Compiler projects using llvm
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=bonaire -O0 -run-pass=legalizer  %s -o - | FileCheck -check-prefix=CI %s
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer  %s -o - | FileCheck -check-prefix=VI %s
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -O0 -run-pass=legalizer  %s -o - | FileCheck -check-prefix=GFX9 %s
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -O0 -run-pass=legalizer  %s -o - | FileCheck -check-prefix=GFX9 %s
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -O0 -run-pass=legalizer  %s -o - | FileCheck -check-prefix=GFX9 %s

---
name: test_load_flat_s1_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s1_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
    ; CI-NEXT: $vgpr0 = COPY [[AND]](s32)
    ; VI-LABEL: name: test_load_flat_s1_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
    ; VI-NEXT: $vgpr0 = COPY [[AND]](s32)
    ; GFX9-LABEL: name: test_load_flat_s1_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
    ; GFX9-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
    ; GFX9-NEXT: $vgpr0 = COPY [[AND]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s1) = G_LOAD %0 :: (load (s1), align 1, addrspace 0)
    %2:_(s32) = G_ZEXT %1
    $vgpr0 = COPY %2
...

---
name: test_load_flat_s2_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s2_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
    ; CI-NEXT: $vgpr0 = COPY [[AND]](s32)
    ; VI-LABEL: name: test_load_flat_s2_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
    ; VI-NEXT: $vgpr0 = COPY [[AND]](s32)
    ; GFX9-LABEL: name: test_load_flat_s2_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
    ; GFX9-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
    ; GFX9-NEXT: $vgpr0 = COPY [[AND]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s2) = G_LOAD %0 :: (load (s2), align 1, addrspace 0)
    %2:_(s32) = G_ZEXT %1
    $vgpr0 = COPY %2
...

---
name: test_load_flat_s8_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s8_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
    ; CI-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    ; VI-LABEL: name: test_load_flat_s8_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
    ; VI-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    ; GFX9-LABEL: name: test_load_flat_s8_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
    ; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s8) = G_LOAD %0 :: (load (s8), align 4, addrspace 0)
    %2:_(s32) = G_ANYEXT %1
    $vgpr0 = COPY %2
...

---
name: test_load_flat_s8_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s8_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    ; VI-LABEL: name: test_load_flat_s8_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    ; GFX9-LABEL: name: test_load_flat_s8_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s8) = G_LOAD %0 :: (load (s8), align 1, addrspace 0)
    %2:_(s32) = G_ANYEXT %1
    $vgpr0 = COPY %2
...

---
name: test_load_flat_s16_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s16_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
    ; CI-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    ; VI-LABEL: name: test_load_flat_s16_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
    ; VI-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    ; GFX9-LABEL: name: test_load_flat_s16_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
    ; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s16) = G_LOAD %0 :: (load (s16), align 4, addrspace 0)
    %2:_(s32) = G_ANYEXT %1
    $vgpr0 = COPY %2
...

---
name: test_load_flat_s16_align2
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s16_align2
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
    ; CI-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    ; VI-LABEL: name: test_load_flat_s16_align2
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
    ; VI-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    ; GFX9-LABEL: name: test_load_flat_s16_align2
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
    ; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s16) = G_LOAD %0 :: (load (s16), align 2, addrspace 0)
    %2:_(s32) = G_ANYEXT %1
    $vgpr0 = COPY %2
...

---
name: test_load_flat_s16_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s16_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: $vgpr0 = COPY [[OR]](s32)
    ; VI-LABEL: name: test_load_flat_s16_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: $vgpr0 = COPY [[OR]](s32)
    ; GFX9-LABEL: name: test_load_flat_s16_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: $vgpr0 = COPY [[OR]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s16) = G_LOAD %0 :: (load (s16), align 1, addrspace 0)
    %2:_(s32) = G_ANYEXT %1
    $vgpr0 = COPY %2
...

---
name: test_load_flat_s32_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s32_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
    ; CI-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    ; VI-LABEL: name: test_load_flat_s32_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
    ; VI-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    ; GFX9-LABEL: name: test_load_flat_s32_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
    ; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s32) = G_LOAD %0 :: (load (s32), align 4, addrspace 0)
    $vgpr0 = COPY %1
...

---
name: test_load_flat_s32_align2
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s32_align2
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: $vgpr0 = COPY [[OR]](s32)
    ; VI-LABEL: name: test_load_flat_s32_align2
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: $vgpr0 = COPY [[OR]](s32)
    ; GFX9-LABEL: name: test_load_flat_s32_align2
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: $vgpr0 = COPY [[OR]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s32) = G_LOAD %0 :: (load (s32), align 2, addrspace 0)
    $vgpr0 = COPY %1
...

---
name: test_load_flat_s32_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s32_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; CI-NEXT: $vgpr0 = COPY [[OR2]](s32)
    ; VI-LABEL: name: test_load_flat_s32_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; VI-NEXT: $vgpr0 = COPY [[OR2]](s32)
    ; GFX9-LABEL: name: test_load_flat_s32_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; GFX9-NEXT: $vgpr0 = COPY [[OR2]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s32) = G_LOAD %0 :: (load (s32), align 1, addrspace 0)
    $vgpr0 = COPY %1
...

---
name: test_load_flat_s48_align8
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s48_align8
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
    ; VI-LABEL: name: test_load_flat_s48_align8
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
    ; GFX9-LABEL: name: test_load_flat_s48_align8
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
    ; GFX9-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s48) = G_LOAD %0 :: (load (s48), align 8, addrspace 0)
    %2:_(s64) = G_ZEXT %1
    $vgpr0_vgpr1 = COPY %2
...

---
name: test_load_flat_s64_align8
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s64_align8
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64))
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
    ; VI-LABEL: name: test_load_flat_s64_align8
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64))
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
    ; GFX9-LABEL: name: test_load_flat_s64_align8
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64))
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s64) = G_LOAD %0 :: (load (s64), align 8, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_s64_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s64_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 4)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
    ; VI-LABEL: name: test_load_flat_s64_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 4)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
    ; GFX9-LABEL: name: test_load_flat_s64_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 4)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s64) = G_LOAD %0 :: (load (s64), align 4, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_s64_align2
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s64_align2
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; CI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
    ; VI-LABEL: name: test_load_flat_s64_align2
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
    ; GFX9-LABEL: name: test_load_flat_s64_align2
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s64) = G_LOAD %0 :: (load (s64), align 2, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_s64_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s64_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; CI-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; CI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; CI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; CI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; CI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; CI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; CI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; CI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; CI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
    ; CI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; CI-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
    ; CI-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
    ; VI-LABEL: name: test_load_flat_s64_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; VI-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; VI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; VI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; VI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; VI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
    ; VI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; VI-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
    ; VI-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
    ; GFX9-LABEL: name: test_load_flat_s64_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; GFX9-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
    ; GFX9-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; GFX9-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; GFX9-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; GFX9-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; GFX9-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; GFX9-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; GFX9-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; GFX9-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
    ; GFX9-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; GFX9-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
    ; GFX9-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s64) = G_LOAD %0 :: (load (s64), align 1, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_s96_align16
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s96_align16
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
    ; VI-LABEL: name: test_load_flat_s96_align16
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
    ; GFX9-LABEL: name: test_load_flat_s96_align16
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s96) = G_LOAD %0 :: (load (s96), align 16, addrspace 0)
    $vgpr0_vgpr1_vgpr2 = COPY %1
...

---
name: test_load_flat_s96_align8
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s96_align8
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 8)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
    ; VI-LABEL: name: test_load_flat_s96_align8
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 8)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
    ; GFX9-LABEL: name: test_load_flat_s96_align8
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 8)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s96) = G_LOAD %0 :: (load (s96), align 8, addrspace 0)
    $vgpr0_vgpr1_vgpr2 = COPY %1
...

---
name: test_load_flat_s96_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s96_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
    ; VI-LABEL: name: test_load_flat_s96_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
    ; GFX9-LABEL: name: test_load_flat_s96_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s96) = G_LOAD %0 :: (load (s96), align 4, addrspace 0)
    $vgpr0_vgpr1_vgpr2 = COPY %1
...

---
name: test_load_flat_s96_align2
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s96_align2
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; CI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C3]](s64)
    ; CI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s16) from unknown-address + 8)
    ; CI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; CI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s16) from unknown-address + 10)
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
    ; VI-LABEL: name: test_load_flat_s96_align2
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; VI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C3]](s64)
    ; VI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s16) from unknown-address + 8)
    ; VI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; VI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s16) from unknown-address + 10)
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
    ; GFX9-LABEL: name: test_load_flat_s96_align2
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; GFX9-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C3]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s16) from unknown-address + 8)
    ; GFX9-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s16) from unknown-address + 10)
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s96) = G_LOAD %0 :: (load (s96), align 2, addrspace 0)
    $vgpr0_vgpr1_vgpr2 = COPY %1
...

---
name: test_load_flat_s96_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s96_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; CI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; CI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; CI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; CI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; CI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; CI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; CI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; CI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; CI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C5]](s64)
    ; CI-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
    ; CI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
    ; CI-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
    ; CI-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
    ; CI-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
    ; CI-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
    ; CI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
    ; CI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; CI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
    ; CI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
    ; CI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
    ; VI-LABEL: name: test_load_flat_s96_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; VI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; VI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; VI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; VI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; VI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; VI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C5]](s64)
    ; VI-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
    ; VI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
    ; VI-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
    ; VI-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
    ; VI-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
    ; VI-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
    ; VI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
    ; VI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; VI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
    ; VI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
    ; VI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
    ; GFX9-LABEL: name: test_load_flat_s96_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; GFX9-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; GFX9-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; GFX9-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; GFX9-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; GFX9-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; GFX9-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; GFX9-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; GFX9-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; GFX9-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; GFX9-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C5]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
    ; GFX9-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
    ; GFX9-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
    ; GFX9-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
    ; GFX9-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
    ; GFX9-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
    ; GFX9-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; GFX9-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
    ; GFX9-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
    ; GFX9-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s96) = G_LOAD %0 :: (load (s96), align 1, addrspace 0)
    $vgpr0_vgpr1_vgpr2 = COPY %1
...

---
name: test_load_flat_s160_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s160_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s32) from unknown-address + 16)
    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
    ; CI-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
    ; VI-LABEL: name: test_load_flat_s160_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s32) from unknown-address + 16)
    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
    ; VI-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
    ; GFX9-LABEL: name: test_load_flat_s160_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s32) from unknown-address + 16)
    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
    ; GFX9-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s160) = G_LOAD %0 :: (load (s160), align 4, addrspace 0)
    S_NOP 0, implicit %1
...

---
name: test_load_flat_s224_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s224_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<3 x s32>) from unknown-address + 16, align 4)
    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
    ; CI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
    ; CI-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
    ; CI-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
    ; VI-LABEL: name: test_load_flat_s224_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<3 x s32>) from unknown-address + 16, align 4)
    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
    ; VI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
    ; VI-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
    ; GFX9-LABEL: name: test_load_flat_s224_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<3 x s32>) from unknown-address + 16, align 4)
    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
    ; GFX9-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s224) = G_LOAD %0 :: (load (s224), align 4, addrspace 0)
     %2:_(s256) = G_IMPLICIT_DEF
     %3:_(s256) = G_INSERT %2, %1, 0
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %3

...

---
name: test_load_flat_s128_align16
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s128_align16
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
    ; VI-LABEL: name: test_load_flat_s128_align16
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
    ; GFX9-LABEL: name: test_load_flat_s128_align16
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s128) = G_LOAD %0 :: (load (s128), align 16, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
...

---
name: test_load_flat_s128_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s128_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
    ; VI-LABEL: name: test_load_flat_s128_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
    ; GFX9-LABEL: name: test_load_flat_s128_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s128) = G_LOAD %0 :: (load (s128), align 4, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
...

---
name: test_load_flat_s128_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s128_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; CI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; CI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; CI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; CI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; CI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; CI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; CI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; CI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; CI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C5]](s64)
    ; CI-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
    ; CI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
    ; CI-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
    ; CI-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
    ; CI-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
    ; CI-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
    ; CI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
    ; CI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; CI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
    ; CI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
    ; CI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
    ; CI-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
    ; CI-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C6]](s64)
    ; CI-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
    ; CI-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
    ; CI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
    ; CI-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
    ; CI-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
    ; CI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
    ; CI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
    ; CI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; CI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
    ; CI-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
    ; CI-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
    ; VI-LABEL: name: test_load_flat_s128_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; VI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; VI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; VI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; VI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; VI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; VI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C5]](s64)
    ; VI-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
    ; VI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
    ; VI-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
    ; VI-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
    ; VI-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
    ; VI-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
    ; VI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
    ; VI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; VI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
    ; VI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
    ; VI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
    ; VI-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
    ; VI-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C6]](s64)
    ; VI-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
    ; VI-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
    ; VI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
    ; VI-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
    ; VI-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
    ; VI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
    ; VI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
    ; VI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; VI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
    ; VI-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
    ; VI-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
    ; GFX9-LABEL: name: test_load_flat_s128_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; GFX9-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; GFX9-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; GFX9-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; GFX9-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; GFX9-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; GFX9-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; GFX9-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; GFX9-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; GFX9-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; GFX9-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C5]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
    ; GFX9-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
    ; GFX9-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
    ; GFX9-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
    ; GFX9-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
    ; GFX9-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
    ; GFX9-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; GFX9-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
    ; GFX9-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
    ; GFX9-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
    ; GFX9-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
    ; GFX9-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C6]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
    ; GFX9-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
    ; GFX9-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
    ; GFX9-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
    ; GFX9-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
    ; GFX9-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
    ; GFX9-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; GFX9-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
    ; GFX9-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
    ; GFX9-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s128) = G_LOAD %0 :: (load (s128), align 1, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
...

---
name: test_load_flat_s256_align32
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s256_align32
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
    ; VI-LABEL: name: test_load_flat_s256_align32
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
    ; GFX9-LABEL: name: test_load_flat_s256_align32
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s256) = G_LOAD %0 :: (load (s256), align 16, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
...

---
name: test_load_flat_p1_align8
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_p1_align8
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1))
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
    ; VI-LABEL: name: test_load_flat_p1_align8
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1))
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
    ; GFX9-LABEL: name: test_load_flat_p1_align8
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1))
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(p1) = G_LOAD %0 :: (load (p1), align 8, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_p1_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_p1_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1), align 4)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
    ; VI-LABEL: name: test_load_flat_p1_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1), align 4)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
    ; GFX9-LABEL: name: test_load_flat_p1_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1), align 4)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(p1) = G_LOAD %0 :: (load (p1), align 4, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_p1_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_p1_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; CI-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; CI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; CI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; CI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; CI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; CI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; CI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; CI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; CI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
    ; CI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; CI-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
    ; CI-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
    ; CI-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
    ; VI-LABEL: name: test_load_flat_p1_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; VI-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; VI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; VI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; VI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; VI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
    ; VI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; VI-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
    ; VI-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
    ; VI-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
    ; GFX9-LABEL: name: test_load_flat_p1_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; GFX9-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
    ; GFX9-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; GFX9-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; GFX9-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; GFX9-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; GFX9-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; GFX9-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; GFX9-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; GFX9-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
    ; GFX9-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; GFX9-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
    ; GFX9-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
    ; GFX9-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(p1) = G_LOAD %0 :: (load (p1), align 1, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_p3_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_p3_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p0) :: (load (p3))
    ; CI-NEXT: $vgpr0 = COPY [[LOAD]](p3)
    ; VI-LABEL: name: test_load_flat_p3_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p0) :: (load (p3))
    ; VI-NEXT: $vgpr0 = COPY [[LOAD]](p3)
    ; GFX9-LABEL: name: test_load_flat_p3_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p0) :: (load (p3))
    ; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](p3)
     %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(p3) = G_LOAD %0 :: (load (p3), align 4, addrspace 0)
    $vgpr0 = COPY %1
...

---
name: test_load_flat_p4_align8
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_p4_align8
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4))
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
    ; VI-LABEL: name: test_load_flat_p4_align8
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4))
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
    ; GFX9-LABEL: name: test_load_flat_p4_align8
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4))
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(p4) = G_LOAD %0 :: (load (p4), align 8, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_p4_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_p4_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 4)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
    ; VI-LABEL: name: test_load_flat_p4_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 4)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
    ; GFX9-LABEL: name: test_load_flat_p4_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 4)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(p4) = G_LOAD %0 :: (load (p4), align 4, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_p4_align2
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_p4_align2
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; CI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
    ; CI-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR2]](s64)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
    ; VI-LABEL: name: test_load_flat_p4_align2
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
    ; VI-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR2]](s64)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
    ; GFX9-LABEL: name: test_load_flat_p4_align2
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
    ; GFX9-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR2]](s64)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(p4) = G_LOAD %0 :: (load (p4), align 2, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_p4_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_p4_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; CI-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; CI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; CI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; CI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; CI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; CI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; CI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; CI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; CI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
    ; CI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; CI-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
    ; CI-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
    ; CI-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR6]](s64)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
    ; VI-LABEL: name: test_load_flat_p4_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; VI-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; VI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; VI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; VI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; VI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
    ; VI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; VI-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
    ; VI-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
    ; VI-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR6]](s64)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
    ; GFX9-LABEL: name: test_load_flat_p4_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; GFX9-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
    ; GFX9-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; GFX9-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; GFX9-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; GFX9-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; GFX9-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; GFX9-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; GFX9-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; GFX9-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
    ; GFX9-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; GFX9-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
    ; GFX9-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
    ; GFX9-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR6]](s64)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(p4) = G_LOAD %0 :: (load (p4), align 1, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_p5_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_p5_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5))
    ; CI-NEXT: $vgpr0 = COPY [[LOAD]](p5)
    ; VI-LABEL: name: test_load_flat_p5_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5))
    ; VI-NEXT: $vgpr0 = COPY [[LOAD]](p5)
    ; GFX9-LABEL: name: test_load_flat_p5_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5))
    ; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](p5)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(p5) = G_LOAD %0 :: (load (p5), align 4, addrspace 0)
    $vgpr0 = COPY %1
...

---
name: test_load_flat_p5_align2
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_p5_align2
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
    ; CI-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
    ; VI-LABEL: name: test_load_flat_p5_align2
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
    ; VI-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
    ; GFX9-LABEL: name: test_load_flat_p5_align2
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
    ; GFX9-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(p5) = G_LOAD %0 :: (load (p5), align 2, addrspace 0)
    $vgpr0 = COPY %1
...

---
name: test_load_flat_p5_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_p5_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; CI-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
    ; CI-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
    ; VI-LABEL: name: test_load_flat_p5_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; VI-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
    ; VI-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
    ; GFX9-LABEL: name: test_load_flat_p5_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; GFX9-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
    ; GFX9-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(p5) = G_LOAD %0 :: (load (p5), align 1, addrspace 0)
    $vgpr0 = COPY %1
...

---
name: test_load_flat_v2s8_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2s8_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
    ; CI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
    ; CI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C1]]
    ; CI-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32)
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C2]]
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
    ; CI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
    ; CI-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
    ; CI-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
    ; VI-LABEL: name: test_load_flat_v2s8_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
    ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
    ; VI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C1]]
    ; VI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C1]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C2]](s16)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
    ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
    ; VI-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
    ; GFX9-LABEL: name: test_load_flat_v2s8_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
    ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
    ; GFX9-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C1]]
    ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
    ; GFX9-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C1]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C2]](s16)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
    ; GFX9-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x s8>) = G_LOAD %0 :: (load (<2 x s8>), align 4, addrspace 0)
    %2:_(s16) = G_BITCAST %1
    %3:_(s32) = G_ANYEXT %2
    $vgpr0 = COPY %3
...

---
name: test_load_flat_v2s8_align2
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2s8_align2
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
    ; CI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C1]]
    ; CI-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32)
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C2]]
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
    ; CI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
    ; CI-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
    ; CI-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
    ; VI-LABEL: name: test_load_flat_v2s8_align2
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
    ; VI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C1]]
    ; VI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C1]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C2]](s16)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
    ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
    ; VI-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
    ; GFX9-LABEL: name: test_load_flat_v2s8_align2
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
    ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
    ; GFX9-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C1]]
    ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
    ; GFX9-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C1]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C2]](s16)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
    ; GFX9-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x s8>) = G_LOAD %0 :: (load (<2 x s8>), align 2, addrspace 0)
    %2:_(s16) = G_BITCAST %1
    %3:_(s32) = G_ANYEXT %2
    $vgpr0 = COPY %3
...

---
name: test_load_flat_v2s8_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2s8_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
    ; CI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
    ; CI-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C3]]
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
    ; CI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
    ; CI-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR1]](s16)
    ; CI-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
    ; VI-LABEL: name: test_load_flat_v2s8_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
    ; VI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
    ; VI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
    ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR1]](s16)
    ; VI-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
    ; GFX9-LABEL: name: test_load_flat_v2s8_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
    ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
    ; GFX9-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
    ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
    ; GFX9-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR1]](s16)
    ; GFX9-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x s8>) = G_LOAD %0 :: (load (<2 x s8>), align 1, addrspace 0)
    %2:_(s16) = G_BITCAST %1
    %3:_(s32) = G_ANYEXT %2
    $vgpr0 = COPY %3
...

---
name: test_load_flat_v3s8_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v3s8_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
    ; CI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
    ; CI-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32)
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C3]]
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
    ; CI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
    ; CI-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
    ; CI-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
    ; CI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C]](s32)
    ; CI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[DEF]], [[C3]]
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[COPY2]](s32)
    ; CI-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[SHL1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[TRUNC3]]
    ; CI-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
    ; CI-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
    ; CI-NEXT: $vgpr0 = COPY [[OR2]](s32)
    ; VI-LABEL: name: test_load_flat_v3s8_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
    ; VI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
    ; VI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
    ; VI-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
    ; VI-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
    ; VI-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
    ; VI-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
    ; VI-NEXT: $vgpr0 = COPY [[OR2]](s32)
    ; GFX9-LABEL: name: test_load_flat_v3s8_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
    ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
    ; GFX9-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
    ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
    ; GFX9-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
    ; GFX9-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
    ; GFX9-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
    ; GFX9-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
    ; GFX9-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
    ; GFX9-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
    ; GFX9-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
    ; GFX9-NEXT: $vgpr0 = COPY [[OR2]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<3 x s8>) = G_LOAD %0 :: (load (<3 x s8>), align 4, addrspace 0)
    %2:_(s24) = G_BITCAST %1
    %3:_(s32) = G_ANYEXT %2
    $vgpr0 = COPY %3
...

---
name: test_load_flat_v3s8_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v3s8_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
    ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
    ; CI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
    ; CI-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
    ; CI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
    ; CI-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
    ; CI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C5]]
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
    ; CI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL2]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
    ; CI-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
    ; CI-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
    ; CI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
    ; CI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[DEF]], [[C5]]
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[COPY2]](s32)
    ; CI-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[SHL3]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[TRUNC3]]
    ; CI-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
    ; CI-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
    ; CI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
    ; CI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
    ; CI-NEXT: $vgpr0 = COPY [[OR4]](s32)
    ; VI-LABEL: name: test_load_flat_v3s8_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
    ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
    ; VI-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
    ; VI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
    ; VI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
    ; VI-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
    ; VI-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
    ; VI-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
    ; VI-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
    ; VI-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
    ; VI-NEXT: $vgpr0 = COPY [[OR4]](s32)
    ; GFX9-LABEL: name: test_load_flat_v3s8_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
    ; GFX9-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
    ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
    ; GFX9-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
    ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
    ; GFX9-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
    ; GFX9-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
    ; GFX9-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
    ; GFX9-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
    ; GFX9-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
    ; GFX9-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
    ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
    ; GFX9-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
    ; GFX9-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
    ; GFX9-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
    ; GFX9-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
    ; GFX9-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
    ; GFX9-NEXT: $vgpr0 = COPY [[OR4]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<3 x s8>) = G_LOAD %0 :: (load (<3 x s8>), align 1, addrspace 0)
    %2:_(s24) = G_BITCAST %1
    %3:_(s32) = G_ANYEXT %2
    $vgpr0 = COPY %3
...

---
name: test_load_flat_v4s8_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v4s8_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
    ; CI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C3]]
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C3]]
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; CI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C3]]
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
    ; CI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C3]]
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C2]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
    ; CI-NEXT: $vgpr0 = COPY [[OR2]](s32)
    ; VI-LABEL: name: test_load_flat_v4s8_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
    ; VI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C3]]
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C3]]
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C3]]
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C3]]
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C2]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
    ; VI-NEXT: $vgpr0 = COPY [[OR2]](s32)
    ; GFX9-LABEL: name: test_load_flat_v4s8_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; GFX9-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C3]]
    ; GFX9-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C3]]
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; GFX9-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C3]]
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
    ; GFX9-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C3]]
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C2]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
    ; GFX9-NEXT: $vgpr0 = COPY [[OR2]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<4 x s8>) = G_LOAD %0 :: (load (<4 x s8>), align 4, addrspace 0)
    %2:_(s32) = G_BITCAST %1
    $vgpr0 = COPY %2
...

---
name: test_load_flat_v4s8_align2
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v4s8_align2
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
    ; CI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
    ; CI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C3]](s32)
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C4]]
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C4]]
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C2]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL1]]
    ; CI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C4]]
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C1]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
    ; CI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C4]]
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C3]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[OR2]], [[SHL3]]
    ; CI-NEXT: $vgpr0 = COPY [[OR3]](s32)
    ; VI-LABEL: name: test_load_flat_v4s8_align2
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
    ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
    ; VI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C3]](s32)
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C4]]
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C4]]
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C2]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL1]]
    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C4]]
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C1]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C4]]
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C3]](s32)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[OR2]], [[SHL3]]
    ; VI-NEXT: $vgpr0 = COPY [[OR3]](s32)
    ; GFX9-LABEL: name: test_load_flat_v4s8_align2
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C3]](s32)
    ; GFX9-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; GFX9-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C4]]
    ; GFX9-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C4]]
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C2]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL1]]
    ; GFX9-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C4]]
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C1]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
    ; GFX9-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C4]]
    ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C3]](s32)
    ; GFX9-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[OR2]], [[SHL3]]
    ; GFX9-NEXT: $vgpr0 = COPY [[OR3]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<4 x s8>) = G_LOAD %0 :: (load (<4 x s8>), align 2, addrspace 0)
    %2:_(s32) = G_BITCAST %1
    $vgpr0 = COPY %2
...

---
name: test_load_flat_v4s8_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v4s8_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C1]](s32)
    ; CI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C3]](s32)
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
    ; CI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C4]](s32)
    ; CI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C5]]
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C5]]
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL3]]
    ; CI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C5]]
    ; CI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C3]](s32)
    ; CI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[OR3]], [[SHL4]]
    ; CI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C5]]
    ; CI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C4]](s32)
    ; CI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[OR4]], [[SHL5]]
    ; CI-NEXT: $vgpr0 = COPY [[OR5]](s32)
    ; VI-LABEL: name: test_load_flat_v4s8_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C1]](s32)
    ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C3]](s32)
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
    ; VI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C4]](s32)
    ; VI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C5]]
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C5]]
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL3]]
    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C5]]
    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C3]](s32)
    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[OR3]], [[SHL4]]
    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C5]]
    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C4]](s32)
    ; VI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[OR4]], [[SHL5]]
    ; VI-NEXT: $vgpr0 = COPY [[OR5]](s32)
    ; GFX9-LABEL: name: test_load_flat_v4s8_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C1]](s32)
    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C3]](s32)
    ; GFX9-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C4]](s32)
    ; GFX9-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; GFX9-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C5]]
    ; GFX9-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C5]]
    ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL3]]
    ; GFX9-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C5]]
    ; GFX9-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C3]](s32)
    ; GFX9-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[OR3]], [[SHL4]]
    ; GFX9-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C5]]
    ; GFX9-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C4]](s32)
    ; GFX9-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[OR4]], [[SHL5]]
    ; GFX9-NEXT: $vgpr0 = COPY [[OR5]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<4 x s8>) = G_LOAD %0 :: (load (<4 x s8>), align 1, addrspace 0)
    %2:_(s32) = G_BITCAST %1
    $vgpr0 = COPY %2
...

---
name: test_load_flat_v8s8_align8
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v8s8_align8
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
    ; CI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
    ; CI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
    ; CI-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C]](s32)
    ; CI-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C1]](s32)
    ; CI-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C2]](s32)
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C3]]
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C3]]
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; CI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C3]]
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
    ; CI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C3]]
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C2]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
    ; CI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C3]]
    ; CI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C3]]
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL3]]
    ; CI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C3]]
    ; CI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C1]](s32)
    ; CI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[OR3]], [[SHL4]]
    ; CI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C3]]
    ; CI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C2]](s32)
    ; CI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[OR4]], [[SHL5]]
    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
    ; VI-LABEL: name: test_load_flat_v8s8_align8
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
    ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
    ; VI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
    ; VI-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C]](s32)
    ; VI-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C1]](s32)
    ; VI-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C2]](s32)
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C3]]
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C3]]
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C3]]
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C3]]
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C2]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C3]]
    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C3]]
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL3]]
    ; VI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C3]]
    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C1]](s32)
    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[OR3]], [[SHL4]]
    ; VI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C3]]
    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C2]](s32)
    ; VI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[OR4]], [[SHL5]]
    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
    ; GFX9-LABEL: name: test_load_flat_v8s8_align8
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C]](s32)
    ; GFX9-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C1]](s32)
    ; GFX9-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C2]](s32)
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; GFX9-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C3]]
    ; GFX9-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C3]]
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; GFX9-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C3]]
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
    ; GFX9-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C3]]
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C2]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
    ; GFX9-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C3]]
    ; GFX9-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C3]]
    ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
    ; GFX9-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL3]]
    ; GFX9-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C3]]
    ; GFX9-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C1]](s32)
    ; GFX9-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[OR3]], [[SHL4]]
    ; GFX9-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C3]]
    ; GFX9-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C2]](s32)
    ; GFX9-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[OR4]], [[SHL5]]
    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<8 x s8>) = G_LOAD %0 :: (load (<8 x s8>), align 8, addrspace 0)
    %2:_(<2 x s32>) = G_BITCAST %1
    $vgpr0_vgpr1 = COPY %2
...

---
name: test_load_flat_v16s8_align16
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v16s8_align16
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
    ; CI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
    ; CI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
    ; CI-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C]](s32)
    ; CI-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C1]](s32)
    ; CI-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C2]](s32)
    ; CI-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C]](s32)
    ; CI-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C1]](s32)
    ; CI-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C2]](s32)
    ; CI-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C]](s32)
    ; CI-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C1]](s32)
    ; CI-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C2]](s32)
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C3]]
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C3]]
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; CI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C3]]
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
    ; CI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C3]]
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C2]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
    ; CI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C3]]
    ; CI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C3]]
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL3]]
    ; CI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C3]]
    ; CI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C1]](s32)
    ; CI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[OR3]], [[SHL4]]
    ; CI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C3]]
    ; CI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C2]](s32)
    ; CI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[OR4]], [[SHL5]]
    ; CI-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[C3]]
    ; CI-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[LSHR6]], [[C3]]
    ; CI-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND9]], [[C]](s32)
    ; CI-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL6]]
    ; CI-NEXT: [[AND10:%[0-9]+]]:_(s32) = G_AND [[LSHR7]], [[C3]]
    ; CI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND10]], [[C1]](s32)
    ; CI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[OR6]], [[SHL7]]
    ; CI-NEXT: [[AND11:%[0-9]+]]:_(s32) = G_AND [[LSHR8]], [[C3]]
    ; CI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C2]](s32)
    ; CI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]]
    ; CI-NEXT: [[AND12:%[0-9]+]]:_(s32) = G_AND [[UV3]], [[C3]]
    ; CI-NEXT: [[AND13:%[0-9]+]]:_(s32) = G_AND [[LSHR9]], [[C3]]
    ; CI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[AND13]], [[C]](s32)
    ; CI-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[AND12]], [[SHL9]]
    ; CI-NEXT: [[AND14:%[0-9]+]]:_(s32) = G_AND [[LSHR10]], [[C3]]
    ; CI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[AND14]], [[C1]](s32)
    ; CI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[OR9]], [[SHL10]]
    ; CI-NEXT: [[AND15:%[0-9]+]]:_(s32) = G_AND [[LSHR11]], [[C3]]
    ; CI-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C2]](s32)
    ; CI-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]]
    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
    ; VI-LABEL: name: test_load_flat_v16s8_align16
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
    ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
    ; VI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
    ; VI-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C]](s32)
    ; VI-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C1]](s32)
    ; VI-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C2]](s32)
    ; VI-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C]](s32)
    ; VI-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C1]](s32)
    ; VI-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C2]](s32)
    ; VI-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C]](s32)
    ; VI-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C1]](s32)
    ; VI-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C2]](s32)
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C3]]
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C3]]
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C3]]
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C3]]
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C2]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C3]]
    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C3]]
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL3]]
    ; VI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C3]]
    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C1]](s32)
    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[OR3]], [[SHL4]]
    ; VI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C3]]
    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C2]](s32)
    ; VI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[OR4]], [[SHL5]]
    ; VI-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[C3]]
    ; VI-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[LSHR6]], [[C3]]
    ; VI-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND9]], [[C]](s32)
    ; VI-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL6]]
    ; VI-NEXT: [[AND10:%[0-9]+]]:_(s32) = G_AND [[LSHR7]], [[C3]]
    ; VI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND10]], [[C1]](s32)
    ; VI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[OR6]], [[SHL7]]
    ; VI-NEXT: [[AND11:%[0-9]+]]:_(s32) = G_AND [[LSHR8]], [[C3]]
    ; VI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C2]](s32)
    ; VI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]]
    ; VI-NEXT: [[AND12:%[0-9]+]]:_(s32) = G_AND [[UV3]], [[C3]]
    ; VI-NEXT: [[AND13:%[0-9]+]]:_(s32) = G_AND [[LSHR9]], [[C3]]
    ; VI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[AND13]], [[C]](s32)
    ; VI-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[AND12]], [[SHL9]]
    ; VI-NEXT: [[AND14:%[0-9]+]]:_(s32) = G_AND [[LSHR10]], [[C3]]
    ; VI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[AND14]], [[C1]](s32)
    ; VI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[OR9]], [[SHL10]]
    ; VI-NEXT: [[AND15:%[0-9]+]]:_(s32) = G_AND [[LSHR11]], [[C3]]
    ; VI-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C2]](s32)
    ; VI-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]]
    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
    ; GFX9-LABEL: name: test_load_flat_v16s8_align16
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C]](s32)
    ; GFX9-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C1]](s32)
    ; GFX9-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C2]](s32)
    ; GFX9-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C]](s32)
    ; GFX9-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C1]](s32)
    ; GFX9-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C2]](s32)
    ; GFX9-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C]](s32)
    ; GFX9-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C1]](s32)
    ; GFX9-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C2]](s32)
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; GFX9-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C3]]
    ; GFX9-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C3]]
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; GFX9-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C3]]
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
    ; GFX9-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C3]]
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C2]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
    ; GFX9-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C3]]
    ; GFX9-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C3]]
    ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
    ; GFX9-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL3]]
    ; GFX9-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C3]]
    ; GFX9-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C1]](s32)
    ; GFX9-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[OR3]], [[SHL4]]
    ; GFX9-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C3]]
    ; GFX9-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C2]](s32)
    ; GFX9-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[OR4]], [[SHL5]]
    ; GFX9-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[C3]]
    ; GFX9-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[LSHR6]], [[C3]]
    ; GFX9-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND9]], [[C]](s32)
    ; GFX9-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL6]]
    ; GFX9-NEXT: [[AND10:%[0-9]+]]:_(s32) = G_AND [[LSHR7]], [[C3]]
    ; GFX9-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND10]], [[C1]](s32)
    ; GFX9-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[OR6]], [[SHL7]]
    ; GFX9-NEXT: [[AND11:%[0-9]+]]:_(s32) = G_AND [[LSHR8]], [[C3]]
    ; GFX9-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C2]](s32)
    ; GFX9-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]]
    ; GFX9-NEXT: [[AND12:%[0-9]+]]:_(s32) = G_AND [[UV3]], [[C3]]
    ; GFX9-NEXT: [[AND13:%[0-9]+]]:_(s32) = G_AND [[LSHR9]], [[C3]]
    ; GFX9-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[AND13]], [[C]](s32)
    ; GFX9-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[AND12]], [[SHL9]]
    ; GFX9-NEXT: [[AND14:%[0-9]+]]:_(s32) = G_AND [[LSHR10]], [[C3]]
    ; GFX9-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[AND14]], [[C1]](s32)
    ; GFX9-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[OR9]], [[SHL10]]
    ; GFX9-NEXT: [[AND15:%[0-9]+]]:_(s32) = G_AND [[LSHR11]], [[C3]]
    ; GFX9-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C2]](s32)
    ; GFX9-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]]
    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<16 x s8>) = G_LOAD %0 :: (load (<16 x s8>), align 16, addrspace 0)
    %2:_(<4 x s32>) = G_BITCAST %1
    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
...

---
name: test_load_flat_v32s8_align32
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v32s8_align32
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
    ; CI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<4 x s32>)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
    ; CI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C3]](s32)
    ; CI-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C1]](s32)
    ; CI-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C2]](s32)
    ; CI-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C3]](s32)
    ; CI-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C1]](s32)
    ; CI-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C2]](s32)
    ; CI-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C3]](s32)
    ; CI-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C1]](s32)
    ; CI-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C2]](s32)
    ; CI-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C3]](s32)
    ; CI-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C1]](s32)
    ; CI-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C2]](s32)
    ; CI-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C3]](s32)
    ; CI-NEXT: [[LSHR15:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C1]](s32)
    ; CI-NEXT: [[LSHR16:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C2]](s32)
    ; CI-NEXT: [[LSHR17:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C3]](s32)
    ; CI-NEXT: [[LSHR18:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C1]](s32)
    ; CI-NEXT: [[LSHR19:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C2]](s32)
    ; CI-NEXT: [[LSHR20:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C3]](s32)
    ; CI-NEXT: [[LSHR21:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C1]](s32)
    ; CI-NEXT: [[LSHR22:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C2]](s32)
    ; CI-NEXT: [[LSHR23:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C3]](s32)
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C4]]
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C4]]
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; CI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C4]]
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C2]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
    ; CI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C4]]
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C3]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
    ; CI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C4]]
    ; CI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C4]]
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C1]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL3]]
    ; CI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C4]]
    ; CI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C2]](s32)
    ; CI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[OR3]], [[SHL4]]
    ; CI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C4]]
    ; CI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C3]](s32)
    ; CI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[OR4]], [[SHL5]]
    ; CI-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[C4]]
    ; CI-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[LSHR6]], [[C4]]
    ; CI-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND9]], [[C1]](s32)
    ; CI-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL6]]
    ; CI-NEXT: [[AND10:%[0-9]+]]:_(s32) = G_AND [[LSHR7]], [[C4]]
    ; CI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND10]], [[C2]](s32)
    ; CI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[OR6]], [[SHL7]]
    ; CI-NEXT: [[AND11:%[0-9]+]]:_(s32) = G_AND [[LSHR8]], [[C4]]
    ; CI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C3]](s32)
    ; CI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]]
    ; CI-NEXT: [[AND12:%[0-9]+]]:_(s32) = G_AND [[UV3]], [[C4]]
    ; CI-NEXT: [[AND13:%[0-9]+]]:_(s32) = G_AND [[LSHR9]], [[C4]]
    ; CI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[AND13]], [[C1]](s32)
    ; CI-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[AND12]], [[SHL9]]
    ; CI-NEXT: [[AND14:%[0-9]+]]:_(s32) = G_AND [[LSHR10]], [[C4]]
    ; CI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[AND14]], [[C2]](s32)
    ; CI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[OR9]], [[SHL10]]
    ; CI-NEXT: [[AND15:%[0-9]+]]:_(s32) = G_AND [[LSHR11]], [[C4]]
    ; CI-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C3]](s32)
    ; CI-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]]
    ; CI-NEXT: [[AND16:%[0-9]+]]:_(s32) = G_AND [[UV4]], [[C4]]
    ; CI-NEXT: [[AND17:%[0-9]+]]:_(s32) = G_AND [[LSHR12]], [[C4]]
    ; CI-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[AND17]], [[C1]](s32)
    ; CI-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[AND16]], [[SHL12]]
    ; CI-NEXT: [[AND18:%[0-9]+]]:_(s32) = G_AND [[LSHR13]], [[C4]]
    ; CI-NEXT: [[SHL13:%[0-9]+]]:_(s32) = G_SHL [[AND18]], [[C2]](s32)
    ; CI-NEXT: [[OR13:%[0-9]+]]:_(s32) = G_OR [[OR12]], [[SHL13]]
    ; CI-NEXT: [[AND19:%[0-9]+]]:_(s32) = G_AND [[LSHR14]], [[C4]]
    ; CI-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[AND19]], [[C3]](s32)
    ; CI-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[OR13]], [[SHL14]]
    ; CI-NEXT: [[AND20:%[0-9]+]]:_(s32) = G_AND [[UV5]], [[C4]]
    ; CI-NEXT: [[AND21:%[0-9]+]]:_(s32) = G_AND [[LSHR15]], [[C4]]
    ; CI-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[AND21]], [[C1]](s32)
    ; CI-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[AND20]], [[SHL15]]
    ; CI-NEXT: [[AND22:%[0-9]+]]:_(s32) = G_AND [[LSHR16]], [[C4]]
    ; CI-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[AND22]], [[C2]](s32)
    ; CI-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[OR15]], [[SHL16]]
    ; CI-NEXT: [[AND23:%[0-9]+]]:_(s32) = G_AND [[LSHR17]], [[C4]]
    ; CI-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[AND23]], [[C3]](s32)
    ; CI-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[OR16]], [[SHL17]]
    ; CI-NEXT: [[AND24:%[0-9]+]]:_(s32) = G_AND [[UV6]], [[C4]]
    ; CI-NEXT: [[AND25:%[0-9]+]]:_(s32) = G_AND [[LSHR18]], [[C4]]
    ; CI-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[AND25]], [[C1]](s32)
    ; CI-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[AND24]], [[SHL18]]
    ; CI-NEXT: [[AND26:%[0-9]+]]:_(s32) = G_AND [[LSHR19]], [[C4]]
    ; CI-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[AND26]], [[C2]](s32)
    ; CI-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[OR18]], [[SHL19]]
    ; CI-NEXT: [[AND27:%[0-9]+]]:_(s32) = G_AND [[LSHR20]], [[C4]]
    ; CI-NEXT: [[SHL20:%[0-9]+]]:_(s32) = G_SHL [[AND27]], [[C3]](s32)
    ; CI-NEXT: [[OR20:%[0-9]+]]:_(s32) = G_OR [[OR19]], [[SHL20]]
    ; CI-NEXT: [[AND28:%[0-9]+]]:_(s32) = G_AND [[UV7]], [[C4]]
    ; CI-NEXT: [[AND29:%[0-9]+]]:_(s32) = G_AND [[LSHR21]], [[C4]]
    ; CI-NEXT: [[SHL21:%[0-9]+]]:_(s32) = G_SHL [[AND29]], [[C1]](s32)
    ; CI-NEXT: [[OR21:%[0-9]+]]:_(s32) = G_OR [[AND28]], [[SHL21]]
    ; CI-NEXT: [[AND30:%[0-9]+]]:_(s32) = G_AND [[LSHR22]], [[C4]]
    ; CI-NEXT: [[SHL22:%[0-9]+]]:_(s32) = G_SHL [[AND30]], [[C2]](s32)
    ; CI-NEXT: [[OR22:%[0-9]+]]:_(s32) = G_OR [[OR21]], [[SHL22]]
    ; CI-NEXT: [[AND31:%[0-9]+]]:_(s32) = G_AND [[LSHR23]], [[C4]]
    ; CI-NEXT: [[SHL23:%[0-9]+]]:_(s32) = G_SHL [[AND31]], [[C3]](s32)
    ; CI-NEXT: [[OR23:%[0-9]+]]:_(s32) = G_OR [[OR22]], [[SHL23]]
    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32), [[OR14]](s32), [[OR17]](s32), [[OR20]](s32), [[OR23]](s32)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<8 x s32>)
    ; VI-LABEL: name: test_load_flat_v32s8_align32
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
    ; VI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<4 x s32>)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
    ; VI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C3]](s32)
    ; VI-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C1]](s32)
    ; VI-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C2]](s32)
    ; VI-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C3]](s32)
    ; VI-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C1]](s32)
    ; VI-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C2]](s32)
    ; VI-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C3]](s32)
    ; VI-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C1]](s32)
    ; VI-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C2]](s32)
    ; VI-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C3]](s32)
    ; VI-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C1]](s32)
    ; VI-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C2]](s32)
    ; VI-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C3]](s32)
    ; VI-NEXT: [[LSHR15:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C1]](s32)
    ; VI-NEXT: [[LSHR16:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C2]](s32)
    ; VI-NEXT: [[LSHR17:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C3]](s32)
    ; VI-NEXT: [[LSHR18:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C1]](s32)
    ; VI-NEXT: [[LSHR19:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C2]](s32)
    ; VI-NEXT: [[LSHR20:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C3]](s32)
    ; VI-NEXT: [[LSHR21:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C1]](s32)
    ; VI-NEXT: [[LSHR22:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C2]](s32)
    ; VI-NEXT: [[LSHR23:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C3]](s32)
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C4]]
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C4]]
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C4]]
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C2]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C4]]
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C3]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C4]]
    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C4]]
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C1]](s32)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL3]]
    ; VI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C4]]
    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C2]](s32)
    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[OR3]], [[SHL4]]
    ; VI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C4]]
    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C3]](s32)
    ; VI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[OR4]], [[SHL5]]
    ; VI-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[C4]]
    ; VI-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[LSHR6]], [[C4]]
    ; VI-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND9]], [[C1]](s32)
    ; VI-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL6]]
    ; VI-NEXT: [[AND10:%[0-9]+]]:_(s32) = G_AND [[LSHR7]], [[C4]]
    ; VI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND10]], [[C2]](s32)
    ; VI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[OR6]], [[SHL7]]
    ; VI-NEXT: [[AND11:%[0-9]+]]:_(s32) = G_AND [[LSHR8]], [[C4]]
    ; VI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C3]](s32)
    ; VI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]]
    ; VI-NEXT: [[AND12:%[0-9]+]]:_(s32) = G_AND [[UV3]], [[C4]]
    ; VI-NEXT: [[AND13:%[0-9]+]]:_(s32) = G_AND [[LSHR9]], [[C4]]
    ; VI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[AND13]], [[C1]](s32)
    ; VI-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[AND12]], [[SHL9]]
    ; VI-NEXT: [[AND14:%[0-9]+]]:_(s32) = G_AND [[LSHR10]], [[C4]]
    ; VI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[AND14]], [[C2]](s32)
    ; VI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[OR9]], [[SHL10]]
    ; VI-NEXT: [[AND15:%[0-9]+]]:_(s32) = G_AND [[LSHR11]], [[C4]]
    ; VI-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C3]](s32)
    ; VI-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]]
    ; VI-NEXT: [[AND16:%[0-9]+]]:_(s32) = G_AND [[UV4]], [[C4]]
    ; VI-NEXT: [[AND17:%[0-9]+]]:_(s32) = G_AND [[LSHR12]], [[C4]]
    ; VI-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[AND17]], [[C1]](s32)
    ; VI-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[AND16]], [[SHL12]]
    ; VI-NEXT: [[AND18:%[0-9]+]]:_(s32) = G_AND [[LSHR13]], [[C4]]
    ; VI-NEXT: [[SHL13:%[0-9]+]]:_(s32) = G_SHL [[AND18]], [[C2]](s32)
    ; VI-NEXT: [[OR13:%[0-9]+]]:_(s32) = G_OR [[OR12]], [[SHL13]]
    ; VI-NEXT: [[AND19:%[0-9]+]]:_(s32) = G_AND [[LSHR14]], [[C4]]
    ; VI-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[AND19]], [[C3]](s32)
    ; VI-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[OR13]], [[SHL14]]
    ; VI-NEXT: [[AND20:%[0-9]+]]:_(s32) = G_AND [[UV5]], [[C4]]
    ; VI-NEXT: [[AND21:%[0-9]+]]:_(s32) = G_AND [[LSHR15]], [[C4]]
    ; VI-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[AND21]], [[C1]](s32)
    ; VI-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[AND20]], [[SHL15]]
    ; VI-NEXT: [[AND22:%[0-9]+]]:_(s32) = G_AND [[LSHR16]], [[C4]]
    ; VI-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[AND22]], [[C2]](s32)
    ; VI-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[OR15]], [[SHL16]]
    ; VI-NEXT: [[AND23:%[0-9]+]]:_(s32) = G_AND [[LSHR17]], [[C4]]
    ; VI-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[AND23]], [[C3]](s32)
    ; VI-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[OR16]], [[SHL17]]
    ; VI-NEXT: [[AND24:%[0-9]+]]:_(s32) = G_AND [[UV6]], [[C4]]
    ; VI-NEXT: [[AND25:%[0-9]+]]:_(s32) = G_AND [[LSHR18]], [[C4]]
    ; VI-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[AND25]], [[C1]](s32)
    ; VI-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[AND24]], [[SHL18]]
    ; VI-NEXT: [[AND26:%[0-9]+]]:_(s32) = G_AND [[LSHR19]], [[C4]]
    ; VI-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[AND26]], [[C2]](s32)
    ; VI-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[OR18]], [[SHL19]]
    ; VI-NEXT: [[AND27:%[0-9]+]]:_(s32) = G_AND [[LSHR20]], [[C4]]
    ; VI-NEXT: [[SHL20:%[0-9]+]]:_(s32) = G_SHL [[AND27]], [[C3]](s32)
    ; VI-NEXT: [[OR20:%[0-9]+]]:_(s32) = G_OR [[OR19]], [[SHL20]]
    ; VI-NEXT: [[AND28:%[0-9]+]]:_(s32) = G_AND [[UV7]], [[C4]]
    ; VI-NEXT: [[AND29:%[0-9]+]]:_(s32) = G_AND [[LSHR21]], [[C4]]
    ; VI-NEXT: [[SHL21:%[0-9]+]]:_(s32) = G_SHL [[AND29]], [[C1]](s32)
    ; VI-NEXT: [[OR21:%[0-9]+]]:_(s32) = G_OR [[AND28]], [[SHL21]]
    ; VI-NEXT: [[AND30:%[0-9]+]]:_(s32) = G_AND [[LSHR22]], [[C4]]
    ; VI-NEXT: [[SHL22:%[0-9]+]]:_(s32) = G_SHL [[AND30]], [[C2]](s32)
    ; VI-NEXT: [[OR22:%[0-9]+]]:_(s32) = G_OR [[OR21]], [[SHL22]]
    ; VI-NEXT: [[AND31:%[0-9]+]]:_(s32) = G_AND [[LSHR23]], [[C4]]
    ; VI-NEXT: [[SHL23:%[0-9]+]]:_(s32) = G_SHL [[AND31]], [[C3]](s32)
    ; VI-NEXT: [[OR23:%[0-9]+]]:_(s32) = G_OR [[OR22]], [[SHL23]]
    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32), [[OR14]](s32), [[OR17]](s32), [[OR20]](s32), [[OR23]](s32)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<8 x s32>)
    ; GFX9-LABEL: name: test_load_flat_v32s8_align32
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
    ; GFX9-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<4 x s32>)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C3]](s32)
    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C1]](s32)
    ; GFX9-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C2]](s32)
    ; GFX9-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C3]](s32)
    ; GFX9-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C1]](s32)
    ; GFX9-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C2]](s32)
    ; GFX9-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C3]](s32)
    ; GFX9-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C1]](s32)
    ; GFX9-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C2]](s32)
    ; GFX9-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C3]](s32)
    ; GFX9-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C1]](s32)
    ; GFX9-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C2]](s32)
    ; GFX9-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C3]](s32)
    ; GFX9-NEXT: [[LSHR15:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C1]](s32)
    ; GFX9-NEXT: [[LSHR16:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C2]](s32)
    ; GFX9-NEXT: [[LSHR17:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C3]](s32)
    ; GFX9-NEXT: [[LSHR18:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C1]](s32)
    ; GFX9-NEXT: [[LSHR19:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C2]](s32)
    ; GFX9-NEXT: [[LSHR20:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C3]](s32)
    ; GFX9-NEXT: [[LSHR21:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C1]](s32)
    ; GFX9-NEXT: [[LSHR22:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C2]](s32)
    ; GFX9-NEXT: [[LSHR23:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C3]](s32)
    ; GFX9-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
    ; GFX9-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C4]]
    ; GFX9-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C4]]
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; GFX9-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C4]]
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C2]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
    ; GFX9-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C4]]
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C3]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
    ; GFX9-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C4]]
    ; GFX9-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C4]]
    ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C1]](s32)
    ; GFX9-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL3]]
    ; GFX9-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C4]]
    ; GFX9-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C2]](s32)
    ; GFX9-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[OR3]], [[SHL4]]
    ; GFX9-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C4]]
    ; GFX9-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C3]](s32)
    ; GFX9-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[OR4]], [[SHL5]]
    ; GFX9-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[C4]]
    ; GFX9-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[LSHR6]], [[C4]]
    ; GFX9-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND9]], [[C1]](s32)
    ; GFX9-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL6]]
    ; GFX9-NEXT: [[AND10:%[0-9]+]]:_(s32) = G_AND [[LSHR7]], [[C4]]
    ; GFX9-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND10]], [[C2]](s32)
    ; GFX9-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[OR6]], [[SHL7]]
    ; GFX9-NEXT: [[AND11:%[0-9]+]]:_(s32) = G_AND [[LSHR8]], [[C4]]
    ; GFX9-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C3]](s32)
    ; GFX9-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]]
    ; GFX9-NEXT: [[AND12:%[0-9]+]]:_(s32) = G_AND [[UV3]], [[C4]]
    ; GFX9-NEXT: [[AND13:%[0-9]+]]:_(s32) = G_AND [[LSHR9]], [[C4]]
    ; GFX9-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[AND13]], [[C1]](s32)
    ; GFX9-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[AND12]], [[SHL9]]
    ; GFX9-NEXT: [[AND14:%[0-9]+]]:_(s32) = G_AND [[LSHR10]], [[C4]]
    ; GFX9-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[AND14]], [[C2]](s32)
    ; GFX9-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[OR9]], [[SHL10]]
    ; GFX9-NEXT: [[AND15:%[0-9]+]]:_(s32) = G_AND [[LSHR11]], [[C4]]
    ; GFX9-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C3]](s32)
    ; GFX9-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]]
    ; GFX9-NEXT: [[AND16:%[0-9]+]]:_(s32) = G_AND [[UV4]], [[C4]]
    ; GFX9-NEXT: [[AND17:%[0-9]+]]:_(s32) = G_AND [[LSHR12]], [[C4]]
    ; GFX9-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[AND17]], [[C1]](s32)
    ; GFX9-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[AND16]], [[SHL12]]
    ; GFX9-NEXT: [[AND18:%[0-9]+]]:_(s32) = G_AND [[LSHR13]], [[C4]]
    ; GFX9-NEXT: [[SHL13:%[0-9]+]]:_(s32) = G_SHL [[AND18]], [[C2]](s32)
    ; GFX9-NEXT: [[OR13:%[0-9]+]]:_(s32) = G_OR [[OR12]], [[SHL13]]
    ; GFX9-NEXT: [[AND19:%[0-9]+]]:_(s32) = G_AND [[LSHR14]], [[C4]]
    ; GFX9-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[AND19]], [[C3]](s32)
    ; GFX9-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[OR13]], [[SHL14]]
    ; GFX9-NEXT: [[AND20:%[0-9]+]]:_(s32) = G_AND [[UV5]], [[C4]]
    ; GFX9-NEXT: [[AND21:%[0-9]+]]:_(s32) = G_AND [[LSHR15]], [[C4]]
    ; GFX9-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[AND21]], [[C1]](s32)
    ; GFX9-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[AND20]], [[SHL15]]
    ; GFX9-NEXT: [[AND22:%[0-9]+]]:_(s32) = G_AND [[LSHR16]], [[C4]]
    ; GFX9-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[AND22]], [[C2]](s32)
    ; GFX9-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[OR15]], [[SHL16]]
    ; GFX9-NEXT: [[AND23:%[0-9]+]]:_(s32) = G_AND [[LSHR17]], [[C4]]
    ; GFX9-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[AND23]], [[C3]](s32)
    ; GFX9-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[OR16]], [[SHL17]]
    ; GFX9-NEXT: [[AND24:%[0-9]+]]:_(s32) = G_AND [[UV6]], [[C4]]
    ; GFX9-NEXT: [[AND25:%[0-9]+]]:_(s32) = G_AND [[LSHR18]], [[C4]]
    ; GFX9-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[AND25]], [[C1]](s32)
    ; GFX9-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[AND24]], [[SHL18]]
    ; GFX9-NEXT: [[AND26:%[0-9]+]]:_(s32) = G_AND [[LSHR19]], [[C4]]
    ; GFX9-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[AND26]], [[C2]](s32)
    ; GFX9-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[OR18]], [[SHL19]]
    ; GFX9-NEXT: [[AND27:%[0-9]+]]:_(s32) = G_AND [[LSHR20]], [[C4]]
    ; GFX9-NEXT: [[SHL20:%[0-9]+]]:_(s32) = G_SHL [[AND27]], [[C3]](s32)
    ; GFX9-NEXT: [[OR20:%[0-9]+]]:_(s32) = G_OR [[OR19]], [[SHL20]]
    ; GFX9-NEXT: [[AND28:%[0-9]+]]:_(s32) = G_AND [[UV7]], [[C4]]
    ; GFX9-NEXT: [[AND29:%[0-9]+]]:_(s32) = G_AND [[LSHR21]], [[C4]]
    ; GFX9-NEXT: [[SHL21:%[0-9]+]]:_(s32) = G_SHL [[AND29]], [[C1]](s32)
    ; GFX9-NEXT: [[OR21:%[0-9]+]]:_(s32) = G_OR [[AND28]], [[SHL21]]
    ; GFX9-NEXT: [[AND30:%[0-9]+]]:_(s32) = G_AND [[LSHR22]], [[C4]]
    ; GFX9-NEXT: [[SHL22:%[0-9]+]]:_(s32) = G_SHL [[AND30]], [[C2]](s32)
    ; GFX9-NEXT: [[OR22:%[0-9]+]]:_(s32) = G_OR [[OR21]], [[SHL22]]
    ; GFX9-NEXT: [[AND31:%[0-9]+]]:_(s32) = G_AND [[LSHR23]], [[C4]]
    ; GFX9-NEXT: [[SHL23:%[0-9]+]]:_(s32) = G_SHL [[AND31]], [[C3]](s32)
    ; GFX9-NEXT: [[OR23:%[0-9]+]]:_(s32) = G_OR [[OR22]], [[SHL23]]
    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32), [[OR14]](s32), [[OR17]](s32), [[OR20]](s32), [[OR23]](s32)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<8 x s32>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<32 x s8>) = G_LOAD %0 :: (load (<32 x s8>), align 32, addrspace 0)
    %2:_(<8 x s32>) = G_BITCAST %1
    $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %2
...

---

name: test_load_flat_v2s16_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2s16_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>))
    ; CI-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
    ; VI-LABEL: name: test_load_flat_v2s16_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>))
    ; VI-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
    ; GFX9-LABEL: name: test_load_flat_v2s16_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>))
    ; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x s16>) = G_LOAD %0 :: (load (<2 x s16>), align 4, addrspace 0)
    $vgpr0 = COPY %1
...

---
name: test_load_flat_v2s16_align2
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2s16_align2
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C1]]
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C1]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C2]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
    ; CI-NEXT: $vgpr0 = COPY [[BITCAST]](<2 x s16>)
    ; VI-LABEL: name: test_load_flat_v2s16_align2
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C1]]
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C1]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C2]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
    ; VI-NEXT: $vgpr0 = COPY [[BITCAST]](<2 x s16>)
    ; GFX9-LABEL: name: test_load_flat_v2s16_align2
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LOAD]](s32), [[LOAD1]](s32)
    ; GFX9-NEXT: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC]](<2 x s16>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x s16>) = G_LOAD %0 :: (load (<2 x s16>), align 2, addrspace 0)
    $vgpr0 = COPY %1
...

---
name: test_load_flat_v2s16_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2s16_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C3]]
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[OR1]], [[C3]]
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C4]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL2]]
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
    ; CI-NEXT: $vgpr0 = COPY [[BITCAST]](<2 x s16>)
    ; VI-LABEL: name: test_load_flat_v2s16_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C3]]
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[OR1]], [[C3]]
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C4]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL2]]
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
    ; VI-NEXT: $vgpr0 = COPY [[BITCAST]](<2 x s16>)
    ; GFX9-LABEL: name: test_load_flat_v2s16_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[OR]](s32), [[OR1]](s32)
    ; GFX9-NEXT: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC]](<2 x s16>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x s16>) = G_LOAD %0 :: (load (<2 x s16>), align 1, addrspace 0)
    $vgpr0 = COPY %1
...

---
name: test_load_flat_v3s16_align8
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v3s16_align8
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
    ; CI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
    ; CI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
    ; CI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
    ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
    ; CI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
    ; CI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
    ; CI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
    ; CI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; CI-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
    ; CI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
    ; CI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
    ; CI-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
    ; CI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
    ; CI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
    ; CI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
    ; VI-LABEL: name: test_load_flat_v3s16_align8
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
    ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
    ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
    ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
    ; VI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
    ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
    ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
    ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; VI-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
    ; VI-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
    ; VI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
    ; GFX9-LABEL: name: test_load_flat_v3s16_align8
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
    ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[BITCAST2]](s32)
    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST3]](s32)
    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 8, addrspace 0)
    %2:_(<3 x s16>) = G_IMPLICIT_DEF
    %3:_(<6 x s16>) = G_CONCAT_VECTORS %1, %2
    $vgpr0_vgpr1_vgpr2 = COPY %3
...

---
name: test_load_flat_v3s16_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v3s16_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C1]](s64)
    ; CI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
    ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
    ; CI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C2]](s32)
    ; CI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C3]]
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C3]]
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C2]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; CI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
    ; CI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C3]]
    ; CI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C3]]
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C2]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
    ; CI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
    ; CI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C3]]
    ; CI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C3]]
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C2]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
    ; CI-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
    ; VI-LABEL: name: test_load_flat_v3s16_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C1]](s64)
    ; VI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
    ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C2]](s32)
    ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C3]]
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C3]]
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C2]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C3]]
    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C3]]
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C2]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
    ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C3]]
    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C3]]
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C2]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
    ; VI-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
    ; GFX9-LABEL: name: test_load_flat_v3s16_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C1]](s64)
    ; GFX9-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C2]](s32)
    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LOAD]](s32), [[LOAD1]](s32)
    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LOAD2]](s32), [[BITCAST]](s32)
    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR]](s32), [[BITCAST1]](s32)
    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 4, addrspace 0)
    %2:_(<3 x s16>) = G_IMPLICIT_DEF
    %3:_(<6 x s16>) = G_CONCAT_VECTORS %1, %2
    $vgpr0_vgpr1_vgpr2 = COPY %3
...

---
name: test_load_flat_v3s16_align2
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v3s16_align2
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C1]](s64)
    ; CI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
    ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
    ; CI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C2]](s32)
    ; CI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C3]]
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C3]]
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C2]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; CI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
    ; CI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C3]]
    ; CI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C3]]
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C2]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
    ; CI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
    ; CI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C3]]
    ; CI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C3]]
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C2]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
    ; CI-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
    ; VI-LABEL: name: test_load_flat_v3s16_align2
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C1]](s64)
    ; VI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
    ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C2]](s32)
    ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C3]]
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C3]]
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C2]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C3]]
    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C3]]
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C2]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
    ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C3]]
    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C3]]
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C2]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
    ; VI-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
    ; GFX9-LABEL: name: test_load_flat_v3s16_align2
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C1]](s64)
    ; GFX9-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C2]](s32)
    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LOAD]](s32), [[LOAD1]](s32)
    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LOAD2]](s32), [[BITCAST]](s32)
    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR]](s32), [[BITCAST1]](s32)
    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 2, addrspace 0)
    %2:_(<3 x s16>) = G_IMPLICIT_DEF
    %3:_(<6 x s16>) = G_CONCAT_VECTORS %1, %2
    $vgpr0_vgpr1_vgpr2 = COPY %3
...

---
name: test_load_flat_v3s16_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v3s16_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C3]](s64)
    ; CI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; CI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; CI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
    ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
    ; CI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C4]](s32)
    ; CI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
    ; CI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C5]]
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[OR1]], [[C5]]
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C4]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL3]]
    ; CI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
    ; CI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C5]]
    ; CI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C5]]
    ; CI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C4]](s32)
    ; CI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL4]]
    ; CI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32)
    ; CI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C5]]
    ; CI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C5]]
    ; CI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C4]](s32)
    ; CI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL5]]
    ; CI-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32)
    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
    ; VI-LABEL: name: test_load_flat_v3s16_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C3]](s64)
    ; VI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; VI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; VI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
    ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C4]](s32)
    ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
    ; VI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C5]]
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[OR1]], [[C5]]
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C4]](s32)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL3]]
    ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C5]]
    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C5]]
    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C4]](s32)
    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL4]]
    ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32)
    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C5]]
    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C5]]
    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C4]](s32)
    ; VI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL5]]
    ; VI-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32)
    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
    ; GFX9-LABEL: name: test_load_flat_v3s16_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C3]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; GFX9-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
    ; GFX9-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C4]](s32)
    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[OR]](s32), [[OR1]](s32)
    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[OR2]](s32), [[BITCAST]](s32)
    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR]](s32), [[BITCAST1]](s32)
    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 1, addrspace 0)
    %2:_(<3 x s16>) = G_IMPLICIT_DEF
    %3:_(<6 x s16>) = G_CONCAT_VECTORS %1, %2
    $vgpr0_vgpr1_vgpr2 = COPY %3
...

---
name: test_load_flat_v4s16_align8
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v4s16_align8
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
    ; VI-LABEL: name: test_load_flat_v4s16_align8
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
    ; GFX9-LABEL: name: test_load_flat_v4s16_align8
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 8, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_v4s16_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v4s16_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 4)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
    ; VI-LABEL: name: test_load_flat_v4s16_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 4)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
    ; GFX9-LABEL: name: test_load_flat_v4s16_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 4)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 4, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_v4s16_align2
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v4s16_align2
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C1]](s64)
    ; CI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C3]]
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C3]]
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C4]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
    ; CI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C3]]
    ; CI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C3]]
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C4]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
    ; CI-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
    ; VI-LABEL: name: test_load_flat_v4s16_align2
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C1]](s64)
    ; VI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C3]]
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C3]]
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C4]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C3]]
    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C3]]
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C4]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
    ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
    ; GFX9-LABEL: name: test_load_flat_v4s16_align2
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C1]](s64)
    ; GFX9-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LOAD]](s32), [[LOAD1]](s32)
    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LOAD2]](s32), [[LOAD3]](s32)
    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 2, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_v4s16_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v4s16_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C3]](s64)
    ; CI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; CI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; CI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
    ; CI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; CI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; CI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; CI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; CI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
    ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C5]]
    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[OR1]], [[C5]]
    ; CI-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C6]](s32)
    ; CI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL4]]
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32)
    ; CI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C5]]
    ; CI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[OR3]], [[C5]]
    ; CI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C6]](s32)
    ; CI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL5]]
    ; CI-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32)
    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
    ; VI-LABEL: name: test_load_flat_v4s16_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C3]](s64)
    ; VI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; VI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; VI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
    ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; VI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; VI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; VI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; VI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C5]]
    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[OR1]], [[C5]]
    ; VI-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C6]](s32)
    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL4]]
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32)
    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C5]]
    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[OR3]], [[C5]]
    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C6]](s32)
    ; VI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL5]]
    ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32)
    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
    ; GFX9-LABEL: name: test_load_flat_v4s16_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C3]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; GFX9-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
    ; GFX9-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
    ; GFX9-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; GFX9-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; GFX9-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[OR]](s32), [[OR1]](s32)
    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[OR2]](s32), [[OR3]](s32)
    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 1, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_v8s16_align8
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v8s16_align8
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
    ; VI-LABEL: name: test_load_flat_v8s16_align8
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
    ; GFX9-LABEL: name: test_load_flat_v8s16_align8
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<8 x s16>) = G_LOAD %0 :: (load (<8 x s16>), align 8, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
...

---
name: test_load_flat_v2s32_align8
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2s32_align8
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
    ; VI-LABEL: name: test_load_flat_v2s32_align8
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
    ; GFX9-LABEL: name: test_load_flat_v2s32_align8
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 8, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_v2s32_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2s32_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
    ; VI-LABEL: name: test_load_flat_v2s32_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
    ; GFX9-LABEL: name: test_load_flat_v2s32_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 4, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...


---
name: test_load_flat_v2s32_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2s32_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
    ; VI-LABEL: name: test_load_flat_v2s32_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
    ; GFX9-LABEL: name: test_load_flat_v2s32_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 4, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_v3s32_align16
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v3s32_align16
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
    ; VI-LABEL: name: test_load_flat_v3s32_align16
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
    ; GFX9-LABEL: name: test_load_flat_v3s32_align16
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<3 x s32>) = G_LOAD %0 :: (load (<3 x s32>), align 16, addrspace 0)
    $vgpr0_vgpr1_vgpr2 = COPY %1
...



---
name: test_load_flat_v3s32_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v3s32_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
    ; VI-LABEL: name: test_load_flat_v3s32_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
    ; GFX9-LABEL: name: test_load_flat_v3s32_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<3 x s32>) = G_LOAD %0 :: (load (<3 x s32>), align 4, addrspace 0)
    $vgpr0_vgpr1_vgpr2 = COPY %1
...

---
name: test_load_flat_v4s32_align16
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v4s32_align16
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
    ; VI-LABEL: name: test_load_flat_v4s32_align16
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
    ; GFX9-LABEL: name: test_load_flat_v4s32_align16
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<4 x s32>) = G_LOAD %0 :: (load (<4 x s32>), align 16, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
...

---
name: test_load_flat_v4s32_align8
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v4s32_align8
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
    ; VI-LABEL: name: test_load_flat_v4s32_align8
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
    ; GFX9-LABEL: name: test_load_flat_v4s32_align8
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<4 x s32>) = G_LOAD %0 :: (load (<4 x s32>), align 8, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
...

---
name: test_load_flat_v4s32_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v4s32_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
    ; VI-LABEL: name: test_load_flat_v4s32_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
    ; GFX9-LABEL: name: test_load_flat_v4s32_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<4 x s32>) = G_LOAD %0 :: (load (<4 x s32>), align 4, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
...

---
name: test_load_flat_v8s32_align32
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v8s32_align32
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
    ; VI-LABEL: name: test_load_flat_v8s32_align32
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
    ; GFX9-LABEL: name: test_load_flat_v8s32_align32
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<8 x s32>) = G_LOAD %0 :: (load (<8 x s32>), align 32, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
...

---
name: test_load_flat_v16s32_align32
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v16s32_align32
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C1]](s64)
    ; CI-NEXT: [[LOAD2:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD1]](p0) :: (load (<4 x s32>) from unknown-address + 32, align 32)
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 48
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[LOAD3:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD2]](p0) :: (load (<4 x s32>) from unknown-address + 48)
    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[CONCAT_VECTORS]](<16 x s32>)
    ; VI-LABEL: name: test_load_flat_v16s32_align32
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C1]](s64)
    ; VI-NEXT: [[LOAD2:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD1]](p0) :: (load (<4 x s32>) from unknown-address + 32, align 32)
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 48
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[LOAD3:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD2]](p0) :: (load (<4 x s32>) from unknown-address + 48)
    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[CONCAT_VECTORS]](<16 x s32>)
    ; GFX9-LABEL: name: test_load_flat_v16s32_align32
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C1]](s64)
    ; GFX9-NEXT: [[LOAD2:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD1]](p0) :: (load (<4 x s32>) from unknown-address + 32, align 32)
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 48
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[LOAD3:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD2]](p0) :: (load (<4 x s32>) from unknown-address + 48)
    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[CONCAT_VECTORS]](<16 x s32>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<16 x s32>) = G_LOAD %0 :: (load (<16 x s32>), align 32, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY %1
...

---
name: test_load_flat_v2s64_align16
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2s64_align16
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>))
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
    ; VI-LABEL: name: test_load_flat_v2s64_align16
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>))
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
    ; GFX9-LABEL: name: test_load_flat_v2s64_align16
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>))
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 16, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
...

---
name: test_load_flat_v2s64_align8
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2s64_align8
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
    ; VI-LABEL: name: test_load_flat_v2s64_align8
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
    ; GFX9-LABEL: name: test_load_flat_v2s64_align8
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 8, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
...

---
name: test_load_flat_v2s64_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2s64_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 4)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
    ; VI-LABEL: name: test_load_flat_v2s64_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 4)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
    ; GFX9-LABEL: name: test_load_flat_v2s64_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 4)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 4, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
...

---
name: test_load_flat_v2s64_align2
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2s64_align2
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; CI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; CI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; CI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s16) from unknown-address + 8)
    ; CI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; CI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s16) from unknown-address + 10)
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD2]]
    ; CI-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR3]](s32)
    ; CI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s16) from unknown-address + 12)
    ; CI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; CI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s16) from unknown-address + 14)
    ; CI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; CI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD3]]
    ; CI-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR4]](s32)
    ; CI-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C3]](s32)
    ; CI-NEXT: [[SHL5:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
    ; CI-NEXT: [[OR5:%[0-9]+]]:_(s64) = G_OR [[SHL5]], [[ZEXT1]]
    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR2]](s64), [[OR5]](s64)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
    ; VI-LABEL: name: test_load_flat_v2s64_align2
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; VI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; VI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s16) from unknown-address + 8)
    ; VI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; VI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s16) from unknown-address + 10)
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD2]]
    ; VI-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR3]](s32)
    ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s16) from unknown-address + 12)
    ; VI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; VI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s16) from unknown-address + 14)
    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD3]]
    ; VI-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR4]](s32)
    ; VI-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C3]](s32)
    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
    ; VI-NEXT: [[OR5:%[0-9]+]]:_(s64) = G_OR [[SHL5]], [[ZEXT1]]
    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR2]](s64), [[OR5]](s64)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
    ; GFX9-LABEL: name: test_load_flat_v2s64_align2
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
    ; GFX9-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; GFX9-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s16) from unknown-address + 8)
    ; GFX9-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s16) from unknown-address + 10)
    ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; GFX9-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD2]]
    ; GFX9-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR3]](s32)
    ; GFX9-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s16) from unknown-address + 12)
    ; GFX9-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s16) from unknown-address + 14)
    ; GFX9-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; GFX9-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD3]]
    ; GFX9-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR4]](s32)
    ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C3]](s32)
    ; GFX9-NEXT: [[SHL5:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
    ; GFX9-NEXT: [[OR5:%[0-9]+]]:_(s64) = G_OR [[SHL5]], [[ZEXT1]]
    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR2]](s64), [[OR5]](s64)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 2, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
...

---
name: test_load_flat_v2s64_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2s64_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; CI-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; CI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; CI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; CI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; CI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; CI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; CI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; CI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; CI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
    ; CI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; CI-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
    ; CI-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
    ; CI-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; CI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C6]](s64)
    ; CI-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
    ; CI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
    ; CI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
    ; CI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
    ; CI-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
    ; CI-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
    ; CI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
    ; CI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; CI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
    ; CI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
    ; CI-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
    ; CI-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
    ; CI-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
    ; CI-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
    ; CI-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
    ; CI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
    ; CI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
    ; CI-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
    ; CI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
    ; CI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
    ; CI-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; CI-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
    ; CI-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
    ; CI-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
    ; CI-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
    ; CI-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
    ; CI-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
    ; CI-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
    ; VI-LABEL: name: test_load_flat_v2s64_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; VI-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; VI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; VI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; VI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; VI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
    ; VI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; VI-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
    ; VI-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
    ; VI-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; VI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C6]](s64)
    ; VI-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
    ; VI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
    ; VI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
    ; VI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
    ; VI-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
    ; VI-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
    ; VI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
    ; VI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; VI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
    ; VI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
    ; VI-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
    ; VI-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
    ; VI-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
    ; VI-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
    ; VI-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
    ; VI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
    ; VI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
    ; VI-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
    ; VI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
    ; VI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
    ; VI-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; VI-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
    ; VI-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
    ; VI-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
    ; VI-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
    ; VI-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
    ; VI-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
    ; VI-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
    ; GFX9-LABEL: name: test_load_flat_v2s64_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; GFX9-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
    ; GFX9-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; GFX9-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; GFX9-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; GFX9-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; GFX9-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; GFX9-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; GFX9-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; GFX9-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
    ; GFX9-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; GFX9-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
    ; GFX9-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
    ; GFX9-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; GFX9-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C6]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
    ; GFX9-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
    ; GFX9-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
    ; GFX9-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
    ; GFX9-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
    ; GFX9-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
    ; GFX9-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; GFX9-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
    ; GFX9-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
    ; GFX9-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
    ; GFX9-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
    ; GFX9-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
    ; GFX9-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
    ; GFX9-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
    ; GFX9-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
    ; GFX9-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
    ; GFX9-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
    ; GFX9-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; GFX9-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
    ; GFX9-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
    ; GFX9-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
    ; GFX9-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
    ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
    ; GFX9-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
    ; GFX9-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 1, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
...

---
name: test_load_flat_v3s64_align32
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v3s64_align32
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 32)
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16, align 16)
    ; CI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
    ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
    ; CI-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
    ; VI-LABEL: name: test_load_flat_v3s64_align32
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 32)
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16, align 16)
    ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
    ; VI-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
    ; GFX9-LABEL: name: test_load_flat_v3s64_align32
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 32)
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16, align 16)
    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
    ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 32, addrspace 0)
    %2:_(<4 x s64>) = G_IMPLICIT_DEF
    %3:_(<4 x s64>) = G_INSERT %2, %1, 0
    $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %3
...

---
name: test_load_flat_v3s64_align8
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v3s64_align8
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16)
    ; CI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
    ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
    ; CI-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
    ; VI-LABEL: name: test_load_flat_v3s64_align8
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16)
    ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
    ; VI-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
    ; GFX9-LABEL: name: test_load_flat_v3s64_align8
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16)
    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
    ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 8, addrspace 0)
    %2:_(<4 x s64>) = G_IMPLICIT_DEF
    %3:_(<4 x s64>) = G_INSERT %2, %1, 0
    $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %3
...

---
name: test_load_flat_v3s64_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v3s64_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; CI-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; CI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; CI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; CI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; CI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; CI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; CI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; CI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; CI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
    ; CI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; CI-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
    ; CI-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
    ; CI-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; CI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C6]](s64)
    ; CI-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
    ; CI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
    ; CI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
    ; CI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
    ; CI-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
    ; CI-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
    ; CI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
    ; CI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; CI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
    ; CI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
    ; CI-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
    ; CI-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
    ; CI-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
    ; CI-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
    ; CI-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
    ; CI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
    ; CI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
    ; CI-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
    ; CI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
    ; CI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
    ; CI-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; CI-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
    ; CI-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
    ; CI-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
    ; CI-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
    ; CI-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
    ; CI-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
    ; CI-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
    ; CI-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; CI-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C7]](s64)
    ; CI-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p0) :: (load (s8) from unknown-address + 16)
    ; CI-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p0) :: (load (s8) from unknown-address + 17)
    ; CI-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
    ; CI-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
    ; CI-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p0) :: (load (s8) from unknown-address + 18)
    ; CI-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
    ; CI-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p0) :: (load (s8) from unknown-address + 19)
    ; CI-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
    ; CI-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
    ; CI-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
    ; CI-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
    ; CI-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
    ; CI-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
    ; CI-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p0) :: (load (s8) from unknown-address + 20)
    ; CI-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p0) :: (load (s8) from unknown-address + 21)
    ; CI-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
    ; CI-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
    ; CI-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p0) :: (load (s8) from unknown-address + 22)
    ; CI-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
    ; CI-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p0) :: (load (s8) from unknown-address + 23)
    ; CI-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
    ; CI-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
    ; CI-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
    ; CI-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
    ; CI-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
    ; CI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
    ; CI-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
    ; CI-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
    ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
    ; CI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[UV3]](s64)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
    ; VI-LABEL: name: test_load_flat_v3s64_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; VI-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; VI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; VI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; VI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; VI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
    ; VI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; VI-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
    ; VI-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
    ; VI-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; VI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C6]](s64)
    ; VI-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
    ; VI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
    ; VI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
    ; VI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
    ; VI-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
    ; VI-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
    ; VI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
    ; VI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; VI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
    ; VI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
    ; VI-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
    ; VI-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
    ; VI-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
    ; VI-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
    ; VI-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
    ; VI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
    ; VI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
    ; VI-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
    ; VI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
    ; VI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
    ; VI-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; VI-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
    ; VI-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
    ; VI-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
    ; VI-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
    ; VI-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
    ; VI-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
    ; VI-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
    ; VI-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; VI-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C7]](s64)
    ; VI-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p0) :: (load (s8) from unknown-address + 16)
    ; VI-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p0) :: (load (s8) from unknown-address + 17)
    ; VI-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
    ; VI-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
    ; VI-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p0) :: (load (s8) from unknown-address + 18)
    ; VI-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
    ; VI-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p0) :: (load (s8) from unknown-address + 19)
    ; VI-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
    ; VI-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
    ; VI-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
    ; VI-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
    ; VI-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
    ; VI-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
    ; VI-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p0) :: (load (s8) from unknown-address + 20)
    ; VI-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p0) :: (load (s8) from unknown-address + 21)
    ; VI-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
    ; VI-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
    ; VI-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p0) :: (load (s8) from unknown-address + 22)
    ; VI-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
    ; VI-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p0) :: (load (s8) from unknown-address + 23)
    ; VI-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
    ; VI-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
    ; VI-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
    ; VI-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
    ; VI-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
    ; VI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
    ; VI-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
    ; VI-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
    ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[UV3]](s64)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
    ; GFX9-LABEL: name: test_load_flat_v3s64_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; GFX9-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
    ; GFX9-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; GFX9-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; GFX9-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; GFX9-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; GFX9-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; GFX9-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; GFX9-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; GFX9-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
    ; GFX9-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; GFX9-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
    ; GFX9-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
    ; GFX9-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; GFX9-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C6]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
    ; GFX9-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
    ; GFX9-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
    ; GFX9-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
    ; GFX9-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
    ; GFX9-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
    ; GFX9-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; GFX9-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
    ; GFX9-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
    ; GFX9-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
    ; GFX9-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
    ; GFX9-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
    ; GFX9-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
    ; GFX9-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
    ; GFX9-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
    ; GFX9-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
    ; GFX9-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
    ; GFX9-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; GFX9-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
    ; GFX9-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
    ; GFX9-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
    ; GFX9-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
    ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
    ; GFX9-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
    ; GFX9-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
    ; GFX9-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; GFX9-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C7]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p0) :: (load (s8) from unknown-address + 16)
    ; GFX9-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p0) :: (load (s8) from unknown-address + 17)
    ; GFX9-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
    ; GFX9-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
    ; GFX9-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p0) :: (load (s8) from unknown-address + 18)
    ; GFX9-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p0) :: (load (s8) from unknown-address + 19)
    ; GFX9-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
    ; GFX9-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
    ; GFX9-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
    ; GFX9-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
    ; GFX9-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
    ; GFX9-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p0) :: (load (s8) from unknown-address + 20)
    ; GFX9-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p0) :: (load (s8) from unknown-address + 21)
    ; GFX9-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
    ; GFX9-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
    ; GFX9-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p0) :: (load (s8) from unknown-address + 22)
    ; GFX9-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p0) :: (load (s8) from unknown-address + 23)
    ; GFX9-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
    ; GFX9-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
    ; GFX9-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
    ; GFX9-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
    ; GFX9-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
    ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
    ; GFX9-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
    ; GFX9-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[UV3]](s64)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 1, addrspace 0)
    %2:_(<4 x s64>) = G_IMPLICIT_DEF
    %3:_(<4 x s64>) = G_INSERT %2, %1, 0
    $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %3
...

---
name: test_load_flat_v4s64_align32
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v4s64_align32
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 32)
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p0) :: (load (<2 x s64>) from unknown-address + 16)
    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
    ; VI-LABEL: name: test_load_flat_v4s64_align32
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 32)
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p0) :: (load (<2 x s64>) from unknown-address + 16)
    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
    ; GFX9-LABEL: name: test_load_flat_v4s64_align32
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 32)
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p0) :: (load (<2 x s64>) from unknown-address + 16)
    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<4 x s64>) = G_LOAD %0 :: (load (<4 x s64>), align 32, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
...

---
name: test_load_flat_v4s64_align8
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v4s64_align8
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p0) :: (load (<2 x s64>) from unknown-address + 16, align 8)
    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
    ; VI-LABEL: name: test_load_flat_v4s64_align8
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p0) :: (load (<2 x s64>) from unknown-address + 16, align 8)
    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
    ; GFX9-LABEL: name: test_load_flat_v4s64_align8
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p0) :: (load (<2 x s64>) from unknown-address + 16, align 8)
    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<4 x s64>) = G_LOAD %0 :: (load (<4 x s64>), align 8, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
...

---
name: test_load_flat_v4s64_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v4s64_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; CI-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; CI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; CI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; CI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; CI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; CI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; CI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; CI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; CI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
    ; CI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; CI-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
    ; CI-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
    ; CI-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; CI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C6]](s64)
    ; CI-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
    ; CI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
    ; CI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
    ; CI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
    ; CI-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
    ; CI-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
    ; CI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
    ; CI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; CI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
    ; CI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
    ; CI-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
    ; CI-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
    ; CI-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
    ; CI-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
    ; CI-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
    ; CI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
    ; CI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
    ; CI-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
    ; CI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
    ; CI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
    ; CI-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; CI-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
    ; CI-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
    ; CI-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
    ; CI-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
    ; CI-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
    ; CI-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
    ; CI-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
    ; CI-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; CI-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C7]](s64)
    ; CI-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p0) :: (load (s8) from unknown-address + 16)
    ; CI-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p0) :: (load (s8) from unknown-address + 17)
    ; CI-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
    ; CI-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
    ; CI-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p0) :: (load (s8) from unknown-address + 18)
    ; CI-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
    ; CI-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p0) :: (load (s8) from unknown-address + 19)
    ; CI-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
    ; CI-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
    ; CI-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
    ; CI-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
    ; CI-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
    ; CI-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
    ; CI-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p0) :: (load (s8) from unknown-address + 20)
    ; CI-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p0) :: (load (s8) from unknown-address + 21)
    ; CI-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
    ; CI-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
    ; CI-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p0) :: (load (s8) from unknown-address + 22)
    ; CI-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
    ; CI-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p0) :: (load (s8) from unknown-address + 23)
    ; CI-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
    ; CI-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
    ; CI-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
    ; CI-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
    ; CI-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
    ; CI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
    ; CI-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
    ; CI-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
    ; CI-NEXT: [[PTR_ADD23:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C6]](s64)
    ; CI-NEXT: [[ZEXTLOAD18:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD23]](p0) :: (load (s8) from unknown-address + 24)
    ; CI-NEXT: [[PTR_ADD24:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD23]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD19:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD24]](p0) :: (load (s8) from unknown-address + 25)
    ; CI-NEXT: [[SHL21:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD19]], [[C1]](s32)
    ; CI-NEXT: [[OR21:%[0-9]+]]:_(s32) = G_OR [[SHL21]], [[ZEXTLOAD18]]
    ; CI-NEXT: [[PTR_ADD25:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD23]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD20:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD25]](p0) :: (load (s8) from unknown-address + 26)
    ; CI-NEXT: [[PTR_ADD26:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD25]], [[C]](s64)
    ; CI-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD26]](p0) :: (load (s8) from unknown-address + 27)
    ; CI-NEXT: [[SHL22:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
    ; CI-NEXT: [[OR22:%[0-9]+]]:_(s32) = G_OR [[SHL22]], [[ZEXTLOAD20]]
    ; CI-NEXT: [[SHL23:%[0-9]+]]:_(s32) = G_SHL [[OR22]], [[C3]](s32)
    ; CI-NEXT: [[OR23:%[0-9]+]]:_(s32) = G_OR [[SHL23]], [[OR21]]
    ; CI-NEXT: [[ZEXT3:%[0-9]+]]:_(s64) = G_ZEXT [[OR23]](s32)
    ; CI-NEXT: [[PTR_ADD27:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD23]], [[C4]](s64)
    ; CI-NEXT: [[ZEXTLOAD21:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD27]](p0) :: (load (s8) from unknown-address + 28)
    ; CI-NEXT: [[PTR_ADD28:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD27]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD22:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD28]](p0) :: (load (s8) from unknown-address + 29)
    ; CI-NEXT: [[SHL24:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD22]], [[C1]](s32)
    ; CI-NEXT: [[OR24:%[0-9]+]]:_(s32) = G_OR [[SHL24]], [[ZEXTLOAD21]]
    ; CI-NEXT: [[PTR_ADD29:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD27]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD23:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD29]](p0) :: (load (s8) from unknown-address + 30)
    ; CI-NEXT: [[PTR_ADD30:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD29]], [[C]](s64)
    ; CI-NEXT: [[LOAD7:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD30]](p0) :: (load (s8) from unknown-address + 31)
    ; CI-NEXT: [[SHL25:%[0-9]+]]:_(s32) = G_SHL [[LOAD7]], [[C1]](s32)
    ; CI-NEXT: [[OR25:%[0-9]+]]:_(s32) = G_OR [[SHL25]], [[ZEXTLOAD23]]
    ; CI-NEXT: [[SHL26:%[0-9]+]]:_(s32) = G_SHL [[OR25]], [[C3]](s32)
    ; CI-NEXT: [[OR26:%[0-9]+]]:_(s32) = G_OR [[SHL26]], [[OR24]]
    ; CI-NEXT: [[ANYEXT3:%[0-9]+]]:_(s64) = G_ANYEXT [[OR26]](s32)
    ; CI-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
    ; CI-NEXT: [[SHL27:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT3]], [[COPY3]](s32)
    ; CI-NEXT: [[OR27:%[0-9]+]]:_(s64) = G_OR [[SHL27]], [[ZEXT3]]
    ; CI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR20]](s64), [[OR27]](s64)
    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s64>), [[BUILD_VECTOR1]](<2 x s64>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
    ; VI-LABEL: name: test_load_flat_v4s64_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; VI-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; VI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; VI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; VI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; VI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
    ; VI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; VI-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
    ; VI-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
    ; VI-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; VI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C6]](s64)
    ; VI-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
    ; VI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
    ; VI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
    ; VI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
    ; VI-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
    ; VI-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
    ; VI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
    ; VI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; VI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
    ; VI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
    ; VI-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
    ; VI-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
    ; VI-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
    ; VI-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
    ; VI-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
    ; VI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
    ; VI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
    ; VI-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
    ; VI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
    ; VI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
    ; VI-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; VI-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
    ; VI-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
    ; VI-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
    ; VI-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
    ; VI-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
    ; VI-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
    ; VI-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
    ; VI-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; VI-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C7]](s64)
    ; VI-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p0) :: (load (s8) from unknown-address + 16)
    ; VI-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p0) :: (load (s8) from unknown-address + 17)
    ; VI-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
    ; VI-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
    ; VI-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p0) :: (load (s8) from unknown-address + 18)
    ; VI-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
    ; VI-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p0) :: (load (s8) from unknown-address + 19)
    ; VI-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
    ; VI-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
    ; VI-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
    ; VI-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
    ; VI-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
    ; VI-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
    ; VI-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p0) :: (load (s8) from unknown-address + 20)
    ; VI-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p0) :: (load (s8) from unknown-address + 21)
    ; VI-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
    ; VI-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
    ; VI-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p0) :: (load (s8) from unknown-address + 22)
    ; VI-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
    ; VI-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p0) :: (load (s8) from unknown-address + 23)
    ; VI-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
    ; VI-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
    ; VI-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
    ; VI-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
    ; VI-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
    ; VI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
    ; VI-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
    ; VI-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
    ; VI-NEXT: [[PTR_ADD23:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C6]](s64)
    ; VI-NEXT: [[ZEXTLOAD18:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD23]](p0) :: (load (s8) from unknown-address + 24)
    ; VI-NEXT: [[PTR_ADD24:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD23]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD19:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD24]](p0) :: (load (s8) from unknown-address + 25)
    ; VI-NEXT: [[SHL21:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD19]], [[C1]](s32)
    ; VI-NEXT: [[OR21:%[0-9]+]]:_(s32) = G_OR [[SHL21]], [[ZEXTLOAD18]]
    ; VI-NEXT: [[PTR_ADD25:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD23]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD20:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD25]](p0) :: (load (s8) from unknown-address + 26)
    ; VI-NEXT: [[PTR_ADD26:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD25]], [[C]](s64)
    ; VI-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD26]](p0) :: (load (s8) from unknown-address + 27)
    ; VI-NEXT: [[SHL22:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
    ; VI-NEXT: [[OR22:%[0-9]+]]:_(s32) = G_OR [[SHL22]], [[ZEXTLOAD20]]
    ; VI-NEXT: [[SHL23:%[0-9]+]]:_(s32) = G_SHL [[OR22]], [[C3]](s32)
    ; VI-NEXT: [[OR23:%[0-9]+]]:_(s32) = G_OR [[SHL23]], [[OR21]]
    ; VI-NEXT: [[ZEXT3:%[0-9]+]]:_(s64) = G_ZEXT [[OR23]](s32)
    ; VI-NEXT: [[PTR_ADD27:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD23]], [[C4]](s64)
    ; VI-NEXT: [[ZEXTLOAD21:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD27]](p0) :: (load (s8) from unknown-address + 28)
    ; VI-NEXT: [[PTR_ADD28:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD27]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD22:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD28]](p0) :: (load (s8) from unknown-address + 29)
    ; VI-NEXT: [[SHL24:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD22]], [[C1]](s32)
    ; VI-NEXT: [[OR24:%[0-9]+]]:_(s32) = G_OR [[SHL24]], [[ZEXTLOAD21]]
    ; VI-NEXT: [[PTR_ADD29:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD27]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD23:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD29]](p0) :: (load (s8) from unknown-address + 30)
    ; VI-NEXT: [[PTR_ADD30:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD29]], [[C]](s64)
    ; VI-NEXT: [[LOAD7:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD30]](p0) :: (load (s8) from unknown-address + 31)
    ; VI-NEXT: [[SHL25:%[0-9]+]]:_(s32) = G_SHL [[LOAD7]], [[C1]](s32)
    ; VI-NEXT: [[OR25:%[0-9]+]]:_(s32) = G_OR [[SHL25]], [[ZEXTLOAD23]]
    ; VI-NEXT: [[SHL26:%[0-9]+]]:_(s32) = G_SHL [[OR25]], [[C3]](s32)
    ; VI-NEXT: [[OR26:%[0-9]+]]:_(s32) = G_OR [[SHL26]], [[OR24]]
    ; VI-NEXT: [[ANYEXT3:%[0-9]+]]:_(s64) = G_ANYEXT [[OR26]](s32)
    ; VI-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
    ; VI-NEXT: [[SHL27:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT3]], [[COPY3]](s32)
    ; VI-NEXT: [[OR27:%[0-9]+]]:_(s64) = G_OR [[SHL27]], [[ZEXT3]]
    ; VI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR20]](s64), [[OR27]](s64)
    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s64>), [[BUILD_VECTOR1]](<2 x s64>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
    ; GFX9-LABEL: name: test_load_flat_v4s64_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; GFX9-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
    ; GFX9-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; GFX9-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; GFX9-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; GFX9-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; GFX9-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; GFX9-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; GFX9-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; GFX9-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
    ; GFX9-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
    ; GFX9-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
    ; GFX9-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
    ; GFX9-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; GFX9-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C6]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
    ; GFX9-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
    ; GFX9-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
    ; GFX9-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
    ; GFX9-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
    ; GFX9-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
    ; GFX9-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; GFX9-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
    ; GFX9-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
    ; GFX9-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
    ; GFX9-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
    ; GFX9-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
    ; GFX9-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
    ; GFX9-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
    ; GFX9-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
    ; GFX9-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
    ; GFX9-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
    ; GFX9-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; GFX9-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
    ; GFX9-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
    ; GFX9-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
    ; GFX9-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
    ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
    ; GFX9-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
    ; GFX9-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
    ; GFX9-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; GFX9-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C7]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p0) :: (load (s8) from unknown-address + 16)
    ; GFX9-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p0) :: (load (s8) from unknown-address + 17)
    ; GFX9-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
    ; GFX9-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
    ; GFX9-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p0) :: (load (s8) from unknown-address + 18)
    ; GFX9-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p0) :: (load (s8) from unknown-address + 19)
    ; GFX9-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
    ; GFX9-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
    ; GFX9-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
    ; GFX9-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
    ; GFX9-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
    ; GFX9-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p0) :: (load (s8) from unknown-address + 20)
    ; GFX9-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p0) :: (load (s8) from unknown-address + 21)
    ; GFX9-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
    ; GFX9-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
    ; GFX9-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p0) :: (load (s8) from unknown-address + 22)
    ; GFX9-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p0) :: (load (s8) from unknown-address + 23)
    ; GFX9-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
    ; GFX9-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
    ; GFX9-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
    ; GFX9-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
    ; GFX9-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
    ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
    ; GFX9-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
    ; GFX9-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
    ; GFX9-NEXT: [[PTR_ADD23:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD15]], [[C6]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD18:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD23]](p0) :: (load (s8) from unknown-address + 24)
    ; GFX9-NEXT: [[PTR_ADD24:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD23]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD19:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD24]](p0) :: (load (s8) from unknown-address + 25)
    ; GFX9-NEXT: [[SHL21:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD19]], [[C1]](s32)
    ; GFX9-NEXT: [[OR21:%[0-9]+]]:_(s32) = G_OR [[SHL21]], [[ZEXTLOAD18]]
    ; GFX9-NEXT: [[PTR_ADD25:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD23]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD20:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD25]](p0) :: (load (s8) from unknown-address + 26)
    ; GFX9-NEXT: [[PTR_ADD26:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD25]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD26]](p0) :: (load (s8) from unknown-address + 27)
    ; GFX9-NEXT: [[SHL22:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
    ; GFX9-NEXT: [[OR22:%[0-9]+]]:_(s32) = G_OR [[SHL22]], [[ZEXTLOAD20]]
    ; GFX9-NEXT: [[SHL23:%[0-9]+]]:_(s32) = G_SHL [[OR22]], [[C3]](s32)
    ; GFX9-NEXT: [[OR23:%[0-9]+]]:_(s32) = G_OR [[SHL23]], [[OR21]]
    ; GFX9-NEXT: [[ZEXT3:%[0-9]+]]:_(s64) = G_ZEXT [[OR23]](s32)
    ; GFX9-NEXT: [[PTR_ADD27:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD23]], [[C4]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD21:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD27]](p0) :: (load (s8) from unknown-address + 28)
    ; GFX9-NEXT: [[PTR_ADD28:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD27]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD22:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD28]](p0) :: (load (s8) from unknown-address + 29)
    ; GFX9-NEXT: [[SHL24:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD22]], [[C1]](s32)
    ; GFX9-NEXT: [[OR24:%[0-9]+]]:_(s32) = G_OR [[SHL24]], [[ZEXTLOAD21]]
    ; GFX9-NEXT: [[PTR_ADD29:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD27]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD23:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD29]](p0) :: (load (s8) from unknown-address + 30)
    ; GFX9-NEXT: [[PTR_ADD30:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD29]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD7:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD30]](p0) :: (load (s8) from unknown-address + 31)
    ; GFX9-NEXT: [[SHL25:%[0-9]+]]:_(s32) = G_SHL [[LOAD7]], [[C1]](s32)
    ; GFX9-NEXT: [[OR25:%[0-9]+]]:_(s32) = G_OR [[SHL25]], [[ZEXTLOAD23]]
    ; GFX9-NEXT: [[SHL26:%[0-9]+]]:_(s32) = G_SHL [[OR25]], [[C3]](s32)
    ; GFX9-NEXT: [[OR26:%[0-9]+]]:_(s32) = G_OR [[SHL26]], [[OR24]]
    ; GFX9-NEXT: [[ANYEXT3:%[0-9]+]]:_(s64) = G_ANYEXT [[OR26]](s32)
    ; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
    ; GFX9-NEXT: [[SHL27:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT3]], [[COPY3]](s32)
    ; GFX9-NEXT: [[OR27:%[0-9]+]]:_(s64) = G_OR [[SHL27]], [[ZEXT3]]
    ; GFX9-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR20]](s64), [[OR27]](s64)
    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s64>), [[BUILD_VECTOR1]](<2 x s64>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<4 x s64>) = G_LOAD %0 :: (load (<4 x s64>), align 1, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
...

---
name: test_load_flat_v2s128_align32
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2s128_align32
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
    ; VI-LABEL: name: test_load_flat_v2s128_align32
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
    ; GFX9-LABEL: name: test_load_flat_v2s128_align32
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x s128>) = G_LOAD %0 :: (load (<2 x s128>), align 32, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
...

---
name: test_load_flat_v2p1_align16
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2p1_align16
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
    ; VI-LABEL: name: test_load_flat_v2p1_align16
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
    ; GFX9-LABEL: name: test_load_flat_v2p1_align16
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 16, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
...

---
name: test_load_flat_v2p1_align8
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2p1_align8
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
    ; VI-LABEL: name: test_load_flat_v2p1_align8
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
    ; GFX9-LABEL: name: test_load_flat_v2p1_align8
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 8, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
...

---
name: test_load_flat_v2p1_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2p1_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
    ; VI-LABEL: name: test_load_flat_v2p1_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
    ; GFX9-LABEL: name: test_load_flat_v2p1_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 4, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
...

---
name: test_load_flat_v2p1_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2p1_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; CI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; CI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; CI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; CI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; CI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; CI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; CI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; CI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; CI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C5]](s64)
    ; CI-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
    ; CI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
    ; CI-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
    ; CI-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
    ; CI-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
    ; CI-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
    ; CI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
    ; CI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; CI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
    ; CI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
    ; CI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
    ; CI-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
    ; CI-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C6]](s64)
    ; CI-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
    ; CI-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
    ; CI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
    ; CI-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
    ; CI-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
    ; CI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
    ; CI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
    ; CI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; CI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
    ; CI-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
    ; CI-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
    ; VI-LABEL: name: test_load_flat_v2p1_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; VI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; VI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; VI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; VI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; VI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; VI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C5]](s64)
    ; VI-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
    ; VI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
    ; VI-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
    ; VI-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
    ; VI-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
    ; VI-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
    ; VI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
    ; VI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; VI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
    ; VI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
    ; VI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
    ; VI-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
    ; VI-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C6]](s64)
    ; VI-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
    ; VI-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
    ; VI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
    ; VI-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
    ; VI-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
    ; VI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
    ; VI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
    ; VI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; VI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
    ; VI-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
    ; VI-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
    ; GFX9-LABEL: name: test_load_flat_v2p1_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; GFX9-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; GFX9-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; GFX9-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; GFX9-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; GFX9-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; GFX9-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; GFX9-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; GFX9-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; GFX9-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; GFX9-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C5]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
    ; GFX9-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
    ; GFX9-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
    ; GFX9-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
    ; GFX9-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
    ; GFX9-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
    ; GFX9-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
    ; GFX9-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
    ; GFX9-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
    ; GFX9-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
    ; GFX9-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
    ; GFX9-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C6]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
    ; GFX9-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
    ; GFX9-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
    ; GFX9-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
    ; GFX9-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
    ; GFX9-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
    ; GFX9-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
    ; GFX9-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
    ; GFX9-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
    ; GFX9-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 1, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
...

---
name: test_load_flat_v2p3_align8
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2p3_align8
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x p3>) = G_LOAD [[COPY]](p0) :: (load (<2 x p3>))
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>)
    ; VI-LABEL: name: test_load_flat_v2p3_align8
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x p3>) = G_LOAD [[COPY]](p0) :: (load (<2 x p3>))
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>)
    ; GFX9-LABEL: name: test_load_flat_v2p3_align8
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x p3>) = G_LOAD [[COPY]](p0) :: (load (<2 x p3>))
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x p3>) = G_LOAD %0 :: (load (<2 x p3>), align 8, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_v2p3_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2p3_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x p3>) = G_LOAD [[COPY]](p0) :: (load (<2 x p3>), align 4)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>)
    ; VI-LABEL: name: test_load_flat_v2p3_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x p3>) = G_LOAD [[COPY]](p0) :: (load (<2 x p3>), align 4)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>)
    ; GFX9-LABEL: name: test_load_flat_v2p3_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x p3>) = G_LOAD [[COPY]](p0) :: (load (<2 x p3>), align 4)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x p3>) = G_LOAD %0 :: (load (<2 x p3>), align 4, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_load_flat_v2p3_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_v2p3_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; CI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; CI-NEXT: [[INTTOPTR:%[0-9]+]]:_(p3) = G_INTTOPTR [[OR2]](s32)
    ; CI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; CI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; CI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; CI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; CI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; CI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; CI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; CI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; CI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; CI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; CI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; CI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; CI-NEXT: [[INTTOPTR1:%[0-9]+]]:_(p3) = G_INTTOPTR [[OR5]](s32)
    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x p3>) = G_BUILD_VECTOR [[INTTOPTR]](p3), [[INTTOPTR1]](p3)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x p3>)
    ; VI-LABEL: name: test_load_flat_v2p3_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; VI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; VI-NEXT: [[INTTOPTR:%[0-9]+]]:_(p3) = G_INTTOPTR [[OR2]](s32)
    ; VI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; VI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; VI-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; VI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; VI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; VI-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; VI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; VI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; VI-NEXT: [[INTTOPTR1:%[0-9]+]]:_(p3) = G_INTTOPTR [[OR5]](s32)
    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x p3>) = G_BUILD_VECTOR [[INTTOPTR]](p3), [[INTTOPTR1]](p3)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x p3>)
    ; GFX9-LABEL: name: test_load_flat_v2p3_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
    ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
    ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
    ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
    ; GFX9-NEXT: [[INTTOPTR:%[0-9]+]]:_(p3) = G_INTTOPTR [[OR2]](s32)
    ; GFX9-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
    ; GFX9-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C4]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
    ; GFX9-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
    ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
    ; GFX9-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
    ; GFX9-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
    ; GFX9-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
    ; GFX9-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
    ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
    ; GFX9-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
    ; GFX9-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
    ; GFX9-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
    ; GFX9-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
    ; GFX9-NEXT: [[INTTOPTR1:%[0-9]+]]:_(p3) = G_INTTOPTR [[OR5]](s32)
    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x p3>) = G_BUILD_VECTOR [[INTTOPTR]](p3), [[INTTOPTR1]](p3)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x p3>)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(<2 x p3>) = G_LOAD %0 :: (load (<2 x p3>), align 1, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_ext_load_flat_s32_from_1_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_ext_load_flat_s32_from_1_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
    ; CI-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    ; VI-LABEL: name: test_ext_load_flat_s32_from_1_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
    ; VI-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    ; GFX9-LABEL: name: test_ext_load_flat_s32_from_1_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
    ; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s32) = G_LOAD %0 :: (load (s8), align 4, addrspace 0)
    $vgpr0 = COPY %1
...

---
name: test_ext_load_flat_s32_from_2_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_ext_load_flat_s32_from_2_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
    ; CI-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    ; VI-LABEL: name: test_ext_load_flat_s32_from_2_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
    ; VI-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    ; GFX9-LABEL: name: test_ext_load_flat_s32_from_2_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
    ; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s32) = G_LOAD %0 :: (load (s16), align 4, addrspace 0)
    $vgpr0 = COPY %1
...

---
name: test_ext_load_flat_s64_from_1_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1


    ; CI-LABEL: name: test_ext_load_flat_s64_from_1_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
    ; CI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
    ; VI-LABEL: name: test_ext_load_flat_s64_from_1_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
    ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
    ; GFX9-LABEL: name: test_ext_load_flat_s64_from_1_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s64) = G_LOAD %0 :: (load (s8), align 4, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_ext_load_flat_s64_from_2_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_ext_load_flat_s64_from_2_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
    ; CI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
    ; VI-LABEL: name: test_ext_load_flat_s64_from_2_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
    ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
    ; GFX9-LABEL: name: test_ext_load_flat_s64_from_2_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s64) = G_LOAD %0 :: (load (s16), align 4, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_ext_load_flat_s64_from_4_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_ext_load_flat_s64_from_4_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
    ; CI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
    ; VI-LABEL: name: test_ext_load_flat_s64_from_4_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
    ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
    ; GFX9-LABEL: name: test_ext_load_flat_s64_from_4_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s64) = G_LOAD %0 :: (load (s32), align 4, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_ext_load_flat_s128_from_4_align4
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_ext_load_flat_s128_from_4_align4
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
    ; CI-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
    ; CI-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
    ; CI-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
    ; CI-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
    ; VI-LABEL: name: test_ext_load_flat_s128_from_4_align4
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
    ; VI-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
    ; VI-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
    ; VI-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
    ; VI-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
    ; GFX9-LABEL: name: test_ext_load_flat_s128_from_4_align4
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
    ; GFX9-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
    ; GFX9-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s128) = G_LOAD %0 :: (load (s32), align 4, addrspace 0)
    $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
...

---
name: test_ext_load_flat_s64_from_2_align2
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_ext_load_flat_s64_from_2_align2
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
    ; CI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
    ; VI-LABEL: name: test_ext_load_flat_s64_from_2_align2
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
    ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
    ; GFX9-LABEL: name: test_ext_load_flat_s64_from_2_align2
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s64) = G_LOAD %0 :: (load (s16), align 4, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

---
name: test_ext_load_flat_s64_from_1_align1
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_ext_load_flat_s64_from_1_align1
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
    ; CI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
    ; CI-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
    ; VI-LABEL: name: test_ext_load_flat_s64_from_1_align1
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
    ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
    ; GFX9-LABEL: name: test_ext_load_flat_s64_from_1_align1
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s64) = G_LOAD %0 :: (load (s8), align 4, addrspace 0)
    $vgpr0_vgpr1 = COPY %1
...

# Make sure there's no crash on very high alignments
---
name: test_load_flat_s32_align536870912
body: |
  bb.0:
    liveins: $vgpr0_vgpr1

    ; CI-LABEL: name: test_load_flat_s32_align536870912
    ; CI: liveins: $vgpr0_vgpr1
    ; CI-NEXT: {{  $}}
    ; CI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 536870912)
    ; CI-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    ; VI-LABEL: name: test_load_flat_s32_align536870912
    ; VI: liveins: $vgpr0_vgpr1
    ; VI-NEXT: {{  $}}
    ; VI-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 536870912)
    ; VI-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    ; GFX9-LABEL: name: test_load_flat_s32_align536870912
    ; GFX9: liveins: $vgpr0_vgpr1
    ; GFX9-NEXT: {{  $}}
    ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
    ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 536870912)
    ; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
    %0:_(p0) = COPY $vgpr0_vgpr1
    %1:_(s32) = G_LOAD %0 :: (load (s16), align 536870912)
    $vgpr0 = COPY %1

...