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB22_2
 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_max_rtn_u64 v[0:1], v2, v[0:1]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB22_2:
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX9-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX9-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s5, v1
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    v_cndmask_b32_e64 v0, 5, 0, vcc
-; GFX9-NEXT:    v_cmp_gt_u64_e32 vcc, s[2:3], v[0:1]
-; GFX9-NEXT:    v_mov_b32_e32 v2, s2
-; GFX9-NEXT:    v_mov_b32_e32 v1, s3
-; GFX9-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
-; GFX9-NEXT:    v_cndmask_b32_e32 v1, 0, v1, vcc
+; GFX9-NEXT:    v_cmp_gt_u64_e32 vcc, s[4:5], v[0:1]
+; GFX9-NEXT:    v_mov_b32_e32 v2, s4
+; GFX9-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1
+; GFX9-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
+; GFX9-NEXT:    v_cndmask_b32_e32 v1, 0, v1, vcc
 ; GFX9-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: umax_i64_constant:
 ; GFX1064:       ; %bb.0: ; %entry
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
 ; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX1064-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX1064-NEXT:    s_cbranch_execz .LBB22_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_max_rtn_u64 v[0:1], v2, v[0:1]
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB22_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1064-NEXT:    v_cndmask_b32_e64 v0, 5, 0, vcc
 ; GFX1064-NEXT:    v_cmp_gt_u64_e32 vcc, s[2:3], v[0:1]
 ; GFX1064-NEXT:    v_cndmask_b32_e64 v0, v0, s2, vcc
 ; GFX1064-NEXT:    v_cndmask_b32_e64 v1, 0, s3, vcc
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1064-NEXT:    s_mov_b32 s2, -1
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: umax_i64_constant:
 ; GFX1032:       ; %bb.0: ; %entry
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1032-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1032-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB22_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_max_rtn_u64 v[0:1], v2, v[0:1]
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB22_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1032-NEXT:    v_cndmask_b32_e64 v0, 5, 0, vcc_lo
 ; GFX1032-NEXT:    v_cmp_gt_u64_e32 vcc_lo, s[2:3], v[0:1]
 ; GFX1032-NEXT:    v_cndmask_b32_e64 v0, v0, s2, vcc_lo
 ; GFX1032-NEXT:    v_cndmask_b32_e64 v1, 0, s3, vcc_lo
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;
 ; GFX1164-LABEL: umax_i64_constant:
 ; GFX1164:       ; %bb.0: ; %entry
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
 ; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX1164-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1164-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX1164-NEXT:    s_cbranch_execz .LBB22_2
 ; GFX1164-NEXT:  ; %bb.1:
 ; GFX1164-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1164-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1164-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_max_rtn_u64 v[0:1], v2, v[0:1]
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB22_2:
 ; GFX1164-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1164-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1164-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1164-NEXT:    v_cndmask_b32_e64 v0, 5, 0, vcc
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1164-NEXT:    v_cmp_gt_u64_e32 vcc, s[2:3], v[0:1]
 ; GFX1164-NEXT:    v_cndmask_b32_e64 v0, v0, s2, vcc
 ; GFX1164-NEXT:    v_cndmask_b32_e64 v1, 0, s3, vcc
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_mov_b32 s2, -1
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: umax_i64_constant:
 ; GFX1132:       ; %bb.0: ; %entry
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1132-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1132-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX1132-NEXT:    s_cbranch_execz .LBB22_2
 ; GFX1132-NEXT:  ; %bb.1:
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1132-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_max_rtn_u64 v[0:1], v2, v[0:1]
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB22_2:
 ; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1132-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1132-NEXT:    v_cndmask_b32_e64 v0, 5, 0, vcc_lo
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_cmp_gt_u64_e32 vcc_lo, s[2:3], v[0:1]
 ; GFX1132-NEXT:    v_cndmask_b32_e64 v0, v0, s2, vcc_lo
 ; GFX1132-NEXT:    v_cndmask_b32_e64 v1, 0, s3, vcc_lo

 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: umin_i32_varying:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_mov_b32_e32 v1, -1
 ; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX8-NEXT:    s_not_b64 exec, exec
 ; GFX8-NEXT:    v_mov_b32_e32 v2, -1

 ; GFX8-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_min_rtn_u32 v0, v0, v3
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB23_2:
 ; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, v1
-; GFX8-NEXT:    v_min_u32_e32 v0, s2, v0
 ; GFX8-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX8-NEXT:    s_mov_b32 s2, -1
-; GFX8-NEXT:    s_nop 0
+; GFX8-NEXT:    v_min_u32_e32 v0, s4, v0
 ; GFX8-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: umin_i32_varying:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_mov_b32_e32 v1, -1
 ; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX9-NEXT:    s_not_b64 exec, exec
 ; GFX9-NEXT:    v_mov_b32_e32 v2, -1

 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_min_rtn_u32 v0, v0, v3
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB23_2:
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v1
-; GFX9-NEXT:    v_min_u32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1
-; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_min_u32_e32 v0, s4, v0
 ; GFX9-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: umin_i32_varying:
 ; GFX1064:       ; %bb.0: ; %entry
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1064-NEXT:    s_not_b64 exec, exec
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, -1
 ; GFX1064-NEXT:    s_not_b64 exec, exec
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1064-NEXT:    v_min_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1064-NEXT:    v_mov_b32_e32 v3, -1
 ; GFX1064-NEXT:    v_min_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf
 ; GFX1064-NEXT:    v_min_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf
 ; GFX1064-NEXT:    v_min_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1064-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX1064-NEXT:    v_min_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX1064-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1064-NEXT:    v_min_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX1064-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1064-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX1064-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX1064-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1064-NEXT:    v_readlane_b32 s5, v1, 31
 ; GFX1064-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX1064-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1064-NEXT:    v_readlane_b32 s7, v1, 63
 ; GFX1064-NEXT:    v_readlane_b32 s6, v1, 47
 ; GFX1064-NEXT:    v_writelane_b32 v3, s5, 32

 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_min_rtn_u32 v0, v0, v4
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB23_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1064-NEXT:    v_min_u32_e32 v0, s3, v0
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: umin_i32_varying:
 ; GFX1032:       ; %bb.0: ; %entry
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, -1
 ; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_min_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX1032-NEXT:    v_mov_b32_e32 v3, -1
 ; GFX1032-NEXT:    v_min_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    v_min_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    v_min_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1032-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_min_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX1032-NEXT:    v_mov_b32_e32 v3, -1
 ; GFX1032-NEXT:    v_readlane_b32 s3, v1, 15
 ; GFX1032-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1032-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1032-NEXT:    v_writelane_b32 v3, s3, 16
 ; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    ; implicit-def: $vgpr0
 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB23_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1032-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_min_rtn_u32 v0, v0, v4
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB23_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1032-NEXT:    v_min_u32_e32 v0, s3, v0
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;

 ; GFX1164-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1164-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_min_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX1164-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX1164-NEXT:    v_min_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
 ; GFX1164-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX1164-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX1164-NEXT:    v_readlane_b32 s4, v1, 15
 ; GFX1164-NEXT:    v_readlane_b32 s5, v1, 31
 ; GFX1164-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX1164-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1164-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX1164-NEXT:    v_readlane_b32 s7, v1, 63
 ; GFX1164-NEXT:    v_readlane_b32 s6, v1, 47

 ; GFX1164-NEXT:    s_mov_b32 s3, s7
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_min_rtn_u32 v0, v0, v4
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB23_2:
 ; GFX1164-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1164-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1164-NEXT:    v_min_u32_e32 v0, s3, v0
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: umin_i32_varying:
 ; GFX1132:       ; %bb.0: ; %entry
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX1132-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, -1
 ; GFX1132-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_min_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX1132-NEXT:    v_mov_b32_e32 v3, -1
 ; GFX1132-NEXT:    v_min_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_min_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    v_min_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX1132-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1132-NEXT:    v_min_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX1132-NEXT:    v_mov_b32_e32 v3, -1
 ; GFX1132-NEXT:    v_readlane_b32 s3, v1, 15
 ; GFX1132-NEXT:    v_readlane_b32 s4, v1, 31
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
 ; GFX1132-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
 ; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
+; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
 ; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1132-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX1132-NEXT:    v_writelane_b32 v3, s3, 16
 ; GFX1132-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_2)
 ; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    ; implicit-def: $vgpr0
 ; GFX1132-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1132-NEXT:    s_cbranch_execz .LBB23_2
 ; GFX1132-NEXT:  ; %bb.1:
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX1132-NEXT:    v_mov_b32_e32 v4, s4
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_min_rtn_u32 v0, v0, v4
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB23_2:
 ; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_min_u32_e32 v0, s3, v0
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
 ; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
   %old = atomicrmw umin i32 addrspace(3)* @local_var32, i32 %lane acq_rel
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @umin_i64_constant(i64 addrspace(1)* %out) {
 ;
 ;
 ; GFX7LESS-LABEL: umin_i64_constant:
 ; GFX7LESS:       ; %bb.0: ; %entry
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
 ; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
 ; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
 ; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX7LESS-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX7LESS-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX7LESS-NEXT:    s_cbranch_execz .LBB24_2
 ; GFX7LESS-NEXT:  ; %bb.1:
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX7LESS-NEXT:    s_mov_b32 m0, -1
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    ds_min_rtn_u64 v[0:1], v2, v[0:1]
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:  .LBB24_2:
 ; GFX7LESS-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
+; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7LESS-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX7LESS-NEXT:    v_readfirstlane_b32 s5, v1
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
 ; GFX7LESS-NEXT:    v_cndmask_b32_e64 v1, 0, -1, vcc
 ; GFX7LESS-NEXT:    v_cndmask_b32_e64 v0, 5, -1, vcc
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX7LESS-NEXT:    v_cmp_lt_u64_e32 vcc, s[4:5], v[0:1]
 ; GFX7LESS-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc
 ; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX7LESS-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7LESS-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX7LESS-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: umin_i64_constant:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB24_2
 ; GFX8-NEXT:  ; %bb.1:
 ; GFX8-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX8-NEXT:    s_mov_b32 m0, -1
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    ds_min_rtn_u64 v[0:1], v2, v[0:1]
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:  .LBB24_2:
 ; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX8-NEXT:    v_readfirstlane_b32 s5, v1
 ; GFX8-NEXT:    v_cndmask_b32_e64 v1, 0, -1, vcc
 ; GFX8-NEXT:    v_cndmask_b32_e64 v0, 5, -1, vcc
 ; GFX8-NEXT:    v_cmp_lt_u64_e32 vcc, s[4:5], v[0:1]
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX8-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s4
+; GFX8-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX8-NEXT:    s_mov_b32 s2, -1
 ; GFX8-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
-; GFX8-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX8-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: umin_i64_constant:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB24_2
 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    ds_min_rtn_u64 v[0:1], v2, v[0:1]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:  .LBB24_2:
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX9-NEXT:    v_readfirstlane_b32 s5, v1
 ; GFX9-NEXT:    v_cndmask_b32_e64 v1, 0, -1, vcc
 ; GFX9-NEXT:    v_cndmask_b32_e64 v0, 5, -1, vcc
 ; GFX9-NEXT:    v_cmp_lt_u64_e32 vcc, s[4:5], v[0:1]
 ; GFX9-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX9-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc
 ; GFX9-NEXT:    v_mov_b32_e32 v2, s4
+; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1
 ; GFX9-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
-; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX1064-LABEL: umin_i64_constant:
 ; GFX1064:       ; %bb.0: ; %entry
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
 ; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX1064-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX1064-NEXT:    s_cbranch_execz .LBB24_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_min_rtn_u64 v[0:1], v2, v[0:1]
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  .LBB24_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1064-NEXT:    v_cndmask_b32_e64 v1, 0, -1, vcc
 ; GFX1064-NEXT:    v_cndmask_b32_e64 v0, 5, -1, vcc
 ; GFX1064-NEXT:    v_cmp_lt_u64_e32 vcc, s[2:3], v[0:1]
 ; GFX1064-NEXT:    v_cndmask_b32_e64 v1, v1, s3, vcc
 ; GFX1064-NEXT:    v_cndmask_b32_e64 v0, v0, s2, vcc
-; GFX1064-NEXT:    s_mov_b32 s2, -1
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
+; GFX1064-NEXT:    s_mov_b32 s2, -1
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: umin_i64_constant:
 ; GFX1032:       ; %bb.0: ; %entry
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1032-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1032-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz .LBB24_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_min_rtn_u64 v[0:1], v2, v[0:1]
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  .LBB24_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1032-NEXT:    v_cndmask_b32_e64 v1, 0, -1, vcc_lo
 ; GFX1032-NEXT:    v_cndmask_b32_e64 v0, 5, -1, vcc_lo
 ; GFX1032-NEXT:    v_cmp_lt_u64_e32 vcc_lo, s[2:3], v[0:1]
 ; GFX1032-NEXT:    v_cndmask_b32_e64 v1, v1, s3, vcc_lo
 ; GFX1032-NEXT:    v_cndmask_b32_e64 v0, v0, s2, vcc_lo
-; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
+; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 ;
 ; GFX1164-LABEL: umin_i64_constant:
 ; GFX1164:       ; %bb.0: ; %entry
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
 ; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX1164-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1164-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX1164-NEXT:    s_cbranch_execz .LBB24_2
 ; GFX1164-NEXT:  ; %bb.1:
 ; GFX1164-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1164-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1164-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1164-NEXT:    ds_min_rtn_u64 v[0:1], v2, v[0:1]
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_gl0_inv
 ; GFX1164-NEXT:  .LBB24_2:
 ; GFX1164-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1164-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1164-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1164-NEXT:    v_cndmask_b32_e64 v1, 0, -1, vcc
 ; GFX1164-NEXT:    v_cndmask_b32_e64 v0, 5, -1, vcc
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1164-NEXT:    v_cmp_lt_u64_e32 vcc, s[2:3], v[0:1]
 ; GFX1164-NEXT:    v_cndmask_b32_e64 v1, v1, s3, vcc
 ; GFX1164-NEXT:    v_cndmask_b32_e64 v0, v0, s2, vcc
-; GFX1164-NEXT:    s_mov_b32 s2, -1
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
+; GFX1164-NEXT:    s_mov_b32 s2, -1
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
 ; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: umin_i64_constant:
 ; GFX1132:       ; %bb.0: ; %entry
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX1132-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1132-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX1132-NEXT:    s_cbranch_execz .LBB24_2
 ; GFX1132-NEXT:  ; %bb.1:
 ; GFX1132-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1132-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0
 ; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1132-NEXT:    ds_min_rtn_u64 v[0:1], v2, v[0:1]
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_gl0_inv
 ; GFX1132-NEXT:  .LBB24_2:
 ; GFX1132-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX1132-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX1132-NEXT:    v_readfirstlane_b32 s3, v1
 ; GFX1132-NEXT:    v_cndmask_b32_e64 v1, 0, -1, vcc_lo
 ; GFX1132-NEXT:    v_cndmask_b32_e64 v0, 5, -1, vcc_lo
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_cmp_lt_u64_e32 vcc_lo, s[2:3], v[0:1]
 ; GFX1132-NEXT:    v_cndmask_b32_e64 v1, v1, s3, vcc_lo
 ; GFX1132-NEXT:    v_cndmask_b32_e64 v0, v0, s2, vcc_lo
-; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
+; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
 ; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %old = atomicrmw umin i64 addrspace(3)* @local_var64, i64 5 acq_rel
   store i64 %old, i64 addrspace(1)* %out
   ret void
 }

llvm/test/CodeGen/AMDGPU/atomic_optimizations_raw_buffer.ll

 declare i32 @llvm.amdgcn.raw.buffer.atomic.add(i32, <4 x i32>, i32, i32, i32)
 declare i32 @llvm.amdgcn.raw.buffer.atomic.sub(i32, <4 x i32>, i32, i32, i32)
 
 ; Show what the atomic optimization pass will do for raw buffers.
 
 define amdgpu_kernel void @add_i32_constant(i32 addrspace(1)* %out, <4 x i32> %inout) {
 ; GFX6-LABEL: add_i32_constant:
 ; GFX6:       ; %bb.0: ; %entry
-; GFX6-NEXT:    s_mov_b64 s[2:3], exec
-; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
-; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
+; GFX6-NEXT:    s_mov_b64 s[4:5], exec
+; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s4, 0
+; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s5, v0
 ; GFX6-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX6-NEXT:    ; implicit-def: $vgpr1
-; GFX6-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX6-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX6-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX6-NEXT:  ; %bb.1:
 ; GFX6-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0xd
-; GFX6-NEXT:    s_bcnt1_i32_b64 s0, s[2:3]
-; GFX6-NEXT:    s_mul_i32 s0, s0, 5
-; GFX6-NEXT:    v_mov_b32_e32 v1, s0
+; GFX6-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX6-NEXT:    s_mul_i32 s4, s4, 5
+; GFX6-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
 ; GFX6-NEXT:  .LBB0_2:
-; GFX6-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX6-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
-; GFX6-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX6-NEXT:    s_mov_b32 s7, 0xf000
-; GFX6-NEXT:    v_mad_u32_u24 v0, v0, 5, s0
-; GFX6-NEXT:    s_mov_b32 s6, -1
+; GFX6-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX6-NEXT:    v_mad_u32_u24 v0, v0, 5, s4
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: add_i32_constant:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[6:7], exec
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX8-NEXT:    s_mov_b64 s[4:5], exec
+; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr1
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX8-NEXT:  ; %bb.1:
 ; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX8-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
-; GFX8-NEXT:    s_mul_i32 s0, s0, 5
-; GFX8-NEXT:    v_mov_b32_e32 v1, s0
+; GFX8-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX8-NEXT:    s_mul_i32 s4, s4, 5
+; GFX8-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
 ; GFX8-NEXT:  .LBB0_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX8-NEXT:    v_mad_u32_u24 v2, v0, 5, s0
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX8-NEXT:    v_mad_u32_u24 v2, v0, 5, s2
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    flat_store_dword v[0:1], v2
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: add_i32_constant:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[6:7], exec
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX9-NEXT:    s_mov_b64 s[4:5], exec
+; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr1
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX9-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
-; GFX9-NEXT:    s_mul_i32 s0, s0, 5
-; GFX9-NEXT:    v_mov_b32_e32 v1, s0
+; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX9-NEXT:    s_mul_i32 s4, s4, 5
+; GFX9-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
 ; GFX9-NEXT:  .LBB0_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX9-NEXT:    v_mad_u32_u24 v0, v0, 5, s0
-; GFX9-NEXT:    v_mov_b32_e32 v1, 0
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX9-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: add_i32_constant:
 ; GFX10W64:       ; %bb.0: ; %entry
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_mov_b64 s[6:7], exec
+; GFX10W64-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr1
-; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX10W64-NEXT:  ; %bb.1:
 ; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W64-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
-; GFX10W64-NEXT:    s_mul_i32 s0, s0, 5
-; GFX10W64-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX10W64-NEXT:    s_mul_i32 s4, s4, 5
+; GFX10W64-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10W64-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
 ; GFX10W64-NEXT:  .LBB0_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W64-NEXT:    v_mad_u32_u24 v0, v0, 5, s0
+; GFX10W64-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W64-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: add_i32_constant:
 ; GFX10W32:       ; %bb.0: ; %entry
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_mov_b32 s5, exec_lo
+; GFX10W32-NEXT:    s_mov_b32 s3, exec_lo
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr1
-; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s5, 0
+; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX10W32-NEXT:  ; %bb.1:
-; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    s_bcnt1_i32_b32 s0, s5
-; GFX10W32-NEXT:    s_mul_i32 s0, s0, 5
-; GFX10W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10W32-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x34
+; GFX10W32-NEXT:    s_bcnt1_i32_b32 s3, s3
+; GFX10W32-NEXT:    s_mul_i32 s3, s3, 5
+; GFX10W32-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
+; GFX10W32-NEXT:    buffer_atomic_add v1, off, s[4:7], 0 glc
 ; GFX10W32-NEXT:  .LBB0_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W32-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W32-NEXT:    v_mad_u32_u24 v0, v0, 5, s0
+; GFX10W32-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W32-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: add_i32_constant:
 ; GFX11W64:       ; %bb.0: ; %entry
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_mov_b64 s[6:7], exec
 ; GFX11W64-NEXT:    s_mov_b64 s[4:5], exec
-; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX11W64-NEXT:    s_mov_b64 s[2:3], exec
+; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX11W64-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX11W64-NEXT:  ; %bb.1:
 ; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W64-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
+; GFX11W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX11W64-NEXT:    s_mul_i32 s0, s0, 5
-; GFX11W64-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11W64-NEXT:    s_mul_i32 s4, s4, 5
+; GFX11W64-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11W64-NEXT:    buffer_atomic_add_u32 v1, off, s[8:11], 0 glc
 ; GFX11W64-NEXT:  .LBB0_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W64-NEXT:    v_mad_u32_u24 v0, v0, 5, s0
+; GFX11W64-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11W64-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: add_i32_constant:
 ; GFX11W32:       ; %bb.0: ; %entry
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_mov_b32 s5, exec_lo
-; GFX11W32-NEXT:    s_mov_b32 s4, exec_lo
-; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s5, 0
+; GFX11W32-NEXT:    s_mov_b32 s3, exec_lo
+; GFX11W32-NEXT:    s_mov_b32 s2, exec_lo
+; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX11W32-NEXT:  ; %bb.1:
-; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    s_bcnt1_i32_b32 s0, s5
+; GFX11W32-NEXT:    s_load_b128 s[4:7], s[0:1], 0x34
+; GFX11W32-NEXT:    s_bcnt1_i32_b32 s3, s3
 ; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX11W32-NEXT:    s_mul_i32 s0, s0, 5
-; GFX11W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11W32-NEXT:    s_mul_i32 s3, s3, 5
+; GFX11W32-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    buffer_atomic_add_u32 v1, off, s[8:11], 0 glc
+; GFX11W32-NEXT:    buffer_atomic_add_u32 v1, off, s[4:7], 0 glc
 ; GFX11W32-NEXT:  .LBB0_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W32-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W32-NEXT:    v_mad_u32_u24 v0, v0, 5, s0
+; GFX11W32-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11W32-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %old = call i32 @llvm.amdgcn.raw.buffer.atomic.add(i32 5, <4 x i32> %inout, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @add_i32_uniform(i32 addrspace(1)* %out, <4 x i32> %inout, i32 %additive) {
 ; GFX6-LABEL: add_i32_uniform:
 ; GFX6:       ; %bb.0: ; %entry
-; GFX6-NEXT:    s_mov_b64 s[2:3], exec
-; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
-; GFX6-NEXT:    s_load_dword s8, s[0:1], 0x11
-; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
+; GFX6-NEXT:    s_mov_b64 s[4:5], exec
+; GFX6-NEXT:    s_load_dword s6, s[0:1], 0x11
+; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s4, 0
+; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s5, v0
 ; GFX6-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX6-NEXT:    ; implicit-def: $vgpr1
-; GFX6-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX6-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX6-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX6-NEXT:  ; %bb.1:
-; GFX6-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0xd
-; GFX6-NEXT:    s_bcnt1_i32_b64 s0, s[2:3]
+; GFX6-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0xd
+; GFX6-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    s_mul_i32 s0, s8, s0
-; GFX6-NEXT:    v_mov_b32_e32 v1, s0
-; GFX6-NEXT:    buffer_atomic_add v1, off, s[12:15], 0 glc
+; GFX6-NEXT:    s_mul_i32 s4, s6, s4
+; GFX6-NEXT:    v_mov_b32_e32 v1, s4
+; GFX6-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
 ; GFX6-NEXT:  .LBB1_2:
-; GFX6-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX6-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
-; GFX6-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX6-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    v_mul_lo_u32 v0, s8, v0
-; GFX6-NEXT:    s_mov_b32 s7, 0xf000
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, s0, v0
-; GFX6-NEXT:    s_mov_b32 s6, -1
-; GFX6-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX6-NEXT:    v_mul_lo_u32 v0, s6, v0
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, s4, v0
+; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: add_i32_uniform:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_load_dword s8, s[0:1], 0x44
+; GFX8-NEXT:    s_load_dword s6, s[0:1], 0x44
 ; GFX8-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr1
-; GFX8-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0x34
-; GFX8-NEXT:    s_bcnt1_i32_b64 s0, s[4:5]
+; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX8-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_mul_i32 s0, s8, s0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s0
-; GFX8-NEXT:    buffer_atomic_add v1, off, s[12:15], 0 glc
+; GFX8-NEXT:    s_mul_i32 s4, s6, s4
+; GFX8-NEXT:    v_mov_b32_e32 v1, s4
+; GFX8-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
 ; GFX8-NEXT:  .LBB1_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mul_lo_u32 v0, s8, v0
+; GFX8-NEXT:    v_mul_lo_u32 v0, s6, v0
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX8-NEXT:    v_add_u32_e32 v2, vcc, s0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX8-NEXT:    v_add_u32_e32 v2, vcc, s2, v0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    flat_store_dword v[0:1], v2
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: add_i32_uniform:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_load_dword s8, s[0:1], 0x44
+; GFX9-NEXT:    s_load_dword s6, s[0:1], 0x44
 ; GFX9-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr1
-; GFX9-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0x34
-; GFX9-NEXT:    s_bcnt1_i32_b64 s0, s[4:5]
+; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_mul_i32 s0, s8, s0
-; GFX9-NEXT:    v_mov_b32_e32 v1, s0
-; GFX9-NEXT:    buffer_atomic_add v1, off, s[12:15], 0 glc
+; GFX9-NEXT:    s_mul_i32 s4, s6, s4
+; GFX9-NEXT:    v_mov_b32_e32 v1, s4
+; GFX9-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
 ; GFX9-NEXT:  .LBB1_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mul_lo_u32 v0, s8, v0
+; GFX9-NEXT:    v_mul_lo_u32 v0, s6, v0
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX9-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-NEXT:    v_add_u32_e32 v0, s0, v0
-; GFX9-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_add_u32_e32 v0, s2, v0
+; GFX9-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: add_i32_uniform:
 ; GFX10W64:       ; %bb.0: ; %entry
-; GFX10W64-NEXT:    s_clause 0x1
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_load_dword s8, s[0:1], 0x44
+; GFX10W64-NEXT:    s_load_dword s6, s[0:1], 0x44
 ; GFX10W64-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr1
 ; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX10W64-NEXT:  ; %bb.1:
-; GFX10W64-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0x34
-; GFX10W64-NEXT:    s_bcnt1_i32_b64 s0, s[4:5]
+; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX10W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    s_mul_i32 s0, s8, s0
-; GFX10W64-NEXT:    v_mov_b32_e32 v1, s0
-; GFX10W64-NEXT:    buffer_atomic_add v1, off, s[12:15], 0 glc
+; GFX10W64-NEXT:    s_mul_i32 s4, s6, s4
+; GFX10W64-NEXT:    v_mov_b32_e32 v1, s4
+; GFX10W64-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
 ; GFX10W64-NEXT:  .LBB1_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], s8, v0, s[0:1]
+; GFX10W64-NEXT:    v_mad_u64_u32 v[0:1], s[2:3], s6, v0, s[2:3]
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W64-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W64-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: add_i32_uniform:
 ; GFX10W32:       ; %bb.0: ; %entry
-; GFX10W32-NEXT:    s_clause 0x1
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_load_dword s4, s[0:1], 0x44
-; GFX10W32-NEXT:    s_mov_b32 s6, exec_lo
+; GFX10W32-NEXT:    s_load_dword s2, s[0:1], 0x44
+; GFX10W32-NEXT:    s_mov_b32 s4, exec_lo
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr1
-; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s5, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX10W32-NEXT:  ; %bb.1:
 ; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    s_bcnt1_i32_b32 s0, s6
+; GFX10W32-NEXT:    s_bcnt1_i32_b32 s4, s4
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    s_mul_i32 s0, s4, s0
-; GFX10W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10W32-NEXT:    s_mul_i32 s4, s2, s4
+; GFX10W32-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX10W32-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
 ; GFX10W32-NEXT:  .LBB1_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s5
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W32-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    v_mad_u64_u32 v[0:1], s0, s4, v0, s[0:1]
+; GFX10W32-NEXT:    v_mad_u64_u32 v[0:1], s2, s2, v0, s[4:5]
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W32-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W32-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: add_i32_uniform:
 ; GFX11W64:       ; %bb.0: ; %entry
-; GFX11W64-NEXT:    s_clause 0x1
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_load_b32 s8, s[0:1], 0x44
+; GFX11W64-NEXT:    s_load_b32 s6, s[0:1], 0x44
 ; GFX11W64-NEXT:    s_mov_b64 s[4:5], exec
-; GFX11W64-NEXT:    s_mov_b64 s[6:7], exec
+; GFX11W64-NEXT:    s_mov_b64 s[2:3], exec
 ; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX11W64-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX11W64-NEXT:  ; %bb.1:
-; GFX11W64-NEXT:    s_load_b128 s[12:15], s[0:1], 0x34
-; GFX11W64-NEXT:    s_bcnt1_i32_b64 s0, s[4:5]
+; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
+; GFX11W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    s_mul_i32 s0, s8, s0
+; GFX11W64-NEXT:    s_mul_i32 s4, s6, s4
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11W64-NEXT:    v_mov_b32_e32 v1, s0
-; GFX11W64-NEXT:    buffer_atomic_add_u32 v1, off, s[12:15], 0 glc
+; GFX11W64-NEXT:    v_mov_b32_e32 v1, s4
+; GFX11W64-NEXT:    buffer_atomic_add_u32 v1, off, s[8:11], 0 glc
 ; GFX11W64-NEXT:  .LBB1_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11W64-NEXT:    v_mad_u64_u32 v[1:2], null, s8, v0, s[0:1]
+; GFX11W64-NEXT:    v_mad_u64_u32 v[1:2], null, s6, v0, s[2:3]
 ; GFX11W64-NEXT:    v_mov_b32_e32 v0, 0
-; GFX11W64-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11W64-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: add_i32_uniform:
 ; GFX11W32:       ; %bb.0: ; %entry
-; GFX11W32-NEXT:    s_clause 0x1
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_load_b32 s4, s[0:1], 0x44
-; GFX11W32-NEXT:    s_mov_b32 s6, exec_lo
-; GFX11W32-NEXT:    s_mov_b32 s5, exec_lo
-; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX11W32-NEXT:    s_load_b32 s2, s[0:1], 0x44
+; GFX11W32-NEXT:    s_mov_b32 s4, exec_lo
+; GFX11W32-NEXT:    s_mov_b32 s3, exec_lo
+; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX11W32-NEXT:  ; %bb.1:
 ; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    s_bcnt1_i32_b32 s0, s6
+; GFX11W32-NEXT:    s_bcnt1_i32_b32 s4, s4
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    s_mul_i32 s0, s4, s0
+; GFX11W32-NEXT:    s_mul_i32 s4, s2, s4
 ; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11W32-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX11W32-NEXT:    buffer_atomic_add_u32 v1, off, s[8:11], 0 glc
 ; GFX11W32-NEXT:  .LBB1_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s5
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W32-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11W32-NEXT:    v_mad_u64_u32 v[1:2], null, s4, v0, s[0:1]
+; GFX11W32-NEXT:    v_mad_u64_u32 v[1:2], null, s2, v0, s[4:5]
 ; GFX11W32-NEXT:    v_mov_b32_e32 v0, 0
-; GFX11W32-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11W32-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %old = call i32 @llvm.amdgcn.raw.buffer.atomic.add(i32 %additive, <4 x i32> %inout, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @add_i32_varying_vdata(i32 addrspace(1)* %out, <4 x i32> %inout) {
 ; GFX6-LABEL: add_i32_varying_vdata:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0xd
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_add v0, off, s[4:7], 0 glc
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: add_i32_varying_vdata:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
-; GFX8-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX8-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX8-NEXT:    s_not_b64 exec, exec
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX8-NEXT:    s_not_b64 exec, exec
-; GFX8-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX8-NEXT:    v_readlane_b32 s6, v2, 63
+; GFX8-NEXT:    v_readlane_b32 s4, v2, 63
 ; GFX8-NEXT:    s_nop 0
 ; GFX8-NEXT:    v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf
-; GFX8-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
 ; GFX8-NEXT:    ; implicit-def: $vgpr0
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX8-NEXT:  ; %bb.1:
 ; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX8-NEXT:    v_mov_b32_e32 v0, s6
+; GFX8-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    buffer_atomic_add v0, off, s[8:11], 0 glc
 ; GFX8-NEXT:  .LBB2_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, v1
+; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v4, s3
-; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s2
+; GFX8-NEXT:    v_mov_b32_e32 v4, s1
+; GFX8-NEXT:    v_mov_b32_e32 v3, s0
 ; GFX8-NEXT:    flat_store_dword v[3:4], v0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: add_i32_varying_vdata:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX9-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX9-NEXT:    s_not_b64 exec, exec
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    s_not_b64 exec, exec
-; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-NEXT:    v_readlane_b32 s6, v2, 63
+; GFX9-NEXT:    v_readlane_b32 s4, v2, 63
 ; GFX9-NEXT:    s_nop 0
 ; GFX9-NEXT:    v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf
-; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
 ; GFX9-NEXT:    ; implicit-def: $vgpr0
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX9-NEXT:    v_mov_b32_e32 v0, s6
+; GFX9-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    buffer_atomic_add v0, off, s[8:11], 0 glc
 ; GFX9-NEXT:  .LBB2_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v1
-; GFX9-NEXT:    v_add_u32_e32 v0, s0, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, 0
+; GFX9-NEXT:    v_add_u32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    global_store_dword v3, v0, s[2:3]
+; GFX9-NEXT:    global_store_dword v3, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: add_i32_varying_vdata:
 ; GFX10W64:       ; %bb.0: ; %entry
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX10W64-NEXT:    s_not_b64 exec, exec
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10W64-NEXT:    s_not_b64 exec, exec
 ; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX10W64-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX10W64-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 15
 ; GFX10W64-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX10W64-NEXT:    v_readlane_b32 s5, v1, 31
+; GFX10W64-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_readlane_b32 s7, v1, 31
-; GFX10W64-NEXT:    v_writelane_b32 v3, s6, 16
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
 ; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 63
-; GFX10W64-NEXT:    v_readlane_b32 s8, v1, 47
-; GFX10W64-NEXT:    v_writelane_b32 v3, s7, 32
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 63
+; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 47
+; GFX10W64-NEXT:    v_writelane_b32 v3, s5, 32
+; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_writelane_b32 v3, s8, 48
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX10W64-NEXT:    v_writelane_b32 v3, s6, 48
+; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX10W64-NEXT:  ; %bb.1:
 ; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W64-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10W64-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10W64-NEXT:    buffer_atomic_add v0, off, s[8:11], 0 glc
 ; GFX10W64-NEXT:  .LBB2_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX10W64-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX10W64-NEXT:    v_mov_b32_e32 v4, 0
-; GFX10W64-NEXT:    v_add_nc_u32_e32 v0, s0, v0
+; GFX10W64-NEXT:    v_add_nc_u32_e32 v0, s2, v0
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    global_store_dword v4, v0, s[2:3]
+; GFX10W64-NEXT:    global_store_dword v4, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: add_i32_varying_vdata:
 ; GFX10W32:       ; %bb.0: ; %entry
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX10W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX10W32-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX10W32-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX10W32-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_or_saveexec_b32 s4, -1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX10W32-NEXT:    v_mov_b32_e32 v3, 0
-; GFX10W32-NEXT:    v_readlane_b32 s6, v1, 31
+; GFX10W32-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX10W32-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX10W32-NEXT:    v_readlane_b32 s5, v1, 15
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10W32-NEXT:    v_readlane_b32 s3, v1, 15
+; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX10W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX10W32-NEXT:    v_writelane_b32 v3, s5, 16
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10W32-NEXT:    s_or_saveexec_b32 s2, -1
+; GFX10W32-NEXT:    v_writelane_b32 v3, s3, 16
+; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX10W32-NEXT:  ; %bb.1:
-; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    v_mov_b32_e32 v0, s6
-; GFX10W32-NEXT:    s_mov_b32 s5, s6
+; GFX10W32-NEXT:    s_mov_b32 s3, s4
+; GFX10W32-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x34
+; GFX10W32-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    buffer_atomic_add v0, off, s[8:11], 0 glc
+; GFX10W32-NEXT:    buffer_atomic_add v0, off, s[4:7], 0 glc
 ; GFX10W32-NEXT:  .LBB2_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10W32-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX10W32-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX10W32-NEXT:    v_mov_b32_e32 v4, 0
-; GFX10W32-NEXT:    v_add_nc_u32_e32 v0, s0, v0
+; GFX10W32-NEXT:    v_add_nc_u32_e32 v0, s2, v0
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    global_store_dword v4, v0, s[2:3]
+; GFX10W32-NEXT:    global_store_dword v4, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: add_i32_varying_vdata:
 ; GFX11W64:       ; %bb.0: ; %entry
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX11W64-NEXT:    s_not_b64 exec, exec
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W64-NEXT:    s_not_b64 exec, exec

 ; GFX11W64-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX11W64-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W64-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX11W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 15
+; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
 ; GFX11W64-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX11W64-NEXT:    v_readlane_b32 s5, v1, 31
+; GFX11W64-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_readlane_b32 s7, v1, 31
-; GFX11W64-NEXT:    v_writelane_b32 v3, s6, 16
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 63
-; GFX11W64-NEXT:    v_readlane_b32 s8, v1, 47
-; GFX11W64-NEXT:    v_writelane_b32 v3, s7, 32
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX11W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 63
+; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 47
+; GFX11W64-NEXT:    v_writelane_b32 v3, s5, 32
+; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
 ; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_writelane_b32 v3, s8, 48
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX11W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX11W64-NEXT:    v_writelane_b32 v3, s6, 48
+; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX11W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr0
-; GFX11W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX11W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX11W64-NEXT:  ; %bb.1:
 ; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W64-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11W64-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11W64-NEXT:    buffer_atomic_add_u32 v0, off, s[8:11], 0 glc
 ; GFX11W64-NEXT:  .LBB2_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX11W64-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX11W64-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W64-NEXT:    v_add_nc_u32_e32 v0, s0, v0
+; GFX11W64-NEXT:    v_add_nc_u32_e32 v0, s2, v0
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    global_store_b32 v4, v0, s[2:3]
+; GFX11W64-NEXT:    global_store_b32 v4, v0, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: add_i32_varying_vdata:
 ; GFX11W32:       ; %bb.0: ; %entry
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX11W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX11W32-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX11W32-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX11W32-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX11W32-NEXT:    v_mov_b32_e32 v3, 0
-; GFX11W32-NEXT:    v_readlane_b32 s6, v1, 31
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
+; GFX11W32-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX11W32-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX11W32-NEXT:    v_readlane_b32 s5, v1, 15
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX11W32-NEXT:    v_readlane_b32 s3, v1, 15
+; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
+; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX11W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11W32-NEXT:    v_writelane_b32 v3, s5, 16
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX11W32-NEXT:    s_or_saveexec_b32 s2, -1
+; GFX11W32-NEXT:    v_writelane_b32 v3, s3, 16
+; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr0
-; GFX11W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX11W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX11W32-NEXT:  ; %bb.1:
-; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    v_mov_b32_e32 v0, s6
-; GFX11W32-NEXT:    s_mov_b32 s5, s6
+; GFX11W32-NEXT:    s_mov_b32 s3, s4
+; GFX11W32-NEXT:    s_load_b128 s[4:7], s[0:1], 0x34
+; GFX11W32-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    buffer_atomic_add_u32 v0, off, s[8:11], 0 glc
+; GFX11W32-NEXT:    buffer_atomic_add_u32 v0, off, s[4:7], 0 glc
 ; GFX11W32-NEXT:  .LBB2_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX11W32-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX11W32-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX11W32-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W32-NEXT:    v_add_nc_u32_e32 v0, s0, v0
+; GFX11W32-NEXT:    v_add_nc_u32_e32 v0, s2, v0
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    global_store_b32 v4, v0, s[2:3]
+; GFX11W32-NEXT:    global_store_b32 v4, v0, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
   %old = call i32 @llvm.amdgcn.raw.buffer.atomic.add(i32 %lane, <4 x i32> %inout, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
entry:
   %old = call i32 @llvm.amdgcn.raw.buffer.atomic.add(i32 1, <4 x i32> %inout, i32 %lane, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @sub_i32_constant(i32 addrspace(1)* %out, <4 x i32> %inout) {
 ; GFX6-LABEL: sub_i32_constant:
 ; GFX6:       ; %bb.0: ; %entry
-; GFX6-NEXT:    s_mov_b64 s[2:3], exec
-; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
-; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
+; GFX6-NEXT:    s_mov_b64 s[4:5], exec
+; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s4, 0
+; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s5, v0
 ; GFX6-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX6-NEXT:    ; implicit-def: $vgpr1
-; GFX6-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX6-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX6-NEXT:    s_cbranch_execz .LBB4_2
 ; GFX6-NEXT:  ; %bb.1:
 ; GFX6-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0xd
-; GFX6-NEXT:    s_bcnt1_i32_b64 s0, s[2:3]
-; GFX6-NEXT:    s_mul_i32 s0, s0, 5
-; GFX6-NEXT:    v_mov_b32_e32 v1, s0
+; GFX6-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX6-NEXT:    s_mul_i32 s4, s4, 5
+; GFX6-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
 ; GFX6-NEXT:  .LBB4_2:
-; GFX6-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX6-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
-; GFX6-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX6-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX6-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
-; GFX6-NEXT:    s_mov_b32 s7, 0xf000
-; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, s0, v0
-; GFX6-NEXT:    s_mov_b32 s6, -1
+; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, s4, v0
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: sub_i32_constant:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[6:7], exec
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX8-NEXT:    s_mov_b64 s[4:5], exec
+; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr1
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB4_2
 ; GFX8-NEXT:  ; %bb.1:
 ; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX8-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
-; GFX8-NEXT:    s_mul_i32 s0, s0, 5
-; GFX8-NEXT:    v_mov_b32_e32 v1, s0
+; GFX8-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX8-NEXT:    s_mul_i32 s4, s4, 5
+; GFX8-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
 ; GFX8-NEXT:  .LBB4_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX8-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
-; GFX8-NEXT:    v_sub_u32_e32 v2, vcc, s0, v0
+; GFX8-NEXT:    v_sub_u32_e32 v2, vcc, s2, v0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    flat_store_dword v[0:1], v2
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sub_i32_constant:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[6:7], exec
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX9-NEXT:    s_mov_b64 s[4:5], exec
+; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr1
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB4_2
 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX9-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
-; GFX9-NEXT:    s_mul_i32 s0, s0, 5
-; GFX9-NEXT:    v_mov_b32_e32 v1, s0
+; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX9-NEXT:    s_mul_i32 s4, s4, 5
+; GFX9-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
 ; GFX9-NEXT:  .LBB4_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX9-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
-; GFX9-NEXT:    v_sub_u32_e32 v0, s0, v0
-; GFX9-NEXT:    v_mov_b32_e32 v1, 0
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_sub_u32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX9-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: sub_i32_constant:
 ; GFX10W64:       ; %bb.0: ; %entry
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_mov_b64 s[6:7], exec
+; GFX10W64-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr1
-; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB4_2
 ; GFX10W64-NEXT:  ; %bb.1:
 ; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W64-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
-; GFX10W64-NEXT:    s_mul_i32 s0, s0, 5
-; GFX10W64-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX10W64-NEXT:    s_mul_i32 s4, s4, 5
+; GFX10W64-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10W64-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
 ; GFX10W64-NEXT:  .LBB4_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W64-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W64-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX10W64-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W64-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: sub_i32_constant:
 ; GFX10W32:       ; %bb.0: ; %entry
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_mov_b32 s5, exec_lo
+; GFX10W32-NEXT:    s_mov_b32 s3, exec_lo
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr1
-; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s5, 0
+; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB4_2
 ; GFX10W32-NEXT:  ; %bb.1:
-; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    s_bcnt1_i32_b32 s0, s5
-; GFX10W32-NEXT:    s_mul_i32 s0, s0, 5
-; GFX10W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10W32-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x34
+; GFX10W32-NEXT:    s_bcnt1_i32_b32 s3, s3
+; GFX10W32-NEXT:    s_mul_i32 s3, s3, 5
+; GFX10W32-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
+; GFX10W32-NEXT:    buffer_atomic_sub v1, off, s[4:7], 0 glc
 ; GFX10W32-NEXT:  .LBB4_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W32-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W32-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W32-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX10W32-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W32-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: sub_i32_constant:
 ; GFX11W64:       ; %bb.0: ; %entry
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_mov_b64 s[6:7], exec
 ; GFX11W64-NEXT:    s_mov_b64 s[4:5], exec
-; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX11W64-NEXT:    s_mov_b64 s[2:3], exec
+; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX11W64-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB4_2
 ; GFX11W64-NEXT:  ; %bb.1:
 ; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W64-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
+; GFX11W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX11W64-NEXT:    s_mul_i32 s0, s0, 5
-; GFX11W64-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11W64-NEXT:    s_mul_i32 s4, s4, 5
+; GFX11W64-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11W64-NEXT:    buffer_atomic_sub_u32 v1, off, s[8:11], 0 glc
 ; GFX11W64-NEXT:  .LBB4_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W64-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W64-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX11W64-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11W64-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: sub_i32_constant:
 ; GFX11W32:       ; %bb.0: ; %entry
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_mov_b32 s5, exec_lo
-; GFX11W32-NEXT:    s_mov_b32 s4, exec_lo
-; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s5, 0
+; GFX11W32-NEXT:    s_mov_b32 s3, exec_lo
+; GFX11W32-NEXT:    s_mov_b32 s2, exec_lo
+; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB4_2
 ; GFX11W32-NEXT:  ; %bb.1:
-; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    s_bcnt1_i32_b32 s0, s5
+; GFX11W32-NEXT:    s_load_b128 s[4:7], s[0:1], 0x34
+; GFX11W32-NEXT:    s_bcnt1_i32_b32 s3, s3
 ; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX11W32-NEXT:    s_mul_i32 s0, s0, 5
-; GFX11W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11W32-NEXT:    s_mul_i32 s3, s3, 5
+; GFX11W32-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    buffer_atomic_sub_u32 v1, off, s[8:11], 0 glc
+; GFX11W32-NEXT:    buffer_atomic_sub_u32 v1, off, s[4:7], 0 glc
 ; GFX11W32-NEXT:  .LBB4_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W32-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W32-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W32-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX11W32-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11W32-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %old = call i32 @llvm.amdgcn.raw.buffer.atomic.sub(i32 5, <4 x i32> %inout, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @sub_i32_uniform(i32 addrspace(1)* %out, <4 x i32> %inout, i32 %subitive) {
 ; GFX6-LABEL: sub_i32_uniform:
 ; GFX6:       ; %bb.0: ; %entry
-; GFX6-NEXT:    s_mov_b64 s[2:3], exec
-; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
-; GFX6-NEXT:    s_load_dword s8, s[0:1], 0x11
-; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
+; GFX6-NEXT:    s_mov_b64 s[4:5], exec
+; GFX6-NEXT:    s_load_dword s6, s[0:1], 0x11
+; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s4, 0
+; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s5, v0
 ; GFX6-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX6-NEXT:    ; implicit-def: $vgpr1
-; GFX6-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX6-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX6-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX6-NEXT:  ; %bb.1:
-; GFX6-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0xd
-; GFX6-NEXT:    s_bcnt1_i32_b64 s0, s[2:3]
+; GFX6-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0xd
+; GFX6-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    s_mul_i32 s0, s8, s0
-; GFX6-NEXT:    v_mov_b32_e32 v1, s0
-; GFX6-NEXT:    buffer_atomic_sub v1, off, s[12:15], 0 glc
+; GFX6-NEXT:    s_mul_i32 s4, s6, s4
+; GFX6-NEXT:    v_mov_b32_e32 v1, s4
+; GFX6-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
 ; GFX6-NEXT:  .LBB5_2:
-; GFX6-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX6-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
-; GFX6-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX6-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    v_mul_lo_u32 v0, s8, v0
-; GFX6-NEXT:    s_mov_b32 s7, 0xf000
-; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, s0, v0
-; GFX6-NEXT:    s_mov_b32 s6, -1
-; GFX6-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX6-NEXT:    v_mul_lo_u32 v0, s6, v0
+; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, s4, v0
+; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: sub_i32_uniform:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_load_dword s8, s[0:1], 0x44
+; GFX8-NEXT:    s_load_dword s6, s[0:1], 0x44
 ; GFX8-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr1
-; GFX8-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0x34
-; GFX8-NEXT:    s_bcnt1_i32_b64 s0, s[4:5]
+; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX8-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_mul_i32 s0, s8, s0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s0
-; GFX8-NEXT:    buffer_atomic_sub v1, off, s[12:15], 0 glc
+; GFX8-NEXT:    s_mul_i32 s4, s6, s4
+; GFX8-NEXT:    v_mov_b32_e32 v1, s4
+; GFX8-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
 ; GFX8-NEXT:  .LBB5_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mul_lo_u32 v0, s8, v0
+; GFX8-NEXT:    v_mul_lo_u32 v0, s6, v0
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX8-NEXT:    v_sub_u32_e32 v2, vcc, s0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX8-NEXT:    v_sub_u32_e32 v2, vcc, s2, v0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    flat_store_dword v[0:1], v2
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sub_i32_uniform:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_load_dword s8, s[0:1], 0x44
+; GFX9-NEXT:    s_load_dword s6, s[0:1], 0x44
 ; GFX9-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr1
-; GFX9-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0x34
-; GFX9-NEXT:    s_bcnt1_i32_b64 s0, s[4:5]
+; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_mul_i32 s0, s8, s0
-; GFX9-NEXT:    v_mov_b32_e32 v1, s0
-; GFX9-NEXT:    buffer_atomic_sub v1, off, s[12:15], 0 glc
+; GFX9-NEXT:    s_mul_i32 s4, s6, s4
+; GFX9-NEXT:    v_mov_b32_e32 v1, s4
+; GFX9-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
 ; GFX9-NEXT:  .LBB5_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mul_lo_u32 v0, s8, v0
+; GFX9-NEXT:    v_mul_lo_u32 v0, s6, v0
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX9-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-NEXT:    v_sub_u32_e32 v0, s0, v0
-; GFX9-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_sub_u32_e32 v0, s2, v0
+; GFX9-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: sub_i32_uniform:
 ; GFX10W64:       ; %bb.0: ; %entry
-; GFX10W64-NEXT:    s_clause 0x1
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_load_dword s8, s[0:1], 0x44
+; GFX10W64-NEXT:    s_load_dword s6, s[0:1], 0x44
 ; GFX10W64-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr1
 ; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX10W64-NEXT:  ; %bb.1:
-; GFX10W64-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0x34
-; GFX10W64-NEXT:    s_bcnt1_i32_b64 s0, s[4:5]
+; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX10W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    s_mul_i32 s0, s8, s0
-; GFX10W64-NEXT:    v_mov_b32_e32 v1, s0
-; GFX10W64-NEXT:    buffer_atomic_sub v1, off, s[12:15], 0 glc
+; GFX10W64-NEXT:    s_mul_i32 s4, s6, s4
+; GFX10W64-NEXT:    v_mov_b32_e32 v1, s4
+; GFX10W64-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
 ; GFX10W64-NEXT:  .LBB5_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    v_mul_lo_u32 v0, s8, v0
+; GFX10W64-NEXT:    v_mul_lo_u32 v0, s6, v0
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W64-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
-; GFX10W64-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W64-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
+; GFX10W64-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: sub_i32_uniform:
 ; GFX10W32:       ; %bb.0: ; %entry
-; GFX10W32-NEXT:    s_clause 0x1
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_load_dword s4, s[0:1], 0x44
-; GFX10W32-NEXT:    s_mov_b32 s6, exec_lo
+; GFX10W32-NEXT:    s_load_dword s2, s[0:1], 0x44
+; GFX10W32-NEXT:    s_mov_b32 s4, exec_lo
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr1
-; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s5, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX10W32-NEXT:  ; %bb.1:
 ; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    s_bcnt1_i32_b32 s0, s6
+; GFX10W32-NEXT:    s_bcnt1_i32_b32 s4, s4
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    s_mul_i32 s0, s4, s0
-; GFX10W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10W32-NEXT:    s_mul_i32 s4, s2, s4
+; GFX10W32-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX10W32-NEXT:    buffer_atomic_sub v1, off, s[8:11], 0 glc
 ; GFX10W32-NEXT:  .LBB5_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s5
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    v_mul_lo_u32 v0, s4, v0
+; GFX10W32-NEXT:    v_mul_lo_u32 v0, s2, v0
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W32-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W32-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
-; GFX10W32-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W32-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
+; GFX10W32-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: sub_i32_uniform:
 ; GFX11W64:       ; %bb.0: ; %entry
-; GFX11W64-NEXT:    s_clause 0x1
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_load_b32 s8, s[0:1], 0x44
+; GFX11W64-NEXT:    s_load_b32 s6, s[0:1], 0x44
 ; GFX11W64-NEXT:    s_mov_b64 s[4:5], exec
-; GFX11W64-NEXT:    s_mov_b64 s[6:7], exec
+; GFX11W64-NEXT:    s_mov_b64 s[2:3], exec
 ; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX11W64-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX11W64-NEXT:  ; %bb.1:
-; GFX11W64-NEXT:    s_load_b128 s[12:15], s[0:1], 0x34
-; GFX11W64-NEXT:    s_bcnt1_i32_b64 s0, s[4:5]
+; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
+; GFX11W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    s_mul_i32 s0, s8, s0
+; GFX11W64-NEXT:    s_mul_i32 s4, s6, s4
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11W64-NEXT:    v_mov_b32_e32 v1, s0
-; GFX11W64-NEXT:    buffer_atomic_sub_u32 v1, off, s[12:15], 0 glc
+; GFX11W64-NEXT:    v_mov_b32_e32 v1, s4
+; GFX11W64-NEXT:    buffer_atomic_sub_u32 v1, off, s[8:11], 0 glc
 ; GFX11W64-NEXT:  .LBB5_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    v_mul_lo_u32 v0, s8, v0
+; GFX11W64-NEXT:    v_mul_lo_u32 v0, s6, v0
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W64-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
-; GFX11W64-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11W64-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
+; GFX11W64-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: sub_i32_uniform:
 ; GFX11W32:       ; %bb.0: ; %entry
-; GFX11W32-NEXT:    s_clause 0x1
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_load_b32 s4, s[0:1], 0x44
-; GFX11W32-NEXT:    s_mov_b32 s6, exec_lo
-; GFX11W32-NEXT:    s_mov_b32 s5, exec_lo
-; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX11W32-NEXT:    s_load_b32 s2, s[0:1], 0x44
+; GFX11W32-NEXT:    s_mov_b32 s4, exec_lo
+; GFX11W32-NEXT:    s_mov_b32 s3, exec_lo
+; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX11W32-NEXT:  ; %bb.1:
 ; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    s_bcnt1_i32_b32 s0, s6
+; GFX11W32-NEXT:    s_bcnt1_i32_b32 s4, s4
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    s_mul_i32 s0, s4, s0
+; GFX11W32-NEXT:    s_mul_i32 s4, s2, s4
 ; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11W32-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX11W32-NEXT:    buffer_atomic_sub_u32 v1, off, s[8:11], 0 glc
 ; GFX11W32-NEXT:  .LBB5_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s5
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    v_mul_lo_u32 v0, s4, v0
+; GFX11W32-NEXT:    v_mul_lo_u32 v0, s2, v0
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W32-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W32-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
-; GFX11W32-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11W32-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
+; GFX11W32-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %old = call i32 @llvm.amdgcn.raw.buffer.atomic.sub(i32 %subitive, <4 x i32> %inout, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @sub_i32_varying_vdata(i32 addrspace(1)* %out, <4 x i32> %inout) {
 ; GFX6-LABEL: sub_i32_varying_vdata:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0xd
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_sub v0, off, s[4:7], 0 glc
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: sub_i32_varying_vdata:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
-; GFX8-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX8-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX8-NEXT:    s_not_b64 exec, exec
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX8-NEXT:    s_not_b64 exec, exec
-; GFX8-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX8-NEXT:    v_readlane_b32 s6, v2, 63
+; GFX8-NEXT:    v_readlane_b32 s4, v2, 63
 ; GFX8-NEXT:    s_nop 0
 ; GFX8-NEXT:    v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf
-; GFX8-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
 ; GFX8-NEXT:    ; implicit-def: $vgpr0
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX8-NEXT:  ; %bb.1:
 ; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX8-NEXT:    v_mov_b32_e32 v0, s6
+; GFX8-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    buffer_atomic_sub v0, off, s[8:11], 0 glc
 ; GFX8-NEXT:  .LBB6_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, v1
+; GFX8-NEXT:    v_sub_u32_e32 v0, vcc, s2, v0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v4, s3
-; GFX8-NEXT:    v_sub_u32_e32 v0, vcc, s0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s2
+; GFX8-NEXT:    v_mov_b32_e32 v4, s1
+; GFX8-NEXT:    v_mov_b32_e32 v3, s0
 ; GFX8-NEXT:    flat_store_dword v[3:4], v0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sub_i32_varying_vdata:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX9-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX9-NEXT:    s_not_b64 exec, exec
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    s_not_b64 exec, exec
-; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-NEXT:    v_readlane_b32 s6, v2, 63
+; GFX9-NEXT:    v_readlane_b32 s4, v2, 63
 ; GFX9-NEXT:    s_nop 0
 ; GFX9-NEXT:    v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf
-; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
 ; GFX9-NEXT:    ; implicit-def: $vgpr0
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX9-NEXT:    v_mov_b32_e32 v0, s6
+; GFX9-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    buffer_atomic_sub v0, off, s[8:11], 0 glc
 ; GFX9-NEXT:  .LBB6_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v1
-; GFX9-NEXT:    v_sub_u32_e32 v0, s0, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, 0
+; GFX9-NEXT:    v_sub_u32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    global_store_dword v3, v0, s[2:3]
+; GFX9-NEXT:    global_store_dword v3, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: sub_i32_varying_vdata:
 ; GFX10W64:       ; %bb.0: ; %entry
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX10W64-NEXT:    s_not_b64 exec, exec
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10W64-NEXT:    s_not_b64 exec, exec
 ; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX10W64-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX10W64-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 15
 ; GFX10W64-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX10W64-NEXT:    v_readlane_b32 s5, v1, 31
+; GFX10W64-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_readlane_b32 s7, v1, 31
-; GFX10W64-NEXT:    v_writelane_b32 v3, s6, 16
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
 ; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 63
-; GFX10W64-NEXT:    v_readlane_b32 s8, v1, 47
-; GFX10W64-NEXT:    v_writelane_b32 v3, s7, 32
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 63
+; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 47
+; GFX10W64-NEXT:    v_writelane_b32 v3, s5, 32
+; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_writelane_b32 v3, s8, 48
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX10W64-NEXT:    v_writelane_b32 v3, s6, 48
+; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX10W64-NEXT:  ; %bb.1:
 ; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W64-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10W64-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10W64-NEXT:    buffer_atomic_sub v0, off, s[8:11], 0 glc
 ; GFX10W64-NEXT:  .LBB6_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX10W64-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX10W64-NEXT:    v_mov_b32_e32 v4, 0
-; GFX10W64-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX10W64-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    global_store_dword v4, v0, s[2:3]
+; GFX10W64-NEXT:    global_store_dword v4, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: sub_i32_varying_vdata:
 ; GFX10W32:       ; %bb.0: ; %entry
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX10W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX10W32-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX10W32-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX10W32-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_or_saveexec_b32 s4, -1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX10W32-NEXT:    v_mov_b32_e32 v3, 0
-; GFX10W32-NEXT:    v_readlane_b32 s6, v1, 31
+; GFX10W32-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX10W32-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX10W32-NEXT:    v_readlane_b32 s5, v1, 15
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10W32-NEXT:    v_readlane_b32 s3, v1, 15
+; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX10W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX10W32-NEXT:    v_writelane_b32 v3, s5, 16
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10W32-NEXT:    s_or_saveexec_b32 s2, -1
+; GFX10W32-NEXT:    v_writelane_b32 v3, s3, 16
+; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX10W32-NEXT:  ; %bb.1:
-; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    v_mov_b32_e32 v0, s6
-; GFX10W32-NEXT:    s_mov_b32 s5, s6
+; GFX10W32-NEXT:    s_mov_b32 s3, s4
+; GFX10W32-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x34
+; GFX10W32-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    buffer_atomic_sub v0, off, s[8:11], 0 glc
+; GFX10W32-NEXT:    buffer_atomic_sub v0, off, s[4:7], 0 glc
 ; GFX10W32-NEXT:  .LBB6_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10W32-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX10W32-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX10W32-NEXT:    v_mov_b32_e32 v4, 0
-; GFX10W32-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX10W32-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    global_store_dword v4, v0, s[2:3]
+; GFX10W32-NEXT:    global_store_dword v4, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: sub_i32_varying_vdata:
 ; GFX11W64:       ; %bb.0: ; %entry
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX11W64-NEXT:    s_not_b64 exec, exec
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W64-NEXT:    s_not_b64 exec, exec

 ; GFX11W64-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX11W64-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W64-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX11W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 15
+; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
 ; GFX11W64-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX11W64-NEXT:    v_readlane_b32 s5, v1, 31
+; GFX11W64-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_readlane_b32 s7, v1, 31
-; GFX11W64-NEXT:    v_writelane_b32 v3, s6, 16
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 63
-; GFX11W64-NEXT:    v_readlane_b32 s8, v1, 47
-; GFX11W64-NEXT:    v_writelane_b32 v3, s7, 32
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX11W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 63
+; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 47
+; GFX11W64-NEXT:    v_writelane_b32 v3, s5, 32
+; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
 ; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_writelane_b32 v3, s8, 48
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX11W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX11W64-NEXT:    v_writelane_b32 v3, s6, 48
+; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX11W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr0
-; GFX11W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX11W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX11W64-NEXT:  ; %bb.1:
 ; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W64-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11W64-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11W64-NEXT:    buffer_atomic_sub_u32 v0, off, s[8:11], 0 glc
 ; GFX11W64-NEXT:  .LBB6_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX11W64-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX11W64-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W64-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX11W64-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    global_store_b32 v4, v0, s[2:3]
+; GFX11W64-NEXT:    global_store_b32 v4, v0, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: sub_i32_varying_vdata:
 ; GFX11W32:       ; %bb.0: ; %entry
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX11W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX11W32-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX11W32-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX11W32-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX11W32-NEXT:    v_mov_b32_e32 v3, 0
-; GFX11W32-NEXT:    v_readlane_b32 s6, v1, 31
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
+; GFX11W32-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX11W32-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX11W32-NEXT:    v_readlane_b32 s5, v1, 15
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX11W32-NEXT:    v_readlane_b32 s3, v1, 15
+; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
+; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX11W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11W32-NEXT:    v_writelane_b32 v3, s5, 16
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX11W32-NEXT:    s_or_saveexec_b32 s2, -1
+; GFX11W32-NEXT:    v_writelane_b32 v3, s3, 16
+; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr0
-; GFX11W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX11W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX11W32-NEXT:  ; %bb.1:
-; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    v_mov_b32_e32 v0, s6
-; GFX11W32-NEXT:    s_mov_b32 s5, s6
+; GFX11W32-NEXT:    s_mov_b32 s3, s4
+; GFX11W32-NEXT:    s_load_b128 s[4:7], s[0:1], 0x34
+; GFX11W32-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    buffer_atomic_sub_u32 v0, off, s[8:11], 0 glc
+; GFX11W32-NEXT:    buffer_atomic_sub_u32 v0, off, s[4:7], 0 glc
 ; GFX11W32-NEXT:  .LBB6_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX11W32-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX11W32-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX11W32-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W32-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX11W32-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    global_store_b32 v4, v0, s[2:3]
+; GFX11W32-NEXT:    global_store_b32 v4, v0, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
   %old = call i32 @llvm.amdgcn.raw.buffer.atomic.sub(i32 %lane, <4 x i32> %inout, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }

llvm/test/CodeGen/AMDGPU/atomic_optimizations_struct_buffer.ll

 declare i32 @llvm.amdgcn.struct.buffer.atomic.add(i32, <4 x i32>, i32, i32, i32, i32)
 declare i32 @llvm.amdgcn.struct.buffer.atomic.sub(i32, <4 x i32>, i32, i32, i32, i32)
 
 ; Show what the atomic optimization pass will do for struct buffers.
 
 define amdgpu_kernel void @add_i32_constant(i32 addrspace(1)* %out, <4 x i32> %inout) {
 ; GFX6-LABEL: add_i32_constant:
 ; GFX6:       ; %bb.0: ; %entry
-; GFX6-NEXT:    s_mov_b64 s[6:7], exec
-; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
-; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s6, 0
-; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s7, v0
+; GFX6-NEXT:    s_mov_b64 s[4:5], exec
+; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s4, 0
+; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s5, v0
 ; GFX6-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX6-NEXT:    ; implicit-def: $vgpr1
 ; GFX6-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX6-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX6-NEXT:  ; %bb.1:
 ; GFX6-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0xd
-; GFX6-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
-; GFX6-NEXT:    s_mul_i32 s0, s0, 5
-; GFX6-NEXT:    v_mov_b32_e32 v1, s0
+; GFX6-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX6-NEXT:    s_mul_i32 s4, s4, 5
+; GFX6-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_add v1, v2, s[8:11], 0 idxen glc
 ; GFX6-NEXT:  .LBB0_2:
 ; GFX6-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
-; GFX6-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX6-NEXT:    s_mov_b32 s7, 0xf000
-; GFX6-NEXT:    v_mad_u32_u24 v0, v0, 5, s0
-; GFX6-NEXT:    s_mov_b32 s6, -1
+; GFX6-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX6-NEXT:    v_mad_u32_u24 v0, v0, 5, s4
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: add_i32_constant:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[6:7], exec
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX8-NEXT:    s_mov_b64 s[4:5], exec
+; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr1
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX8-NEXT:  ; %bb.1:
 ; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX8-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
-; GFX8-NEXT:    s_mul_i32 s0, s0, 5
-; GFX8-NEXT:    v_mov_b32_e32 v1, s0
+; GFX8-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX8-NEXT:    s_mul_i32 s4, s4, 5
+; GFX8-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    buffer_atomic_add v1, v2, s[8:11], 0 idxen glc
 ; GFX8-NEXT:  .LBB0_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX8-NEXT:    v_mad_u32_u24 v2, v0, 5, s0
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX8-NEXT:    v_mad_u32_u24 v2, v0, 5, s2
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    flat_store_dword v[0:1], v2
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: add_i32_constant:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[6:7], exec
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX9-NEXT:    s_mov_b64 s[4:5], exec
+; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr1
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX9-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
-; GFX9-NEXT:    s_mul_i32 s0, s0, 5
-; GFX9-NEXT:    v_mov_b32_e32 v1, s0
+; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX9-NEXT:    s_mul_i32 s4, s4, 5
+; GFX9-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    buffer_atomic_add v1, v2, s[8:11], 0 idxen glc
 ; GFX9-NEXT:  .LBB0_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX9-NEXT:    v_mad_u32_u24 v0, v0, 5, s0
-; GFX9-NEXT:    v_mov_b32_e32 v1, 0
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX9-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: add_i32_constant:
 ; GFX10W64:       ; %bb.0: ; %entry
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_mov_b64 s[6:7], exec
+; GFX10W64-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr1
-; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX10W64-NEXT:  ; %bb.1:
 ; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W64-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
+; GFX10W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX10W64-NEXT:    v_mov_b32_e32 v2, 0
-; GFX10W64-NEXT:    s_mul_i32 s0, s0, 5
-; GFX10W64-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10W64-NEXT:    s_mul_i32 s4, s4, 5
+; GFX10W64-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10W64-NEXT:    buffer_atomic_add v1, v2, s[8:11], 0 idxen glc
 ; GFX10W64-NEXT:  .LBB0_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W64-NEXT:    v_mad_u32_u24 v0, v0, 5, s0
+; GFX10W64-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W64-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: add_i32_constant:
 ; GFX10W32:       ; %bb.0: ; %entry
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_mov_b32 s5, exec_lo
+; GFX10W32-NEXT:    s_mov_b32 s3, exec_lo
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr1
-; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s5, 0
+; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX10W32-NEXT:  ; %bb.1:
-; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    s_bcnt1_i32_b32 s0, s5
+; GFX10W32-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x34
+; GFX10W32-NEXT:    s_bcnt1_i32_b32 s3, s3
 ; GFX10W32-NEXT:    v_mov_b32_e32 v2, 0
-; GFX10W32-NEXT:    s_mul_i32 s0, s0, 5
-; GFX10W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10W32-NEXT:    s_mul_i32 s3, s3, 5
+; GFX10W32-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    buffer_atomic_add v1, v2, s[8:11], 0 idxen glc
+; GFX10W32-NEXT:    buffer_atomic_add v1, v2, s[4:7], 0 idxen glc
 ; GFX10W32-NEXT:  .LBB0_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W32-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W32-NEXT:    v_mad_u32_u24 v0, v0, 5, s0
+; GFX10W32-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W32-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: add_i32_constant:
 ; GFX11W64:       ; %bb.0: ; %entry
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_mov_b64 s[6:7], exec
 ; GFX11W64-NEXT:    s_mov_b64 s[4:5], exec
-; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX11W64-NEXT:    s_mov_b64 s[2:3], exec
+; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX11W64-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX11W64-NEXT:  ; %bb.1:
 ; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W64-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
+; GFX11W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX11W64-NEXT:    v_mov_b32_e32 v2, 0
-; GFX11W64-NEXT:    s_mul_i32 s0, s0, 5
+; GFX11W64-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11W64-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11W64-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11W64-NEXT:    buffer_atomic_add_u32 v1, v2, s[8:11], 0 idxen glc
 ; GFX11W64-NEXT:  .LBB0_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W64-NEXT:    v_mad_u32_u24 v0, v0, 5, s0
+; GFX11W64-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11W64-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: add_i32_constant:
 ; GFX11W32:       ; %bb.0: ; %entry
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_mov_b32 s5, exec_lo
-; GFX11W32-NEXT:    s_mov_b32 s4, exec_lo
-; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s5, 0
+; GFX11W32-NEXT:    s_mov_b32 s3, exec_lo
+; GFX11W32-NEXT:    s_mov_b32 s2, exec_lo
+; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB0_2
 ; GFX11W32-NEXT:  ; %bb.1:
-; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    s_bcnt1_i32_b32 s0, s5
+; GFX11W32-NEXT:    s_load_b128 s[4:7], s[0:1], 0x34
+; GFX11W32-NEXT:    s_bcnt1_i32_b32 s3, s3
 ; GFX11W32-NEXT:    v_mov_b32_e32 v2, 0
-; GFX11W32-NEXT:    s_mul_i32 s0, s0, 5
+; GFX11W32-NEXT:    s_mul_i32 s3, s3, 5
 ; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11W32-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    buffer_atomic_add_u32 v1, v2, s[8:11], 0 idxen glc
+; GFX11W32-NEXT:    buffer_atomic_add_u32 v1, v2, s[4:7], 0 idxen glc
 ; GFX11W32-NEXT:  .LBB0_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W32-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W32-NEXT:    v_mad_u32_u24 v0, v0, 5, s0
+; GFX11W32-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11W32-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %old = call i32 @llvm.amdgcn.struct.buffer.atomic.add(i32 5, <4 x i32> %inout, i32 0, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @add_i32_uniform(i32 addrspace(1)* %out, <4 x i32> %inout, i32 %additive) {
 ; GFX6-LABEL: add_i32_uniform:
 ; GFX6:       ; %bb.0: ; %entry
-; GFX6-NEXT:    s_mov_b64 s[2:3], exec
-; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
-; GFX6-NEXT:    s_load_dword s8, s[0:1], 0x11
-; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
+; GFX6-NEXT:    s_mov_b64 s[4:5], exec
+; GFX6-NEXT:    s_load_dword s6, s[0:1], 0x11
+; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s4, 0
+; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s5, v0
 ; GFX6-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX6-NEXT:    ; implicit-def: $vgpr1
-; GFX6-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX6-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX6-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX6-NEXT:  ; %bb.1:
-; GFX6-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0xd
-; GFX6-NEXT:    s_bcnt1_i32_b64 s0, s[2:3]
+; GFX6-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0xd
+; GFX6-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    s_mul_i32 s0, s8, s0
-; GFX6-NEXT:    v_mov_b32_e32 v1, s0
+; GFX6-NEXT:    s_mul_i32 s4, s6, s4
+; GFX6-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v2, 0
-; GFX6-NEXT:    buffer_atomic_add v1, v2, s[12:15], 0 idxen glc
+; GFX6-NEXT:    buffer_atomic_add v1, v2, s[8:11], 0 idxen glc
 ; GFX6-NEXT:  .LBB1_2:
-; GFX6-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX6-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
-; GFX6-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX6-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    v_mul_lo_u32 v0, s8, v0
-; GFX6-NEXT:    s_mov_b32 s7, 0xf000
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, s0, v0
-; GFX6-NEXT:    s_mov_b32 s6, -1
-; GFX6-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX6-NEXT:    v_mul_lo_u32 v0, s6, v0
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, s4, v0
+; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: add_i32_uniform:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_load_dword s8, s[0:1], 0x44
-; GFX8-NEXT:    s_mov_b64 s[6:7], exec
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX8-NEXT:    s_load_dword s6, s[0:1], 0x44
+; GFX8-NEXT:    s_mov_b64 s[4:5], exec
+; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr1
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0x34
-; GFX8-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
+; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX8-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_mul_i32 s0, s8, s0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s0
+; GFX8-NEXT:    s_mul_i32 s4, s6, s4
+; GFX8-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0
-; GFX8-NEXT:    buffer_atomic_add v1, v2, s[12:15], 0 idxen glc
+; GFX8-NEXT:    buffer_atomic_add v1, v2, s[8:11], 0 idxen glc
 ; GFX8-NEXT:  .LBB1_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mul_lo_u32 v0, s8, v0
+; GFX8-NEXT:    v_mul_lo_u32 v0, s6, v0
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX8-NEXT:    v_add_u32_e32 v2, vcc, s0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX8-NEXT:    v_add_u32_e32 v2, vcc, s2, v0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    flat_store_dword v[0:1], v2
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: add_i32_uniform:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_load_dword s8, s[0:1], 0x44
-; GFX9-NEXT:    s_mov_b64 s[6:7], exec
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX9-NEXT:    s_load_dword s6, s[0:1], 0x44
+; GFX9-NEXT:    s_mov_b64 s[4:5], exec
+; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr1
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0x34
-; GFX9-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
+; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_mul_i32 s0, s8, s0
-; GFX9-NEXT:    v_mov_b32_e32 v1, s0
+; GFX9-NEXT:    s_mul_i32 s4, s6, s4
+; GFX9-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-NEXT:    buffer_atomic_add v1, v2, s[12:15], 0 idxen glc
+; GFX9-NEXT:    buffer_atomic_add v1, v2, s[8:11], 0 idxen glc
 ; GFX9-NEXT:  .LBB1_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mul_lo_u32 v0, s8, v0
+; GFX9-NEXT:    v_mul_lo_u32 v0, s6, v0
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX9-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-NEXT:    v_add_u32_e32 v0, s0, v0
-; GFX9-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_add_u32_e32 v0, s2, v0
+; GFX9-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: add_i32_uniform:
 ; GFX10W64:       ; %bb.0: ; %entry
-; GFX10W64-NEXT:    s_clause 0x1
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_load_dword s8, s[0:1], 0x44
-; GFX10W64-NEXT:    s_mov_b64 s[6:7], exec
+; GFX10W64-NEXT:    s_load_dword s6, s[0:1], 0x44
+; GFX10W64-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr1
-; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX10W64-NEXT:  ; %bb.1:
-; GFX10W64-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0x34
-; GFX10W64-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
+; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX10W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX10W64-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    s_mul_i32 s0, s8, s0
-; GFX10W64-NEXT:    v_mov_b32_e32 v1, s0
-; GFX10W64-NEXT:    buffer_atomic_add v1, v2, s[12:15], 0 idxen glc
+; GFX10W64-NEXT:    s_mul_i32 s4, s6, s4
+; GFX10W64-NEXT:    v_mov_b32_e32 v1, s4
+; GFX10W64-NEXT:    buffer_atomic_add v1, v2, s[8:11], 0 idxen glc
 ; GFX10W64-NEXT:  .LBB1_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], s8, v0, s[0:1]
+; GFX10W64-NEXT:    v_mad_u64_u32 v[0:1], s[2:3], s6, v0, s[2:3]
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W64-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W64-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: add_i32_uniform:
 ; GFX10W32:       ; %bb.0: ; %entry
-; GFX10W32-NEXT:    s_clause 0x1
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_load_dword s4, s[0:1], 0x44
-; GFX10W32-NEXT:    s_mov_b32 s6, exec_lo
+; GFX10W32-NEXT:    s_load_dword s2, s[0:1], 0x44
+; GFX10W32-NEXT:    s_mov_b32 s4, exec_lo
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr1
-; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s5, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX10W32-NEXT:  ; %bb.1:
 ; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    s_bcnt1_i32_b32 s0, s6
+; GFX10W32-NEXT:    s_bcnt1_i32_b32 s4, s4
 ; GFX10W32-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    s_mul_i32 s0, s4, s0
-; GFX10W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10W32-NEXT:    s_mul_i32 s4, s2, s4
+; GFX10W32-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX10W32-NEXT:    buffer_atomic_add v1, v2, s[8:11], 0 idxen glc
 ; GFX10W32-NEXT:  .LBB1_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s5
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W32-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    v_mad_u64_u32 v[0:1], s0, s4, v0, s[0:1]
+; GFX10W32-NEXT:    v_mad_u64_u32 v[0:1], s2, s2, v0, s[4:5]
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W32-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W32-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: add_i32_uniform:
 ; GFX11W64:       ; %bb.0: ; %entry
-; GFX11W64-NEXT:    s_clause 0x1
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_load_b32 s8, s[0:1], 0x44
-; GFX11W64-NEXT:    s_mov_b64 s[6:7], exec
+; GFX11W64-NEXT:    s_load_b32 s6, s[0:1], 0x44
 ; GFX11W64-NEXT:    s_mov_b64 s[4:5], exec
-; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX11W64-NEXT:    s_mov_b64 s[2:3], exec
+; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX11W64-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX11W64-NEXT:  ; %bb.1:
-; GFX11W64-NEXT:    s_load_b128 s[12:15], s[0:1], 0x34
-; GFX11W64-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
+; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
+; GFX11W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX11W64-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    s_mul_i32 s0, s8, s0
+; GFX11W64-NEXT:    s_mul_i32 s4, s6, s4
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11W64-NEXT:    v_mov_b32_e32 v1, s0
-; GFX11W64-NEXT:    buffer_atomic_add_u32 v1, v2, s[12:15], 0 idxen glc
+; GFX11W64-NEXT:    v_mov_b32_e32 v1, s4
+; GFX11W64-NEXT:    buffer_atomic_add_u32 v1, v2, s[8:11], 0 idxen glc
 ; GFX11W64-NEXT:  .LBB1_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11W64-NEXT:    v_mad_u64_u32 v[1:2], null, s8, v0, s[0:1]
+; GFX11W64-NEXT:    v_mad_u64_u32 v[1:2], null, s6, v0, s[2:3]
 ; GFX11W64-NEXT:    v_mov_b32_e32 v0, 0
-; GFX11W64-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11W64-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: add_i32_uniform:
 ; GFX11W32:       ; %bb.0: ; %entry
-; GFX11W32-NEXT:    s_clause 0x1
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_load_b32 s4, s[0:1], 0x44
-; GFX11W32-NEXT:    s_mov_b32 s6, exec_lo
-; GFX11W32-NEXT:    s_mov_b32 s5, exec_lo
-; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX11W32-NEXT:    s_load_b32 s2, s[0:1], 0x44
+; GFX11W32-NEXT:    s_mov_b32 s4, exec_lo
+; GFX11W32-NEXT:    s_mov_b32 s3, exec_lo
+; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX11W32-NEXT:  ; %bb.1:
 ; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    s_bcnt1_i32_b32 s0, s6
+; GFX11W32-NEXT:    s_bcnt1_i32_b32 s4, s4
 ; GFX11W32-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    s_mul_i32 s0, s4, s0
+; GFX11W32-NEXT:    s_mul_i32 s4, s2, s4
 ; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11W32-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX11W32-NEXT:    buffer_atomic_add_u32 v1, v2, s[8:11], 0 idxen glc
 ; GFX11W32-NEXT:  .LBB1_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s5
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W32-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11W32-NEXT:    v_mad_u64_u32 v[1:2], null, s4, v0, s[0:1]
+; GFX11W32-NEXT:    v_mad_u64_u32 v[1:2], null, s2, v0, s[4:5]
 ; GFX11W32-NEXT:    v_mov_b32_e32 v0, 0
-; GFX11W32-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11W32-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %old = call i32 @llvm.amdgcn.struct.buffer.atomic.add(i32 %additive, <4 x i32> %inout, i32 0, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @add_i32_varying_vdata(i32 addrspace(1)* %out, <4 x i32> %inout) {
 ; GFX6-LABEL: add_i32_varying_vdata:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0xd
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
 ; GFX6-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_add v0, v1, s[4:7], 0 idxen glc
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: add_i32_varying_vdata:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
-; GFX8-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX8-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX8-NEXT:    s_not_b64 exec, exec
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX8-NEXT:    s_not_b64 exec, exec
-; GFX8-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX8-NEXT:    v_readlane_b32 s6, v2, 63
+; GFX8-NEXT:    v_readlane_b32 s4, v2, 63
 ; GFX8-NEXT:    s_nop 0
 ; GFX8-NEXT:    v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf
-; GFX8-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
 ; GFX8-NEXT:    ; implicit-def: $vgpr0
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX8-NEXT:  ; %bb.1:
 ; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX8-NEXT:    v_mov_b32_e32 v0, s6
+; GFX8-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    buffer_atomic_add v0, v3, s[8:11], 0 idxen glc
 ; GFX8-NEXT:  .LBB2_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, v1
+; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v4, s3
-; GFX8-NEXT:    v_add_u32_e32 v0, vcc, s0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s2
+; GFX8-NEXT:    v_mov_b32_e32 v4, s1
+; GFX8-NEXT:    v_mov_b32_e32 v3, s0
 ; GFX8-NEXT:    flat_store_dword v[3:4], v0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: add_i32_varying_vdata:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX9-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX9-NEXT:    s_not_b64 exec, exec
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    s_not_b64 exec, exec
-; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-NEXT:    v_readlane_b32 s6, v2, 63
+; GFX9-NEXT:    v_readlane_b32 s4, v2, 63
 ; GFX9-NEXT:    s_nop 0
 ; GFX9-NEXT:    v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf
-; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
 ; GFX9-NEXT:    ; implicit-def: $vgpr0
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX9-NEXT:    v_mov_b32_e32 v0, s6
+; GFX9-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX9-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    buffer_atomic_add v0, v3, s[8:11], 0 idxen glc
 ; GFX9-NEXT:  .LBB2_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v1
-; GFX9-NEXT:    v_add_u32_e32 v0, s0, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, 0
+; GFX9-NEXT:    v_add_u32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    global_store_dword v3, v0, s[2:3]
+; GFX9-NEXT:    global_store_dword v3, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: add_i32_varying_vdata:
 ; GFX10W64:       ; %bb.0: ; %entry
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX10W64-NEXT:    s_not_b64 exec, exec
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10W64-NEXT:    s_not_b64 exec, exec
 ; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX10W64-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX10W64-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 15
 ; GFX10W64-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX10W64-NEXT:    v_readlane_b32 s5, v1, 31
+; GFX10W64-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_readlane_b32 s7, v1, 31
-; GFX10W64-NEXT:    v_writelane_b32 v3, s6, 16
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
 ; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 63
-; GFX10W64-NEXT:    v_readlane_b32 s8, v1, 47
-; GFX10W64-NEXT:    v_writelane_b32 v3, s7, 32
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 63
+; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 47
+; GFX10W64-NEXT:    v_writelane_b32 v3, s5, 32
+; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_writelane_b32 v3, s8, 48
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX10W64-NEXT:    v_writelane_b32 v3, s6, 48
+; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX10W64-NEXT:  ; %bb.1:
 ; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W64-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10W64-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX10W64-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10W64-NEXT:    buffer_atomic_add v0, v4, s[8:11], 0 idxen glc
 ; GFX10W64-NEXT:  .LBB2_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX10W64-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX10W64-NEXT:    v_mov_b32_e32 v4, 0
-; GFX10W64-NEXT:    v_add_nc_u32_e32 v0, s0, v0
+; GFX10W64-NEXT:    v_add_nc_u32_e32 v0, s2, v0
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    global_store_dword v4, v0, s[2:3]
+; GFX10W64-NEXT:    global_store_dword v4, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: add_i32_varying_vdata:
 ; GFX10W32:       ; %bb.0: ; %entry
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX10W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX10W32-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX10W32-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX10W32-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_or_saveexec_b32 s4, -1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX10W32-NEXT:    v_mov_b32_e32 v3, 0
-; GFX10W32-NEXT:    v_readlane_b32 s6, v1, 31
+; GFX10W32-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX10W32-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX10W32-NEXT:    v_readlane_b32 s5, v1, 15
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10W32-NEXT:    v_readlane_b32 s3, v1, 15
+; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX10W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX10W32-NEXT:    v_writelane_b32 v3, s5, 16
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10W32-NEXT:    s_or_saveexec_b32 s2, -1
+; GFX10W32-NEXT:    v_writelane_b32 v3, s3, 16
+; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX10W32-NEXT:  ; %bb.1:
-; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10W32-NEXT:    s_mov_b32 s3, s4
+; GFX10W32-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x34
+; GFX10W32-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX10W32-NEXT:    v_mov_b32_e32 v4, 0
-; GFX10W32-NEXT:    s_mov_b32 s5, s6
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    buffer_atomic_add v0, v4, s[8:11], 0 idxen glc
+; GFX10W32-NEXT:    buffer_atomic_add v0, v4, s[4:7], 0 idxen glc
 ; GFX10W32-NEXT:  .LBB2_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10W32-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX10W32-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX10W32-NEXT:    v_mov_b32_e32 v4, 0
-; GFX10W32-NEXT:    v_add_nc_u32_e32 v0, s0, v0
+; GFX10W32-NEXT:    v_add_nc_u32_e32 v0, s2, v0
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    global_store_dword v4, v0, s[2:3]
+; GFX10W32-NEXT:    global_store_dword v4, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: add_i32_varying_vdata:
 ; GFX11W64:       ; %bb.0: ; %entry
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX11W64-NEXT:    s_not_b64 exec, exec
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W64-NEXT:    s_not_b64 exec, exec

 ; GFX11W64-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX11W64-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W64-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX11W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 15
+; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
 ; GFX11W64-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX11W64-NEXT:    v_readlane_b32 s5, v1, 31
+; GFX11W64-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_readlane_b32 s7, v1, 31
-; GFX11W64-NEXT:    v_writelane_b32 v3, s6, 16
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 63
-; GFX11W64-NEXT:    v_readlane_b32 s8, v1, 47
-; GFX11W64-NEXT:    v_writelane_b32 v3, s7, 32
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX11W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 63
+; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 47
+; GFX11W64-NEXT:    v_writelane_b32 v3, s5, 32
+; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
 ; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_writelane_b32 v3, s8, 48
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX11W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX11W64-NEXT:    v_writelane_b32 v3, s6, 48
+; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX11W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr0
-; GFX11W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX11W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX11W64-NEXT:  ; %bb.1:
 ; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W64-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11W64-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX11W64-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11W64-NEXT:    buffer_atomic_add_u32 v0, v4, s[8:11], 0 idxen glc
 ; GFX11W64-NEXT:  .LBB2_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX11W64-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX11W64-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W64-NEXT:    v_add_nc_u32_e32 v0, s0, v0
+; GFX11W64-NEXT:    v_add_nc_u32_e32 v0, s2, v0
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    global_store_b32 v4, v0, s[2:3]
+; GFX11W64-NEXT:    global_store_b32 v4, v0, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: add_i32_varying_vdata:
 ; GFX11W32:       ; %bb.0: ; %entry
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX11W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX11W32-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX11W32-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX11W32-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX11W32-NEXT:    v_mov_b32_e32 v3, 0
-; GFX11W32-NEXT:    v_readlane_b32 s6, v1, 31
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
+; GFX11W32-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX11W32-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX11W32-NEXT:    v_readlane_b32 s5, v1, 15
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX11W32-NEXT:    v_readlane_b32 s3, v1, 15
+; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
+; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX11W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11W32-NEXT:    v_writelane_b32 v3, s5, 16
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX11W32-NEXT:    s_or_saveexec_b32 s2, -1
+; GFX11W32-NEXT:    v_writelane_b32 v3, s3, 16
+; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr0
-; GFX11W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX11W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB2_2
 ; GFX11W32-NEXT:  ; %bb.1:
-; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11W32-NEXT:    s_mov_b32 s3, s4
+; GFX11W32-NEXT:    s_load_b128 s[4:7], s[0:1], 0x34
+; GFX11W32-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX11W32-NEXT:    v_mov_b32_e32 v4, 0
-; GFX11W32-NEXT:    s_mov_b32 s5, s6
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    buffer_atomic_add_u32 v0, v4, s[8:11], 0 idxen glc
+; GFX11W32-NEXT:    buffer_atomic_add_u32 v0, v4, s[4:7], 0 idxen glc
 ; GFX11W32-NEXT:  .LBB2_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX11W32-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX11W32-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX11W32-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W32-NEXT:    v_add_nc_u32_e32 v0, s0, v0
+; GFX11W32-NEXT:    v_add_nc_u32_e32 v0, s2, v0
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    global_store_b32 v4, v0, s[2:3]
+; GFX11W32-NEXT:    global_store_b32 v4, v0, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
   %old = call i32 @llvm.amdgcn.struct.buffer.atomic.add(i32 %lane, <4 x i32> %inout, i32 0, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
entry:
   %old = call i32 @llvm.amdgcn.struct.buffer.atomic.add(i32 1, <4 x i32> %inout, i32 0, i32 %lane, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @sub_i32_constant(i32 addrspace(1)* %out, <4 x i32> %inout) {
 ; GFX6-LABEL: sub_i32_constant:
 ; GFX6:       ; %bb.0: ; %entry
-; GFX6-NEXT:    s_mov_b64 s[6:7], exec
-; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
-; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s6, 0
-; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s7, v0
+; GFX6-NEXT:    s_mov_b64 s[4:5], exec
+; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s4, 0
+; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s5, v0
 ; GFX6-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX6-NEXT:    ; implicit-def: $vgpr1
 ; GFX6-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX6-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX6-NEXT:  ; %bb.1:
 ; GFX6-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0xd
-; GFX6-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
-; GFX6-NEXT:    s_mul_i32 s0, s0, 5
-; GFX6-NEXT:    v_mov_b32_e32 v1, s0
+; GFX6-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX6-NEXT:    s_mul_i32 s4, s4, 5
+; GFX6-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_sub v1, v2, s[8:11], 0 idxen glc
 ; GFX6-NEXT:  .LBB5_2:
 ; GFX6-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
-; GFX6-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX6-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX6-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
-; GFX6-NEXT:    s_mov_b32 s7, 0xf000
-; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, s0, v0
-; GFX6-NEXT:    s_mov_b32 s6, -1
+; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, s4, v0
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: sub_i32_constant:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[6:7], exec
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX8-NEXT:    s_mov_b64 s[4:5], exec
+; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr1
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX8-NEXT:  ; %bb.1:
 ; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX8-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
-; GFX8-NEXT:    s_mul_i32 s0, s0, 5
-; GFX8-NEXT:    v_mov_b32_e32 v1, s0
+; GFX8-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX8-NEXT:    s_mul_i32 s4, s4, 5
+; GFX8-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    buffer_atomic_sub v1, v2, s[8:11], 0 idxen glc
 ; GFX8-NEXT:  .LBB5_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX8-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
-; GFX8-NEXT:    v_sub_u32_e32 v2, vcc, s0, v0
+; GFX8-NEXT:    v_sub_u32_e32 v2, vcc, s2, v0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    flat_store_dword v[0:1], v2
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sub_i32_constant:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[6:7], exec
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX9-NEXT:    s_mov_b64 s[4:5], exec
+; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr1
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX9-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
-; GFX9-NEXT:    s_mul_i32 s0, s0, 5
-; GFX9-NEXT:    v_mov_b32_e32 v1, s0
+; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX9-NEXT:    s_mul_i32 s4, s4, 5
+; GFX9-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    buffer_atomic_sub v1, v2, s[8:11], 0 idxen glc
 ; GFX9-NEXT:  .LBB5_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX9-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
-; GFX9-NEXT:    v_sub_u32_e32 v0, s0, v0
-; GFX9-NEXT:    v_mov_b32_e32 v1, 0
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_sub_u32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX9-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: sub_i32_constant:
 ; GFX10W64:       ; %bb.0: ; %entry
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_mov_b64 s[6:7], exec
+; GFX10W64-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr1
-; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX10W64-NEXT:  ; %bb.1:
 ; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W64-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
+; GFX10W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX10W64-NEXT:    v_mov_b32_e32 v2, 0
-; GFX10W64-NEXT:    s_mul_i32 s0, s0, 5
-; GFX10W64-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10W64-NEXT:    s_mul_i32 s4, s4, 5
+; GFX10W64-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10W64-NEXT:    buffer_atomic_sub v1, v2, s[8:11], 0 idxen glc
 ; GFX10W64-NEXT:  .LBB5_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W64-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W64-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX10W64-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W64-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: sub_i32_constant:
 ; GFX10W32:       ; %bb.0: ; %entry
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_mov_b32 s5, exec_lo
+; GFX10W32-NEXT:    s_mov_b32 s3, exec_lo
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr1
-; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s5, 0
+; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX10W32-NEXT:  ; %bb.1:
-; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    s_bcnt1_i32_b32 s0, s5
+; GFX10W32-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x34
+; GFX10W32-NEXT:    s_bcnt1_i32_b32 s3, s3
 ; GFX10W32-NEXT:    v_mov_b32_e32 v2, 0
-; GFX10W32-NEXT:    s_mul_i32 s0, s0, 5
-; GFX10W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10W32-NEXT:    s_mul_i32 s3, s3, 5
+; GFX10W32-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    buffer_atomic_sub v1, v2, s[8:11], 0 idxen glc
+; GFX10W32-NEXT:    buffer_atomic_sub v1, v2, s[4:7], 0 idxen glc
 ; GFX10W32-NEXT:  .LBB5_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W32-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W32-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W32-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX10W32-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W32-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: sub_i32_constant:
 ; GFX11W64:       ; %bb.0: ; %entry
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_mov_b64 s[6:7], exec
 ; GFX11W64-NEXT:    s_mov_b64 s[4:5], exec
-; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX11W64-NEXT:    s_mov_b64 s[2:3], exec
+; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX11W64-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX11W64-NEXT:  ; %bb.1:
 ; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W64-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
+; GFX11W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX11W64-NEXT:    v_mov_b32_e32 v2, 0
-; GFX11W64-NEXT:    s_mul_i32 s0, s0, 5
+; GFX11W64-NEXT:    s_mul_i32 s4, s4, 5
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11W64-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11W64-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11W64-NEXT:    buffer_atomic_sub_u32 v1, v2, s[8:11], 0 idxen glc
 ; GFX11W64-NEXT:  .LBB5_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W64-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W64-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX11W64-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11W64-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: sub_i32_constant:
 ; GFX11W32:       ; %bb.0: ; %entry
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_mov_b32 s5, exec_lo
-; GFX11W32-NEXT:    s_mov_b32 s4, exec_lo
-; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s5, 0
+; GFX11W32-NEXT:    s_mov_b32 s3, exec_lo
+; GFX11W32-NEXT:    s_mov_b32 s2, exec_lo
+; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX11W32-NEXT:  ; %bb.1:
-; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    s_bcnt1_i32_b32 s0, s5
+; GFX11W32-NEXT:    s_load_b128 s[4:7], s[0:1], 0x34
+; GFX11W32-NEXT:    s_bcnt1_i32_b32 s3, s3
 ; GFX11W32-NEXT:    v_mov_b32_e32 v2, 0
-; GFX11W32-NEXT:    s_mul_i32 s0, s0, 5
+; GFX11W32-NEXT:    s_mul_i32 s3, s3, 5
 ; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11W32-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    buffer_atomic_sub_u32 v1, v2, s[8:11], 0 idxen glc
+; GFX11W32-NEXT:    buffer_atomic_sub_u32 v1, v2, s[4:7], 0 idxen glc
 ; GFX11W32-NEXT:  .LBB5_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W32-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W32-NEXT:    v_mul_u32_u24_e32 v0, 5, v0
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W32-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX11W32-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11W32-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %old = call i32 @llvm.amdgcn.struct.buffer.atomic.sub(i32 5, <4 x i32> %inout, i32 0, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @sub_i32_uniform(i32 addrspace(1)* %out, <4 x i32> %inout, i32 %subitive) {
 ; GFX6-LABEL: sub_i32_uniform:
 ; GFX6:       ; %bb.0: ; %entry
-; GFX6-NEXT:    s_mov_b64 s[2:3], exec
-; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
-; GFX6-NEXT:    s_load_dword s8, s[0:1], 0x11
-; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
+; GFX6-NEXT:    s_mov_b64 s[4:5], exec
+; GFX6-NEXT:    s_load_dword s6, s[0:1], 0x11
+; GFX6-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s4, 0
+; GFX6-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s5, v0
 ; GFX6-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX6-NEXT:    ; implicit-def: $vgpr1
-; GFX6-NEXT:    s_and_saveexec_b64 s[6:7], vcc
+; GFX6-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX6-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX6-NEXT:  ; %bb.1:
-; GFX6-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0xd
-; GFX6-NEXT:    s_bcnt1_i32_b64 s0, s[2:3]
+; GFX6-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0xd
+; GFX6-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    s_mul_i32 s0, s8, s0
-; GFX6-NEXT:    v_mov_b32_e32 v1, s0
+; GFX6-NEXT:    s_mul_i32 s4, s6, s4
+; GFX6-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v2, 0
-; GFX6-NEXT:    buffer_atomic_sub v1, v2, s[12:15], 0 idxen glc
+; GFX6-NEXT:    buffer_atomic_sub v1, v2, s[8:11], 0 idxen glc
 ; GFX6-NEXT:  .LBB6_2:
-; GFX6-NEXT:    s_or_b64 exec, exec, s[6:7]
+; GFX6-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
-; GFX6-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX6-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    v_mul_lo_u32 v0, s8, v0
-; GFX6-NEXT:    s_mov_b32 s7, 0xf000
-; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, s0, v0
-; GFX6-NEXT:    s_mov_b32 s6, -1
-; GFX6-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX6-NEXT:    v_mul_lo_u32 v0, s6, v0
+; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, s4, v0
+; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: sub_i32_uniform:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_load_dword s8, s[0:1], 0x44
-; GFX8-NEXT:    s_mov_b64 s[6:7], exec
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX8-NEXT:    s_load_dword s6, s[0:1], 0x44
+; GFX8-NEXT:    s_mov_b64 s[4:5], exec
+; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX8-NEXT:    ; implicit-def: $vgpr1
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0x34
-; GFX8-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
+; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX8-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_mul_i32 s0, s8, s0
-; GFX8-NEXT:    v_mov_b32_e32 v1, s0
+; GFX8-NEXT:    s_mul_i32 s4, s6, s4
+; GFX8-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0
-; GFX8-NEXT:    buffer_atomic_sub v1, v2, s[12:15], 0 idxen glc
+; GFX8-NEXT:    buffer_atomic_sub v1, v2, s[8:11], 0 idxen glc
 ; GFX8-NEXT:  .LBB6_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mul_lo_u32 v0, s8, v0
+; GFX8-NEXT:    v_mul_lo_u32 v0, s6, v0
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX8-NEXT:    v_sub_u32_e32 v2, vcc, s0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v0, s2
-; GFX8-NEXT:    v_mov_b32_e32 v1, s3
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX8-NEXT:    v_sub_u32_e32 v2, vcc, s2, v0
+; GFX8-NEXT:    v_mov_b32_e32 v0, s0
+; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    flat_store_dword v[0:1], v2
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sub_i32_uniform:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_load_dword s8, s[0:1], 0x44
-; GFX9-NEXT:    s_mov_b64 s[6:7], exec
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX9-NEXT:    s_load_dword s6, s[0:1], 0x44
+; GFX9-NEXT:    s_mov_b64 s[4:5], exec
+; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX9-NEXT:    ; implicit-def: $vgpr1
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0x34
-; GFX9-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
+; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_mul_i32 s0, s8, s0
-; GFX9-NEXT:    v_mov_b32_e32 v1, s0
+; GFX9-NEXT:    s_mul_i32 s4, s6, s4
+; GFX9-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-NEXT:    buffer_atomic_sub v1, v2, s[12:15], 0 idxen glc
+; GFX9-NEXT:    buffer_atomic_sub v1, v2, s[8:11], 0 idxen glc
 ; GFX9-NEXT:  .LBB6_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mul_lo_u32 v0, s8, v0
+; GFX9-NEXT:    v_mul_lo_u32 v0, s6, v0
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v1
-; GFX9-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-NEXT:    v_sub_u32_e32 v0, s0, v0
-; GFX9-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_sub_u32_e32 v0, s2, v0
+; GFX9-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: sub_i32_uniform:
 ; GFX10W64:       ; %bb.0: ; %entry
-; GFX10W64-NEXT:    s_clause 0x1
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_load_dword s8, s[0:1], 0x44
-; GFX10W64-NEXT:    s_mov_b64 s[6:7], exec
+; GFX10W64-NEXT:    s_load_dword s6, s[0:1], 0x44
+; GFX10W64-NEXT:    s_mov_b64 s[4:5], exec
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr1
-; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX10W64-NEXT:  ; %bb.1:
-; GFX10W64-NEXT:    s_load_dwordx4 s[12:15], s[0:1], 0x34
-; GFX10W64-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
+; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX10W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX10W64-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    s_mul_i32 s0, s8, s0
-; GFX10W64-NEXT:    v_mov_b32_e32 v1, s0
-; GFX10W64-NEXT:    buffer_atomic_sub v1, v2, s[12:15], 0 idxen glc
+; GFX10W64-NEXT:    s_mul_i32 s4, s6, s4
+; GFX10W64-NEXT:    v_mov_b32_e32 v1, s4
+; GFX10W64-NEXT:    buffer_atomic_sub v1, v2, s[8:11], 0 idxen glc
 ; GFX10W64-NEXT:  .LBB6_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    v_mul_lo_u32 v0, s8, v0
+; GFX10W64-NEXT:    v_mul_lo_u32 v0, s6, v0
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W64-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
-; GFX10W64-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W64-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
+; GFX10W64-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: sub_i32_uniform:
 ; GFX10W32:       ; %bb.0: ; %entry
-; GFX10W32-NEXT:    s_clause 0x1
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_load_dword s4, s[0:1], 0x44
-; GFX10W32-NEXT:    s_mov_b32 s6, exec_lo
+; GFX10W32-NEXT:    s_load_dword s2, s[0:1], 0x44
+; GFX10W32-NEXT:    s_mov_b32 s4, exec_lo
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr1
-; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s5, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX10W32-NEXT:  ; %bb.1:
 ; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    s_bcnt1_i32_b32 s0, s6
+; GFX10W32-NEXT:    s_bcnt1_i32_b32 s4, s4
 ; GFX10W32-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    s_mul_i32 s0, s4, s0
-; GFX10W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX10W32-NEXT:    s_mul_i32 s4, s2, s4
+; GFX10W32-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX10W32-NEXT:    buffer_atomic_sub v1, v2, s[8:11], 0 idxen glc
 ; GFX10W32-NEXT:  .LBB6_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s5
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    v_mul_lo_u32 v0, s4, v0
+; GFX10W32-NEXT:    v_mul_lo_u32 v0, s2, v0
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10W32-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10W32-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
-; GFX10W32-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10W32-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
+; GFX10W32-NEXT:    global_store_dword v1, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: sub_i32_uniform:
 ; GFX11W64:       ; %bb.0: ; %entry
-; GFX11W64-NEXT:    s_clause 0x1
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_load_b32 s8, s[0:1], 0x44
-; GFX11W64-NEXT:    s_mov_b64 s[6:7], exec
+; GFX11W64-NEXT:    s_load_b32 s6, s[0:1], 0x44
 ; GFX11W64-NEXT:    s_mov_b64 s[4:5], exec
-; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX11W64-NEXT:    s_mov_b64 s[2:3], exec
+; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s7, v0
+; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
 ; GFX11W64-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX11W64-NEXT:  ; %bb.1:
-; GFX11W64-NEXT:    s_load_b128 s[12:15], s[0:1], 0x34
-; GFX11W64-NEXT:    s_bcnt1_i32_b64 s0, s[6:7]
+; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
+; GFX11W64-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX11W64-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    s_mul_i32 s0, s8, s0
+; GFX11W64-NEXT:    s_mul_i32 s4, s6, s4
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11W64-NEXT:    v_mov_b32_e32 v1, s0
-; GFX11W64-NEXT:    buffer_atomic_sub_u32 v1, v2, s[12:15], 0 idxen glc
+; GFX11W64-NEXT:    v_mov_b32_e32 v1, s4
+; GFX11W64-NEXT:    buffer_atomic_sub_u32 v1, v2, s[8:11], 0 idxen glc
 ; GFX11W64-NEXT:  .LBB6_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    v_mul_lo_u32 v0, s8, v0
+; GFX11W64-NEXT:    v_mul_lo_u32 v0, s6, v0
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W64-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
-; GFX11W64-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11W64-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
+; GFX11W64-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: sub_i32_uniform:
 ; GFX11W32:       ; %bb.0: ; %entry
-; GFX11W32-NEXT:    s_clause 0x1
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_load_b32 s4, s[0:1], 0x44
-; GFX11W32-NEXT:    s_mov_b32 s6, exec_lo
-; GFX11W32-NEXT:    s_mov_b32 s5, exec_lo
-; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s6, 0
+; GFX11W32-NEXT:    s_load_b32 s2, s[0:1], 0x44
+; GFX11W32-NEXT:    s_mov_b32 s4, exec_lo
+; GFX11W32-NEXT:    s_mov_b32 s3, exec_lo
+; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_cmpx_eq_u32_e32 0, v0
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX11W32-NEXT:  ; %bb.1:
 ; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    s_bcnt1_i32_b32 s0, s6
+; GFX11W32-NEXT:    s_bcnt1_i32_b32 s4, s4
 ; GFX11W32-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    s_mul_i32 s0, s4, s0
+; GFX11W32-NEXT:    s_mul_i32 s4, s2, s4
 ; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11W32-NEXT:    v_mov_b32_e32 v1, s0
+; GFX11W32-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX11W32-NEXT:    buffer_atomic_sub_u32 v1, v2, s[8:11], 0 idxen glc
 ; GFX11W32-NEXT:  .LBB6_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s5
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s3
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    v_mul_lo_u32 v0, s4, v0
+; GFX11W32-NEXT:    v_mul_lo_u32 v0, s2, v0
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX11W32-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W32-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
-; GFX11W32-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11W32-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
+; GFX11W32-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %old = call i32 @llvm.amdgcn.struct.buffer.atomic.sub(i32 %subitive, <4 x i32> %inout, i32 0, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @sub_i32_varying_vdata(i32 addrspace(1)* %out, <4 x i32> %inout) {
 ; GFX6-LABEL: sub_i32_varying_vdata:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0xd
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
 ; GFX6-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_sub v0, v1, s[4:7], 0 idxen glc
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX8-LABEL: sub_i32_varying_vdata:
 ; GFX8:       ; %bb.0: ; %entry
-; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX8-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0
-; GFX8-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX8-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX8-NEXT:    s_not_b64 exec, exec
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX8-NEXT:    s_not_b64 exec, exec
-; GFX8-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX8-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
 ; GFX8-NEXT:    s_nop 1
 ; GFX8-NEXT:    v_add_u32_dpp v2, vcc, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX8-NEXT:    v_readlane_b32 s6, v2, 63
+; GFX8-NEXT:    v_readlane_b32 s4, v2, 63
 ; GFX8-NEXT:    s_nop 0
 ; GFX8-NEXT:    v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf
-; GFX8-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX8-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
 ; GFX8-NEXT:    ; implicit-def: $vgpr0
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX8-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX8-NEXT:  ; %bb.1:
 ; GFX8-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX8-NEXT:    v_mov_b32_e32 v0, s6
+; GFX8-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    buffer_atomic_sub v0, v3, s[8:11], 0 idxen glc
 ; GFX8-NEXT:  .LBB7_2:
-; GFX8-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX8-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, v1
+; GFX8-NEXT:    v_sub_u32_e32 v0, vcc, s2, v0
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v4, s3
-; GFX8-NEXT:    v_sub_u32_e32 v0, vcc, s0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v3, s2
+; GFX8-NEXT:    v_mov_b32_e32 v4, s1
+; GFX8-NEXT:    v_mov_b32_e32 v3, s0
 ; GFX8-NEXT:    flat_store_dword v[3:4], v0
 ; GFX8-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sub_i32_varying_vdata:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
 ; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
 ; GFX9-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX9-NEXT:    s_not_b64 exec, exec
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    s_not_b64 exec, exec
-; GFX9-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GFX9-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
 ; GFX9-NEXT:    s_nop 1
 ; GFX9-NEXT:    v_add_u32_dpp v2, v2, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-NEXT:    v_readlane_b32 s6, v2, 63
+; GFX9-NEXT:    v_readlane_b32 s4, v2, 63
 ; GFX9-NEXT:    s_nop 0
 ; GFX9-NEXT:    v_mov_b32_dpp v1, v2 wave_shr:1 row_mask:0xf bank_mask:0xf
-; GFX9-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX9-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
 ; GFX9-NEXT:    ; implicit-def: $vgpr0
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX9-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX9-NEXT:  ; %bb.1:
 ; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX9-NEXT:    v_mov_b32_e32 v0, s6
+; GFX9-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX9-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    buffer_atomic_sub v0, v3, s[8:11], 0 idxen glc
 ; GFX9-NEXT:  .LBB7_2:
-; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v1
-; GFX9-NEXT:    v_sub_u32_e32 v0, s0, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, 0
+; GFX9-NEXT:    v_sub_u32_e32 v0, s2, v0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    global_store_dword v3, v0, s[2:3]
+; GFX9-NEXT:    global_store_dword v3, v0, s[0:1]
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10W64-LABEL: sub_i32_varying_vdata:
 ; GFX10W64:       ; %bb.0: ; %entry
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX10W64-NEXT:    s_not_b64 exec, exec
 ; GFX10W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10W64-NEXT:    s_not_b64 exec, exec
 ; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W64-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX10W64-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX10W64-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX10W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 15
 ; GFX10W64-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX10W64-NEXT:    v_readlane_b32 s5, v1, 31
+; GFX10W64-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX10W64-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_readlane_b32 s7, v1, 31
-; GFX10W64-NEXT:    v_writelane_b32 v3, s6, 16
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
 ; GFX10W64-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 63
-; GFX10W64-NEXT:    v_readlane_b32 s8, v1, 47
-; GFX10W64-NEXT:    v_writelane_b32 v3, s7, 32
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX10W64-NEXT:    v_readlane_b32 s4, v1, 63
+; GFX10W64-NEXT:    v_readlane_b32 s6, v1, 47
+; GFX10W64-NEXT:    v_writelane_b32 v3, s5, 32
+; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX10W64-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX10W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX10W64-NEXT:    v_writelane_b32 v3, s8, 48
-; GFX10W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX10W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX10W64-NEXT:    v_writelane_b32 v3, s6, 48
+; GFX10W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX10W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX10W64-NEXT:    ; implicit-def: $vgpr0
-; GFX10W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX10W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX10W64-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX10W64-NEXT:  ; %bb.1:
 ; GFX10W64-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W64-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10W64-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX10W64-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10W64-NEXT:    buffer_atomic_sub v0, v4, s[8:11], 0 idxen glc
 ; GFX10W64-NEXT:  .LBB7_2:
 ; GFX10W64-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX10W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX10W64-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W64-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10W64-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX10W64-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX10W64-NEXT:    v_mov_b32_e32 v4, 0
-; GFX10W64-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX10W64-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX10W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W64-NEXT:    global_store_dword v4, v0, s[2:3]
+; GFX10W64-NEXT:    global_store_dword v4, v0, s[0:1]
 ; GFX10W64-NEXT:    s_endpgm
 ;
 ; GFX10W32-LABEL: sub_i32_varying_vdata:
 ; GFX10W32:       ; %bb.0: ; %entry
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX10W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX10W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX10W32-NEXT:    s_or_saveexec_b32 s2, -1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX10W32-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX10W32-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX10W32-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
-; GFX10W32-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10W32-NEXT:    s_or_saveexec_b32 s4, -1
 ; GFX10W32-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX10W32-NEXT:    v_mov_b32_e32 v3, 0
-; GFX10W32-NEXT:    v_readlane_b32 s6, v1, 31
+; GFX10W32-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX10W32-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX10W32-NEXT:    v_readlane_b32 s5, v1, 15
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10W32-NEXT:    v_readlane_b32 s3, v1, 15
+; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX10W32-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX10W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX10W32-NEXT:    v_writelane_b32 v3, s5, 16
-; GFX10W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX10W32-NEXT:    s_or_saveexec_b32 s2, -1
+; GFX10W32-NEXT:    v_writelane_b32 v3, s3, 16
+; GFX10W32-NEXT:    s_mov_b32 exec_lo, s2
 ; GFX10W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX10W32-NEXT:    ; implicit-def: $vgpr0
-; GFX10W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX10W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX10W32-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX10W32-NEXT:  ; %bb.1:
-; GFX10W32-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
-; GFX10W32-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10W32-NEXT:    s_mov_b32 s3, s4
+; GFX10W32-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x34
+; GFX10W32-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX10W32-NEXT:    v_mov_b32_e32 v4, 0
-; GFX10W32-NEXT:    s_mov_b32 s5, s6
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    buffer_atomic_sub v0, v4, s[8:11], 0 idxen glc
+; GFX10W32-NEXT:    buffer_atomic_sub v0, v4, s[4:7], 0 idxen glc
 ; GFX10W32-NEXT:  .LBB7_2:
 ; GFX10W32-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX10W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX10W32-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX10W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX10W32-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10W32-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX10W32-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX10W32-NEXT:    v_mov_b32_e32 v4, 0
-; GFX10W32-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX10W32-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX10W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10W32-NEXT:    global_store_dword v4, v0, s[2:3]
+; GFX10W32-NEXT:    global_store_dword v4, v0, s[0:1]
 ; GFX10W32-NEXT:    s_endpgm
 ;
 ; GFX11W64-LABEL: sub_i32_varying_vdata:
 ; GFX11W64:       ; %bb.0: ; %entry
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX11W64-NEXT:    s_not_b64 exec, exec
 ; GFX11W64-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W64-NEXT:    s_not_b64 exec, exec

 ; GFX11W64-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX11W64-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
 ; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W64-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX11W64-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xc bank_mask:0xf
-; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 15
+; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
 ; GFX11W64-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
+; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 15
+; GFX11W64-NEXT:    v_readlane_b32 s5, v1, 31
+; GFX11W64-NEXT:    v_writelane_b32 v3, s4, 16
 ; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX11W64-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_readlane_b32 s7, v1, 31
-; GFX11W64-NEXT:    v_writelane_b32 v3, s6, 16
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11W64-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 63
-; GFX11W64-NEXT:    v_readlane_b32 s8, v1, 47
-; GFX11W64-NEXT:    v_writelane_b32 v3, s7, 32
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX11W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX11W64-NEXT:    v_readlane_b32 s4, v1, 63
+; GFX11W64-NEXT:    v_readlane_b32 s6, v1, 47
+; GFX11W64-NEXT:    v_writelane_b32 v3, s5, 32
+; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX11W64-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
 ; GFX11W64-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX11W64-NEXT:    s_or_saveexec_b64 s[4:5], -1
-; GFX11W64-NEXT:    v_writelane_b32 v3, s8, 48
-; GFX11W64-NEXT:    s_mov_b64 exec, s[4:5]
+; GFX11W64-NEXT:    s_or_saveexec_b64 s[2:3], -1
+; GFX11W64-NEXT:    v_writelane_b32 v3, s6, 48
+; GFX11W64-NEXT:    s_mov_b64 exec, s[2:3]
 ; GFX11W64-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX11W64-NEXT:    ; implicit-def: $vgpr0
-; GFX11W64-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX11W64-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX11W64-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX11W64-NEXT:  ; %bb.1:
 ; GFX11W64-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W64-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11W64-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX11W64-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11W64-NEXT:    buffer_atomic_sub_u32 v0, v4, s[8:11], 0 idxen glc
 ; GFX11W64-NEXT:  .LBB7_2:
-; GFX11W64-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX11W64-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX11W64-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W64-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W64-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX11W64-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX11W64-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX11W64-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX11W64-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W64-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX11W64-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX11W64-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W64-NEXT:    global_store_b32 v4, v0, s[2:3]
+; GFX11W64-NEXT:    global_store_b32 v4, v0, s[0:1]
 ; GFX11W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W64-NEXT:    s_endpgm
 ;
 ; GFX11W32-LABEL: sub_i32_varying_vdata:
 ; GFX11W32:       ; %bb.0: ; %entry
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX11W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX11W32-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11W32-NEXT:    s_not_b32 exec_lo, exec_lo
 ; GFX11W32-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX11W32-NEXT:    v_mov_b32_e32 v3, 0
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_mov_b32_e32 v2, v1
 ; GFX11W32-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
-; GFX11W32-NEXT:    s_load_b64 s[2:3], s[0:1], 0x24
-; GFX11W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11W32-NEXT:    v_add_nc_u32_dpp v1, v2, v1 quad_perm:[0,1,2,3] row_mask:0xa bank_mask:0xf
-; GFX11W32-NEXT:    v_mov_b32_e32 v3, 0
-; GFX11W32-NEXT:    v_readlane_b32 s6, v1, 31
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
+; GFX11W32-NEXT:    v_readlane_b32 s4, v1, 31
 ; GFX11W32-NEXT:    v_mov_b32_dpp v3, v1 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX11W32-NEXT:    v_readlane_b32 s5, v1, 15
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX11W32-NEXT:    v_readlane_b32 s3, v1, 15
+; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
+; GFX11W32-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX11W32-NEXT:    s_or_saveexec_b32 s4, -1
-; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11W32-NEXT:    v_writelane_b32 v3, s5, 16
-; GFX11W32-NEXT:    s_mov_b32 exec_lo, s4
+; GFX11W32-NEXT:    s_or_saveexec_b32 s2, -1
+; GFX11W32-NEXT:    v_writelane_b32 v3, s3, 16
+; GFX11W32-NEXT:    s_mov_b32 exec_lo, s2
+; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
 ; GFX11W32-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
 ; GFX11W32-NEXT:    ; implicit-def: $vgpr0
-; GFX11W32-NEXT:    s_and_saveexec_b32 s4, vcc_lo
+; GFX11W32-NEXT:    s_and_saveexec_b32 s2, vcc_lo
 ; GFX11W32-NEXT:    s_cbranch_execz .LBB7_2
 ; GFX11W32-NEXT:  ; %bb.1:
-; GFX11W32-NEXT:    s_load_b128 s[8:11], s[0:1], 0x34
-; GFX11W32-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11W32-NEXT:    s_mov_b32 s3, s4
+; GFX11W32-NEXT:    s_load_b128 s[4:7], s[0:1], 0x34
+; GFX11W32-NEXT:    v_mov_b32_e32 v0, s3
 ; GFX11W32-NEXT:    v_mov_b32_e32 v4, 0
-; GFX11W32-NEXT:    s_mov_b32 s5, s6
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    buffer_atomic_sub_u32 v0, v4, s[8:11], 0 idxen glc
+; GFX11W32-NEXT:    buffer_atomic_sub_u32 v0, v4, s[4:7], 0 idxen glc
 ; GFX11W32-NEXT:  .LBB7_2:
-; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX11W32-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX11W32-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11W32-NEXT:    s_waitcnt vmcnt(0)
-; GFX11W32-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX11W32-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX11W32-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX11W32-NEXT:    v_mov_b32_e32 v4, 0
 ; GFX11W32-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11W32-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
+; GFX11W32-NEXT:    v_sub_nc_u32_e32 v0, s2, v0
 ; GFX11W32-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11W32-NEXT:    global_store_b32 v4, v0, s[2:3]
+; GFX11W32-NEXT:    global_store_b32 v4, v0, s[0:1]
 ; GFX11W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11W32-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
   %old = call i32 @llvm.amdgcn.struct.buffer.atomic.sub(i32 %lane, <4 x i32> %inout, i32 0, i32 0, i32 0, i32 0)
   store i32 %old, i32 addrspace(1)* %out
   ret void
 }

llvm/test/CodeGen/AMDGPU/branch-relaxation-inst-size-gfx10.ll

 ; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs -amdgpu-s-branch-bits=4 < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GFX10 %s
 ; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs -amdgpu-s-branch-bits=4 < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GFX9 %s
 ; RUN: llc -march=amdgcn -mcpu=gfx1100 -verify-machineinstrs -amdgpu-s-branch-bits=4 < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GFX10 %s
 
 ; Make sure the code size estimate for inline asm is 12-bytes per
 ; instruction, rather than 8 in previous generations.
 
 ; GCN-LABEL: {{^}}long_forward_branch_gfx10only:
-; GFX9: s_load_dwordx2
 ; GFX9: s_cmp_eq_u32
 ; GFX9-NEXT: s_cbranch_scc1
 
 ; GFX10: s_cmp_eq_u32
 ; GFX10-NEXT: s_cbranch_scc0
 ; GFX10: s_getpc_b64
 ; GFX10: s_add_u32
 ; GFX10: s_addc_u32

llvm/test/CodeGen/AMDGPU/madak.ll

define amdgpu_kernel void @no_madak_src1_modifier_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in.a, float addrspace(1)* noalias %in.b) #0 {
   store float %madak, float addrspace(1)* %out.gep, align 4
   ret void
 }
 
 ; SIFoldOperands should not fold the SGPR copy into the instruction before GFX10
 ; because the implicit immediate already uses the constant bus.
 ; On GFX10+ we can use two scalar operands.
 ; GCN-LABEL: {{^}}madak_constant_bus_violation:
-; GCN:       s_load_dword [[SGPR0:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, {{0x12|0x48}}
-
 ; GCN:       {{buffer|flat|global}}_load_dword [[VGPR:v[0-9]+]]
+; GCN:       s_load_dword [[SGPR0:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, {{0x12|0x48}}
 ; MAD:       v_mov_b32_e32 [[MADAK:v[0-9]+]], 0x42280000
 ; MAD:       v_mac_f32_e64 [[MADAK]], [[SGPR0]], 0.5
 ; GFX10:     v_mov_b32_e32 [[SGPR0_VCOPY:v[0-9]+]], [[SGPR0]]
 ; GFX10-MAD: v_madak_f32 [[MADAK:v[0-9]+]], 0.5, [[SGPR0_VCOPY]], 0x42280000
 ; GFX10-FMA: v_fmaak_f32 [[MADAK:v[0-9]+]], 0.5, [[SGPR0_VCOPY]], 0x42280000
 ; GFX940-FMA: v_fmac_f32_e64 [[MADAK:v[0-9]+]], [[SGPR0]], 0.5
 ; GCN:       v_mul_f32_e32 [[MUL:v[0-9]+]], [[MADAK]], [[VGPR]]
 ; GFX6:      buffer_store_dword [[MUL]]

llvm/test/CodeGen/AMDGPU/non-entry-alloca.ll

 ; FLATSCR-NEXT:    s_mov_b32 s33, 0
 ; FLATSCR-NEXT:    s_waitcnt lgkmcnt(0)
 ; FLATSCR-NEXT:    s_cmp_lg_u32 s4, 0
 ; FLATSCR-NEXT:    s_cbranch_scc1 .LBB0_3
 ; FLATSCR-NEXT:  ; %bb.1: ; %bb.0
 ; FLATSCR-NEXT:    s_cmp_lg_u32 s5, 0
 ; FLATSCR-NEXT:    s_cbranch_scc1 .LBB0_3
 ; FLATSCR-NEXT:  ; %bb.2: ; %bb.1
-; FLATSCR-NEXT:    v_mov_b32_e32 v1, 0
 ; FLATSCR-NEXT:    s_add_i32 s2, s32, 0x1000
+; FLATSCR-NEXT:    v_mov_b32_e32 v1, 0
 ; FLATSCR-NEXT:    v_mov_b32_e32 v2, 1
 ; FLATSCR-NEXT:    s_lshl_b32 s3, s6, 2
 ; FLATSCR-NEXT:    s_mov_b32 s32, s2
 ; FLATSCR-NEXT:    scratch_store_dwordx2 off, v[1:2], s2
 ; FLATSCR-NEXT:    s_add_i32 s2, s2, s3
 ; FLATSCR-NEXT:    scratch_load_dword v2, off, s2
 ; FLATSCR-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x0
 ; FLATSCR-NEXT:    s_waitcnt vmcnt(0)

llvm/test/CodeGen/AMDGPU/sdiv64.ll

 ; GCN-NEXT:    v_bfe_i32 v0, v0, 0, 24
 ; GCN-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; GCN-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GCN-NEXT:    buffer_store_short v1, off, s[0:3], 0 offset:4
 ; GCN-NEXT:    s_endpgm
 ;
 ; GCN-IR-LABEL: s_test_sdiv24_48:
 ; GCN-IR:       ; %bb.0: ; %_udiv-special-cases
-; GCN-IR-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-IR-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
+; GCN-IR-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0xb
 ; GCN-IR-NEXT:    s_mov_b32 s15, 0
 ; GCN-IR-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-IR-NEXT:    s_sext_i32_i16 s5, s5
+; GCN-IR-NEXT:    s_ashr_i64 s[2:3], s[4:5], 24
 ; GCN-IR-NEXT:    s_sext_i32_i16 s7, s7
-; GCN-IR-NEXT:    s_sext_i32_i16 s1, s1
-; GCN-IR-NEXT:    s_ashr_i64 s[2:3], s[6:7], 24
-; GCN-IR-NEXT:    s_ashr_i64 s[0:1], s[0:1], 24
 ; GCN-IR-NEXT:    s_lshl_b64 s[2:3], s[2:3], 16
-; GCN-IR-NEXT:    s_lshl_b64 s[6:7], s[0:1], 16
-; GCN-IR-NEXT:    s_ashr_i32 s0, s3, 31
-; GCN-IR-NEXT:    s_ashr_i64 s[8:9], s[2:3], 16
-; GCN-IR-NEXT:    s_mov_b32 s1, s0
-; GCN-IR-NEXT:    s_ashr_i64 s[10:11], s[6:7], 16
-; GCN-IR-NEXT:    s_ashr_i32 s2, s7, 31
-; GCN-IR-NEXT:    s_xor_b64 s[6:7], s[0:1], s[8:9]
+; GCN-IR-NEXT:    s_ashr_i64 s[4:5], s[6:7], 24
+; GCN-IR-NEXT:    s_ashr_i64 s[6:7], s[2:3], 16
+; GCN-IR-NEXT:    s_ashr_i32 s2, s3, 31
+; GCN-IR-NEXT:    s_lshl_b64 s[4:5], s[4:5], 16
 ; GCN-IR-NEXT:    s_mov_b32 s3, s2
-; GCN-IR-NEXT:    s_sub_u32 s12, s6, s0
-; GCN-IR-NEXT:    s_subb_u32 s13, s7, s0
-; GCN-IR-NEXT:    s_xor_b64 s[6:7], s[2:3], s[10:11]
-; GCN-IR-NEXT:    s_sub_u32 s6, s6, s2
-; GCN-IR-NEXT:    s_subb_u32 s7, s7, s2
+; GCN-IR-NEXT:    s_ashr_i64 s[8:9], s[4:5], 16
+; GCN-IR-NEXT:    s_ashr_i32 s4, s5, 31
+; GCN-IR-NEXT:    s_xor_b64 s[6:7], s[2:3], s[6:7]
+; GCN-IR-NEXT:    s_mov_b32 s5, s4
+; GCN-IR-NEXT:    s_sub_u32 s12, s6, s2
+; GCN-IR-NEXT:    s_subb_u32 s13, s7, s2
+; GCN-IR-NEXT:    s_xor_b64 s[6:7], s[4:5], s[8:9]
+; GCN-IR-NEXT:    s_sub_u32 s6, s6, s4
+; GCN-IR-NEXT:    s_subb_u32 s7, s7, s4
 ; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[8:9], s[6:7], 0
 ; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[10:11], s[12:13], 0
-; GCN-IR-NEXT:    s_or_b64 s[16:17], s[8:9], s[10:11]
+; GCN-IR-NEXT:    s_or_b64 s[10:11], s[8:9], s[10:11]
 ; GCN-IR-NEXT:    s_flbit_i32_b32 s8, s6
 ; GCN-IR-NEXT:    s_add_i32 s8, s8, 32
 ; GCN-IR-NEXT:    s_flbit_i32_b32 s9, s7
 ; GCN-IR-NEXT:    s_min_u32 s14, s8, s9
 ; GCN-IR-NEXT:    s_flbit_i32_b32 s8, s12
 ; GCN-IR-NEXT:    s_add_i32 s8, s8, 32
 ; GCN-IR-NEXT:    s_flbit_i32_b32 s9, s13
 ; GCN-IR-NEXT:    s_min_u32 s18, s8, s9
-; GCN-IR-NEXT:    s_sub_u32 s10, s14, s18
-; GCN-IR-NEXT:    s_subb_u32 s11, 0, 0
-; GCN-IR-NEXT:    v_cmp_gt_u64_e64 s[20:21], s[10:11], 63
-; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[22:23], s[10:11], 63
-; GCN-IR-NEXT:    s_or_b64 s[20:21], s[16:17], s[20:21]
-; GCN-IR-NEXT:    s_and_b64 s[16:17], s[20:21], exec
-; GCN-IR-NEXT:    s_cselect_b32 s17, 0, s13
-; GCN-IR-NEXT:    s_cselect_b32 s16, 0, s12
+; GCN-IR-NEXT:    s_sub_u32 s16, s14, s18
+; GCN-IR-NEXT:    s_subb_u32 s17, 0, 0
+; GCN-IR-NEXT:    v_cmp_gt_u64_e64 s[20:21], s[16:17], 63
+; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[22:23], s[16:17], 63
+; GCN-IR-NEXT:    s_or_b64 s[20:21], s[10:11], s[20:21]
+; GCN-IR-NEXT:    s_and_b64 s[10:11], s[20:21], exec
+; GCN-IR-NEXT:    s_cselect_b32 s11, 0, s13
+; GCN-IR-NEXT:    s_cselect_b32 s10, 0, s12
 ; GCN-IR-NEXT:    s_or_b64 s[20:21], s[20:21], s[22:23]
 ; GCN-IR-NEXT:    s_mov_b64 s[8:9], 0
 ; GCN-IR-NEXT:    s_andn2_b64 vcc, exec, s[20:21]
 ; GCN-IR-NEXT:    s_cbranch_vccz .LBB9_5
 ; GCN-IR-NEXT:  ; %bb.1: ; %udiv-bb1
-; GCN-IR-NEXT:    s_add_u32 s16, s10, 1
-; GCN-IR-NEXT:    s_addc_u32 s17, s11, 0
-; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[20:21], s[16:17], 0
-; GCN-IR-NEXT:    s_sub_i32 s10, 63, s10
-; GCN-IR-NEXT:    s_andn2_b64 vcc, exec, s[20:21]
-; GCN-IR-NEXT:    s_lshl_b64 s[10:11], s[12:13], s10
+; GCN-IR-NEXT:    s_add_u32 s20, s16, 1
+; GCN-IR-NEXT:    s_addc_u32 s21, s17, 0
+; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[10:11], s[20:21], 0
+; GCN-IR-NEXT:    s_sub_i32 s16, 63, s16
+; GCN-IR-NEXT:    s_andn2_b64 vcc, exec, s[10:11]
+; GCN-IR-NEXT:    s_lshl_b64 s[10:11], s[12:13], s16
 ; GCN-IR-NEXT:    s_cbranch_vccz .LBB9_4
 ; GCN-IR-NEXT:  ; %bb.2: ; %udiv-preheader
-; GCN-IR-NEXT:    s_lshr_b64 s[16:17], s[12:13], s16
+; GCN-IR-NEXT:    s_lshr_b64 s[16:17], s[12:13], s20
 ; GCN-IR-NEXT:    s_add_u32 s19, s6, -1
 ; GCN-IR-NEXT:    s_addc_u32 s20, s7, -1
 ; GCN-IR-NEXT:    s_not_b64 s[8:9], s[14:15]
 ; GCN-IR-NEXT:    s_add_u32 s12, s8, s18
 ; GCN-IR-NEXT:    s_addc_u32 s13, s9, 0
 ; GCN-IR-NEXT:    s_mov_b64 s[14:15], 0
 ; GCN-IR-NEXT:    s_mov_b32 s9, 0
 ; GCN-IR-NEXT:  .LBB9_3: ; %udiv-do-while

 ; GCN-IR-NEXT:    s_add_u32 s12, s12, 1
 ; GCN-IR-NEXT:    s_addc_u32 s13, s13, 0
 ; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[22:23], s[12:13], 0
 ; GCN-IR-NEXT:    s_mov_b64 s[14:15], s[8:9]
 ; GCN-IR-NEXT:    s_and_b64 vcc, exec, s[22:23]
 ; GCN-IR-NEXT:    s_cbranch_vccz .LBB9_3
 ; GCN-IR-NEXT:  .LBB9_4: ; %Flow3
 ; GCN-IR-NEXT:    s_lshl_b64 s[6:7], s[10:11], 1
-; GCN-IR-NEXT:    s_or_b64 s[16:17], s[8:9], s[6:7]
+; GCN-IR-NEXT:    s_or_b64 s[10:11], s[8:9], s[6:7]
 ; GCN-IR-NEXT:  .LBB9_5: ; %udiv-end
-; GCN-IR-NEXT:    s_xor_b64 s[0:1], s[2:3], s[0:1]
-; GCN-IR-NEXT:    s_xor_b64 s[2:3], s[16:17], s[0:1]
+; GCN-IR-NEXT:    s_load_dwordx2 s[12:13], s[0:1], 0x9
+; GCN-IR-NEXT:    s_xor_b64 s[0:1], s[4:5], s[2:3]
+; GCN-IR-NEXT:    s_xor_b64 s[2:3], s[10:11], s[0:1]
 ; GCN-IR-NEXT:    s_sub_u32 s0, s2, s0
 ; GCN-IR-NEXT:    s_subb_u32 s1, s3, s1
-; GCN-IR-NEXT:    s_mov_b32 s7, 0xf000
-; GCN-IR-NEXT:    s_mov_b32 s6, -1
+; GCN-IR-NEXT:    s_mov_b32 s15, 0xf000
+; GCN-IR-NEXT:    s_mov_b32 s14, -1
 ; GCN-IR-NEXT:    v_mov_b32_e32 v0, s1
-; GCN-IR-NEXT:    buffer_store_short v0, off, s[4:7], 0 offset:4
+; GCN-IR-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-IR-NEXT:    buffer_store_short v0, off, s[12:15], 0 offset:4
 ; GCN-IR-NEXT:    s_waitcnt expcnt(0)
 ; GCN-IR-NEXT:    v_mov_b32_e32 v0, s0
-; GCN-IR-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN-IR-NEXT:    buffer_store_dword v0, off, s[12:15], 0
 ; GCN-IR-NEXT:    s_endpgm
   %1 = ashr i48 %x, 24
   %2 = ashr i48 %y, 24
   %result = sdiv i48 %1, %2
   store i48 %result, i48 addrspace(1)* %out
   ret void
 }
 

llvm/test/CodeGen/AMDGPU/sgpr-control-flow.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=SI %s
 ;
 ; Most SALU instructions ignore control flow, so we need to make sure
 ; they don't overwrite values from other blocks.
 
 ; If the branch decision is made based on a value in an SGPR then all
 ; threads will execute the same code paths, so we don't need to worry
 ; about instructions in different blocks overwriting each other.
 
 define amdgpu_kernel void @sgpr_if_else_salu_br(i32 addrspace(1)* %out, i32 %a, i32 %b, i32 %c, i32 %d, i32 %e) {
 ; SI-LABEL: sgpr_if_else_salu_br:
 ; SI:       ; %bb.0: ; %entry
-; SI-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0xb
-; SI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
-; SI-NEXT:    s_load_dword s0, s[0:1], 0xf
+; SI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0xb
+; SI-NEXT:    s_load_dword s2, s[0:1], 0xf
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
-; SI-NEXT:    s_cmp_lg_u32 s8, 0
+; SI-NEXT:    s_cmp_lg_u32 s4, 0
 ; SI-NEXT:    s_cbranch_scc0 .LBB0_4
 ; SI-NEXT:  ; %bb.1: ; %else
-; SI-NEXT:    s_add_i32 s2, s11, s0
+; SI-NEXT:    s_add_i32 s7, s7, s2
 ; SI-NEXT:    s_cbranch_execnz .LBB0_3
 ; SI-NEXT:  .LBB0_2: ; %if
-; SI-NEXT:    s_sub_i32 s2, s9, s10
+; SI-NEXT:    s_sub_i32 s7, s5, s6
 ; SI-NEXT:  .LBB0_3: ; %endif
-; SI-NEXT:    s_add_i32 s0, s2, s8
-; SI-NEXT:    s_mov_b32 s7, 0xf000
-; SI-NEXT:    s_mov_b32 s6, -1
-; SI-NEXT:    v_mov_b32_e32 v0, s0
-; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_add_i32 s4, s7, s4
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; SI-NEXT:    s_endpgm
 ; SI-NEXT:  .LBB0_4:
-; SI-NEXT:    ; implicit-def: $sgpr2
+; SI-NEXT:    ; implicit-def: $sgpr7
 ; SI-NEXT:    s_branch .LBB0_2
 
 entry:
   %0 = icmp eq i32 %a, 0
   br i1 %0, label %if, label %else
 
 if:
   %1 = sub i32 %b, %c
   br label %endif
 
 else:
   %2 = add i32 %d, %e
   br label %endif
 
 endif:
   %3 = phi i32 [%1, %if], [%2, %else]
   %4 = add i32 %3, %a
   store i32 %4, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @sgpr_if_else_salu_br_opt(i32 addrspace(1)* %out, [8 x i32], i32 %a, [8 x i32], i32 %b, [8 x i32], i32 %c, [8 x i32], i32 %d, [8 x i32], i32 %e) {
 ; SI-LABEL: sgpr_if_else_salu_br_opt:
 ; SI:       ; %bb.0: ; %entry
-; SI-NEXT:    s_load_dword s6, s[0:1], 0x13
-; SI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
+; SI-NEXT:    s_load_dword s4, s[0:1], 0x13
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
-; SI-NEXT:    s_cmp_lg_u32 s6, 0
+; SI-NEXT:    s_cmp_lg_u32 s4, 0
 ; SI-NEXT:    s_cbranch_scc0 .LBB1_4
 ; SI-NEXT:  ; %bb.1: ; %else
 ; SI-NEXT:    s_load_dword s2, s[0:1], 0x2e
 ; SI-NEXT:    s_load_dword s3, s[0:1], 0x37
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
-; SI-NEXT:    s_add_i32 s7, s2, s3
+; SI-NEXT:    s_add_i32 s5, s2, s3
 ; SI-NEXT:    s_cbranch_execnz .LBB1_3
 ; SI-NEXT:  .LBB1_2: ; %if
 ; SI-NEXT:    s_load_dword s2, s[0:1], 0x1c
-; SI-NEXT:    s_load_dword s0, s[0:1], 0x25
+; SI-NEXT:    s_load_dword s3, s[0:1], 0x25
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
-; SI-NEXT:    s_add_i32 s7, s2, s0
+; SI-NEXT:    s_add_i32 s5, s2, s3
 ; SI-NEXT:  .LBB1_3: ; %endif
-; SI-NEXT:    s_add_i32 s0, s7, s6
-; SI-NEXT:    s_mov_b32 s7, 0xf000
-; SI-NEXT:    s_mov_b32 s6, -1
-; SI-NEXT:    v_mov_b32_e32 v0, s0
-; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_add_i32 s4, s5, s4
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; SI-NEXT:    s_endpgm
 ; SI-NEXT:  .LBB1_4:
-; SI-NEXT:    ; implicit-def: $sgpr7
+; SI-NEXT:    ; implicit-def: $sgpr5
 ; SI-NEXT:    s_branch .LBB1_2
 
 entry:
   %cmp0 = icmp eq i32 %a, 0
   br i1 %cmp0, label %if, label %else
 
 if:
   %add0 = add i32 %b, %c
endif:
   ret void
 }
 
 ; The two S_ADD instructions should write to different registers, since
 ; different threads will take different control flow paths.
 define amdgpu_kernel void @sgpr_if_else_valu_br(i32 addrspace(1)* %out, float %a, i32 %b, i32 %c, i32 %d, i32 %e) {
 ; SI-LABEL: sgpr_if_else_valu_br:
 ; SI:       ; %bb.0: ; %entry
-; SI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
-; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0xc
+; SI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0xc
 ; SI-NEXT:    v_cvt_f32_u32_e32 v0, v0
 ; SI-NEXT:    ; implicit-def: $sgpr8
 ; SI-NEXT:    v_cmp_lg_f32_e32 vcc, 0, v0
-; SI-NEXT:    s_and_saveexec_b64 s[6:7], vcc
-; SI-NEXT:    s_xor_b64 s[6:7], exec, s[6:7]
+; SI-NEXT:    s_and_saveexec_b64 s[2:3], vcc
+; SI-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
 ; SI-NEXT:    s_cbranch_execz .LBB2_2
 ; SI-NEXT:  ; %bb.1: ; %else
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
-; SI-NEXT:    s_add_i32 s8, s2, s3
+; SI-NEXT:    s_add_i32 s8, s6, s7
 ; SI-NEXT:  .LBB2_2: ; %Flow
-; SI-NEXT:    s_waitcnt lgkmcnt(0)
-; SI-NEXT:    s_or_saveexec_b64 s[2:3], s[6:7]
+; SI-NEXT:    s_or_saveexec_b64 s[2:3], s[2:3]
 ; SI-NEXT:    v_mov_b32_e32 v0, s8
 ; SI-NEXT:    s_xor_b64 exec, exec, s[2:3]
+; SI-NEXT:    s_cbranch_execz .LBB2_4
 ; SI-NEXT:  ; %bb.3: ; %if
-; SI-NEXT:    s_add_i32 s0, s0, s1
-; SI-NEXT:    v_mov_b32_e32 v0, s0
-; SI-NEXT:  ; %bb.4: ; %endif
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_add_i32 s4, s4, s5
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:  .LBB2_4: ; %endif
 ; SI-NEXT:    s_or_b64 exec, exec, s[2:3]
-; SI-NEXT:    s_mov_b32 s7, 0xf000
-; SI-NEXT:    s_mov_b32 s6, -1
-; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; SI-NEXT:    s_endpgm
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #0
   %tid_f = uitofp i32 %tid to float
   %tmp1 = fcmp ueq float %tid_f, 0.0
   br i1 %tmp1, label %if, label %else
 
 if:

llvm/test/CodeGen/AMDGPU/si-unify-exit-multiple-unreachables.ll

 ; UNIFY-NEXT:    br i1 %cmp2, label %if.then3, label %UnifiedReturnBlock
 ; UNIFY-LABEL: if.then3:
 ; UNIFY-NEXT:    %cmp1.i7 = icmp eq i32 %a.load, 0
 ; UNIFY-NEXT:    br i1 %cmp1.i7, label %if.end6.sink.split, label %cond.false.i8
 ; UNIFY-LABEL: cond.false.i8:
 ; UNIFY-NEXT:    call void @llvm.trap()
 ; UNIFY-NEXT:    br label %UnifiedUnreachableBlock
 ; UNIFY-LABEL: if.end6.sink.split:
+; UNIFY-NEXT:    %x.kernarg.offset = getelementptr inbounds i8, i8 addrspace(4)* %kernel.kernarg.segment, i64 8
+; UNIFY-NEXT:    %x.kernarg.offset.cast = bitcast i8 addrspace(4)* %x.kernarg.offset to i32 addrspace(1)* addrspace(4)*
+; UNIFY-NEXT:    %x.load = load i32 addrspace(1)*, i32 addrspace(1)* addrspace(4)* %x.kernarg.offset.cast, align 8, !invariant.load !0
 ; UNIFY-NEXT:    %idxprom = sext i32 %tid to i64
 ; UNIFY-NEXT:    %x1 = getelementptr inbounds i32, i32 addrspace(1)* %x.load, i64 %idxprom
 ; UNIFY-NEXT:    store i32 %a.load, i32 addrspace(1)* %x1, align 4
 ; UNIFY-NEXT:    br label %UnifiedReturnBlock
 ; UNIFY-LABEL: UnifiedUnreachableBlock:
 ; UNIFY-NEXT:    call void @llvm.amdgcn.unreachable()
 ; UNIFY-NEXT:    br label %UnifiedReturnBlock
 ; UNIFY-LABEL: UnifiedReturnBlock:

llvm/test/CodeGen/AMDGPU/srem64.ll

 ; GCN-NEXT:    v_bfe_i32 v0, v0, 0, 24
 ; GCN-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; GCN-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GCN-NEXT:    buffer_store_short v1, off, s[0:3], 0 offset:4
 ; GCN-NEXT:    s_endpgm
 ;
 ; GCN-IR-LABEL: s_test_srem24_48:
 ; GCN-IR:       ; %bb.0: ; %_udiv-special-cases
-; GCN-IR-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-IR-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
+; GCN-IR-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0xb
+; GCN-IR-NEXT:    s_mov_b32 s13, 0
 ; GCN-IR-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-IR-NEXT:    s_sext_i32_i16 s5, s5
 ; GCN-IR-NEXT:    s_sext_i32_i16 s7, s7
-; GCN-IR-NEXT:    s_sext_i32_i16 s1, s1
-; GCN-IR-NEXT:    s_ashr_i64 s[2:3], s[6:7], 24
-; GCN-IR-NEXT:    s_ashr_i64 s[0:1], s[0:1], 24
+; GCN-IR-NEXT:    s_ashr_i64 s[2:3], s[4:5], 24
+; GCN-IR-NEXT:    s_ashr_i64 s[4:5], s[6:7], 24
 ; GCN-IR-NEXT:    s_lshl_b64 s[2:3], s[2:3], 16
-; GCN-IR-NEXT:    s_lshl_b64 s[6:7], s[0:1], 16
-; GCN-IR-NEXT:    s_ashr_i32 s0, s3, 31
-; GCN-IR-NEXT:    s_ashr_i32 s12, s7, 31
-; GCN-IR-NEXT:    s_ashr_i64 s[8:9], s[2:3], 16
-; GCN-IR-NEXT:    s_ashr_i64 s[10:11], s[6:7], 16
-; GCN-IR-NEXT:    s_mov_b32 s1, s0
-; GCN-IR-NEXT:    s_mov_b32 s13, s12
-; GCN-IR-NEXT:    s_xor_b64 s[2:3], s[8:9], s[0:1]
-; GCN-IR-NEXT:    s_xor_b64 s[6:7], s[10:11], s[12:13]
-; GCN-IR-NEXT:    s_sub_u32 s2, s2, s0
-; GCN-IR-NEXT:    s_subb_u32 s3, s3, s0
-; GCN-IR-NEXT:    s_sub_u32 s6, s6, s12
-; GCN-IR-NEXT:    s_subb_u32 s7, s7, s12
+; GCN-IR-NEXT:    s_lshl_b64 s[4:5], s[4:5], 16
+; GCN-IR-NEXT:    s_ashr_i64 s[6:7], s[2:3], 16
+; GCN-IR-NEXT:    s_ashr_i32 s2, s3, 31
+; GCN-IR-NEXT:    s_ashr_i32 s10, s5, 31
+; GCN-IR-NEXT:    s_ashr_i64 s[8:9], s[4:5], 16
+; GCN-IR-NEXT:    s_mov_b32 s3, s2
+; GCN-IR-NEXT:    s_mov_b32 s11, s10
+; GCN-IR-NEXT:    s_xor_b64 s[4:5], s[6:7], s[2:3]
+; GCN-IR-NEXT:    s_xor_b64 s[6:7], s[8:9], s[10:11]
+; GCN-IR-NEXT:    s_sub_u32 s4, s4, s2
+; GCN-IR-NEXT:    s_subb_u32 s5, s5, s2
+; GCN-IR-NEXT:    s_sub_u32 s6, s6, s10
+; GCN-IR-NEXT:    s_subb_u32 s7, s7, s10
 ; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[8:9], s[6:7], 0
-; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[10:11], s[2:3], 0
-; GCN-IR-NEXT:    s_mov_b32 s13, 0
+; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[10:11], s[4:5], 0
 ; GCN-IR-NEXT:    s_or_b64 s[10:11], s[8:9], s[10:11]
 ; GCN-IR-NEXT:    s_flbit_i32_b32 s8, s6
 ; GCN-IR-NEXT:    s_add_i32 s8, s8, 32
 ; GCN-IR-NEXT:    s_flbit_i32_b32 s9, s7
 ; GCN-IR-NEXT:    s_min_u32 s12, s8, s9
-; GCN-IR-NEXT:    s_flbit_i32_b32 s8, s2
+; GCN-IR-NEXT:    s_flbit_i32_b32 s8, s4
 ; GCN-IR-NEXT:    s_add_i32 s8, s8, 32
-; GCN-IR-NEXT:    s_flbit_i32_b32 s9, s3
+; GCN-IR-NEXT:    s_flbit_i32_b32 s9, s5
 ; GCN-IR-NEXT:    s_min_u32 s16, s8, s9
 ; GCN-IR-NEXT:    s_sub_u32 s14, s12, s16
 ; GCN-IR-NEXT:    s_subb_u32 s15, 0, 0
 ; GCN-IR-NEXT:    v_cmp_gt_u64_e64 s[18:19], s[14:15], 63
 ; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[20:21], s[14:15], 63
 ; GCN-IR-NEXT:    s_or_b64 s[18:19], s[10:11], s[18:19]
 ; GCN-IR-NEXT:    s_and_b64 s[10:11], s[18:19], exec
-; GCN-IR-NEXT:    s_cselect_b32 s11, 0, s3
-; GCN-IR-NEXT:    s_cselect_b32 s10, 0, s2
+; GCN-IR-NEXT:    s_cselect_b32 s11, 0, s5
+; GCN-IR-NEXT:    s_cselect_b32 s10, 0, s4
 ; GCN-IR-NEXT:    s_or_b64 s[18:19], s[18:19], s[20:21]
 ; GCN-IR-NEXT:    s_mov_b64 s[8:9], 0
 ; GCN-IR-NEXT:    s_andn2_b64 vcc, exec, s[18:19]
 ; GCN-IR-NEXT:    s_cbranch_vccz .LBB9_5
 ; GCN-IR-NEXT:  ; %bb.1: ; %udiv-bb1
 ; GCN-IR-NEXT:    s_add_u32 s18, s14, 1
 ; GCN-IR-NEXT:    s_addc_u32 s19, s15, 0
 ; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[10:11], s[18:19], 0
 ; GCN-IR-NEXT:    s_sub_i32 s14, 63, s14
 ; GCN-IR-NEXT:    s_andn2_b64 vcc, exec, s[10:11]
-; GCN-IR-NEXT:    s_lshl_b64 s[10:11], s[2:3], s14
+; GCN-IR-NEXT:    s_lshl_b64 s[10:11], s[4:5], s14
 ; GCN-IR-NEXT:    s_cbranch_vccz .LBB9_4
 ; GCN-IR-NEXT:  ; %bb.2: ; %udiv-preheader
-; GCN-IR-NEXT:    s_lshr_b64 s[14:15], s[2:3], s18
+; GCN-IR-NEXT:    s_lshr_b64 s[14:15], s[4:5], s18
 ; GCN-IR-NEXT:    s_add_u32 s18, s6, -1
 ; GCN-IR-NEXT:    s_addc_u32 s19, s7, -1
 ; GCN-IR-NEXT:    s_not_b64 s[8:9], s[12:13]
 ; GCN-IR-NEXT:    s_add_u32 s12, s8, s16
 ; GCN-IR-NEXT:    s_addc_u32 s13, s9, 0
 ; GCN-IR-NEXT:    s_mov_b64 s[16:17], 0
 ; GCN-IR-NEXT:    s_mov_b32 s9, 0
 ; GCN-IR-NEXT:  .LBB9_3: ; %udiv-do-while

 ; GCN-IR-NEXT:    s_and_b64 vcc, exec, s[20:21]
 ; GCN-IR-NEXT:    s_cbranch_vccz .LBB9_3
 ; GCN-IR-NEXT:  .LBB9_4: ; %Flow3
 ; GCN-IR-NEXT:    s_lshl_b64 s[10:11], s[10:11], 1
 ; GCN-IR-NEXT:    s_or_b64 s[10:11], s[8:9], s[10:11]
 ; GCN-IR-NEXT:  .LBB9_5: ; %udiv-end
 ; GCN-IR-NEXT:    v_mov_b32_e32 v0, s10
 ; GCN-IR-NEXT:    v_mul_hi_u32 v0, s6, v0
-; GCN-IR-NEXT:    s_mul_i32 s8, s6, s11
-; GCN-IR-NEXT:    s_mul_i32 s7, s7, s10
-; GCN-IR-NEXT:    s_mul_i32 s6, s6, s10
-; GCN-IR-NEXT:    v_add_i32_e32 v0, vcc, s8, v0
-; GCN-IR-NEXT:    v_add_i32_e32 v0, vcc, s7, v0
-; GCN-IR-NEXT:    v_mov_b32_e32 v1, s6
-; GCN-IR-NEXT:    v_mov_b32_e32 v2, s3
-; GCN-IR-NEXT:    v_sub_i32_e32 v1, vcc, s2, v1
+; GCN-IR-NEXT:    s_load_dwordx2 s[12:13], s[0:1], 0x9
+; GCN-IR-NEXT:    s_mul_i32 s0, s6, s11
+; GCN-IR-NEXT:    v_mov_b32_e32 v2, s5
+; GCN-IR-NEXT:    v_add_i32_e32 v0, vcc, s0, v0
+; GCN-IR-NEXT:    s_mul_i32 s0, s7, s10
+; GCN-IR-NEXT:    v_add_i32_e32 v0, vcc, s0, v0
+; GCN-IR-NEXT:    s_mul_i32 s0, s6, s10
+; GCN-IR-NEXT:    v_mov_b32_e32 v1, s0
+; GCN-IR-NEXT:    v_sub_i32_e32 v1, vcc, s4, v1
 ; GCN-IR-NEXT:    v_subb_u32_e32 v0, vcc, v2, v0, vcc
-; GCN-IR-NEXT:    v_xor_b32_e32 v1, s0, v1
-; GCN-IR-NEXT:    v_xor_b32_e32 v0, s1, v0
-; GCN-IR-NEXT:    v_mov_b32_e32 v2, s1
-; GCN-IR-NEXT:    v_subrev_i32_e32 v1, vcc, s0, v1
+; GCN-IR-NEXT:    v_xor_b32_e32 v1, s2, v1
+; GCN-IR-NEXT:    v_xor_b32_e32 v0, s3, v0
+; GCN-IR-NEXT:    v_mov_b32_e32 v2, s3
+; GCN-IR-NEXT:    v_subrev_i32_e32 v1, vcc, s2, v1
+; GCN-IR-NEXT:    s_mov_b32 s15, 0xf000
+; GCN-IR-NEXT:    s_mov_b32 s14, -1
 ; GCN-IR-NEXT:    v_subb_u32_e32 v0, vcc, v0, v2, vcc
-; GCN-IR-NEXT:    s_mov_b32 s7, 0xf000
-; GCN-IR-NEXT:    s_mov_b32 s6, -1
-; GCN-IR-NEXT:    buffer_store_short v0, off, s[4:7], 0 offset:4
-; GCN-IR-NEXT:    buffer_store_dword v1, off, s[4:7], 0
+; GCN-IR-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-IR-NEXT:    buffer_store_short v0, off, s[12:15], 0 offset:4
+; GCN-IR-NEXT:    buffer_store_dword v1, off, s[12:15], 0
 ; GCN-IR-NEXT:    s_endpgm
   %1 = ashr i48 %x, 24
   %2 = ashr i48 %y, 24
   %result = srem i48 %1, %2
   store i48 %result, i48 addrspace(1)* %out
   ret void
 }
 

llvm/test/CodeGen/AMDGPU/subreg-coalescer-undef-use.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -march=amdgcn -mcpu=tahiti -amdgpu-dce-in-ra=0 -o - %s | FileCheck %s
 ; Don't crash when the use of an undefined value is only detected by the
 ; register coalescer because it is hidden with subregister insert/extract.
 target triple="amdgcn--"
 
 define amdgpu_kernel void @foobar(float %a0, float %a1, float addrspace(1)* %out) nounwind {
 ; CHECK-LABEL: foobar:
 ; CHECK:       ; %bb.0: ; %entry
-; CHECK-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; CHECK-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; CHECK-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, -1, 0
 ; CHECK-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; CHECK-NEXT:    s_mov_b32 s6, -1
+; CHECK-NEXT:    s_mov_b32 s2, -1
 ; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
-; CHECK-NEXT:    v_mov_b32_e32 v0, s0
-; CHECK-NEXT:    v_mov_b32_e32 v1, s1
-; CHECK-NEXT:    v_mov_b32_e32 v2, s2
-; CHECK-NEXT:    v_mov_b32_e32 v3, s3
-; CHECK-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; CHECK-NEXT:    v_mov_b32_e32 v0, s4
+; CHECK-NEXT:    v_mov_b32_e32 v1, s5
+; CHECK-NEXT:    v_mov_b32_e32 v2, s6
+; CHECK-NEXT:    v_mov_b32_e32 v3, s7
+; CHECK-NEXT:    s_and_saveexec_b64 s[6:7], vcc
 ; CHECK-NEXT:  ; %bb.1: ; %ift
-; CHECK-NEXT:    s_mov_b32 s0, s1
-; CHECK-NEXT:    v_mov_b32_e32 v0, s0
-; CHECK-NEXT:    v_mov_b32_e32 v1, s1
-; CHECK-NEXT:    v_mov_b32_e32 v2, s2
-; CHECK-NEXT:    v_mov_b32_e32 v3, s3
+; CHECK-NEXT:    s_mov_b32 s4, s5
+; CHECK-NEXT:    v_mov_b32_e32 v0, s4
+; CHECK-NEXT:    v_mov_b32_e32 v1, s5
+; CHECK-NEXT:    v_mov_b32_e32 v2, s6
+; CHECK-NEXT:    v_mov_b32_e32 v3, s7
 ; CHECK-NEXT:  ; %bb.2: ; %ife
-; CHECK-NEXT:    s_or_b64 exec, exec, s[4:5]
-; CHECK-NEXT:    s_mov_b32 s7, 0xf000
-; CHECK-NEXT:    s_mov_b32 s4, s2
-; CHECK-NEXT:    s_mov_b32 s5, s3
-; CHECK-NEXT:    buffer_store_dword v1, off, s[4:7], 0
+; CHECK-NEXT:    s_or_b64 exec, exec, s[6:7]
+; CHECK-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
+; CHECK-NEXT:    s_mov_b32 s3, 0xf000
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    buffer_store_dword v1, off, s[0:3], 0
 ; CHECK-NEXT:    s_endpgm
 
 ; FIXME: The change related to the fact that
 ; DetectDeadLanes pass hit "Copy across incompatible class" SGPR -> VGPR in analysis
 ; and hence it cannot derive the fact that the vector element in the "ift" block is unused.
 ; Such a copies appear because the float4 vectors and their elements in the test are uniform
 ; but the PHI node in "ife" block is divergent because of the CF dependency (divergent branch in bb0)
 entry:

llvm/test/CodeGen/AMDGPU/udiv64.ll

 ; GCN-NEXT:    v_cndmask_b32_e32 v0, v1, v4, vcc
 ; GCN-NEXT:    v_cndmask_b32_e32 v1, v2, v5, vcc
 ; GCN-NEXT:    buffer_store_short v1, off, s[0:3], 0 offset:4
 ; GCN-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; GCN-IR-LABEL: s_test_udiv24_i48:
 ; GCN-IR:       ; %bb.0: ; %_udiv-special-cases
-; GCN-IR-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-IR-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
+; GCN-IR-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0xb
 ; GCN-IR-NEXT:    s_mov_b32 s11, 0
 ; GCN-IR-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-IR-NEXT:    s_and_b32 s3, s7, 0xffff
-; GCN-IR-NEXT:    s_and_b32 s2, s6, 0xff000000
-; GCN-IR-NEXT:    s_and_b32 s1, s1, 0xffff
-; GCN-IR-NEXT:    s_and_b32 s0, s0, 0xff000000
+; GCN-IR-NEXT:    s_and_b32 s3, s5, 0xffff
+; GCN-IR-NEXT:    s_and_b32 s2, s4, 0xff000000
+; GCN-IR-NEXT:    s_and_b32 s5, s7, 0xffff
+; GCN-IR-NEXT:    s_and_b32 s4, s6, 0xff000000
 ; GCN-IR-NEXT:    s_lshr_b64 s[8:9], s[2:3], 24
-; GCN-IR-NEXT:    s_lshr_b64 s[0:1], s[0:1], 24
+; GCN-IR-NEXT:    s_lshr_b64 s[2:3], s[4:5], 24
 ; GCN-IR-NEXT:    s_and_b32 s9, s9, 0xffff
-; GCN-IR-NEXT:    s_and_b32 s1, s1, 0xffff
-; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[2:3], s[0:1], 0
+; GCN-IR-NEXT:    s_and_b32 s3, s3, 0xffff
+; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[4:5], s[2:3], 0
 ; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[6:7], s[8:9], 0
-; GCN-IR-NEXT:    s_or_b64 s[12:13], s[2:3], s[6:7]
-; GCN-IR-NEXT:    s_flbit_i32_b32 s2, s0
-; GCN-IR-NEXT:    s_add_i32 s2, s2, 32
-; GCN-IR-NEXT:    s_flbit_i32_b32 s3, s1
-; GCN-IR-NEXT:    s_min_u32 s10, s2, s3
-; GCN-IR-NEXT:    s_flbit_i32_b32 s2, s8
-; GCN-IR-NEXT:    s_add_i32 s2, s2, 32
-; GCN-IR-NEXT:    s_flbit_i32_b32 s3, s9
-; GCN-IR-NEXT:    s_min_u32 s14, s2, s3
-; GCN-IR-NEXT:    s_sub_u32 s6, s10, s14
-; GCN-IR-NEXT:    s_subb_u32 s7, 0, 0
-; GCN-IR-NEXT:    v_cmp_gt_u64_e64 s[16:17], s[6:7], 63
-; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[18:19], s[6:7], 63
-; GCN-IR-NEXT:    s_or_b64 s[16:17], s[12:13], s[16:17]
-; GCN-IR-NEXT:    s_and_b64 s[12:13], s[16:17], exec
-; GCN-IR-NEXT:    s_cselect_b32 s13, 0, s9
-; GCN-IR-NEXT:    s_cselect_b32 s12, 0, s8
+; GCN-IR-NEXT:    s_or_b64 s[6:7], s[4:5], s[6:7]
+; GCN-IR-NEXT:    s_flbit_i32_b32 s4, s2
+; GCN-IR-NEXT:    s_add_i32 s4, s4, 32
+; GCN-IR-NEXT:    s_flbit_i32_b32 s5, s3
+; GCN-IR-NEXT:    s_min_u32 s10, s4, s5
+; GCN-IR-NEXT:    s_flbit_i32_b32 s4, s8
+; GCN-IR-NEXT:    s_add_i32 s4, s4, 32
+; GCN-IR-NEXT:    s_flbit_i32_b32 s5, s9
+; GCN-IR-NEXT:    s_min_u32 s14, s4, s5
+; GCN-IR-NEXT:    s_sub_u32 s12, s10, s14
+; GCN-IR-NEXT:    s_subb_u32 s13, 0, 0
+; GCN-IR-NEXT:    v_cmp_gt_u64_e64 s[16:17], s[12:13], 63
+; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[18:19], s[12:13], 63
+; GCN-IR-NEXT:    s_or_b64 s[16:17], s[6:7], s[16:17]
+; GCN-IR-NEXT:    s_and_b64 s[6:7], s[16:17], exec
+; GCN-IR-NEXT:    s_cselect_b32 s7, 0, s9
+; GCN-IR-NEXT:    s_cselect_b32 s6, 0, s8
 ; GCN-IR-NEXT:    s_or_b64 s[16:17], s[16:17], s[18:19]
-; GCN-IR-NEXT:    s_mov_b64 s[2:3], 0
+; GCN-IR-NEXT:    s_mov_b64 s[4:5], 0
 ; GCN-IR-NEXT:    s_andn2_b64 vcc, exec, s[16:17]
 ; GCN-IR-NEXT:    s_cbranch_vccz .LBB7_5
 ; GCN-IR-NEXT:  ; %bb.1: ; %udiv-bb1
-; GCN-IR-NEXT:    s_add_u32 s12, s6, 1
-; GCN-IR-NEXT:    s_addc_u32 s13, s7, 0
-; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[16:17], s[12:13], 0
-; GCN-IR-NEXT:    s_sub_i32 s6, 63, s6
-; GCN-IR-NEXT:    s_andn2_b64 vcc, exec, s[16:17]
-; GCN-IR-NEXT:    s_lshl_b64 s[6:7], s[8:9], s6
+; GCN-IR-NEXT:    s_add_u32 s16, s12, 1
+; GCN-IR-NEXT:    s_addc_u32 s17, s13, 0
+; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[6:7], s[16:17], 0
+; GCN-IR-NEXT:    s_sub_i32 s12, 63, s12
+; GCN-IR-NEXT:    s_andn2_b64 vcc, exec, s[6:7]
+; GCN-IR-NEXT:    s_lshl_b64 s[6:7], s[8:9], s12
 ; GCN-IR-NEXT:    s_cbranch_vccz .LBB7_4
 ; GCN-IR-NEXT:  ; %bb.2: ; %udiv-preheader
-; GCN-IR-NEXT:    s_lshr_b64 s[12:13], s[8:9], s12
-; GCN-IR-NEXT:    s_add_u32 s15, s0, -1
-; GCN-IR-NEXT:    s_addc_u32 s16, s1, -1
-; GCN-IR-NEXT:    s_not_b64 s[2:3], s[10:11]
-; GCN-IR-NEXT:    s_add_u32 s8, s2, s14
-; GCN-IR-NEXT:    s_addc_u32 s9, s3, 0
+; GCN-IR-NEXT:    s_lshr_b64 s[12:13], s[8:9], s16
+; GCN-IR-NEXT:    s_add_u32 s15, s2, -1
+; GCN-IR-NEXT:    s_addc_u32 s16, s3, -1
+; GCN-IR-NEXT:    s_not_b64 s[4:5], s[10:11]
+; GCN-IR-NEXT:    s_add_u32 s8, s4, s14
+; GCN-IR-NEXT:    s_addc_u32 s9, s5, 0
 ; GCN-IR-NEXT:    s_mov_b64 s[10:11], 0
-; GCN-IR-NEXT:    s_mov_b32 s3, 0
+; GCN-IR-NEXT:    s_mov_b32 s5, 0
 ; GCN-IR-NEXT:  .LBB7_3: ; %udiv-do-while
 ; GCN-IR-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GCN-IR-NEXT:    s_lshl_b64 s[12:13], s[12:13], 1
-; GCN-IR-NEXT:    s_lshr_b32 s2, s7, 31
+; GCN-IR-NEXT:    s_lshr_b32 s4, s7, 31
 ; GCN-IR-NEXT:    s_lshl_b64 s[6:7], s[6:7], 1
-; GCN-IR-NEXT:    s_or_b64 s[12:13], s[12:13], s[2:3]
+; GCN-IR-NEXT:    s_or_b64 s[12:13], s[12:13], s[4:5]
 ; GCN-IR-NEXT:    s_or_b64 s[6:7], s[10:11], s[6:7]
-; GCN-IR-NEXT:    s_sub_u32 s2, s15, s12
-; GCN-IR-NEXT:    s_subb_u32 s2, s16, s13
-; GCN-IR-NEXT:    s_ashr_i32 s10, s2, 31
+; GCN-IR-NEXT:    s_sub_u32 s4, s15, s12
+; GCN-IR-NEXT:    s_subb_u32 s4, s16, s13
+; GCN-IR-NEXT:    s_ashr_i32 s10, s4, 31
 ; GCN-IR-NEXT:    s_mov_b32 s11, s10
-; GCN-IR-NEXT:    s_and_b32 s2, s10, 1
-; GCN-IR-NEXT:    s_and_b64 s[10:11], s[10:11], s[0:1]
+; GCN-IR-NEXT:    s_and_b32 s4, s10, 1
+; GCN-IR-NEXT:    s_and_b64 s[10:11], s[10:11], s[2:3]
 ; GCN-IR-NEXT:    s_sub_u32 s12, s12, s10
 ; GCN-IR-NEXT:    s_subb_u32 s13, s13, s11
 ; GCN-IR-NEXT:    s_add_u32 s8, s8, 1
 ; GCN-IR-NEXT:    s_addc_u32 s9, s9, 0
 ; GCN-IR-NEXT:    v_cmp_eq_u64_e64 s[18:19], s[8:9], 0
-; GCN-IR-NEXT:    s_mov_b64 s[10:11], s[2:3]
+; GCN-IR-NEXT:    s_mov_b64 s[10:11], s[4:5]
 ; GCN-IR-NEXT:    s_and_b64 vcc, exec, s[18:19]
 ; GCN-IR-NEXT:    s_cbranch_vccz .LBB7_3
 ; GCN-IR-NEXT:  .LBB7_4: ; %Flow3
-; GCN-IR-NEXT:    s_lshl_b64 s[0:1], s[6:7], 1
-; GCN-IR-NEXT:    s_or_b64 s[12:13], s[2:3], s[0:1]
+; GCN-IR-NEXT:    s_lshl_b64 s[2:3], s[6:7], 1
+; GCN-IR-NEXT:    s_or_b64 s[6:7], s[4:5], s[2:3]
 ; GCN-IR-NEXT:  .LBB7_5: ; %udiv-end
-; GCN-IR-NEXT:    s_mov_b32 s7, 0xf000
-; GCN-IR-NEXT:    s_mov_b32 s6, -1
-; GCN-IR-NEXT:    v_mov_b32_e32 v0, s13
-; GCN-IR-NEXT:    buffer_store_short v0, off, s[4:7], 0 offset:4
-; GCN-IR-NEXT:    s_waitcnt expcnt(0)
-; GCN-IR-NEXT:    v_mov_b32_e32 v0, s12
-; GCN-IR-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN-IR-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN-IR-NEXT:    s_mov_b32 s3, 0xf000
+; GCN-IR-NEXT:    s_mov_b32 s2, -1
+; GCN-IR-NEXT:    v_mov_b32_e32 v0, s7
+; GCN-IR-NEXT:    v_mov_b32_e32 v1, s6
+; GCN-IR-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-IR-NEXT:    buffer_store_short v0, off, s[0:3], 0 offset:4
+; GCN-IR-NEXT:    buffer_store_dword v1, off, s[0:3], 0
 ; GCN-IR-NEXT:    s_endpgm
   %1 = lshr i48 %x, 24
   %2 = lshr i48 %y, 24
   %result = udiv i48 %1, %2
   store i48 %result, i48 addrspace(1)* %out
   ret void
 }