test/CodeGen/X86/vector-umax-range.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-- -mcpu=x86-64    | FileCheck %s --check-prefixes=SSE2
; RUN: llc < %s -mtriple=x86_64-- -mcpu=x86-64-v2 | FileCheck %s --check-prefixes=SSE42
; RUN: llc < %s -mtriple=x86_64-- -mcpu=btver2    | FileCheck %s --check-prefixes=AVX,AVX1
; RUN: llc < %s -mtriple=x86_64-- -mcpu=x86-64-v3 | FileCheck %s --check-prefixes=AVX,AVX2
; RUN: llc < %s -mtriple=x86_64-- -mcpu=knl       | FileCheck %s --check-prefixes=AVX,AVX512F
; RUN: llc < %s -mtriple=x86_64-- -mcpu=x86-64-v4 | FileCheck %s --check-prefixes=AVX,AVX512VL

define <8 x i16> @umax_v8i16_as_umax_v8i8(<8 x i16> %a0, <8 x i16> %a1) nounwind {
; SSE2-LABEL: umax_v8i16_as_umax_v8i8:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psrlw $8, %xmm0
; SSE2-NEXT:    psrlw $11, %xmm1
; SSE2-NEXT:    pmaxsw %xmm1, %xmm0
; SSE2-NEXT:    retq
;
; SSE42-LABEL: umax_v8i16_as_umax_v8i8:
; SSE42:       # %bb.0:
; SSE42-NEXT:    psrlw $8, %xmm0
; SSE42-NEXT:    psrlw $11, %xmm1
; SSE42-NEXT:    pmaxuw %xmm1, %xmm0
; SSE42-NEXT:    retq
;
; AVX-LABEL: umax_v8i16_as_umax_v8i8:
; AVX:       # %bb.0:
; AVX-NEXT:    vpsrlw $8, %xmm0, %xmm0
; AVX-NEXT:    vpsrlw $11, %xmm1, %xmm1
; AVX-NEXT:    vpmaxuw %xmm1, %xmm0, %xmm0
; AVX-NEXT:    retq
  %x0 = lshr <8 x i16> %a0, splat (i16 8)
  %x1 = lshr <8 x i16> %a1, splat (i16 11)
  %r = call <8 x i16> @llvm.umax.v8i16(<8 x i16> %x0, <8 x i16> %x1)
  ret <8 x i16> %r
}

define <4 x i32> @umax_v4i32_as_umax_v4i8(<4 x i32> %a0, <4 x i32> %a1) nounwind {
; SSE2-LABEL: umax_v4i32_as_umax_v4i8:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psrld $30, %xmm0
; SSE2-NEXT:    psrld $29, %xmm1
; SSE2-NEXT:    pmaxub %xmm1, %xmm0
; SSE2-NEXT:    retq
;
; SSE42-LABEL: umax_v4i32_as_umax_v4i8:
; SSE42:       # %bb.0:
; SSE42-NEXT:    psrld $30, %xmm0
; SSE42-NEXT:    psrld $29, %xmm1
; SSE42-NEXT:    pmaxud %xmm1, %xmm0
; SSE42-NEXT:    retq
;
; AVX-LABEL: umax_v4i32_as_umax_v4i8:
; AVX:       # %bb.0:
; AVX-NEXT:    vpsrld $30, %xmm0, %xmm0
; AVX-NEXT:    vpsrld $29, %xmm1, %xmm1
; AVX-NEXT:    vpmaxud %xmm1, %xmm0, %xmm0
; AVX-NEXT:    retq
  %x0 = lshr <4 x i32> %a0, splat (i32 30)
  %x1 = lshr <4 x i32> %a1, splat (i32 29)
  %r = call <4 x i32> @llvm.umax.v4i32(<4 x i32> %x0, <4 x i32> %x1)
  ret <4 x i32> %r
}

define <4 x i32> @umax_v4i32_as_umax_v4i16(<4 x i32> %a0, <4 x i32> %a1) nounwind {
; SSE2-LABEL: umax_v4i32_as_umax_v4i16:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psrld $16, %xmm0
; SSE2-NEXT:    psrld $16, %xmm1
; SSE2-NEXT:    psubusw %xmm0, %xmm1
; SSE2-NEXT:    paddw %xmm1, %xmm0
; SSE2-NEXT:    retq
;
; SSE42-LABEL: umax_v4i32_as_umax_v4i16:
; SSE42:       # %bb.0:
; SSE42-NEXT:    psrld $16, %xmm0
; SSE42-NEXT:    psrld $16, %xmm1
; SSE42-NEXT:    pmaxud %xmm1, %xmm0
; SSE42-NEXT:    retq
;
; AVX-LABEL: umax_v4i32_as_umax_v4i16:
; AVX:       # %bb.0:
; AVX-NEXT:    vpsrld $16, %xmm0, %xmm0
; AVX-NEXT:    vpsrld $16, %xmm1, %xmm1
; AVX-NEXT:    vpmaxud %xmm1, %xmm0, %xmm0
; AVX-NEXT:    retq
  %x0 = lshr <4 x i32> %a0, splat (i32 16)
  %x1 = lshr <4 x i32> %a1, splat (i32 16)
  %r = call <4 x i32> @llvm.umax.v4i32(<4 x i32> %x0, <4 x i32> %x1)
  ret <4 x i32> %r
}

define <2 x i64> @umax_v2i64_as_umax_v2i16(<2 x i64> %a0, <2 x i64> %a1) nounwind {
; SSE2-LABEL: umax_v2i64_as_umax_v2i16:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psrlq $49, %xmm0
; SSE2-NEXT:    psrlq $63, %xmm1
; SSE2-NEXT:    pmaxsw %xmm1, %xmm0
; SSE2-NEXT:    retq
;
; SSE42-LABEL: umax_v2i64_as_umax_v2i16:
; SSE42:       # %bb.0:
; SSE42-NEXT:    psrlq $49, %xmm0
; SSE42-NEXT:    psrlq $63, %xmm1
; SSE42-NEXT:    pmaxuw %xmm1, %xmm0
; SSE42-NEXT:    retq
;
; AVX1-LABEL: umax_v2i64_as_umax_v2i16:
; AVX1:       # %bb.0:
; AVX1-NEXT:    vpsrlq $49, %xmm0, %xmm0
; AVX1-NEXT:    vpsrlq $63, %xmm1, %xmm1
; AVX1-NEXT:    vpmaxuw %xmm1, %xmm0, %xmm0
; AVX1-NEXT:    retq
;
; AVX2-LABEL: umax_v2i64_as_umax_v2i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsrlq $49, %xmm0, %xmm0
; AVX2-NEXT:    vpsrlq $63, %xmm1, %xmm1
; AVX2-NEXT:    vpmaxuw %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    retq
;
; AVX512F-LABEL: umax_v2i64_as_umax_v2i16:
; AVX512F:       # %bb.0:
; AVX512F-NEXT:    vpsrlq $49, %xmm0, %xmm0
; AVX512F-NEXT:    vpsrlq $63, %xmm1, %xmm1
; AVX512F-NEXT:    vpmaxuq %zmm1, %zmm0, %zmm0
; AVX512F-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
; AVX512F-NEXT:    retq
;
; AVX512VL-LABEL: umax_v2i64_as_umax_v2i16:
; AVX512VL:       # %bb.0:
; AVX512VL-NEXT:    vpsrlq $49, %xmm0, %xmm0
; AVX512VL-NEXT:    vpsrlq $63, %xmm1, %xmm1
; AVX512VL-NEXT:    vpmaxuq %xmm1, %xmm0, %xmm0
; AVX512VL-NEXT:    retq
  %x0 = lshr <2 x i64> %a0, splat (i64 49)
  %x1 = lshr <2 x i64> %a1, splat (i64 63)
  %r = call <2 x i64> @llvm.umax.v2i64(<2 x i64> %x0, <2 x i64> %x1)
  ret <2 x i64> %r
}

define <2 x i64> @umax_v2i64_as_umax_v2i32(<2 x i64> %a0, <2 x i64> %a1) nounwind {
; SSE2-LABEL: umax_v2i64_as_umax_v2i32:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psrlq $33, %xmm0
; SSE2-NEXT:    psrlq $43, %xmm1
; SSE2-NEXT:    movdqa %xmm0, %xmm2
; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
; SSE2-NEXT:    pand %xmm2, %xmm0
; SSE2-NEXT:    pandn %xmm1, %xmm2
; SSE2-NEXT:    por %xmm2, %xmm0
; SSE2-NEXT:    retq
;
; SSE42-LABEL: umax_v2i64_as_umax_v2i32:
; SSE42:       # %bb.0:
; SSE42-NEXT:    psrlq $33, %xmm0
; SSE42-NEXT:    psrlq $43, %xmm1
; SSE42-NEXT:    pmaxud %xmm1, %xmm0
; SSE42-NEXT:    retq
;
; AVX1-LABEL: umax_v2i64_as_umax_v2i32:
; AVX1:       # %bb.0:
; AVX1-NEXT:    vpsrlq $33, %xmm0, %xmm0
; AVX1-NEXT:    vpsrlq $43, %xmm1, %xmm1
; AVX1-NEXT:    vpmaxud %xmm1, %xmm0, %xmm0
; AVX1-NEXT:    retq
;
; AVX2-LABEL: umax_v2i64_as_umax_v2i32:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsrlq $33, %xmm0, %xmm0
; AVX2-NEXT:    vpsrlq $43, %xmm1, %xmm1
; AVX2-NEXT:    vpmaxud %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    retq
;
; AVX512F-LABEL: umax_v2i64_as_umax_v2i32:
; AVX512F:       # %bb.0:
; AVX512F-NEXT:    vpsrlq $33, %xmm0, %xmm0
; AVX512F-NEXT:    vpsrlq $43, %xmm1, %xmm1
; AVX512F-NEXT:    vpmaxuq %zmm1, %zmm0, %zmm0
; AVX512F-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
; AVX512F-NEXT:    retq
;
; AVX512VL-LABEL: umax_v2i64_as_umax_v2i32:
; AVX512VL:       # %bb.0:
; AVX512VL-NEXT:    vpsrlq $33, %xmm0, %xmm0
; AVX512VL-NEXT:    vpsrlq $43, %xmm1, %xmm1
; AVX512VL-NEXT:    vpmaxuq %xmm1, %xmm0, %xmm0
; AVX512VL-NEXT:    retq
  %x0 = lshr <2 x i64> %a0, splat (i64 33)
  %x1 = lshr <2 x i64> %a1, splat (i64 43)
  %r = call <2 x i64> @llvm.umax.v2i64(<2 x i64> %x0, <2 x i64> %x1)
  ret <2 x i64> %r
}

define <4 x i64> @umax_v4i64_as_umax_v4i16(<4 x i64> %a0) nounwind {
; SSE2-LABEL: umax_v4i64_as_umax_v4i16:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psrlq $48, %xmm1
; SSE2-NEXT:    psrlq $48, %xmm0
; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [65530,65530]
; SSE2-NEXT:    movdqa %xmm2, %xmm3
; SSE2-NEXT:    psubusw %xmm0, %xmm3
; SSE2-NEXT:    paddw %xmm3, %xmm0
; SSE2-NEXT:    psubusw %xmm1, %xmm2
; SSE2-NEXT:    paddw %xmm2, %xmm1
; SSE2-NEXT:    retq
;
; SSE42-LABEL: umax_v4i64_as_umax_v4i16:
; SSE42:       # %bb.0:
; SSE42-NEXT:    psrlq $48, %xmm1
; SSE42-NEXT:    psrlq $48, %xmm0
; SSE42-NEXT:    movdqa {{.*#+}} xmm2 = [65530,65530]
; SSE42-NEXT:    pmaxuw %xmm2, %xmm0
; SSE42-NEXT:    pmaxuw %xmm2, %xmm1
; SSE42-NEXT:    retq
;
; AVX1-LABEL: umax_v4i64_as_umax_v4i16:
; AVX1:       # %bb.0:
; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [65530,65530]
; AVX1-NEXT:    vpsrlq $48, %xmm0, %xmm1
; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT:    vpsrlq $48, %xmm0, %xmm0
; AVX1-NEXT:    vpmaxuw %xmm2, %xmm0, %xmm0
; AVX1-NEXT:    vpmaxuw %xmm2, %xmm1, %xmm1
; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-NEXT:    retq
;
; AVX2-LABEL: umax_v4i64_as_umax_v4i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsrlq $48, %ymm0, %ymm0
; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [65530,65530,65530,65530]
; AVX2-NEXT:    vpmaxuw %ymm1, %ymm0, %ymm0
; AVX2-NEXT:    retq
;
; AVX512F-LABEL: umax_v4i64_as_umax_v4i16:
; AVX512F:       # %bb.0:
; AVX512F-NEXT:    vpsrlq $48, %ymm0, %ymm0
; AVX512F-NEXT:    vpmaxuq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm0, %zmm0
; AVX512F-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
; AVX512F-NEXT:    retq
;
; AVX512VL-LABEL: umax_v4i64_as_umax_v4i16:
; AVX512VL:       # %bb.0:
; AVX512VL-NEXT:    vpsrlq $48, %ymm0, %ymm0
; AVX512VL-NEXT:    vpmaxuq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %ymm0, %ymm0
; AVX512VL-NEXT:    retq
  %x0 = lshr <4 x i64> %a0, splat (i64 48)
  %r = call <4 x i64> @llvm.umax.v4i64(<4 x i64> %x0, <4 x i64> splat (i64 65530))
  ret <4 x i64> %r
}