[AMDGPU] Correct rmw atomics s_waitcnt generation

[t-tye]

Summary

The AMD GPU SIMemoryLegalizer was using the ordering address space
rather than the instruction address space when determining the
s_waitcnt to generate to ensure that a read-modify-write atomic has
completed. This resulted in additional unnecessary counters being
waited on.

[rampitec]

LGTM. Thanks Tony!

llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp

 //===- SIMemoryLegalizer.cpp ----------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 /// \file
if (MOI.getOrdering() == AtomicOrdering::Release ||
                                    MOI.getIsCrossAddressSpaceOrdering(),
                                    Position::BEFORE);
 
     if (MOI.getOrdering() == AtomicOrdering::Acquire ||
         MOI.getOrdering() == AtomicOrdering::AcquireRelease ||
         MOI.getOrdering() == AtomicOrdering::SequentiallyConsistent ||
         MOI.getFailureOrdering() == AtomicOrdering::Acquire ||
         MOI.getFailureOrdering() == AtomicOrdering::SequentiallyConsistent) {
       Changed |= CC->insertWait(MI, MOI.getScope(),
-                                MOI.getOrderingAddrSpace(),
+                                MOI.getInstrAddrSpace(),
                                 isAtomicRet(*MI) ? SIMemOp::LOAD :
                                                    SIMemOp::STORE,
                                 MOI.getIsCrossAddressSpaceOrdering(),
                                 Position::AFTER);
       Changed |= CC->insertAcquire(MI, MOI.getScope(),
                                    MOI.getOrderingAddrSpace(),
                                    Position::AFTER);
     }

llvm/test/CodeGen/AMDGPU/GlobalISel/mubuf-global.ll

 ; GFX6-NEXT:    s_mov_b32 s0, s2
 ; GFX6-NEXT:    s_mov_b32 s1, s3
 ; GFX6-NEXT:    v_mov_b32_e32 v0, 2
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_movk_i32 s4, 0x3ffc
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_add v0, off, s[0:3], s4 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_waitcnt expcnt(0)
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX7-LABEL: mubuf_atomicrmw_sgpr_ptr_offset4095:
 ; GFX7:       ; %bb.0:
 ; GFX7-NEXT:    s_mov_b32 s0, s2
 ; GFX7-NEXT:    s_mov_b32 s1, s3
 ; GFX7-NEXT:    v_mov_b32_e32 v0, 2
 ; GFX7-NEXT:    s_mov_b32 s2, -1
 ; GFX7-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7-NEXT:    s_movk_i32 s4, 0x3ffc
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_atomic_add v0, off, s[0:3], s4 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    ; return to shader part epilog
   %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4095
   %result = atomicrmw add i32 addrspace(1)* %gep, i32 2 seq_cst
   %cast = bitcast i32 %result to float
   ret float %cast
 }
 
 define amdgpu_ps float @mubuf_atomicrmw_sgpr_ptr_offset4294967296(i32 addrspace(1)* inreg %ptr) {
 ; GFX6-LABEL: mubuf_atomicrmw_sgpr_ptr_offset4294967296:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_mov_b32 s4, 0
 ; GFX6-NEXT:    s_mov_b32 s5, 4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX6-NEXT:    s_mov_b32 s0, s2
 ; GFX6-NEXT:    s_mov_b32 s1, s3
 ; GFX6-NEXT:    v_mov_b32_e32 v0, 2
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_mov_b32 s2, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_add v0, v[1:2], s[0:3], 0 addr64 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_waitcnt expcnt(0)
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX7-LABEL: mubuf_atomicrmw_sgpr_ptr_offset4294967296:
 ; GFX7:       ; %bb.0:
 ; GFX7-NEXT:    s_mov_b32 s4, 0
 ; GFX7-NEXT:    s_mov_b32 s5, 4
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX7-NEXT:    s_mov_b32 s0, s2
 ; GFX7-NEXT:    s_mov_b32 s1, s3
 ; GFX7-NEXT:    v_mov_b32_e32 v0, 2
 ; GFX7-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7-NEXT:    s_mov_b32 s2, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s5
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_atomic_add v0, v[1:2], s[0:3], 0 addr64 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    ; return to shader part epilog
   %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4294967296
   %result = atomicrmw add i32 addrspace(1)* %gep, i32 2 seq_cst
   %cast = bitcast i32 %result to float
   ret float %cast
 }
 
 define amdgpu_ps float @mubuf_atomicrmw_vgpr_ptr_offset4095(i32 addrspace(1)* %ptr) {
 ; GFX6-LABEL: mubuf_atomicrmw_vgpr_ptr_offset4095:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    v_mov_b32_e32 v2, 2
 ; GFX6-NEXT:    s_mov_b32 s2, 0
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_mov_b64 s[0:1], 0
 ; GFX6-NEXT:    s_movk_i32 s4, 0x3ffc
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_add v2, v[0:1], s[0:3], s4 addr64 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    v_mov_b32_e32 v0, v2
 ; GFX6-NEXT:    s_waitcnt expcnt(0)
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX7-LABEL: mubuf_atomicrmw_vgpr_ptr_offset4095:
 ; GFX7:       ; %bb.0:
 ; GFX7-NEXT:    v_mov_b32_e32 v2, 2
 ; GFX7-NEXT:    s_mov_b32 s2, 0
 ; GFX7-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7-NEXT:    s_mov_b64 s[0:1], 0
 ; GFX7-NEXT:    s_movk_i32 s4, 0x3ffc
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_atomic_add v2, v[0:1], s[0:3], s4 addr64 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    v_mov_b32_e32 v0, v2
 ; GFX7-NEXT:    ; return to shader part epilog
   %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4095
   %result = atomicrmw add i32 addrspace(1)* %gep, i32 2 seq_cst
   %cast = bitcast i32 %result to float
   ret float %cast
 }
 
 define amdgpu_ps float @mubuf_atomicrmw_vgpr_ptr_offset4294967296(i32 addrspace(1)* %ptr) {
 ; GFX6-LABEL: mubuf_atomicrmw_vgpr_ptr_offset4294967296:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_mov_b32 s0, 0
 ; GFX6-NEXT:    s_mov_b32 s1, 4
 ; GFX6-NEXT:    v_mov_b32_e32 v2, 2
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_mov_b32 s2, s0
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_add v2, v[0:1], s[0:3], 0 addr64 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    v_mov_b32_e32 v0, v2
 ; GFX6-NEXT:    s_waitcnt expcnt(0)
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX7-LABEL: mubuf_atomicrmw_vgpr_ptr_offset4294967296:
 ; GFX7:       ; %bb.0:
 ; GFX7-NEXT:    s_mov_b32 s0, 0
 ; GFX7-NEXT:    s_mov_b32 s1, 4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, 2
 ; GFX7-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7-NEXT:    s_mov_b32 s2, s0
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_atomic_add v2, v[0:1], s[0:3], 0 addr64 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    v_mov_b32_e32 v0, v2
 ; GFX7-NEXT:    ; return to shader part epilog
   %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4294967296
   %result = atomicrmw add i32 addrspace(1)* %gep, i32 2 seq_cst
   %cast = bitcast i32 %result to float
   ret float %cast
 }
 
 define amdgpu_ps float @mubuf_atomicrmw_sgpr_ptr_vgpr_offset(i32 addrspace(1)* inreg %ptr, i32 %voffset) {
 ; GFX6-LABEL: mubuf_atomicrmw_sgpr_ptr_vgpr_offset:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; GFX6-NEXT:    v_lshl_b64 v[0:1], v[0:1], 2
 ; GFX6-NEXT:    s_mov_b32 s0, s2
 ; GFX6-NEXT:    s_mov_b32 s1, s3
 ; GFX6-NEXT:    v_mov_b32_e32 v2, 2
 ; GFX6-NEXT:    s_mov_b32 s2, 0
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_add v2, v[0:1], s[0:3], 0 addr64 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    v_mov_b32_e32 v0, v2
 ; GFX6-NEXT:    s_waitcnt expcnt(0)
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX7-LABEL: mubuf_atomicrmw_sgpr_ptr_vgpr_offset:
 ; GFX7:       ; %bb.0:
 ; GFX7-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; GFX7-NEXT:    v_lshl_b64 v[0:1], v[0:1], 2
 ; GFX7-NEXT:    s_mov_b32 s0, s2
 ; GFX7-NEXT:    s_mov_b32 s1, s3
 ; GFX7-NEXT:    v_mov_b32_e32 v2, 2
 ; GFX7-NEXT:    s_mov_b32 s2, 0
 ; GFX7-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_atomic_add v2, v[0:1], s[0:3], 0 addr64 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    v_mov_b32_e32 v0, v2
 ; GFX7-NEXT:    ; return to shader part epilog
   %gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 %voffset
   %result = atomicrmw add i32 addrspace(1)* %gep, i32 2 seq_cst
   %cast = bitcast i32 %result to float
   ret float %cast
 }
 
 define amdgpu_ps float @mubuf_cmpxchg_sgpr_ptr_offset4095(i32 addrspace(1)* inreg %ptr, i32 %old, i32 %in) {
 ; GFX6-LABEL: mubuf_cmpxchg_sgpr_ptr_offset4095:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_mov_b32 s0, s2
 ; GFX6-NEXT:    s_mov_b32 s1, s3
 ; GFX6-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_movk_i32 s4, 0x3ffc
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[1:2], off, s[0:3], s4 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    v_mov_b32_e32 v0, v1
 ; GFX6-NEXT:    s_waitcnt expcnt(0)
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX7-LABEL: mubuf_cmpxchg_sgpr_ptr_offset4095:
 ; GFX7:       ; %bb.0:
 ; GFX7-NEXT:    s_mov_b32 s0, s2
 ; GFX7-NEXT:    s_mov_b32 s1, s3
 ; GFX7-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX7-NEXT:    s_mov_b32 s2, -1
 ; GFX7-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7-NEXT:    s_movk_i32 s4, 0x3ffc
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_atomic_cmpswap v[1:2], off, s[0:3], s4 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    v_mov_b32_e32 v0, v1
 ; GFX7-NEXT:    ; return to shader part epilog
   %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4095
   %result.struct = cmpxchg i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst seq_cst
   %result = extractvalue { i32, i1 } %result.struct, 0
   %cast = bitcast i32 %result to float
   ret float %cast
 }
 
 define amdgpu_ps float @mubuf_cmpxchg_sgpr_ptr_offset4294967296(i32 addrspace(1)* inreg %ptr, i32 %old, i32 %in) {
 ; GFX6-LABEL: mubuf_cmpxchg_sgpr_ptr_offset4294967296:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_mov_b32 s4, 0
 ; GFX6-NEXT:    s_mov_b32 s5, 4
 ; GFX6-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX6-NEXT:    s_mov_b32 s0, s2
 ; GFX6-NEXT:    s_mov_b32 s1, s3
 ; GFX6-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_mov_b32 s2, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v4, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[1:2], v[3:4], s[0:3], 0 addr64 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    v_mov_b32_e32 v0, v1
 ; GFX6-NEXT:    s_waitcnt expcnt(0)
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX7-LABEL: mubuf_cmpxchg_sgpr_ptr_offset4294967296:
 ; GFX7:       ; %bb.0:
 ; GFX7-NEXT:    s_mov_b32 s4, 0
 ; GFX7-NEXT:    s_mov_b32 s5, 4
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX7-NEXT:    s_mov_b32 s0, s2
 ; GFX7-NEXT:    s_mov_b32 s1, s3
 ; GFX7-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX7-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7-NEXT:    s_mov_b32 s2, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v4, s5
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_atomic_cmpswap v[1:2], v[3:4], s[0:3], 0 addr64 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    v_mov_b32_e32 v0, v1
 ; GFX7-NEXT:    ; return to shader part epilog
   %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4294967296
   %result.struct = cmpxchg i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst seq_cst
   %result = extractvalue { i32, i1 } %result.struct, 0
   %cast = bitcast i32 %result to float
   ret float %cast
 }
 
 define amdgpu_ps float @mubuf_cmpxchg_vgpr_ptr_offset4095(i32 addrspace(1)* %ptr, i32 %old, i32 %in) {
 ; GFX6-LABEL: mubuf_cmpxchg_vgpr_ptr_offset4095:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    v_mov_b32_e32 v4, v2
 ; GFX6-NEXT:    s_mov_b32 s2, 0
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_mov_b64 s[0:1], 0
 ; GFX6-NEXT:    s_movk_i32 s4, 0x3ffc
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[3:4], v[0:1], s[0:3], s4 addr64 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX6-NEXT:    s_waitcnt expcnt(0)
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX7-LABEL: mubuf_cmpxchg_vgpr_ptr_offset4095:
 ; GFX7:       ; %bb.0:
 ; GFX7-NEXT:    v_mov_b32_e32 v4, v2
 ; GFX7-NEXT:    s_mov_b32 s2, 0
 ; GFX7-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7-NEXT:    s_mov_b64 s[0:1], 0
 ; GFX7-NEXT:    s_movk_i32 s4, 0x3ffc
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_atomic_cmpswap v[3:4], v[0:1], s[0:3], s4 addr64 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX7-NEXT:    ; return to shader part epilog
   %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4095
   %result.struct = cmpxchg i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst seq_cst
   %result = extractvalue { i32, i1 } %result.struct, 0
   %cast = bitcast i32 %result to float
   ret float %cast
 }
 
 define amdgpu_ps float @mubuf_cmpxchg_vgpr_ptr_offset4294967296(i32 addrspace(1)* %ptr, i32 %old, i32 %in) {
 ; GFX6-LABEL: mubuf_cmpxchg_vgpr_ptr_offset4294967296:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_mov_b32 s0, 0
 ; GFX6-NEXT:    v_mov_b32_e32 v4, v2
 ; GFX6-NEXT:    s_mov_b32 s1, 4
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_mov_b32 s2, s0
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[3:4], v[0:1], s[0:3], 0 addr64 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX6-NEXT:    s_waitcnt expcnt(0)
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX7-LABEL: mubuf_cmpxchg_vgpr_ptr_offset4294967296:
 ; GFX7:       ; %bb.0:
 ; GFX7-NEXT:    s_mov_b32 s0, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v4, v2
 ; GFX7-NEXT:    s_mov_b32 s1, 4
 ; GFX7-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7-NEXT:    s_mov_b32 s2, s0
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_atomic_cmpswap v[3:4], v[0:1], s[0:3], 0 addr64 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX7-NEXT:    ; return to shader part epilog
   %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4294967296
   %result.struct = cmpxchg i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst seq_cst
   %result = extractvalue { i32, i1 } %result.struct, 0
   %cast = bitcast i32 %result to float
   ret float %cast
 }
 
 define amdgpu_ps float @mubuf_cmpxchg_sgpr_ptr_vgpr_offset(i32 addrspace(1)* inreg %ptr, i32 %voffset, i32 %old, i32 %in) {
 ; GFX6-LABEL: mubuf_cmpxchg_sgpr_ptr_vgpr_offset:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX6-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; GFX6-NEXT:    v_lshl_b64 v[0:1], v[0:1], 2
 ; GFX6-NEXT:    s_mov_b32 s0, s2
 ; GFX6-NEXT:    s_mov_b32 s1, s3
 ; GFX6-NEXT:    s_mov_b32 s2, 0
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[2:3], v[0:1], s[0:3], 0 addr64 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    v_mov_b32_e32 v0, v2
 ; GFX6-NEXT:    s_waitcnt expcnt(0)
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX7-LABEL: mubuf_cmpxchg_sgpr_ptr_vgpr_offset:
 ; GFX7:       ; %bb.0:
 ; GFX7-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX7-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; GFX7-NEXT:    v_lshl_b64 v[0:1], v[0:1], 2
 ; GFX7-NEXT:    s_mov_b32 s0, s2
 ; GFX7-NEXT:    s_mov_b32 s1, s3
 ; GFX7-NEXT:    s_mov_b32 s2, 0
 ; GFX7-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    buffer_atomic_cmpswap v[2:3], v[0:1], s[0:3], 0 addr64 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1
 ; GFX7-NEXT:    v_mov_b32_e32 v0, v2
 ; GFX7-NEXT:    ; return to shader part epilog
   %gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 %voffset
   %result.struct = cmpxchg i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst seq_cst
   %result = extractvalue { i32, i1 } %result.struct, 0
   %cast = bitcast i32 %result to float
   ret float %cast
 }

llvm/test/CodeGen/AMDGPU/atomic_optimizations_local_pointer.ll

 ; GFX1064-NEXT:    s_cbranch_execz BB0_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, local_var32@abs32@lo
 ; GFX1064-NEXT:    v_mul_u32_u24_e64 v1, s2, 5
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_add_rtn_u32 v1, v2, v1
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB0_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; 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

 ; GFX1032-NEXT:    s_cbranch_execz BB0_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    s_bcnt1_i32_b32 s2, s2
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, local_var32@abs32@lo
 ; GFX1032-NEXT:    v_mul_u32_u24_e64 v1, s2, 5
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_add_rtn_u32 v1, v2, v1
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB0_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
 ; 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

 ; GFX1064-NEXT:    s_bcnt1_i32_b64 s3, s[6:7]
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, local_var32@abs32@lo
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    s_mul_i32 s3, s2, s3
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, s3
 ; 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 vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB1_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    v_mul_lo_u32 v0, s2, v0
 ; GFX1064-NEXT:    v_readfirstlane_b32 s0, v1
 ; GFX1064-NEXT:    s_mov_b32 s7, 0x31016000

 ; GFX1032-NEXT:    s_bcnt1_i32_b32 s1, s3
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, local_var32@abs32@lo
 ; 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_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 vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB1_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s0
 ; 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

 ; GFX1064-NEXT:    s_cbranch_execz BB2_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; GFX1064-NEXT:    v_mov_b32_e32 v4, s7
 ; GFX1064-NEXT:    s_mov_b32 s3, s7
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_add_rtn_u32 v0, v7, v4
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB2_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; 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

 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz BB2_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; 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, v7, v4
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB2_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
 ; 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

 ; GFX1064-NEXT:    s_cbranch_execz BB3_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; GFX1064-NEXT:    v_mov_b32_e32 v4, s7
 ; GFX1064-NEXT:    s_mov_b32 s3, s7
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_add_rtn_u32 v0, v7, v4
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB3_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; 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

 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz BB3_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; 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, v7, v4
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB3_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
 ; 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

 ; GFX1064-NEXT:    s_cbranch_execz BB4_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; GFX1064-NEXT:    v_mov_b32_e32 v4, s7
 ; GFX1064-NEXT:    s_mov_b32 s3, s7
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_add_rtn_u32 v0, v7, v4
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB4_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; 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

 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz BB4_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; 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, v7, v4
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB4_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
 ; 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

 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX1064-NEXT:    v_mov_b32_e32 v3, local_var64@abs32@lo
 ; GFX1064-NEXT:    v_mul_hi_u32_u24_e64 v2, s4, 5
 ; GFX1064-NEXT:    v_mul_u32_u24_e64 v1, s4, 5
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_add_rtn_u64 v[1:2], v3, v[1:2]
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB5_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
 ; GFX1064-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v2
 ; GFX1064-NEXT:    v_mad_u64_u32 v[0:1], s[2:3], v0, 5, s[2:3]
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000

 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    s_bcnt1_i32_b32 s3, s3
 ; GFX1032-NEXT:    v_mov_b32_e32 v3, local_var64@abs32@lo
 ; GFX1032-NEXT:    v_mul_hi_u32_u24_e64 v2, s3, 5
 ; GFX1032-NEXT:    v_mul_u32_u24_e64 v1, s3, 5
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_add_rtn_u64 v[1:2], v3, v[1:2]
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB5_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s2
 ; GFX1032-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v2
 ; GFX1032-NEXT:    v_mad_u64_u32 v[0:1], s2, v0, 5, s[2:3]
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000

 ; GFX1064-NEXT:    s_mul_hi_u32 s8, s2, s6
 ; GFX1064-NEXT:    s_mul_i32 s6, s2, s6
 ; GFX1064-NEXT:    s_add_i32 s8, s8, s7
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, s6
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, s8
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_add_rtn_u64 v[1:2], v3, v[1:2]
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB6_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    v_mul_lo_u32 v3, s3, v0
 ; GFX1064-NEXT:    v_mul_hi_u32 v4, s2, v0
 ; GFX1064-NEXT:    v_mul_lo_u32 v0, s2, v0

 ; GFX1032-NEXT:    s_mul_hi_u32 s7, s2, s5
 ; GFX1032-NEXT:    s_mul_i32 s5, s2, s5
 ; GFX1032-NEXT:    s_add_i32 s7, s7, s6
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, s7
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_add_rtn_u64 v[1:2], v3, v[1:2]
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB6_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s4
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    v_mul_lo_u32 v3, s3, v0
 ; GFX1032-NEXT:    v_mul_hi_u32 v4, s2, v0
 ; GFX1032-NEXT:    v_mul_lo_u32 v0, s2, v0

 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, local_var64@abs32@lo
 ; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1064-NEXT:    s_mov_b32 s2, -1
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_add_rtn_u64 v[0:1], v2, v[0:1]
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: add_i64_varying:
 ; GFX1032:       ; %bb.0: ; %entry
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, local_var64@abs32@lo
 ; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_add_rtn_u64 v[0:1], v2, v[0:1]
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
   %zext = zext i32 %lane to i64
   %old = atomicrmw add i64 addrspace(3)* @local_var64, i64 %zext acq_rel
   store i64 %old, i64 addrspace(1)* %out

 ; GFX1064-NEXT:    s_cbranch_execz BB8_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, local_var32@abs32@lo
 ; GFX1064-NEXT:    v_mul_u32_u24_e64 v1, s2, 5
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_sub_rtn_u32 v1, v2, v1
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB8_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; 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

 ; GFX1032-NEXT:    s_cbranch_execz BB8_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    s_bcnt1_i32_b32 s2, s2
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, local_var32@abs32@lo
 ; GFX1032-NEXT:    v_mul_u32_u24_e64 v1, s2, 5
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_sub_rtn_u32 v1, v2, v1
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB8_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
 ; 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

 ; GFX1064-NEXT:    s_bcnt1_i32_b64 s3, s[6:7]
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, local_var32@abs32@lo
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    s_mul_i32 s3, s2, s3
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, s3
 ; 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 vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB9_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    v_mul_lo_u32 v0, s2, v0
 ; GFX1064-NEXT:    v_readfirstlane_b32 s0, v1
 ; GFX1064-NEXT:    s_mov_b32 s7, 0x31016000

 ; GFX1032-NEXT:    s_bcnt1_i32_b32 s1, s3
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, local_var32@abs32@lo
 ; 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_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 vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB9_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s0
 ; 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

 ; GFX1064-NEXT:    s_cbranch_execz BB10_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; GFX1064-NEXT:    v_mov_b32_e32 v4, s7
 ; GFX1064-NEXT:    s_mov_b32 s3, s7
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_sub_rtn_u32 v0, v7, v4
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB10_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; 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

 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz BB10_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; 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, v7, v4
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB10_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
 ; 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

 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
 ; GFX1064-NEXT:    v_mov_b32_e32 v3, local_var64@abs32@lo
 ; GFX1064-NEXT:    v_mul_hi_u32_u24_e64 v2, s4, 5
 ; GFX1064-NEXT:    v_mul_u32_u24_e64 v1, s4, 5
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_sub_rtn_u64 v[1:2], v3, v[1:2]
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB11_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
 ; GFX1064-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX1064-NEXT:    v_mul_u32_u24_e32 v1, 5, v0
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v2
 ; GFX1064-NEXT:    v_mul_hi_u32_u24_e32 v2, 5, v0

 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    s_bcnt1_i32_b32 s3, s3
 ; GFX1032-NEXT:    v_mov_b32_e32 v3, local_var64@abs32@lo
 ; GFX1032-NEXT:    v_mul_hi_u32_u24_e64 v2, s3, 5
 ; GFX1032-NEXT:    v_mul_u32_u24_e64 v1, s3, 5
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_sub_rtn_u64 v[1:2], v3, v[1:2]
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB11_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s2
 ; GFX1032-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX1032-NEXT:    v_mul_u32_u24_e32 v1, 5, v0
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v2
 ; GFX1032-NEXT:    v_mul_hi_u32_u24_e32 v2, 5, v0

 ; GFX1064-NEXT:    s_mul_hi_u32 s8, s2, s6
 ; GFX1064-NEXT:    s_mul_i32 s6, s2, s6
 ; GFX1064-NEXT:    s_add_i32 s8, s8, s7
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, s6
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, s8
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_sub_rtn_u64 v[1:2], v3, v[1:2]
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB12_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    v_mul_lo_u32 v3, s3, v0
 ; GFX1064-NEXT:    v_mul_hi_u32 v4, s2, v0
 ; GFX1064-NEXT:    v_mul_lo_u32 v0, s2, v0

 ; GFX1032-NEXT:    s_mul_hi_u32 s7, s2, s5
 ; GFX1032-NEXT:    s_mul_i32 s5, s2, s5
 ; GFX1032-NEXT:    s_add_i32 s7, s7, s6
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, s7
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_sub_rtn_u64 v[1:2], v3, v[1:2]
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB12_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s4
 ; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    v_mul_lo_u32 v3, s3, v0
 ; GFX1032-NEXT:    v_mul_hi_u32 v4, s2, v0
 ; GFX1032-NEXT:    v_mul_lo_u32 v0, s2, v0

 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, local_var64@abs32@lo
 ; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1064-NEXT:    s_mov_b32 s2, -1
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_sub_rtn_u64 v[0:1], v2, v[0:1]
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX1064-NEXT:    s_endpgm
 ;
 ; GFX1032-LABEL: sub_i64_varying:
 ; GFX1032:       ; %bb.0: ; %entry
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, local_var64@abs32@lo
 ; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1032-NEXT:    s_mov_b32 s2, -1
 ; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1032-NEXT:    ds_sub_rtn_u64 v[0:1], v2, v[0:1]
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GFX1032-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
   %zext = zext i32 %lane to i64
   %old = atomicrmw sub i64 addrspace(3)* @local_var64, i64 %zext acq_rel
   store i64 %old, i64 addrspace(1)* %out

 ; GFX1064-NEXT:    s_cbranch_execz BB14_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; GFX1064-NEXT:    v_mov_b32_e32 v4, s7
 ; GFX1064-NEXT:    s_mov_b32 s3, s7
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_and_rtn_b32 v0, v7, v4
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB14_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; 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

 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz BB14_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; 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, v7, v4
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB14_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
 ; 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

 ; GFX1064-NEXT:    s_cbranch_execz BB15_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; GFX1064-NEXT:    v_mov_b32_e32 v4, s7
 ; GFX1064-NEXT:    s_mov_b32 s3, s7
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_or_rtn_b32 v0, v7, v4
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB15_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; 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

 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz BB15_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; 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, v7, v4
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB15_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
 ; 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

 ; GFX1064-NEXT:    s_cbranch_execz BB16_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; GFX1064-NEXT:    v_mov_b32_e32 v4, s7
 ; GFX1064-NEXT:    s_mov_b32 s3, s7
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_xor_rtn_b32 v0, v7, v4
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB16_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; 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

 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz BB16_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; 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, v7, v4
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB16_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
 ; 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

 ; GFX1064-NEXT:    s_cbranch_execz BB17_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; GFX1064-NEXT:    v_mov_b32_e32 v4, s7
 ; GFX1064-NEXT:    s_mov_b32 s3, s7
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_max_rtn_i32 v0, v7, v4
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB17_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, v1
 ; GFX1064-NEXT:    v_max_i32_e32 v0, s3, v0
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000

 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz BB17_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; 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, v7, v4
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB17_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, v1
 ; GFX1032-NEXT:    v_max_i32_e32 v0, s3, v0
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000

 ; GFX1064-NEXT:    s_cbranch_execz BB18_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, local_var64@abs32@lo
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 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 vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB18_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
 ; 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

 ; GFX1032-NEXT:    s_cbranch_execz BB18_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, local_var64@abs32@lo
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 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 vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB18_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s2
 ; 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

 ; GFX1064-NEXT:    s_cbranch_execz BB19_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; GFX1064-NEXT:    v_mov_b32_e32 v4, s7
 ; GFX1064-NEXT:    s_mov_b32 s3, s7
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_min_rtn_i32 v0, v7, v4
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB19_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX1064-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, v1
 ; GFX1064-NEXT:    v_min_i32_e32 v0, s3, v0
 ; GFX1064-NEXT:    s_mov_b32 s3, 0x31016000

 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz BB19_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; 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, v7, v4
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB19_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
 ; GFX1032-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, v1
 ; GFX1032-NEXT:    v_min_i32_e32 v0, s3, v0
 ; GFX1032-NEXT:    s_mov_b32 s3, 0x31016000

 ; GFX1064-NEXT:    s_cbranch_execz BB20_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, local_var64@abs32@lo
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 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 vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB20_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
 ; 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

 ; GFX1032-NEXT:    s_cbranch_execz BB20_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, local_var64@abs32@lo
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 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 vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB20_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s2
 ; 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

 ; GFX1064-NEXT:    s_cbranch_execz BB21_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; GFX1064-NEXT:    v_mov_b32_e32 v4, s7
 ; GFX1064-NEXT:    s_mov_b32 s3, s7
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_max_rtn_u32 v0, v7, v4
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB21_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; 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

 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz BB21_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; 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, v7, v4
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB21_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
 ; 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

 ; GFX1064-NEXT:    s_cbranch_execz BB22_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, local_var64@abs32@lo
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 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 vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB22_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
 ; 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

 ; GFX1032-NEXT:    s_cbranch_execz BB22_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, local_var64@abs32@lo
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 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 vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB22_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s2
 ; 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

 ; GFX1064-NEXT:    s_cbranch_execz BB23_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; GFX1064-NEXT:    v_mov_b32_e32 v4, s7
 ; GFX1064-NEXT:    s_mov_b32 s3, s7
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX1064-NEXT:    ds_min_rtn_u32 v0, v7, v4
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB23_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; 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

 ; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
 ; GFX1032-NEXT:    s_cbranch_execz BB23_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v7, local_var32@abs32@lo
 ; 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, v7, v4
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB23_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s3
 ; 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

 ; GFX1064-NEXT:    s_cbranch_execz BB24_2
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1064-NEXT:    v_mov_b32_e32 v2, local_var64@abs32@lo
 ; GFX1064-NEXT:    v_mov_b32_e32 v1, 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 vmcnt(0) lgkmcnt(0)
+; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1064-NEXT:    buffer_gl0_inv
 ; GFX1064-NEXT:  BB24_2:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
 ; 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

 ; GFX1032-NEXT:    s_cbranch_execz BB24_2
 ; GFX1032-NEXT:  ; %bb.1:
 ; GFX1032-NEXT:    v_mov_b32_e32 v0, 5
 ; GFX1032-NEXT:    v_mov_b32_e32 v2, local_var64@abs32@lo
 ; GFX1032-NEXT:    v_mov_b32_e32 v1, 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 vmcnt(0) lgkmcnt(0)
+; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1032-NEXT:    buffer_gl0_inv
 ; GFX1032-NEXT:  BB24_2:
 ; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s2
 ; 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

llvm/test/CodeGen/AMDGPU/atomicrmw-nand.ll

 ; GCN-NEXT:  BB1_1: ; %atomicrmw.start
 ; GCN-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    v_mov_b32_e32 v3, v2
 ; GCN-NEXT:    v_not_b32_e32 v2, v3
 ; GCN-NEXT:    v_or_b32_e32 v2, -5, v2
 ; GCN-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GCN-NEXT:    global_atomic_cmpswap v2, v[0:1], v[2:3], off glc
-; GCN-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    buffer_wbinvl1_vol
 ; GCN-NEXT:    v_cmp_eq_u32_e32 vcc, v2, v3
 ; GCN-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GCN-NEXT:    s_andn2_b64 exec, exec, s[4:5]
 ; GCN-NEXT:    s_cbranch_execnz BB1_1
 ; GCN-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GCN-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GCN-NEXT:    v_mov_b32_e32 v0, v2

llvm/test/CodeGen/AMDGPU/global-saddr-atomics.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 < %s | FileCheck -check-prefixes=GCN,GFX9 %s
 ; RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -mattr=+wavefrontsize64 < %s | FileCheck -check-prefixes=GCN,GFX10 %s
 
 ; Test using saddr addressing mode of global_* flat atomic instructions.
 
 define amdgpu_ps void @global_xchg_saddr_i32_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_xchg_saddr_i32_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_swap v0, v1, s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_xchg_saddr_i32_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_swap v0, v1, s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %unused = atomicrmw xchg i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   ret void
 }
 
 ; Maximum positive offset on gfx10
 define amdgpu_ps void @global_xchg_saddr_i32_nortn_offset_2047(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_xchg_saddr_i32_nortn_offset_2047:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_swap v0, v1, s[2:3] offset:2047
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_xchg_saddr_i32_nortn_offset_2047:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_swap v0, v1, s[2:3] offset:2047
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 2047
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %unused = atomicrmw xchg i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   ret void
 }
 
 ; Maximum negative offset on gfx10
 define amdgpu_ps void @global_xchg_saddr_i32_nortn_offset_neg2048(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_xchg_saddr_i32_nortn_offset_neg2048:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_swap v0, v1, s[2:3] offset:-2048
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_xchg_saddr_i32_nortn_offset_neg2048:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_swap v0, v1, s[2:3] offset:-2048
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -2048
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %unused = atomicrmw xchg i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps float @global_xchg_saddr_i32_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_xchg_saddr_i32_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_swap v0, v0, v1, s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_xchg_saddr_i32_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_swap v0, v0, v1, s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %rtn = atomicrmw xchg i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps float @global_xchg_saddr_i32_rtn_2048(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_xchg_saddr_i32_rtn_2048:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_swap v0, v0, v1, s[2:3] offset:2048 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_xchg_saddr_i32_rtn_2048:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_add_co_u32_e64 v0, s[0:1], s2, v0
 ; GFX10-NEXT:    v_add_co_ci_u32_e64 v3, s[0:1], s3, 0, s[0:1]
 ; GFX10-NEXT:    v_add_co_u32_e64 v2, vcc, 0x800, v0
 ; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc, 0, v3, vcc
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_swap v0, v[2:3], v1, off glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 2048
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %rtn = atomicrmw xchg i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps float @global_xchg_saddr_i32_rtn_neg2048(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_xchg_saddr_i32_rtn_neg2048:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_swap v0, v0, v1, s[2:3] offset:-2048 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_xchg_saddr_i32_rtn_neg2048:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_swap v0, v0, v1, s[2:3] offset:-2048 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -2048
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %rtn = atomicrmw xchg i32 addrspace(1)* %cast.gep1, i32 %data seq_cst

 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    ds_read_b64 v[2:3], v2
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_readfirstlane_b32 s0, v2
 ; GFX9-NEXT:    v_readfirstlane_b32 s1, v3
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    s_nop 3
 ; GFX9-NEXT:    global_atomic_swap v0, v0, v1, s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_xchg_saddr_uniform_ptr_in_vgprs_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-NEXT:    ds_read_b64 v[2:3], v2
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    v_readfirstlane_b32 s0, v2
 ; GFX10-NEXT:    v_readfirstlane_b32 s1, v3
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    s_nop 2
 ; GFX10-NEXT:    global_atomic_swap v0, v0, v1, s[0:1] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %sbase = load i8 addrspace(1)*, i8 addrspace(1)* addrspace(3)* @ptr.in.lds
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %rtn = atomicrmw xchg i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 ; Base pointer is uniform, but also in VGPRs, with imm offset
 define amdgpu_ps float @global_xchg_saddr_uniform_ptr_in_vgprs_rtn_immoffset(i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_xchg_saddr_uniform_ptr_in_vgprs_rtn_immoffset:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    ds_read_b64 v[2:3], v2
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_readfirstlane_b32 s0, v2
 ; GFX9-NEXT:    v_readfirstlane_b32 s1, v3
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    s_nop 3
 ; GFX9-NEXT:    global_atomic_swap v0, v0, v1, s[0:1] offset:42 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_xchg_saddr_uniform_ptr_in_vgprs_rtn_immoffset:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-NEXT:    ds_read_b64 v[2:3], v2
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    v_readfirstlane_b32 s0, v2
 ; GFX10-NEXT:    v_readfirstlane_b32 s1, v3
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    s_nop 2
 ; GFX10-NEXT:    global_atomic_swap v0, v0, v1, s[0:1] offset:42 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %sbase = load i8 addrspace(1)*, i8 addrspace(1)* addrspace(3)* @ptr.in.lds
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 42
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*

 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    ds_read_b64 v[2:3], v2
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_readfirstlane_b32 s0, v2
 ; GFX9-NEXT:    v_readfirstlane_b32 s1, v3
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    s_nop 3
 ; GFX9-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_xchg_saddr_uniform_ptr_in_vgprs_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-NEXT:    ds_read_b64 v[2:3], v2
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    v_readfirstlane_b32 s0, v2
 ; GFX10-NEXT:    v_readfirstlane_b32 s1, v3
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    s_nop 2
 ; GFX10-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %sbase = load i8 addrspace(1)*, i8 addrspace(1)* addrspace(3)* @ptr.in.lds
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %unused = atomicrmw xchg i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   ret void
 }
 
 ; Base pointer is uniform, but also in VGPRs, with imm offset
 define amdgpu_ps void @global_xchg_saddr_uniform_ptr_in_vgprs_nortn_immoffset(i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_xchg_saddr_uniform_ptr_in_vgprs_nortn_immoffset:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    ds_read_b64 v[2:3], v2
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_readfirstlane_b32 s0, v2
 ; GFX9-NEXT:    v_readfirstlane_b32 s1, v3
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    s_nop 3
 ; GFX9-NEXT:    global_atomic_swap v0, v1, s[0:1] offset:42
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_xchg_saddr_uniform_ptr_in_vgprs_nortn_immoffset:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-NEXT:    ds_read_b64 v[2:3], v2
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    v_readfirstlane_b32 s0, v2
 ; GFX10-NEXT:    v_readfirstlane_b32 s1, v3
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    s_nop 2
 ; GFX10-NEXT:    global_atomic_swap v0, v1, s[0:1] offset:42
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %sbase = load i8 addrspace(1)*, i8 addrspace(1)* addrspace(3)* @ptr.in.lds
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 42

 ; atomicrmw xchg
 ; --------------------------------------------------------------------------------
 
 define amdgpu_ps <2 x float> @global_xchg_saddr_i64_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_xchg_saddr_i64_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_swap_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_xchg_saddr_i64_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_swap_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %rtn = atomicrmw xchg i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps <2 x float> @global_xchg_saddr_i64_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_xchg_saddr_i64_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_swap_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_xchg_saddr_i64_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_swap_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %rtn = atomicrmw xchg i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps void @global_xchg_saddr_i64_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_xchg_saddr_i64_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_swap_x2 v0, v[1:2], s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_xchg_saddr_i64_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_swap_x2 v0, v[1:2], s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %unused = atomicrmw xchg i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_xchg_saddr_i64_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_xchg_saddr_i64_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_swap_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_xchg_saddr_i64_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_swap_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %unused = atomicrmw xchg i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   ret void
 }
 
 ; --------------------------------------------------------------------------------
 ; atomicrmw add
 ; --------------------------------------------------------------------------------
 
 define amdgpu_ps float @global_add_saddr_i32_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_add_saddr_i32_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_add v0, v0, v1, s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_add_saddr_i32_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_add v0, v0, v1, s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %rtn = atomicrmw add i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps float @global_add_saddr_i32_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_add_saddr_i32_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_add v0, v0, v1, s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_add_saddr_i32_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_add v0, v0, v1, s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %rtn = atomicrmw add i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps void @global_add_saddr_i32_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_add_saddr_i32_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_add v0, v1, s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_add_saddr_i32_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_add v0, v1, s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %unused = atomicrmw add i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_add_saddr_i32_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_add_saddr_i32_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_add v0, v1, s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_add_saddr_i32_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_add v0, v1, s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %unused = atomicrmw add i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps <2 x float> @global_add_saddr_i64_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_add_saddr_i64_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_add_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_add_saddr_i64_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_add_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %rtn = atomicrmw add i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps <2 x float> @global_add_saddr_i64_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_add_saddr_i64_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_add_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_add_saddr_i64_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_add_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %rtn = atomicrmw add i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps void @global_add_saddr_i64_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_add_saddr_i64_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_add_x2 v0, v[1:2], s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_add_saddr_i64_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_add_x2 v0, v[1:2], s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %unused = atomicrmw add i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_add_saddr_i64_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_add_saddr_i64_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_add_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_add_saddr_i64_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_add_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %unused = atomicrmw add i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   ret void
 }
 
 ; --------------------------------------------------------------------------------
 ; atomicrmw sub
 ; --------------------------------------------------------------------------------
 
 define amdgpu_ps float @global_sub_saddr_i32_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_sub_saddr_i32_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_sub v0, v0, v1, s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_sub_saddr_i32_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_sub v0, v0, v1, s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %rtn = atomicrmw sub i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps float @global_sub_saddr_i32_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_sub_saddr_i32_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_sub v0, v0, v1, s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_sub_saddr_i32_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_sub v0, v0, v1, s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %rtn = atomicrmw sub i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps void @global_sub_saddr_i32_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_sub_saddr_i32_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_sub v0, v1, s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_sub_saddr_i32_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_sub v0, v1, s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %unused = atomicrmw sub i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_sub_saddr_i32_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_sub_saddr_i32_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_sub v0, v1, s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_sub_saddr_i32_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_sub v0, v1, s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %unused = atomicrmw sub i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps <2 x float> @global_sub_saddr_i64_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_sub_saddr_i64_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_sub_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_sub_saddr_i64_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_sub_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %rtn = atomicrmw sub i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps <2 x float> @global_sub_saddr_i64_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_sub_saddr_i64_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_sub_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_sub_saddr_i64_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_sub_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %rtn = atomicrmw sub i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps void @global_sub_saddr_i64_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_sub_saddr_i64_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_sub_x2 v0, v[1:2], s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_sub_saddr_i64_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_sub_x2 v0, v[1:2], s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %unused = atomicrmw sub i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_sub_saddr_i64_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_sub_saddr_i64_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_sub_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_sub_saddr_i64_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_sub_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %unused = atomicrmw sub i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   ret void
 }
 
 ; --------------------------------------------------------------------------------
 ; atomicrmw and
 ; --------------------------------------------------------------------------------
 
 define amdgpu_ps float @global_and_saddr_i32_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_and_saddr_i32_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_and v0, v0, v1, s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_and_saddr_i32_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_and v0, v0, v1, s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %rtn = atomicrmw and i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps float @global_and_saddr_i32_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_and_saddr_i32_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_and v0, v0, v1, s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_and_saddr_i32_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_and v0, v0, v1, s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %rtn = atomicrmw and i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps void @global_and_saddr_i32_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_and_saddr_i32_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_and v0, v1, s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_and_saddr_i32_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_and v0, v1, s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %unused = atomicrmw and i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_and_saddr_i32_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_and_saddr_i32_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_and v0, v1, s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_and_saddr_i32_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_and v0, v1, s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %unused = atomicrmw and i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps <2 x float> @global_and_saddr_i64_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_and_saddr_i64_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_and_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_and_saddr_i64_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_and_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %rtn = atomicrmw and i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps <2 x float> @global_and_saddr_i64_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_and_saddr_i64_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_and_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_and_saddr_i64_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_and_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %rtn = atomicrmw and i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps void @global_and_saddr_i64_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_and_saddr_i64_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_and_x2 v0, v[1:2], s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_and_saddr_i64_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_and_x2 v0, v[1:2], s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %unused = atomicrmw and i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_and_saddr_i64_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_and_saddr_i64_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_and_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_and_saddr_i64_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_and_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %unused = atomicrmw and i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   ret void
 }
 
 ; --------------------------------------------------------------------------------
 ; atomicrmw or
 ; --------------------------------------------------------------------------------
 
 define amdgpu_ps float @global_or_saddr_i32_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_or_saddr_i32_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_or v0, v0, v1, s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_or_saddr_i32_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_or v0, v0, v1, s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %rtn = atomicrmw or i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps float @global_or_saddr_i32_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_or_saddr_i32_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_or v0, v0, v1, s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_or_saddr_i32_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_or v0, v0, v1, s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %rtn = atomicrmw or i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps void @global_or_saddr_i32_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_or_saddr_i32_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_or v0, v1, s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_or_saddr_i32_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_or v0, v1, s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %unused = atomicrmw or i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_or_saddr_i32_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_or_saddr_i32_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_or v0, v1, s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_or_saddr_i32_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_or v0, v1, s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %unused = atomicrmw or i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps <2 x float> @global_or_saddr_i64_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_or_saddr_i64_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_or_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_or_saddr_i64_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_or_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %rtn = atomicrmw or i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps <2 x float> @global_or_saddr_i64_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_or_saddr_i64_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_or_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_or_saddr_i64_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_or_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %rtn = atomicrmw or i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps void @global_or_saddr_i64_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_or_saddr_i64_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_or_x2 v0, v[1:2], s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_or_saddr_i64_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_or_x2 v0, v[1:2], s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %unused = atomicrmw or i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_or_saddr_i64_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_or_saddr_i64_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_or_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_or_saddr_i64_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_or_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %unused = atomicrmw or i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   ret void
 }
 
 ; --------------------------------------------------------------------------------
 ; atomicrmw xor
 ; --------------------------------------------------------------------------------
 
 define amdgpu_ps float @global_xor_saddr_i32_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_xor_saddr_i32_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_xor v0, v0, v1, s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_xor_saddr_i32_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_xor v0, v0, v1, s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %rtn = atomicrmw xor i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps float @global_xor_saddr_i32_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_xor_saddr_i32_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_xor v0, v0, v1, s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_xor_saddr_i32_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_xor v0, v0, v1, s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %rtn = atomicrmw xor i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps void @global_xor_saddr_i32_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_xor_saddr_i32_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_xor v0, v1, s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_xor_saddr_i32_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_xor v0, v1, s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %unused = atomicrmw xor i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_xor_saddr_i32_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_xor_saddr_i32_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_xor v0, v1, s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_xor_saddr_i32_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_xor v0, v1, s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %unused = atomicrmw xor i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps <2 x float> @global_xor_saddr_i64_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_xor_saddr_i64_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_xor_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_xor_saddr_i64_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_xor_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %rtn = atomicrmw xor i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps <2 x float> @global_xor_saddr_i64_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_xor_saddr_i64_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_xor_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_xor_saddr_i64_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_xor_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %rtn = atomicrmw xor i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps void @global_xor_saddr_i64_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_xor_saddr_i64_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_xor_x2 v0, v[1:2], s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_xor_saddr_i64_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_xor_x2 v0, v[1:2], s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %unused = atomicrmw xor i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_xor_saddr_i64_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_xor_saddr_i64_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_xor_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_xor_saddr_i64_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_xor_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %unused = atomicrmw xor i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   ret void
 }
 
 ; --------------------------------------------------------------------------------
 ; atomicrmw max
 ; --------------------------------------------------------------------------------
 
 define amdgpu_ps float @global_max_saddr_i32_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_max_saddr_i32_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_smax v0, v0, v1, s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_max_saddr_i32_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_smax v0, v0, v1, s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %rtn = atomicrmw max i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps float @global_max_saddr_i32_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_max_saddr_i32_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_smax v0, v0, v1, s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_max_saddr_i32_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_smax v0, v0, v1, s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %rtn = atomicrmw max i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps void @global_max_saddr_i32_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_max_saddr_i32_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_smax v0, v1, s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_max_saddr_i32_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_smax v0, v1, s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %unused = atomicrmw max i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_max_saddr_i32_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_max_saddr_i32_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_smax v0, v1, s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_max_saddr_i32_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_smax v0, v1, s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %unused = atomicrmw max i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps <2 x float> @global_max_saddr_i64_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_max_saddr_i64_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_smax_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_max_saddr_i64_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_smax_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %rtn = atomicrmw max i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps <2 x float> @global_max_saddr_i64_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_max_saddr_i64_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_smax_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_max_saddr_i64_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_smax_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %rtn = atomicrmw max i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps void @global_max_saddr_i64_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_max_saddr_i64_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_smax_x2 v0, v[1:2], s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_max_saddr_i64_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_smax_x2 v0, v[1:2], s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %unused = atomicrmw max i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_max_saddr_i64_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_max_saddr_i64_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_smax_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_max_saddr_i64_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_smax_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %unused = atomicrmw max i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   ret void
 }
 
 ; --------------------------------------------------------------------------------
 ; atomicrmw min
 ; --------------------------------------------------------------------------------
 
 define amdgpu_ps float @global_min_saddr_i32_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_min_saddr_i32_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_smin v0, v0, v1, s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_min_saddr_i32_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_smin v0, v0, v1, s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %rtn = atomicrmw min i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps float @global_min_saddr_i32_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_min_saddr_i32_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_smin v0, v0, v1, s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_min_saddr_i32_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_smin v0, v0, v1, s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %rtn = atomicrmw min i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps void @global_min_saddr_i32_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_min_saddr_i32_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_smin v0, v1, s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_min_saddr_i32_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_smin v0, v1, s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %unused = atomicrmw min i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_min_saddr_i32_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_min_saddr_i32_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_smin v0, v1, s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_min_saddr_i32_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_smin v0, v1, s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %unused = atomicrmw min i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps <2 x float> @global_min_saddr_i64_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_min_saddr_i64_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_smin_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_min_saddr_i64_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_smin_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %rtn = atomicrmw min i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps <2 x float> @global_min_saddr_i64_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_min_saddr_i64_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_smin_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_min_saddr_i64_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_smin_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %rtn = atomicrmw min i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps void @global_min_saddr_i64_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_min_saddr_i64_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_smin_x2 v0, v[1:2], s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_min_saddr_i64_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_smin_x2 v0, v[1:2], s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %unused = atomicrmw min i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_min_saddr_i64_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_min_saddr_i64_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_smin_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_min_saddr_i64_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_smin_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %unused = atomicrmw min i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   ret void
 }
 
 ; --------------------------------------------------------------------------------
 ; atomicrmw umax
 ; --------------------------------------------------------------------------------
 
 define amdgpu_ps float @global_umax_saddr_i32_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_umax_saddr_i32_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_umax v0, v0, v1, s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_umax_saddr_i32_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_umax v0, v0, v1, s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %rtn = atomicrmw umax i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps float @global_umax_saddr_i32_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_umax_saddr_i32_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_umax v0, v0, v1, s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_umax_saddr_i32_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_umax v0, v0, v1, s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %rtn = atomicrmw umax i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps void @global_umax_saddr_i32_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_umax_saddr_i32_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_umax v0, v1, s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_umax_saddr_i32_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_umax v0, v1, s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %unused = atomicrmw umax i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_umax_saddr_i32_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_umax_saddr_i32_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_umax v0, v1, s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_umax_saddr_i32_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_umax v0, v1, s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %unused = atomicrmw umax i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps <2 x float> @global_umax_saddr_i64_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_umax_saddr_i64_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_umax_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_umax_saddr_i64_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_umax_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %rtn = atomicrmw umax i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps <2 x float> @global_umax_saddr_i64_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_umax_saddr_i64_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_umax_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_umax_saddr_i64_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_umax_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %rtn = atomicrmw umax i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps void @global_umax_saddr_i64_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_umax_saddr_i64_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_umax_x2 v0, v[1:2], s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_umax_saddr_i64_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_umax_x2 v0, v[1:2], s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %unused = atomicrmw umax i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_umax_saddr_i64_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_umax_saddr_i64_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_umax_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_umax_saddr_i64_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_umax_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %unused = atomicrmw umax i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   ret void
 }
 
 ; --------------------------------------------------------------------------------
 ; atomicrmw umin
 ; --------------------------------------------------------------------------------
 
 define amdgpu_ps float @global_umin_saddr_i32_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_umin_saddr_i32_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_umin v0, v0, v1, s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_umin_saddr_i32_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_umin v0, v0, v1, s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %rtn = atomicrmw umin i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps float @global_umin_saddr_i32_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_umin_saddr_i32_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_umin v0, v0, v1, s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_umin_saddr_i32_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_umin v0, v0, v1, s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %rtn = atomicrmw umin i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps void @global_umin_saddr_i32_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_umin_saddr_i32_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_umin v0, v1, s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_umin_saddr_i32_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_umin v0, v1, s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %unused = atomicrmw umin i32 addrspace(1)* %cast.gep0, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_umin_saddr_i32_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX9-LABEL: global_umin_saddr_i32_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_umin v0, v1, s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_umin_saddr_i32_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_umin v0, v1, s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %unused = atomicrmw umin i32 addrspace(1)* %cast.gep1, i32 %data seq_cst
   ret void
 }
 
 define amdgpu_ps <2 x float> @global_umin_saddr_i64_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_umin_saddr_i64_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_umin_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_umin_saddr_i64_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_umin_x2 v[0:1], v0, v[1:2], s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %rtn = atomicrmw umin i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps <2 x float> @global_umin_saddr_i64_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_umin_saddr_i64_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_umin_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_umin_saddr_i64_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_umin_x2 v[0:1], v0, v[1:2], s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %rtn = atomicrmw umin i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps void @global_umin_saddr_i64_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_umin_saddr_i64_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_umin_x2 v0, v[1:2], s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_umin_saddr_i64_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_umin_x2 v0, v[1:2], s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %unused = atomicrmw umin i64 addrspace(1)* %cast.gep0, i64 %data seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_umin_saddr_i64_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %data) {
 ; GFX9-LABEL: global_umin_saddr_i64_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_umin_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_umin_saddr_i64_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_umin_x2 v0, v[1:2], s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %unused = atomicrmw umin i64 addrspace(1)* %cast.gep1, i64 %data seq_cst
   ret void
 }
 
 ; --------------------------------------------------------------------------------
 ; cmpxchg
 ; --------------------------------------------------------------------------------
 
 define amdgpu_ps float @global_cmpxchg_saddr_i32_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %cmp, i32 %data) {
 ; GFX9-LABEL: global_cmpxchg_saddr_i32_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_cmpswap v0, v0, v[2:3], s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_cmpxchg_saddr_i32_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_cmpswap v0, v0, v[2:3], s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %cmpxchg = cmpxchg i32 addrspace(1)* %cast.gep0, i32 %cmp, i32 %data seq_cst seq_cst
   %rtn = extractvalue { i32, i1 } %cmpxchg, 0
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps float @global_cmpxchg_saddr_i32_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %cmp, i32 %data) {
 ; GFX9-LABEL: global_cmpxchg_saddr_i32_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_cmpswap v0, v0, v[2:3], s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_cmpxchg_saddr_i32_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_cmpswap v0, v0, v[2:3], s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %cmpxchg = cmpxchg i32 addrspace(1)* %cast.gep1, i32 %cmp, i32 %data seq_cst seq_cst
   %rtn = extractvalue { i32, i1 } %cmpxchg, 0
   %cast.rtn = bitcast i32 %rtn to float
   ret float %cast.rtn
 }
 
 define amdgpu_ps void @global_cmpxchg_saddr_i32_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %cmp, i32 %data) {
 ; GFX9-LABEL: global_cmpxchg_saddr_i32_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_cmpswap v0, v[2:3], s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_cmpxchg_saddr_i32_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_cmpswap v0, v[2:3], s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i32 addrspace(1)*
   %unused = cmpxchg i32 addrspace(1)* %cast.gep0, i32 %cmp, i32 %data seq_cst seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_cmpxchg_saddr_i32_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i32 %cmp, i32 %data) {
 ; GFX9-LABEL: global_cmpxchg_saddr_i32_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_cmpswap v0, v[2:3], s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_cmpxchg_saddr_i32_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_cmpswap v0, v[2:3], s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i32 addrspace(1)*
   %unused = cmpxchg i32 addrspace(1)* %cast.gep1, i32 %cmp, i32 %data seq_cst seq_cst
   ret void
 }
 
 define amdgpu_ps <2 x float> @global_cmpxchg_saddr_i64_rtn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %cmp, i64 %data) {
 ; GFX9-LABEL: global_cmpxchg_saddr_i64_rtn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_mov_b32_e32 v6, v2
 ; GFX9-NEXT:    v_mov_b32_e32 v5, v1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_cmpswap_x2 v[0:1], v0, v[3:6], s[2:3] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_cmpxchg_saddr_i64_rtn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_mov_b32_e32 v6, v2
 ; GFX10-NEXT:    v_mov_b32_e32 v5, v1
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_cmpswap_x2 v[0:1], v0, v[3:6], s[2:3] glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %cmpxchg = cmpxchg i64 addrspace(1)* %cast.gep0, i64 %cmp, i64 %data seq_cst seq_cst
   %rtn = extractvalue { i64, i1 } %cmpxchg, 0
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps <2 x float> @global_cmpxchg_saddr_i64_rtn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %cmp, i64 %data) {
 ; GFX9-LABEL: global_cmpxchg_saddr_i64_rtn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_mov_b32_e32 v6, v2
 ; GFX9-NEXT:    v_mov_b32_e32 v5, v1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_cmpswap_x2 v[0:1], v0, v[3:6], s[2:3] offset:-128 glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: global_cmpxchg_saddr_i64_rtn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_mov_b32_e32 v6, v2
 ; GFX10-NEXT:    v_mov_b32_e32 v5, v1
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_cmpswap_x2 v[0:1], v0, v[3:6], s[2:3] offset:-128 glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    ; return to shader part epilog
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*
   %cmpxchg = cmpxchg i64 addrspace(1)* %cast.gep1, i64 %cmp, i64 %data seq_cst seq_cst
   %rtn = extractvalue { i64, i1 } %cmpxchg, 0
   %cast.rtn = bitcast i64 %rtn to <2 x float>
   ret <2 x float> %cast.rtn
 }
 
 define amdgpu_ps void @global_cmpxchg_saddr_i64_nortn(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %cmp, i64 %data) {
 ; GFX9-LABEL: global_cmpxchg_saddr_i64_nortn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_mov_b32_e32 v6, v2
 ; GFX9-NEXT:    v_mov_b32_e32 v5, v1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_cmpswap_x2 v0, v[3:6], s[2:3]
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_cmpxchg_saddr_i64_nortn:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_mov_b32_e32 v6, v2
 ; GFX10-NEXT:    v_mov_b32_e32 v5, v1
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_cmpswap_x2 v0, v[3:6], s[2:3]
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %cast.gep0 = bitcast i8 addrspace(1)* %gep0 to i64 addrspace(1)*
   %unused = cmpxchg i64 addrspace(1)* %cast.gep0, i64 %cmp, i64 %data seq_cst seq_cst
   ret void
 }
 
 define amdgpu_ps void @global_cmpxchg_saddr_i64_nortn_neg128(i8 addrspace(1)* inreg %sbase, i32 %voffset, i64 %cmp, i64 %data) {
 ; GFX9-LABEL: global_cmpxchg_saddr_i64_nortn_neg128:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_mov_b32_e32 v6, v2
 ; GFX9-NEXT:    v_mov_b32_e32 v5, v1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_atomic_cmpswap_x2 v0, v[3:6], s[2:3] offset:-128
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    buffer_wbinvl1
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: global_cmpxchg_saddr_i64_nortn_neg128:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_mov_b32_e32 v6, v2
 ; GFX10-NEXT:    v_mov_b32_e32 v5, v1
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    global_atomic_cmpswap_x2 v0, v[3:6], s[2:3] offset:-128
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    buffer_gl0_inv
 ; GFX10-NEXT:    buffer_gl1_inv
 ; GFX10-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, i8 addrspace(1)* %sbase, i64 %zext.offset
   %gep1 = getelementptr inbounds i8, i8 addrspace(1)* %gep0, i64 -128
   %cast.gep1 = bitcast i8 addrspace(1)* %gep1 to i64 addrspace(1)*

llvm/test/CodeGen/AMDGPU/memory-legalizer-global-agent.ll

 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dword s4, s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    buffer_atomic_swap v0, off, s[0:3], 0
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_acquire_atomicrmw:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dword s2, s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    flat_atomic_swap v[0:1], v2
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_acquire_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_acquire_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-CU-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_acquire_atomicrmw:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dword s0, s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_swap v0, off, s[4:7], 0
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in syncscope("agent") acquire
   ret void
 }
 
 define amdgpu_kernel void @global_agent_release_atomicrmw(

 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dword s4, s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_swap v0, off, s[0:3], 0
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_acq_rel_atomicrmw:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dword s2, s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_swap v[0:1], v2
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_acq_rel_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_acq_rel_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_acq_rel_atomicrmw:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dword s0, s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_swap v0, off, s[4:7], 0
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in syncscope("agent") acq_rel
   ret void
 }
 
 define amdgpu_kernel void @global_agent_seq_cst_atomicrmw(
 ; GFX6-LABEL: global_agent_seq_cst_atomicrmw:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dword s4, s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_swap v0, off, s[0:3], 0
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_seq_cst_atomicrmw:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dword s2, s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_swap v[0:1], v2
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_seq_cst_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_seq_cst_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_seq_cst_atomicrmw:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dword s0, s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_swap v0, off, s[4:7], 0
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in syncscope("agent") seq_cst
   ret void
 }
 
 define amdgpu_kernel void @global_agent_acquire_ret_atomicrmw(
 ; GFX6-LABEL: global_agent_acquire_ret_atomicrmw:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dword s4, s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    buffer_atomic_swap v0, off, s[0:3], 0 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_acquire_ret_atomicrmw:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dword s2, s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    flat_atomic_swap v2, v[0:1], v2 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_acquire_ret_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    global_atomic_swap v1, v0, v1, s[0:1] glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_acquire_ret_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-CU-NEXT:    global_atomic_swap v1, v0, v1, s[0:1] glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_acquire_ret_atomicrmw:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dword s0, s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_swap v0, off, s[4:7], 0 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in syncscope("agent") acquire
   store i32 %val, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_agent_acq_rel_ret_atomicrmw(
 ; GFX6-LABEL: global_agent_acq_rel_ret_atomicrmw:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dword s4, s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_swap v0, off, s[0:3], 0 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_acq_rel_ret_atomicrmw:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dword s2, s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_swap v2, v[0:1], v2 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_acq_rel_ret_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_swap v1, v0, v1, s[0:1] glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_acq_rel_ret_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_swap v1, v0, v1, s[0:1] glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_acq_rel_ret_atomicrmw:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dword s0, s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_swap v0, off, s[4:7], 0 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in syncscope("agent") acq_rel
   store i32 %val, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_agent_seq_cst_ret_atomicrmw(
 ; GFX6-LABEL: global_agent_seq_cst_ret_atomicrmw:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dword s4, s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_swap v0, off, s[0:3], 0 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_seq_cst_ret_atomicrmw:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dword s2, s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_swap v2, v[0:1], v2 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_seq_cst_ret_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_swap v1, v0, v1, s[0:1] glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_seq_cst_ret_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_swap v1, v0, v1, s[0:1] glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_seq_cst_ret_atomicrmw:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dword s0, s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_swap v0, off, s[4:7], 0 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in syncscope("agent") seq_cst
   store i32 %val, i32 addrspace(1)* %out, align 4
   ret void
 }

 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_acquire_monotonic_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_acquire_monotonic_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_acquire_monotonic_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_acquire_monotonic_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("agent") acquire monotonic
   ret void
 }
 

 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_acq_rel_monotonic_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_acq_rel_monotonic_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_acq_rel_monotonic_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_acq_rel_monotonic_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("agent") acq_rel monotonic
   ret void
 }
 
 define amdgpu_kernel void @global_agent_seq_cst_monotonic_cmpxchg(
 ; GFX6-LABEL: global_agent_seq_cst_monotonic_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_seq_cst_monotonic_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_seq_cst_monotonic_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_seq_cst_monotonic_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_seq_cst_monotonic_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("agent") seq_cst monotonic
   ret void
 }
 
 define amdgpu_kernel void @global_agent_acquire_acquire_cmpxchg(
 ; GFX6-LABEL: global_agent_acquire_acquire_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_acquire_acquire_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_acquire_acquire_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_acquire_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_acquire_acquire_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("agent") acquire acquire
   ret void
 }
 
 define amdgpu_kernel void @global_agent_release_acquire_cmpxchg(
 ; GFX6-LABEL: global_agent_release_acquire_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_release_acquire_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_release_acquire_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_release_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_release_acquire_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("agent") release acquire
   ret void
 }
 
 define amdgpu_kernel void @global_agent_acq_rel_acquire_cmpxchg(
 ; GFX6-LABEL: global_agent_acq_rel_acquire_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_acq_rel_acquire_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_acq_rel_acquire_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_acq_rel_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_acq_rel_acquire_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("agent") acq_rel acquire
   ret void
 }
 
 define amdgpu_kernel void @global_agent_seq_cst_acquire_cmpxchg(
 ; GFX6-LABEL: global_agent_seq_cst_acquire_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_seq_cst_acquire_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_seq_cst_acquire_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_seq_cst_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_seq_cst_acquire_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("agent") seq_cst acquire
   ret void
 }
 
 define amdgpu_kernel void @global_agent_seq_cst_seq_cst_cmpxchg(
 ; GFX6-LABEL: global_agent_seq_cst_seq_cst_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_seq_cst_seq_cst_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_seq_cst_seq_cst_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_seq_cst_seq_cst_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_seq_cst_seq_cst_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("agent") seq_cst seq_cst
   ret void
 }
 
 define amdgpu_kernel void @global_agent_acquire_monotonic_ret_cmpxchg(
 ; GFX6-LABEL: global_agent_acquire_monotonic_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_acquire_monotonic_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_acquire_monotonic_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_acquire_monotonic_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_acquire_monotonic_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("agent") acquire monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_agent_acq_rel_monotonic_ret_cmpxchg(
 ; GFX6-LABEL: global_agent_acq_rel_monotonic_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_acq_rel_monotonic_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_acq_rel_monotonic_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_acq_rel_monotonic_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_acq_rel_monotonic_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("agent") acq_rel monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_agent_seq_cst_monotonic_ret_cmpxchg(
 ; GFX6-LABEL: global_agent_seq_cst_monotonic_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_seq_cst_monotonic_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_seq_cst_monotonic_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_seq_cst_monotonic_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_seq_cst_monotonic_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("agent") seq_cst monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_agent_acquire_acquire_ret_cmpxchg(
 ; GFX6-LABEL: global_agent_acquire_acquire_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_acquire_acquire_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_acquire_acquire_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_acquire_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_acquire_acquire_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("agent") acquire acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_agent_release_acquire_ret_cmpxchg(
 ; GFX6-LABEL: global_agent_release_acquire_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_release_acquire_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_release_acquire_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_release_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_release_acquire_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("agent") release acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_agent_acq_rel_acquire_ret_cmpxchg(
 ; GFX6-LABEL: global_agent_acq_rel_acquire_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_acq_rel_acquire_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_acq_rel_acquire_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_acq_rel_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_acq_rel_acquire_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("agent") acq_rel acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_agent_seq_cst_acquire_ret_cmpxchg(
 ; GFX6-LABEL: global_agent_seq_cst_acquire_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_seq_cst_acquire_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_seq_cst_acquire_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_seq_cst_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_seq_cst_acquire_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("agent") seq_cst acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_agent_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX6-LABEL: global_agent_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_agent_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_agent_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_agent_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_agent_seq_cst_seq_cst_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("agent") seq_cst seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4

llvm/test/CodeGen/AMDGPU/memory-legalizer-global-system.ll

 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dword s4, s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    buffer_atomic_swap v0, off, s[0:3], 0
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_acquire_atomicrmw:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dword s2, s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    flat_atomic_swap v[0:1], v2
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_acquire_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_acquire_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-CU-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_acquire_atomicrmw:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dword s0, s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_swap v0, off, s[4:7], 0
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in acquire
   ret void
 }
 
 define amdgpu_kernel void @global_system_release_atomicrmw(

 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dword s4, s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_swap v0, off, s[0:3], 0
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_acq_rel_atomicrmw:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dword s2, s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_swap v[0:1], v2
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_acq_rel_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_acq_rel_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_acq_rel_atomicrmw:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dword s0, s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_swap v0, off, s[4:7], 0
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in acq_rel
   ret void
 }
 
 define amdgpu_kernel void @global_system_seq_cst_atomicrmw(
 ; GFX6-LABEL: global_system_seq_cst_atomicrmw:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dword s4, s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_swap v0, off, s[0:3], 0
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_seq_cst_atomicrmw:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dword s2, s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_swap v[0:1], v2
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_seq_cst_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_seq_cst_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_seq_cst_atomicrmw:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dword s0, s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_swap v0, off, s[4:7], 0
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in seq_cst
   ret void
 }
 
 define amdgpu_kernel void @global_system_acquire_ret_atomicrmw(
 ; GFX6-LABEL: global_system_acquire_ret_atomicrmw:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dword s4, s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    buffer_atomic_swap v0, off, s[0:3], 0 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_acquire_ret_atomicrmw:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dword s2, s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    flat_atomic_swap v2, v[0:1], v2 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_acquire_ret_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    global_atomic_swap v1, v0, v1, s[0:1] glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_acquire_ret_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-CU-NEXT:    global_atomic_swap v1, v0, v1, s[0:1] glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_acquire_ret_atomicrmw:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dword s0, s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_swap v0, off, s[4:7], 0 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in acquire
   store i32 %val, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_system_acq_rel_ret_atomicrmw(
 ; GFX6-LABEL: global_system_acq_rel_ret_atomicrmw:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dword s4, s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_swap v0, off, s[0:3], 0 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_acq_rel_ret_atomicrmw:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dword s2, s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_swap v2, v[0:1], v2 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_acq_rel_ret_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_swap v1, v0, v1, s[0:1] glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_acq_rel_ret_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_swap v1, v0, v1, s[0:1] glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_acq_rel_ret_atomicrmw:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dword s0, s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_swap v0, off, s[4:7], 0 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in acq_rel
   store i32 %val, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_system_seq_cst_ret_atomicrmw(
 ; GFX6-LABEL: global_system_seq_cst_ret_atomicrmw:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dword s4, s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_swap v0, off, s[0:3], 0 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_seq_cst_ret_atomicrmw:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dword s2, s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_swap v2, v[0:1], v2 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_seq_cst_ret_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_swap v1, v0, v1, s[0:1] glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_seq_cst_ret_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_swap v1, v0, v1, s[0:1] glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_seq_cst_ret_atomicrmw:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dword s0, s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_swap v0, off, s[4:7], 0 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in seq_cst
   store i32 %val, i32 addrspace(1)* %out, align 4
   ret void
 }

 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_acquire_monotonic_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_acquire_monotonic_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_acquire_monotonic_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_acquire_monotonic_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in acquire monotonic
   ret void
 }
 

 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_acq_rel_monotonic_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_acq_rel_monotonic_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_acq_rel_monotonic_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_acq_rel_monotonic_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in acq_rel monotonic
   ret void
 }
 
 define amdgpu_kernel void @global_system_seq_cst_monotonic_cmpxchg(
 ; GFX6-LABEL: global_system_seq_cst_monotonic_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_seq_cst_monotonic_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_seq_cst_monotonic_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_seq_cst_monotonic_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_seq_cst_monotonic_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst monotonic
   ret void
 }
 
 define amdgpu_kernel void @global_system_acquire_acquire_cmpxchg(
 ; GFX6-LABEL: global_system_acquire_acquire_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_acquire_acquire_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_acquire_acquire_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_acquire_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_acquire_acquire_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in acquire acquire
   ret void
 }
 
 define amdgpu_kernel void @global_system_release_acquire_cmpxchg(
 ; GFX6-LABEL: global_system_release_acquire_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_release_acquire_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_release_acquire_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_release_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_release_acquire_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in release acquire
   ret void
 }
 
 define amdgpu_kernel void @global_system_acq_rel_acquire_cmpxchg(
 ; GFX6-LABEL: global_system_acq_rel_acquire_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_acq_rel_acquire_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_acq_rel_acquire_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_acq_rel_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_acq_rel_acquire_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in acq_rel acquire
   ret void
 }
 
 define amdgpu_kernel void @global_system_seq_cst_acquire_cmpxchg(
 ; GFX6-LABEL: global_system_seq_cst_acquire_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_seq_cst_acquire_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_seq_cst_acquire_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_seq_cst_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_seq_cst_acquire_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst acquire
   ret void
 }
 
 define amdgpu_kernel void @global_system_seq_cst_seq_cst_cmpxchg(
 ; GFX6-LABEL: global_system_seq_cst_seq_cst_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_seq_cst_seq_cst_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_seq_cst_seq_cst_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_seq_cst_seq_cst_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_seq_cst_seq_cst_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst seq_cst
   ret void
 }
 
 define amdgpu_kernel void @global_system_acquire_monotonic_ret_cmpxchg(
 ; GFX6-LABEL: global_system_acquire_monotonic_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_acquire_monotonic_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_acquire_monotonic_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_acquire_monotonic_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_acquire_monotonic_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in acquire monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_system_acq_rel_monotonic_ret_cmpxchg(
 ; GFX6-LABEL: global_system_acq_rel_monotonic_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_acq_rel_monotonic_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_acq_rel_monotonic_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_acq_rel_monotonic_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_acq_rel_monotonic_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in acq_rel monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_system_seq_cst_monotonic_ret_cmpxchg(
 ; GFX6-LABEL: global_system_seq_cst_monotonic_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_seq_cst_monotonic_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_seq_cst_monotonic_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_seq_cst_monotonic_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_seq_cst_monotonic_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_system_acquire_acquire_ret_cmpxchg(
 ; GFX6-LABEL: global_system_acquire_acquire_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_acquire_acquire_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_acquire_acquire_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_acquire_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_acquire_acquire_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in acquire acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_system_release_acquire_ret_cmpxchg(
 ; GFX6-LABEL: global_system_release_acquire_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_release_acquire_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_release_acquire_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_release_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_release_acquire_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in release acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_system_acq_rel_acquire_ret_cmpxchg(
 ; GFX6-LABEL: global_system_acq_rel_acquire_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_acq_rel_acquire_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_acq_rel_acquire_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_acq_rel_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_acq_rel_acquire_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in acq_rel acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_system_seq_cst_acquire_ret_cmpxchg(
 ; GFX6-LABEL: global_system_seq_cst_acquire_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_seq_cst_acquire_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_seq_cst_acquire_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_seq_cst_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_seq_cst_acquire_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_system_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX6-LABEL: global_system_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_wbinvl1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_system_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    buffer_wbinvl1_vol
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_system_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    buffer_gl1_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_system_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    buffer_gl0_inv
 ; GFX10-CU-NEXT:    buffer_gl1_inv
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_system_seq_cst_seq_cst_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4

llvm/test/CodeGen/AMDGPU/memory-legalizer-global-workgroup.ll

 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dword s4, s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    buffer_atomic_swap v0, off, s[0:3], 0
-; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_acquire_atomicrmw:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dword s2, s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    flat_atomic_swap v[0:1], v2
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_acquire_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_acquire_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-CU-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_acquire_atomicrmw:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dword s0, s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_swap v0, off, s[4:7], 0
-; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in syncscope("workgroup") acquire
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_release_atomicrmw(

 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dword s4, s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_swap v0, off, s[0:3], 0
-; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_acq_rel_atomicrmw:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dword s2, s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_swap v[0:1], v2
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_acq_rel_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_acq_rel_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_acq_rel_atomicrmw:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dword s0, s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_swap v0, off, s[4:7], 0
-; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in syncscope("workgroup") acq_rel
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_seq_cst_atomicrmw(
 ; GFX6-LABEL: global_workgroup_seq_cst_atomicrmw:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dword s4, s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_swap v0, off, s[0:3], 0
-; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_seq_cst_atomicrmw:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dword s2, s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_swap v[0:1], v2
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_seq_cst_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_seq_cst_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    global_atomic_swap v0, v1, s[0:1]
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_seq_cst_atomicrmw:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dword s0, s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_swap v0, off, s[4:7], 0
-; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in syncscope("workgroup") seq_cst
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_acquire_ret_atomicrmw(
 ; GFX6-LABEL: global_workgroup_acquire_ret_atomicrmw:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dword s4, s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    buffer_atomic_swap v0, off, s[0:3], 0 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_acquire_ret_atomicrmw:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dword s2, s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    flat_atomic_swap v2, v[0:1], v2 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_acquire_ret_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    global_atomic_swap v1, v0, v1, s[0:1] glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_acquire_ret_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-CU-NEXT:    global_atomic_swap v1, v0, v1, s[0:1] glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_acquire_ret_atomicrmw:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dword s0, s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_swap v0, off, s[4:7], 0 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in syncscope("workgroup") acquire
   store i32 %val, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_acq_rel_ret_atomicrmw(
 ; GFX6-LABEL: global_workgroup_acq_rel_ret_atomicrmw:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dword s4, s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_swap v0, off, s[0:3], 0 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_acq_rel_ret_atomicrmw:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dword s2, s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_swap v2, v[0:1], v2 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_acq_rel_ret_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_swap v1, v0, v1, s[0:1] glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_acq_rel_ret_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    global_atomic_swap v1, v0, v1, s[0:1] glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_acq_rel_ret_atomicrmw:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dword s0, s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_swap v0, off, s[4:7], 0 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in syncscope("workgroup") acq_rel
   store i32 %val, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_seq_cst_ret_atomicrmw(
 ; GFX6-LABEL: global_workgroup_seq_cst_ret_atomicrmw:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dword s4, s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_swap v0, off, s[0:3], 0 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_seq_cst_ret_atomicrmw:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dword s2, s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_swap v2, v[0:1], v2 glc
-; GFX7-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_seq_cst_ret_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_swap v1, v0, v1, s[0:1] glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_seq_cst_ret_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dword s2, s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    global_atomic_swap v1, v0, v1, s[0:1] glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    global_store_dword v0, v1, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_seq_cst_ret_atomicrmw:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dword s0, s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_swap v0, off, s[4:7], 0 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in syncscope("workgroup") seq_cst
   store i32 %val, i32 addrspace(1)* %out, align 4
   ret void
 }

 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_acquire_monotonic_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_acquire_monotonic_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_acquire_monotonic_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_acquire_monotonic_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("workgroup") acquire monotonic
   ret void
 }
 

 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_acq_rel_monotonic_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_acq_rel_monotonic_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_acq_rel_monotonic_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_acq_rel_monotonic_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("workgroup") acq_rel monotonic
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_seq_cst_monotonic_cmpxchg(
 ; GFX6-LABEL: global_workgroup_seq_cst_monotonic_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_seq_cst_monotonic_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_seq_cst_monotonic_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_seq_cst_monotonic_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_seq_cst_monotonic_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("workgroup") seq_cst monotonic
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_acquire_acquire_cmpxchg(
 ; GFX6-LABEL: global_workgroup_acquire_acquire_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_acquire_acquire_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_acquire_acquire_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_acquire_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_acquire_acquire_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("workgroup") acquire acquire
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_release_acquire_cmpxchg(
 ; GFX6-LABEL: global_workgroup_release_acquire_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_release_acquire_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_release_acquire_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_release_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_release_acquire_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("workgroup") release acquire
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_acq_rel_acquire_cmpxchg(
 ; GFX6-LABEL: global_workgroup_acq_rel_acquire_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_acq_rel_acquire_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_acq_rel_acquire_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_acq_rel_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_acq_rel_acquire_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("workgroup") acq_rel acquire
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_seq_cst_acquire_cmpxchg(
 ; GFX6-LABEL: global_workgroup_seq_cst_acquire_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_seq_cst_acquire_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_seq_cst_acquire_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_seq_cst_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_seq_cst_acquire_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("workgroup") seq_cst acquire
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_seq_cst_seq_cst_cmpxchg(
 ; GFX6-LABEL: global_workgroup_seq_cst_seq_cst_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16
-; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_seq_cst_seq_cst_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s0, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v[0:1], v[2:3]
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_seq_cst_seq_cst_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_seq_cst_seq_cst_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v2, v[0:1], s[0:1] offset:16
-; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_seq_cst_seq_cst_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16
-; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("workgroup") seq_cst seq_cst
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_acquire_monotonic_ret_cmpxchg(
 ; GFX6-LABEL: global_workgroup_acquire_monotonic_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_acquire_monotonic_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_acquire_monotonic_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_acquire_monotonic_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_acquire_monotonic_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("workgroup") acquire monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_acq_rel_monotonic_ret_cmpxchg(
 ; GFX6-LABEL: global_workgroup_acq_rel_monotonic_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_acq_rel_monotonic_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_acq_rel_monotonic_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_acq_rel_monotonic_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_acq_rel_monotonic_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("workgroup") acq_rel monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_seq_cst_monotonic_ret_cmpxchg(
 ; GFX6-LABEL: global_workgroup_seq_cst_monotonic_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_seq_cst_monotonic_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_seq_cst_monotonic_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_seq_cst_monotonic_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_seq_cst_monotonic_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("workgroup") seq_cst monotonic
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_acquire_acquire_ret_cmpxchg(
 ; GFX6-LABEL: global_workgroup_acquire_acquire_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_acquire_acquire_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_acquire_acquire_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_acquire_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_acquire_acquire_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("workgroup") acquire acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_release_acquire_ret_cmpxchg(
 ; GFX6-LABEL: global_workgroup_release_acquire_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_release_acquire_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_release_acquire_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_release_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_release_acquire_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("workgroup") release acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_acq_rel_acquire_ret_cmpxchg(
 ; GFX6-LABEL: global_workgroup_acq_rel_acquire_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_acq_rel_acquire_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_acq_rel_acquire_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_acq_rel_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_acq_rel_acquire_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("workgroup") acq_rel acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_seq_cst_acquire_ret_cmpxchg(
 ; GFX6-LABEL: global_workgroup_seq_cst_acquire_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_seq_cst_acquire_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_seq_cst_acquire_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_seq_cst_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_seq_cst_acquire_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("workgroup") seq_cst acquire
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4
   ret void
 }
 
 define amdgpu_kernel void @global_workgroup_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX6-LABEL: global_workgroup_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX6:       ; %bb.0: ; %entry
 ; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX6-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x2
 ; GFX6-NEXT:    s_mov_b32 s3, 0x100f000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 offset:16 glc
-; GFX6-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX7-LABEL: global_workgroup_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX7:       ; %bb.0: ; %entry
 ; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX7-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    s_add_u32 s4, s0, 16
 ; GFX7-NEXT:    s_addc_u32 s5, s1, 0
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s5
 ; GFX7-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
-; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX7-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX7-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; GFX7-NEXT:    flat_store_dword v[0:1], v2
 ; GFX7-NEXT:    s_endpgm
 ;
 ; GFX10-WGP-LABEL: global_workgroup_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_clause 0x1
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: global_workgroup_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_clause 0x1
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
-; GFX10-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-CU-NEXT:    global_store_dword v2, v0, s[0:1]
 ; GFX10-CU-NEXT:    s_endpgm
 ;
 ; SKIP-CACHE-INV-LABEL: global_workgroup_seq_cst_seq_cst_ret_cmpxchg:
 ; SKIP-CACHE-INV:       ; %bb.0: ; %entry
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
 ; SKIP-CACHE-INV-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xb
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
 ; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
 ; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
 ; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
-; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
 ; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SKIP-CACHE-INV-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
   %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in syncscope("workgroup") seq_cst seq_cst
   %val0 = extractvalue { i32, i1 } %val, 0
   store i32 %val0, i32 addrspace(1)* %out, align 4

llvm/test/CodeGen/AMDGPU/memory-legalizer-local-agent.ll

 ;
 ; GFX10-WGP-LABEL: local_agent_acquire_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-WGP-NEXT:    ds_wrxchg_rtn_b32 v0, v0, v1
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_acquire_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_wrxchg_rtn_b32 v0, v0, v1
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_acq_rel_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_wrxchg_rtn_b32 v0, v0, v1
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_seq_cst_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ;
 ; GFX10-WGP-LABEL: local_agent_acquire_ret_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-WGP-NEXT:    ds_wrxchg_rtn_b32 v1, v0, v1
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_acquire_ret_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_wrxchg_rtn_b32 v1, v0, v1
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_acq_rel_ret_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_wrxchg_rtn_b32 v1, v0, v1
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_seq_cst_ret_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_acquire_monotonic_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_acq_rel_monotonic_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_seq_cst_monotonic_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_acquire_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_release_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_acq_rel_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_seq_cst_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_seq_cst_seq_cst_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-LABEL: local_agent_acquire_monotonic_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_acquire_monotonic_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_acq_rel_monotonic_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_seq_cst_monotonic_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-LABEL: local_agent_acquire_acquire_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_acquire_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_release_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_acq_rel_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_seq_cst_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_agent_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

llvm/test/CodeGen/AMDGPU/memory-legalizer-local-system.ll

 ;
 ; GFX10-WGP-LABEL: local_system_acquire_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-WGP-NEXT:    ds_wrxchg_rtn_b32 v0, v0, v1
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_acquire_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_wrxchg_rtn_b32 v0, v0, v1
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_acq_rel_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_wrxchg_rtn_b32 v0, v0, v1
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_seq_cst_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ;
 ; GFX10-WGP-LABEL: local_system_acquire_ret_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-WGP-NEXT:    ds_wrxchg_rtn_b32 v1, v0, v1
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_acquire_ret_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_wrxchg_rtn_b32 v1, v0, v1
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_acq_rel_ret_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_wrxchg_rtn_b32 v1, v0, v1
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_seq_cst_ret_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_acquire_monotonic_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_acq_rel_monotonic_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_seq_cst_monotonic_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_acquire_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_release_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_acq_rel_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_seq_cst_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_seq_cst_seq_cst_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-LABEL: local_system_acquire_monotonic_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_acquire_monotonic_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_acq_rel_monotonic_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_seq_cst_monotonic_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-LABEL: local_system_acquire_acquire_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_acquire_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_release_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_acq_rel_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_seq_cst_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_system_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

llvm/test/CodeGen/AMDGPU/memory-legalizer-local-workgroup.ll

 ;
 ; GFX10-WGP-LABEL: local_workgroup_acquire_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-WGP-NEXT:    ds_wrxchg_rtn_b32 v0, v0, v1
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_acquire_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_wrxchg_rtn_b32 v0, v0, v1
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_acq_rel_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_wrxchg_rtn_b32 v0, v0, v1
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_seq_cst_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ;
 ; GFX10-WGP-LABEL: local_workgroup_acquire_ret_atomicrmw:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-WGP-NEXT:    ds_wrxchg_rtn_b32 v1, v0, v1
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_acquire_ret_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_wrxchg_rtn_b32 v1, v0, v1
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_acq_rel_ret_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_wrxchg_rtn_b32 v1, v0, v1
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_seq_cst_ret_atomicrmw:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_acquire_monotonic_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_acq_rel_monotonic_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_seq_cst_monotonic_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_acquire_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_release_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_acq_rel_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_seq_cst_acquire_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_b32 v0, v1, v2 offset:16
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_seq_cst_seq_cst_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0

 ; GFX10-WGP-LABEL: local_workgroup_acquire_monotonic_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_acquire_monotonic_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_acq_rel_monotonic_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_seq_cst_monotonic_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-LABEL: local_workgroup_acquire_acquire_ret_cmpxchg:
 ; GFX10-WGP:       ; %bb.0: ; %entry
 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_acquire_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_release_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_acq_rel_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_seq_cst_acquire_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX10-WGP-NEXT:    v_mov_b32_e32 v2, s1
 ; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX10-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-WGP-NEXT:    ds_cmpst_rtn_b32 v1, v0, v1, v2 offset:16
-; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-WGP-NEXT:    buffer_gl0_inv
 ; GFX10-WGP-NEXT:    ds_write_b32 v0, v1
 ; GFX10-WGP-NEXT:    s_endpgm
 ;
 ; GFX10-CU-LABEL: local_workgroup_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX10-CU:       ; %bb.0: ; %entry
 ; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)