# RUN: llc -mtriple=x86_64-linux-gnu -mattr=+sse2                        -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=ALL --check-prefix=NOVL --check-prefix=SSE2
# RUN: llc -mtriple=x86_64-linux-gnu -mattr=+avx                         -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=ALL --check-prefix=NOVL --check-prefix=AVX1
# RUN: llc -mtriple=x86_64-linux-gnu -mattr=+avx512f,+avx512vl           -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=ALL --check-prefix=AVX512VL
# RUN: llc -mtriple=x86_64-linux-gnu -mattr=+avx512f,+avx512vl,+avx512bw -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=ALL --check-prefix=AVX512BWVL

--- |
  define <16 x i8> @test_sub_v16i8(<16 x i8> %arg1, <16 x i8> %arg2) {
    %ret = sub <16 x i8> %arg1, %arg2
    ret <16 x i8> %ret
  }

  define <8 x i16> @test_sub_v8i16(<8 x i16> %arg1, <8 x i16> %arg2) {
    %ret = sub <8 x i16> %arg1, %arg2
    ret <8 x i16> %ret
  }

  define <4 x i32> @test_sub_v4i32(<4 x i32> %arg1, <4 x i32> %arg2) {
    %ret = sub <4 x i32> %arg1, %arg2
    ret <4 x i32> %ret
  }

  define <2 x i64> @test_sub_v2i64(<2 x i64> %arg1, <2 x i64> %arg2) {
    %ret = sub <2 x i64> %arg1, %arg2
    ret <2 x i64> %ret
  }

...
---
name:            test_sub_v16i8
# ALL-LABEL: name:  test_sub_v16i8
alignment:       16
legalized:       true
regBankSelected: true
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# SSE2:                %2:vr128 = PSUBBrr %0, %1
#
# AVX1:                %2:vr128 = VPSUBBrr %0, %1
#
# AVX512VL:            %2:vr128 = VPSUBBrr %0, %1
#
# AVX512BWVL:          %2:vr128x = VPSUBBZ128rr %0, %1
body:             |
  bb.1 (%ir-block.0):
    liveins: $xmm0, $xmm1

    %0(<16 x s8>) = COPY $xmm0
    %1(<16 x s8>) = COPY $xmm1
    %2(<16 x s8>) = G_SUB %0, %1
    $xmm0 = COPY %2(<16 x s8>)
    RET 0, implicit $xmm0

...
---
name:            test_sub_v8i16
# ALL-LABEL: name:  test_sub_v8i16
alignment:       16
legalized:       true
regBankSelected: true
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# SSE2:                %2:vr128 = PSUBWrr %0, %1
#
# AVX1:                %2:vr128 = VPSUBWrr %0, %1
#
# AVX512VL:            %2:vr128 = VPSUBWrr %0, %1
#
# AVX512BWVL:          %2:vr128x = VPSUBWZ128rr %0, %1
body:             |
  bb.1 (%ir-block.0):
    liveins: $xmm0, $xmm1

    %0(<8 x s16>) = COPY $xmm0
    %1(<8 x s16>) = COPY $xmm1
    %2(<8 x s16>) = G_SUB %0, %1
    $xmm0 = COPY %2(<8 x s16>)
    RET 0, implicit $xmm0

...
---
name:            test_sub_v4i32
# ALL-LABEL: name:  test_sub_v4i32
alignment:       16
legalized:       true
regBankSelected: true
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# SSE2:                %2:vr128 = PSUBDrr %0, %1
#
# AVX1:                %2:vr128 = VPSUBDrr %0, %1
#
# AVX512VL:            %2:vr128x = VPSUBDZ128rr %0, %1
#
# AVX512BWVL:          %2:vr128x = VPSUBDZ128rr %0, %1
body:             |
  bb.1 (%ir-block.0):
    liveins: $xmm0, $xmm1

    %0(<4 x s32>) = COPY $xmm0
    %1(<4 x s32>) = COPY $xmm1
    %2(<4 x s32>) = G_SUB %0, %1
    $xmm0 = COPY %2(<4 x s32>)
    RET 0, implicit $xmm0

...
---
name:            test_sub_v2i64
# ALL-LABEL: name:  test_sub_v2i64
alignment:       16
legalized:       true
regBankSelected: true
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# SSE2:                %2:vr128 = PSUBQrr %0, %1
#
# AVX1:                %2:vr128 = VPSUBQrr %0, %1
#
# AVX512VL:            %2:vr128x = VPSUBQZ128rr %0, %1
#
# AVX512BWVL:          %2:vr128x = VPSUBQZ128rr %0, %1
body:             |
  bb.1 (%ir-block.0):
    liveins: $xmm0, $xmm1

    %0(<2 x s64>) = COPY $xmm0
    %1(<2 x s64>) = COPY $xmm1
    %2(<2 x s64>) = G_SUB %0, %1
    $xmm0 = COPY %2(<2 x s64>)
    RET 0, implicit $xmm0

...