# RUN: llc -mtriple=x86_64-linux-gnu -global-isel -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s

--- |
  define <8 x i16> @test_mul_v8i16(<8 x i16> %arg1, <8 x i16> %arg2) #0 {
    %ret = mul <8 x i16> %arg1, %arg2
    ret <8 x i16> %ret
  }

  define <8 x i16> @test_mul_v8i16_avx(<8 x i16> %arg1, <8 x i16> %arg2) #1 {
    %ret = mul <8 x i16> %arg1, %arg2
    ret <8 x i16> %ret
  }

  define <8 x i16> @test_mul_v8i16_avx512bwvl(<8 x i16> %arg1, <8 x i16> %arg2) #2 {
    %ret = mul <8 x i16> %arg1, %arg2
    ret <8 x i16> %ret
  }

  define <4 x i32> @test_mul_v4i32(<4 x i32> %arg1, <4 x i32> %arg2) #3 {
    %ret = mul <4 x i32> %arg1, %arg2
    ret <4 x i32> %ret
  }

  define <4 x i32> @test_mul_v4i32_avx(<4 x i32> %arg1, <4 x i32> %arg2) #1 {
    %ret = mul <4 x i32> %arg1, %arg2
    ret <4 x i32> %ret
  }

  define <4 x i32> @test_mul_v4i32_avx512vl(<4 x i32> %arg1, <4 x i32> %arg2) #4 {
    %ret = mul <4 x i32> %arg1, %arg2
    ret <4 x i32> %ret
  }

  define <2 x i64> @test_mul_v2i64(<2 x i64> %arg1, <2 x i64> %arg2) #5 {
    %ret = mul <2 x i64> %arg1, %arg2
    ret <2 x i64> %ret
  }

  define <16 x i16> @test_mul_v16i16(<16 x i16> %arg1, <16 x i16> %arg2) #6 {
    %ret = mul <16 x i16> %arg1, %arg2
    ret <16 x i16> %ret
  }

  define <16 x i16> @test_mul_v16i16_avx512bwvl(<16 x i16> %arg1, <16 x i16> %arg2) #2 {
    %ret = mul <16 x i16> %arg1, %arg2
    ret <16 x i16> %ret
  }

  define <8 x i32> @test_mul_v8i32(<8 x i32> %arg1, <8 x i32> %arg2) #6 {
    %ret = mul <8 x i32> %arg1, %arg2
    ret <8 x i32> %ret
  }

  define <8 x i32> @test_mul_v8i32_avx512vl(<8 x i32> %arg1, <8 x i32> %arg2) #4 {
    %ret = mul <8 x i32> %arg1, %arg2
    ret <8 x i32> %ret
  }

  define <4 x i64> @test_mul_v4i64(<4 x i64> %arg1, <4 x i64> %arg2) #5 {
    %ret = mul <4 x i64> %arg1, %arg2
    ret <4 x i64> %ret
  }

  define <32 x i16> @test_mul_v32i16(<32 x i16> %arg1, <32 x i16> %arg2) #7 {
    %ret = mul <32 x i16> %arg1, %arg2
    ret <32 x i16> %ret
  }

  define <16 x i32> @test_mul_v16i32(<16 x i32> %arg1, <16 x i32> %arg2) #8 {
    %ret = mul <16 x i32> %arg1, %arg2
    ret <16 x i32> %ret
  }

  define <8 x i64> @test_mul_v8i64(<8 x i64> %arg1, <8 x i64> %arg2) #9 {
    %ret = mul <8 x i64> %arg1, %arg2
    ret <8 x i64> %ret
  }

  attributes #0 = { "target-features"="+sse2" }
  attributes #1 = { "target-features"="+avx" }
  attributes #2 = { "target-features"="+avx512vl,+avx512f,+avx512bw" }
  attributes #3 = { "target-features"="+sse4.1" }
  attributes #4 = { "target-features"="+avx512vl,+avx512f" }
  attributes #5 = { "target-features"="+avx2,+avx512vl,+avx512f,+avx512dq" }
  attributes #6 = { "target-features"="+avx2" }
  attributes #7 = { "target-features"="+avx512f,+avx512bw" }
  attributes #8 = { "target-features"="+avx512f" }
  attributes #9 = { "target-features"="+avx512f,+avx512dq" }

...
---
name:            test_mul_v8i16
# CHECK-LABEL: name:  test_mul_v8i16
alignment:       4
legalized:       true
regBankSelected: true
# CHECK:      registers:
# CHECK-NEXT:   - { id: 0, class: vr128 }
# CHECK-NEXT:   - { id: 1, class: vr128 }
# CHECK-NEXT:   - { id: 2, class: vr128 }
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# CHECK:          %2 = PMULLWrr %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_MUL %0, %1
    %xmm0 = COPY %2(<8 x s16>)
    RET 0, implicit %xmm0

...
---
name:            test_mul_v8i16_avx
# CHECK-LABEL: name:  test_mul_v8i16_avx
alignment:       4
legalized:       true
regBankSelected: true
# CHECK:      registers:
# CHECK-NEXT:   - { id: 0, class: vr128 }
# CHECK-NEXT:   - { id: 1, class: vr128 }
# CHECK-NEXT:   - { id: 2, class: vr128 }
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# CHECK:          %2 = VPMULLWrr %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_MUL %0, %1
    %xmm0 = COPY %2(<8 x s16>)
    RET 0, implicit %xmm0

...
---
name:            test_mul_v8i16_avx512bwvl
# CHECK-LABEL: name:  test_mul_v8i16_avx512bwvl
alignment:       4
legalized:       true
regBankSelected: true
# CHECK:      registers:
# CHECK-NEXT:   - { id: 0, class: vr128x }
# CHECK-NEXT:   - { id: 1, class: vr128x }
# CHECK-NEXT:   - { id: 2, class: vr128x }
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# CHECK:          %2 = VPMULLWZ128rr %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_MUL %0, %1
    %xmm0 = COPY %2(<8 x s16>)
    RET 0, implicit %xmm0

...
---
name:            test_mul_v4i32
# CHECK-LABEL: name:  test_mul_v4i32
alignment:       4
legalized:       true
regBankSelected: true
# CHECK:      registers:
# CHECK-NEXT:   - { id: 0, class: vr128 }
# CHECK-NEXT:   - { id: 1, class: vr128 }
# CHECK-NEXT:   - { id: 2, class: vr128 }
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# CHECK:          %2 = PMULLDrr %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_MUL %0, %1
    %xmm0 = COPY %2(<4 x s32>)
    RET 0, implicit %xmm0

...
---
name:            test_mul_v4i32_avx
# CHECK-LABEL: name:  test_mul_v4i32_avx
alignment:       4
legalized:       true
regBankSelected: true
# CHECK:      registers:
# CHECK-NEXT:   - { id: 0, class: vr128 }
# CHECK-NEXT:   - { id: 1, class: vr128 }
# CHECK-NEXT:   - { id: 2, class: vr128 }
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# CHECK:          %2 = VPMULLDrr %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_MUL %0, %1
    %xmm0 = COPY %2(<4 x s32>)
    RET 0, implicit %xmm0

...
---
name:            test_mul_v4i32_avx512vl
# CHECK-LABEL: name:  test_mul_v4i32_avx512vl
alignment:       4
legalized:       true
regBankSelected: true
# CHECK:      registers:
# CHECK-NEXT:   - { id: 0, class: vr128x }
# CHECK-NEXT:   - { id: 1, class: vr128x }
# CHECK-NEXT:   - { id: 2, class: vr128x }
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# CHECK:          %2 = VPMULLDZ128rr %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_MUL %0, %1
    %xmm0 = COPY %2(<4 x s32>)
    RET 0, implicit %xmm0

...
---
name:            test_mul_v2i64
# CHECK-LABEL: name:  test_mul_v2i64
alignment:       4
legalized:       true
regBankSelected: true
# CHECK:      registers:
# CHECK-NEXT:   - { id: 0, class: vr128x }
# CHECK-NEXT:   - { id: 1, class: vr128x }
# CHECK-NEXT:   - { id: 2, class: vr128x }
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# CHECK:          %2 = VPMULLQZ128rr %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_MUL %0, %1
    %xmm0 = COPY %2(<2 x s64>)
    RET 0, implicit %xmm0

...
---
name:            test_mul_v16i16
# CHECK-LABEL: name:  test_mul_v16i16
alignment:       4
legalized:       true
regBankSelected: true
# CHECK:      registers:
# CHECK-NEXT:   - { id: 0, class: vr256 }
# CHECK-NEXT:   - { id: 1, class: vr256 }
# CHECK-NEXT:   - { id: 2, class: vr256 }
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# CHECK:          %2 = VPMULLWYrr %0, %1
body:             |
  bb.1 (%ir-block.0):
    liveins: %ymm0, %ymm1

    %0(<16 x s16>) = COPY %ymm0
    %1(<16 x s16>) = COPY %ymm1
    %2(<16 x s16>) = G_MUL %0, %1
    %ymm0 = COPY %2(<16 x s16>)
    RET 0, implicit %ymm0

...
---
name:            test_mul_v16i16_avx512bwvl
# CHECK-LABEL: name:  test_mul_v16i16_avx512bwvl
alignment:       4
legalized:       true
regBankSelected: true
# CHECK:      registers:
# CHECK-NEXT:   - { id: 0, class: vr256x }
# CHECK-NEXT:   - { id: 1, class: vr256x }
# CHECK-NEXT:   - { id: 2, class: vr256x }
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# CHECK:          %2 = VPMULLWZ256rr %0, %1
body:             |
  bb.1 (%ir-block.0):
    liveins: %ymm0, %ymm1

    %0(<16 x s16>) = COPY %ymm0
    %1(<16 x s16>) = COPY %ymm1
    %2(<16 x s16>) = G_MUL %0, %1
    %ymm0 = COPY %2(<16 x s16>)
    RET 0, implicit %ymm0

...
---
name:            test_mul_v8i32
# CHECK-LABEL: name:  test_mul_v8i32
alignment:       4
legalized:       true
regBankSelected: true
# CHECK:      registers:
# CHECK-NEXT:   - { id: 0, class: vr256 }
# CHECK-NEXT:   - { id: 1, class: vr256 }
# CHECK-NEXT:   - { id: 2, class: vr256 }
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# CHECK:          %2 = VPMULLDYrr %0, %1
body:             |
  bb.1 (%ir-block.0):
    liveins: %ymm0, %ymm1

    %0(<8 x s32>) = COPY %ymm0
    %1(<8 x s32>) = COPY %ymm1
    %2(<8 x s32>) = G_MUL %0, %1
    %ymm0 = COPY %2(<8 x s32>)
    RET 0, implicit %ymm0

...
---
name:            test_mul_v8i32_avx512vl
# CHECK-LABEL: name:  test_mul_v8i32_avx512vl
alignment:       4
legalized:       true
regBankSelected: true
# CHECK:      registers:
# CHECK-NEXT:   - { id: 0, class: vr256x }
# CHECK-NEXT:   - { id: 1, class: vr256x }
# CHECK-NEXT:   - { id: 2, class: vr256x }
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# CHECK:          %2 = VPMULLDZ256rr %0, %1
body:             |
  bb.1 (%ir-block.0):
    liveins: %ymm0, %ymm1

    %0(<8 x s32>) = COPY %ymm0
    %1(<8 x s32>) = COPY %ymm1
    %2(<8 x s32>) = G_MUL %0, %1
    %ymm0 = COPY %2(<8 x s32>)
    RET 0, implicit %ymm0

...
---
name:            test_mul_v4i64
# CHECK-LABEL: name:  test_mul_v4i64
alignment:       4
legalized:       true
regBankSelected: true
# CHECK:      registers:
# CHECK-NEXT:   - { id: 0, class: vr256x }
# CHECK-NEXT:   - { id: 1, class: vr256x }
# CHECK-NEXT:   - { id: 2, class: vr256x }
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# CHECK:          %2 = VPMULLQZ256rr %0, %1
body:             |
  bb.1 (%ir-block.0):
    liveins: %ymm0, %ymm1

    %0(<4 x s64>) = COPY %ymm0
    %1(<4 x s64>) = COPY %ymm1
    %2(<4 x s64>) = G_MUL %0, %1
    %ymm0 = COPY %2(<4 x s64>)
    RET 0, implicit %ymm0

...
---
name:            test_mul_v32i16
# CHECK-LABEL: name:  test_mul_v32i16
alignment:       4
legalized:       true
regBankSelected: true
# CHECK:      registers:
# CHECK-NEXT:   - { id: 0, class: vr512 }
# CHECK-NEXT:   - { id: 1, class: vr512 }
# CHECK-NEXT:   - { id: 2, class: vr512 }
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# CHECK:          %2 = VPMULLWZrr %0, %1
body:             |
  bb.1 (%ir-block.0):
    liveins: %zmm0, %zmm1

    %0(<32 x s16>) = COPY %zmm0
    %1(<32 x s16>) = COPY %zmm1
    %2(<32 x s16>) = G_MUL %0, %1
    %zmm0 = COPY %2(<32 x s16>)
    RET 0, implicit %zmm0

...
---
name:            test_mul_v16i32
# CHECK-LABEL: name:  test_mul_v16i32
alignment:       4
legalized:       true
regBankSelected: true
# CHECK:      registers:
# CHECK-NEXT:   - { id: 0, class: vr512 }
# CHECK-NEXT:   - { id: 1, class: vr512 }
# CHECK-NEXT:   - { id: 2, class: vr512 }
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# CHECK:          %2 = VPMULLDZrr %0, %1
body:             |
  bb.1 (%ir-block.0):
    liveins: %zmm0, %zmm1

    %0(<16 x s32>) = COPY %zmm0
    %1(<16 x s32>) = COPY %zmm1
    %2(<16 x s32>) = G_MUL %0, %1
    %zmm0 = COPY %2(<16 x s32>)
    RET 0, implicit %zmm0

...
---
name:            test_mul_v8i64
# CHECK-LABEL: name:  test_mul_v8i64
alignment:       4
legalized:       true
regBankSelected: true
# CHECK:      registers:
# CHECK-NEXT:   - { id: 0, class: vr512 }
# CHECK-NEXT:   - { id: 1, class: vr512 }
# CHECK-NEXT:   - { id: 2, class: vr512 }
registers:
  - { id: 0, class: vecr }
  - { id: 1, class: vecr }
  - { id: 2, class: vecr }
# CHECK:          %2 = VPMULLQZrr %0, %1
body:             |
  bb.1 (%ir-block.0):
    liveins: %zmm0, %zmm1

    %0(<8 x s64>) = COPY %zmm0
    %1(<8 x s64>) = COPY %zmm1
    %2(<8 x s64>) = G_MUL %0, %1
    %zmm0 = COPY %2(<8 x s64>)
    RET 0, implicit %zmm0

...