# RUN: llc -march=amdgcn -verify-machineinstrs -run-pass si-insert-waits -o - %s | FileCheck %s
--- |
  define amdgpu_ps <4 x float> @exp_done_waitcnt(<4 x i32> inreg, <4 x i32> inreg, i32 inreg %w, float %v) {
    %a = load volatile float, float addrspace(1)* undef
    %b = load volatile float, float addrspace(1)* undef
    %c = load volatile float, float addrspace(1)* undef
    %d = load volatile float, float addrspace(1)* undef
    call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %a, float %b, float %c, float %d)
    ret <4 x float> <float 5.000000e-01, float 1.000000e+00, float 2.000000e+00, float 4.000000e+00>
  }

  declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)

  attributes #0 = { readnone }
  attributes #1 = { nounwind }

...
---

# CHECK-LABEL: name: exp_done_waitcnt{{$}}
# CHECK: EXP_DONE
# CHECK-NEXT: S_WAITCNT 3855
# CHECK: %vgpr0 = V_MOV_B32
# CHECK: %vgpr1 = V_MOV_B32
# CHECK: %vgpr2 = V_MOV_B32
# CHECK: %vgpr3 = V_MOV_B32
name:            exp_done_waitcnt
alignment:       0
exposesReturnsTwice: false
legalized:       false
regBankSelected: false
selected:        false
tracksRegLiveness: true
frameInfo:
  isFrameAddressTaken: false
  isReturnAddressTaken: false
  hasStackMap:     false
  hasPatchPoint:   false
  stackSize:       0
  offsetAdjustment: 0
  maxAlignment:    0
  adjustsStack:    false
  hasCalls:        false
  maxCallFrameSize: 0
  hasOpaqueSPAdjustment: false
  hasVAStart:      false
  hasMustTailInVarArgFunc: false
body:             |
  bb.0 (%ir-block.2):
    %sgpr3 = S_MOV_B32 61440
    %sgpr2 = S_MOV_B32 -1
    %vgpr0 = BUFFER_LOAD_DWORD_OFFSET %sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 0, 0, 0, implicit %exec :: (volatile load 4 from `float addrspace(1)* undef`)
    %vgpr1 = BUFFER_LOAD_DWORD_OFFSET %sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 0, 0, 0, implicit %exec :: (volatile load 4 from `float addrspace(1)* undef`)
    %vgpr2 = BUFFER_LOAD_DWORD_OFFSET %sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 0, 0, 0, implicit %exec :: (volatile load 4 from `float addrspace(1)* undef`)
    %vgpr3 = BUFFER_LOAD_DWORD_OFFSET killed %sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 0, 0, 0, implicit %exec :: (volatile load 4 from `float addrspace(1)* undef`)
    EXP_DONE 0, killed %vgpr0, killed %vgpr1, killed %vgpr2, killed %vgpr3, -1, -1, 15, implicit %exec
    %vgpr0 = V_MOV_B32_e32 1056964608, implicit %exec
    %vgpr1 = V_MOV_B32_e32 1065353216, implicit %exec
    %vgpr2 = V_MOV_B32_e32 1073741824, implicit %exec
    %vgpr3 = V_MOV_B32_e32 1082130432, implicit %exec
    SI_RETURN killed %vgpr0, killed %vgpr1, killed %vgpr2, killed %vgpr3

...