# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
# RUN: llc -march=amdgcn -mcpu=fiji -run-pass=regbankselect -regbankselect-fast  -verify-machineinstrs %s -o - | FileCheck -check-prefix=WAVE64 %s
# RUN: llc -march=amdgcn -mcpu=fiji -run-pass=regbankselect -regbankselect-greedy  -verify-machineinstrs %s -o - | FileCheck -check-prefix=WAVE64 %s
# RUN: llc -march=amdgcn -mcpu=gfx1010 -run-pass=regbankselect -regbankselect-fast  -verify-machineinstrs %s -o - | FileCheck -check-prefix=WAVE32 %s
# RUN: llc -march=amdgcn -mcpu=gfx1010 -run-pass=regbankselect -regbankselect-greedy  -verify-machineinstrs %s -o - | FileCheck -check-prefix=WAVE32 %s

---
name: buffer_load_sss
legalized: true
tracksRegLiveness: true
body: |
  bb.0:
    liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, $sgpr5

    ; WAVE64-LABEL: name: buffer_load_sss
    ; WAVE64: liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, $sgpr5
    ; WAVE64: [[COPY:%[0-9]+]]:sgpr(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
    ; WAVE64: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
    ; WAVE64: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr5
    ; WAVE64: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32)
    ; WAVE64: [[INT:%[0-9]+]]:vgpr(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), [[COPY]](<4 x s32>), [[COPY1]](s32), [[COPY2]](s32), 0, 0
    ; WAVE32-LABEL: name: buffer_load_sss
    ; WAVE32: liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, $sgpr5
    ; WAVE32: [[COPY:%[0-9]+]]:sgpr(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
    ; WAVE32: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
    ; WAVE32: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr5
    ; WAVE32: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32)
    ; WAVE32: [[INT:%[0-9]+]]:vgpr(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), [[COPY]](<4 x s32>), [[COPY1]](s32), [[COPY2]](s32), 0, 0
    %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
    %1:_(s32) = COPY $sgpr4
    %2:_(s32) = COPY $sgpr5
    %3:_(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), %0, %1, %2, 0, 0

...

---
name: buffer_load_ssv
legalized: true
tracksRegLiveness: true
body: |
  bb.0:
    liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, $vgpr5

    ; WAVE64-LABEL: name: buffer_load_ssv
    ; WAVE64: liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, $vgpr5
    ; WAVE64: [[COPY:%[0-9]+]]:sgpr(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
    ; WAVE64: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
    ; WAVE64: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr5
    ; WAVE64: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32)
    ; WAVE64: [[INT:%[0-9]+]]:vgpr(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), [[COPY]](<4 x s32>), [[COPY1]](s32), [[COPY2]](s32), 0, 0
    ; WAVE32-LABEL: name: buffer_load_ssv
    ; WAVE32: liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, $vgpr5
    ; WAVE32: [[COPY:%[0-9]+]]:sgpr(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
    ; WAVE32: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
    ; WAVE32: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr5
    ; WAVE32: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32)
    ; WAVE32: [[INT:%[0-9]+]]:vgpr(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), [[COPY]](<4 x s32>), [[COPY1]](s32), [[COPY2]](s32), 0, 0
    %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
    %1:_(s32) = COPY $sgpr4
    %2:_(s32) = COPY $vgpr5
    %3:_(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), %0, %1, %2, 0, 0

...

---
name: buffer_load_svs
legalized: true
tracksRegLiveness: true
body: |
  bb.0:
    liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, $sgpr5

    ; WAVE64-LABEL: name: buffer_load_svs
    ; WAVE64: liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, $sgpr5
    ; WAVE64: [[COPY:%[0-9]+]]:sgpr(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
    ; WAVE64: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
    ; WAVE64: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr5
    ; WAVE64: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32)
    ; WAVE64: [[INT:%[0-9]+]]:vgpr(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), [[COPY]](<4 x s32>), [[COPY1]](s32), [[COPY2]](s32), 0, 0
    ; WAVE32-LABEL: name: buffer_load_svs
    ; WAVE32: liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, $sgpr5
    ; WAVE32: [[COPY:%[0-9]+]]:sgpr(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
    ; WAVE32: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
    ; WAVE32: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr5
    ; WAVE32: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32)
    ; WAVE32: [[INT:%[0-9]+]]:vgpr(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), [[COPY]](<4 x s32>), [[COPY1]](s32), [[COPY2]](s32), 0, 0
    %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
    %1:_(s32) = COPY $sgpr4
    %2:_(s32) = COPY $sgpr5
    %3:_(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), %0, %1, %2, 0, 0

...

---
name: buffer_load_vss
legalized: true
tracksRegLiveness: true
body: |
  bb.0:
    liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $sgpr4, $sgpr5

    ; WAVE64-LABEL: name: buffer_load_vss
    ; WAVE64: successors: %bb.1(0x80000000)
    ; WAVE64: liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $sgpr4, $sgpr5
    ; WAVE64: [[COPY:%[0-9]+]]:vgpr(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
    ; WAVE64: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
    ; WAVE64: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr5
    ; WAVE64: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32)
    ; WAVE64: [[DEF:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
    ; WAVE64: [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
    ; WAVE64: [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
    ; WAVE64: [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
    ; WAVE64: .1:
    ; WAVE64: successors: %bb.2(0x40000000), %bb.1(0x40000000)
    ; WAVE64: [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.0, %10, %bb.1
    ; WAVE64: [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF]](<4 x s32>), %bb.0, %3(<4 x s32>), %bb.1
    ; WAVE64: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
    ; WAVE64: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
    ; WAVE64: [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
    ; WAVE64: [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
    ; WAVE64: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
    ; WAVE64: [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
    ; WAVE64: [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
    ; WAVE64: [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
    ; WAVE64: [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
    ; WAVE64: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
    ; WAVE64: [[INT:%[0-9]+]]:vgpr(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), [[BUILD_VECTOR]](<4 x s32>), [[COPY1]](s32), [[COPY2]](s32), 0, 0
    ; WAVE64: [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
    ; WAVE64: $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
    ; WAVE64: S_CBRANCH_EXECNZ %bb.1, implicit $exec
    ; WAVE64: .2:
    ; WAVE64: successors: %bb.3(0x80000000)
    ; WAVE64: $exec = S_MOV_B64_term [[S_MOV_B64_term]]
    ; WAVE64: .3:
    ; WAVE32-LABEL: name: buffer_load_vss
    ; WAVE32: successors: %bb.1(0x80000000)
    ; WAVE32: liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $sgpr4, $sgpr5
    ; WAVE32: [[COPY:%[0-9]+]]:vgpr(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
    ; WAVE32: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
    ; WAVE32: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr5
    ; WAVE32: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32)
    ; WAVE32: [[DEF:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
    ; WAVE32: [[DEF1:%[0-9]+]]:sreg_32_xm0_xexec = IMPLICIT_DEF
    ; WAVE32: [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
    ; WAVE32: [[S_MOV_B32_term:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32_term $exec_lo
    ; WAVE32: .1:
    ; WAVE32: successors: %bb.2(0x40000000), %bb.1(0x40000000)
    ; WAVE32: [[PHI:%[0-9]+]]:sreg_32_xm0_xexec = PHI [[DEF1]], %bb.0, %10, %bb.1
    ; WAVE32: [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF]](<4 x s32>), %bb.0, %3(<4 x s32>), %bb.1
    ; WAVE32: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
    ; WAVE32: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
    ; WAVE32: [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
    ; WAVE32: [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
    ; WAVE32: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
    ; WAVE32: [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
    ; WAVE32: [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
    ; WAVE32: [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
    ; WAVE32: [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
    ; WAVE32: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
    ; WAVE32: [[INT:%[0-9]+]]:vgpr(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), [[BUILD_VECTOR]](<4 x s32>), [[COPY1]](s32), [[COPY2]](s32), 0, 0
    ; WAVE32: [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_]], implicit-def $exec, implicit-def $scc, implicit $exec
    ; WAVE32: $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
    ; WAVE32: S_CBRANCH_EXECNZ %bb.1, implicit $exec
    ; WAVE32: .2:
    ; WAVE32: successors: %bb.3(0x80000000)
    ; WAVE32: $exec_lo = S_MOV_B32_term [[S_MOV_B32_term]]
    ; WAVE32: .3:
    %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
    %1:_(s32) = COPY $sgpr4
    %2:_(s32) = COPY $sgpr5
    %3:_(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), %0, %1, %2, 0, 0

...

---
name: buffer_load_vvs
legalized: true
tracksRegLiveness: true
body: |
  bb.0:
    liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4, $sgpr5

    ; WAVE64-LABEL: name: buffer_load_vvs
    ; WAVE64: successors: %bb.1(0x80000000)
    ; WAVE64: liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4, $sgpr5
    ; WAVE64: [[COPY:%[0-9]+]]:vgpr(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
    ; WAVE64: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr4
    ; WAVE64: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr5
    ; WAVE64: [[DEF:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
    ; WAVE64: [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
    ; WAVE64: [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
    ; WAVE64: [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
    ; WAVE64: .1:
    ; WAVE64: successors: %bb.2(0x40000000), %bb.1(0x40000000)
    ; WAVE64: [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.0, %9, %bb.1
    ; WAVE64: [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF]](<4 x s32>), %bb.0, %3(<4 x s32>), %bb.1
    ; WAVE64: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
    ; WAVE64: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
    ; WAVE64: [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
    ; WAVE64: [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
    ; WAVE64: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
    ; WAVE64: [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
    ; WAVE64: [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
    ; WAVE64: [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
    ; WAVE64: [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
    ; WAVE64: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
    ; WAVE64: [[INT:%[0-9]+]]:vgpr(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), [[BUILD_VECTOR]](<4 x s32>), [[COPY1]](s32), [[COPY2]](s32), 0, 0
    ; WAVE64: [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
    ; WAVE64: $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
    ; WAVE64: S_CBRANCH_EXECNZ %bb.1, implicit $exec
    ; WAVE64: .2:
    ; WAVE64: successors: %bb.3(0x80000000)
    ; WAVE64: $exec = S_MOV_B64_term [[S_MOV_B64_term]]
    ; WAVE64: .3:
    ; WAVE32-LABEL: name: buffer_load_vvs
    ; WAVE32: successors: %bb.1(0x80000000)
    ; WAVE32: liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4, $sgpr5
    ; WAVE32: [[COPY:%[0-9]+]]:vgpr(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
    ; WAVE32: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr4
    ; WAVE32: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr5
    ; WAVE32: [[DEF:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
    ; WAVE32: [[DEF1:%[0-9]+]]:sreg_32_xm0_xexec = IMPLICIT_DEF
    ; WAVE32: [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
    ; WAVE32: [[S_MOV_B32_term:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32_term $exec_lo
    ; WAVE32: .1:
    ; WAVE32: successors: %bb.2(0x40000000), %bb.1(0x40000000)
    ; WAVE32: [[PHI:%[0-9]+]]:sreg_32_xm0_xexec = PHI [[DEF1]], %bb.0, %9, %bb.1
    ; WAVE32: [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF]](<4 x s32>), %bb.0, %3(<4 x s32>), %bb.1
    ; WAVE32: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
    ; WAVE32: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
    ; WAVE32: [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
    ; WAVE32: [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
    ; WAVE32: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
    ; WAVE32: [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
    ; WAVE32: [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
    ; WAVE32: [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
    ; WAVE32: [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
    ; WAVE32: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
    ; WAVE32: [[INT:%[0-9]+]]:vgpr(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), [[BUILD_VECTOR]](<4 x s32>), [[COPY1]](s32), [[COPY2]](s32), 0, 0
    ; WAVE32: [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_]], implicit-def $exec, implicit-def $scc, implicit $exec
    ; WAVE32: $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
    ; WAVE32: S_CBRANCH_EXECNZ %bb.1, implicit $exec
    ; WAVE32: .2:
    ; WAVE32: successors: %bb.3(0x80000000)
    ; WAVE32: $exec_lo = S_MOV_B32_term [[S_MOV_B32_term]]
    ; WAVE32: .3:
    %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
    %1:_(s32) = COPY $vgpr4
    %2:_(s32) = COPY $sgpr5
    %3:_(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), %0, %1, %2, 0, 0

...

---
name: buffer_load_svv
legalized: true
tracksRegLiveness: true
body: |
  bb.0:
    liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $vgpr4, $vgpr5

    ; WAVE64-LABEL: name: buffer_load_svv
    ; WAVE64: liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $vgpr4, $vgpr5
    ; WAVE64: [[COPY:%[0-9]+]]:sgpr(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
    ; WAVE64: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr4
    ; WAVE64: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr5
    ; WAVE64: [[INT:%[0-9]+]]:vgpr(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), [[COPY]](<4 x s32>), [[COPY1]](s32), [[COPY2]](s32), 0, 0
    ; WAVE32-LABEL: name: buffer_load_svv
    ; WAVE32: liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $vgpr4, $vgpr5
    ; WAVE32: [[COPY:%[0-9]+]]:sgpr(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
    ; WAVE32: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr4
    ; WAVE32: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr5
    ; WAVE32: [[INT:%[0-9]+]]:vgpr(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), [[COPY]](<4 x s32>), [[COPY1]](s32), [[COPY2]](s32), 0, 0
    %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
    %1:_(s32) = COPY $vgpr4
    %2:_(s32) = COPY $vgpr5
    %3:_(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), %0, %1, %2, 0, 0

...

---
name: buffer_load_vsv
legalized: true
tracksRegLiveness: true
body: |
  bb.0:
    liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $sgpr4, $vgpr5

    ; WAVE64-LABEL: name: buffer_load_vsv
    ; WAVE64: successors: %bb.1(0x80000000)
    ; WAVE64: liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $sgpr4, $vgpr5
    ; WAVE64: [[COPY:%[0-9]+]]:vgpr(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
    ; WAVE64: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
    ; WAVE64: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr5
    ; WAVE64: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32)
    ; WAVE64: [[DEF:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
    ; WAVE64: [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
    ; WAVE64: [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
    ; WAVE64: [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
    ; WAVE64: .1:
    ; WAVE64: successors: %bb.2(0x40000000), %bb.1(0x40000000)
    ; WAVE64: [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.0, %10, %bb.1
    ; WAVE64: [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF]](<4 x s32>), %bb.0, %3(<4 x s32>), %bb.1
    ; WAVE64: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
    ; WAVE64: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
    ; WAVE64: [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
    ; WAVE64: [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
    ; WAVE64: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
    ; WAVE64: [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
    ; WAVE64: [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
    ; WAVE64: [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
    ; WAVE64: [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
    ; WAVE64: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
    ; WAVE64: [[INT:%[0-9]+]]:vgpr(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), [[BUILD_VECTOR]](<4 x s32>), [[COPY1]](s32), [[COPY2]](s32), 0, 0
    ; WAVE64: [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
    ; WAVE64: $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
    ; WAVE64: S_CBRANCH_EXECNZ %bb.1, implicit $exec
    ; WAVE64: .2:
    ; WAVE64: successors: %bb.3(0x80000000)
    ; WAVE64: $exec = S_MOV_B64_term [[S_MOV_B64_term]]
    ; WAVE64: .3:
    ; WAVE32-LABEL: name: buffer_load_vsv
    ; WAVE32: successors: %bb.1(0x80000000)
    ; WAVE32: liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $sgpr4, $vgpr5
    ; WAVE32: [[COPY:%[0-9]+]]:vgpr(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
    ; WAVE32: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
    ; WAVE32: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr5
    ; WAVE32: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32)
    ; WAVE32: [[DEF:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
    ; WAVE32: [[DEF1:%[0-9]+]]:sreg_32_xm0_xexec = IMPLICIT_DEF
    ; WAVE32: [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
    ; WAVE32: [[S_MOV_B32_term:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32_term $exec_lo
    ; WAVE32: .1:
    ; WAVE32: successors: %bb.2(0x40000000), %bb.1(0x40000000)
    ; WAVE32: [[PHI:%[0-9]+]]:sreg_32_xm0_xexec = PHI [[DEF1]], %bb.0, %10, %bb.1
    ; WAVE32: [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF]](<4 x s32>), %bb.0, %3(<4 x s32>), %bb.1
    ; WAVE32: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
    ; WAVE32: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
    ; WAVE32: [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
    ; WAVE32: [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
    ; WAVE32: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
    ; WAVE32: [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
    ; WAVE32: [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
    ; WAVE32: [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
    ; WAVE32: [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
    ; WAVE32: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
    ; WAVE32: [[INT:%[0-9]+]]:vgpr(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), [[BUILD_VECTOR]](<4 x s32>), [[COPY1]](s32), [[COPY2]](s32), 0, 0
    ; WAVE32: [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_]], implicit-def $exec, implicit-def $scc, implicit $exec
    ; WAVE32: $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
    ; WAVE32: S_CBRANCH_EXECNZ %bb.1, implicit $exec
    ; WAVE32: .2:
    ; WAVE32: successors: %bb.3(0x80000000)
    ; WAVE32: $exec_lo = S_MOV_B32_term [[S_MOV_B32_term]]
    ; WAVE32: .3:
    %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
    %1:_(s32) = COPY $sgpr4
    %2:_(s32) = COPY $vgpr5
    %3:_(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), %0, %1, %2, 0, 0

...

---
name: buffer_load_vvv
legalized: true
tracksRegLiveness: true
body: |
  bb.0:
    liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4, $vgpr5

    ; WAVE64-LABEL: name: buffer_load_vvv
    ; WAVE64: successors: %bb.1(0x80000000)
    ; WAVE64: liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4, $vgpr5
    ; WAVE64: [[COPY:%[0-9]+]]:vgpr(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
    ; WAVE64: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr4
    ; WAVE64: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr5
    ; WAVE64: [[DEF:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
    ; WAVE64: [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
    ; WAVE64: [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
    ; WAVE64: [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
    ; WAVE64: .1:
    ; WAVE64: successors: %bb.2(0x40000000), %bb.1(0x40000000)
    ; WAVE64: [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.0, %9, %bb.1
    ; WAVE64: [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF]](<4 x s32>), %bb.0, %3(<4 x s32>), %bb.1
    ; WAVE64: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
    ; WAVE64: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
    ; WAVE64: [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
    ; WAVE64: [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
    ; WAVE64: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
    ; WAVE64: [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
    ; WAVE64: [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
    ; WAVE64: [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
    ; WAVE64: [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
    ; WAVE64: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
    ; WAVE64: [[INT:%[0-9]+]]:vgpr(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), [[BUILD_VECTOR]](<4 x s32>), [[COPY1]](s32), [[COPY2]](s32), 0, 0
    ; WAVE64: [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
    ; WAVE64: $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
    ; WAVE64: S_CBRANCH_EXECNZ %bb.1, implicit $exec
    ; WAVE64: .2:
    ; WAVE64: successors: %bb.3(0x80000000)
    ; WAVE64: $exec = S_MOV_B64_term [[S_MOV_B64_term]]
    ; WAVE64: .3:
    ; WAVE32-LABEL: name: buffer_load_vvv
    ; WAVE32: successors: %bb.1(0x80000000)
    ; WAVE32: liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4, $vgpr5
    ; WAVE32: [[COPY:%[0-9]+]]:vgpr(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
    ; WAVE32: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr4
    ; WAVE32: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr5
    ; WAVE32: [[DEF:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
    ; WAVE32: [[DEF1:%[0-9]+]]:sreg_32_xm0_xexec = IMPLICIT_DEF
    ; WAVE32: [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
    ; WAVE32: [[S_MOV_B32_term:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32_term $exec_lo
    ; WAVE32: .1:
    ; WAVE32: successors: %bb.2(0x40000000), %bb.1(0x40000000)
    ; WAVE32: [[PHI:%[0-9]+]]:sreg_32_xm0_xexec = PHI [[DEF1]], %bb.0, %9, %bb.1
    ; WAVE32: [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF]](<4 x s32>), %bb.0, %3(<4 x s32>), %bb.1
    ; WAVE32: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
    ; WAVE32: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
    ; WAVE32: [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
    ; WAVE32: [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
    ; WAVE32: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
    ; WAVE32: [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
    ; WAVE32: [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
    ; WAVE32: [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
    ; WAVE32: [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
    ; WAVE32: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
    ; WAVE32: [[INT:%[0-9]+]]:vgpr(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), [[BUILD_VECTOR]](<4 x s32>), [[COPY1]](s32), [[COPY2]](s32), 0, 0
    ; WAVE32: [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_]], implicit-def $exec, implicit-def $scc, implicit $exec
    ; WAVE32: $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
    ; WAVE32: S_CBRANCH_EXECNZ %bb.1, implicit $exec
    ; WAVE32: .2:
    ; WAVE32: successors: %bb.3(0x80000000)
    ; WAVE32: $exec_lo = S_MOV_B32_term [[S_MOV_B32_term]]
    ; WAVE32: .3:
    %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
    %1:_(s32) = COPY $vgpr4
    %2:_(s32) = COPY $vgpr5
    %3:_(<4 x s32>) = G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.buffer.load), %0, %1, %2, 0, 0

...