# RUN: llc -mtriple=x86_64-linux-gnu                                  -global-isel -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=ALL --check-prefix=NO_AVX512VL --check-prefix=NO_AVX512F --check-prefix=SSE
# RUN: llc -mtriple=x86_64-linux-gnu -mattr=+avx                      -global-isel -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=ALL --check-prefix=NO_AVX512VL --check-prefix=NO_AVX512F --check-prefix=AVX
# RUN: llc -mtriple=x86_64-linux-gnu -mattr=+avx512f                  -global-isel -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=ALL --check-prefix=NO_AVX512VL --check-prefix=AVX512ALL --check-prefix=AVX512F
# RUN: llc -mtriple=x86_64-linux-gnu -mattr=+avx512f -mattr=+avx512vl -global-isel -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=ALL --check-prefix=AVX512ALL --check-prefix=AVX512VL

--- |
  define <4 x i32> @test_load_v4i32_noalign(<4 x i32>* %p1) {
    %r = load <4 x i32>, <4 x i32>* %p1, align 1
    ret <4 x i32> %r
  }

  define <4 x i32> @test_load_v4i32_align(<4 x i32>* %p1) {
    %r = load <4 x i32>, <4 x i32>* %p1, align 16
    ret <4 x i32> %r
  }

  define <4 x i32>* @test_store_v4i32_align(<4 x i32> %val, <4 x i32>* %p1) {
    store <4 x i32> %val, <4 x i32>* %p1, align 16
    ret <4 x i32>* %p1
  }

  define <4 x i32>* @test_store_v4i32_noalign(<4 x i32> %val, <4 x i32>* %p1) {
    store <4 x i32> %val, <4 x i32>* %p1, align 1
    ret <4 x i32>* %p1
  }

...
---
# ALL-LABEL: name:            test_load_v4i32_noalign
name:            test_load_v4i32_noalign
alignment:       4
legalized:       true
regBankSelected: true
registers:
# ALL:          - { id: 0, class: gr64 }
# NO_AVX512F:   - { id: 1, class: vr128 }
# AVX512ALL:    - { id: 1, class: vr128x }
  - { id: 0, class: gpr }
  - { id: 1, class: vecr }
# ALL:      %0 = COPY %rdi
# SSE:      %1 = MOVUPSrm %0, 1, _, 0, _ :: (load 16 from %ir.p1, align 1)
# AVX:      %1 = VMOVUPSrm %0, 1, _, 0, _ :: (load 16 from %ir.p1, align 1)
# AVX512F:  %1 = VMOVUPSZ128rm_NOVLX %0, 1, _, 0, _ :: (load 16 from %ir.p1, align 1)
# AVX512VL: %1 = VMOVUPSZ128rm %0, 1, _, 0, _ :: (load 16 from %ir.p1, align 1)
# ALL: %xmm0 = COPY %1
body:             |
  bb.1 (%ir-block.0):
    liveins: %rdi

    %0(p0) = COPY %rdi
    %1(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.p1, align 1)
    %xmm0 = COPY %1(<4 x s32>)
    RET 0, implicit %xmm0

...
---
# ALL-LABEL: name:            test_load_v4i32_align
name:            test_load_v4i32_align
alignment:       4
legalized:       true
regBankSelected: true
registers:
# ALL:   - { id: 0, class: gr64 }
# NO_AVX512F:   - { id: 1, class: vr128 }
# AVX512ALL:    - { id: 1, class: vr128x }
  - { id: 0, class: gpr }
  - { id: 1, class: vecr }
# ALL:      %0 = COPY %rdi
# SSE:      %1 = MOVAPSrm %0, 1, _, 0, _ :: (load 16 from %ir.p1)
# AVX:      %1 = VMOVAPSrm %0, 1, _, 0, _ :: (load 16 from %ir.p1)
# AVX512F:  %1 = VMOVAPSZ128rm_NOVLX %0, 1, _, 0, _ :: (load 16 from %ir.p1)
# AVX512VL: %1 = VMOVAPSZ128rm %0, 1, _, 0, _ :: (load 16 from %ir.p1)
# ALL: %xmm0 = COPY %1
body:             |
  bb.1 (%ir-block.0):
    liveins: %rdi

    %0(p0) = COPY %rdi
    %1(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.p1)
    %xmm0 = COPY %1(<4 x s32>)
    RET 0, implicit %xmm0

...
---
# ALL-LABEL: name:            test_store_v4i32_align
name:            test_store_v4i32_align
alignment:       4
legalized:       true
regBankSelected: true
registers:
# NO_AVX512F:   - { id: 0, class: vr128 }
# AVX512ALL:    - { id: 0, class: vr128x }
# ALL:   - { id: 1, class: gr64 }
  - { id: 0, class: vecr }
  - { id: 1, class: gpr }
# ALL:       %0 = COPY %xmm0
# ALL:       %1 = COPY %rdi
# SSE:       MOVAPSmr %1, 1, _, 0, _, %0 :: (store 16 into %ir.p1)
# AVX:       VMOVAPSmr %1, 1, _, 0, _, %0 :: (store 16 into %ir.p1)
# AVX512F:   VMOVAPSZ128mr_NOVLX %1, 1, _, 0, _, %0 :: (store 16 into %ir.p1)
# AVX512VL:  VMOVAPSZ128mr %1, 1, _, 0, _, %0 :: (store 16 into %ir.p1)
# ALL:       %rax = COPY %1
body:             |
  bb.1 (%ir-block.0):
    liveins: %rdi, %xmm0

    %0(<4 x s32>) = COPY %xmm0
    %1(p0) = COPY %rdi
    G_STORE %0(<4 x s32>), %1(p0) :: (store 16 into %ir.p1, align 16)
    %rax = COPY %1(p0)
    RET 0, implicit %rax

...
---
# ALL-LABEL: name:            test_store_v4i32_noalign
name:            test_store_v4i32_noalign
alignment:       4
legalized:       true
regBankSelected: true
registers:
# NO_AVX512F:   - { id: 0, class: vr128 }
# AVX512ALL:    - { id: 0, class: vr128x }
# ALL:   - { id: 1, class: gr64 }
  - { id: 0, class: vecr }
  - { id: 1, class: gpr }
# ALL:       %0 = COPY %xmm0
# ALL:       %1 = COPY %rdi
# SSE:       MOVUPSmr %1, 1, _, 0, _, %0 :: (store 16 into %ir.p1, align 1)
# AVX:       VMOVUPSmr %1, 1, _, 0, _, %0 :: (store 16 into %ir.p1, align 1)
# AVX512F:   VMOVUPSZ128mr_NOVLX %1, 1, _, 0, _, %0 :: (store 16 into %ir.p1, align 1)
# AVX512VL:  VMOVUPSZ128mr %1, 1, _, 0, _, %0 :: (store 16 into %ir.p1, align 1)
# ALL:       %rax = COPY %1
body:             |
  bb.1 (%ir-block.0):
    liveins: %rdi, %xmm0

    %0(<4 x s32>) = COPY %xmm0
    %1(p0) = COPY %rdi
    G_STORE %0(<4 x s32>), %1(p0) :: (store 16 into %ir.p1, align 1)
    %rax = COPY %1(p0)
    RET 0, implicit %rax

...