; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mattr=+sve -force-streaming-compatible < %s | FileCheck %s
; RUN: llc -mattr=+sme -force-streaming < %s | FileCheck %s
; RUN: llc -force-streaming-compatible < %s | FileCheck %s --check-prefix=NONEON-NOSVE


target triple = "aarch64-unknown-linux-gnu"

define <4 x i8> @shuffle_ext_byone_v4i8(<4 x i8> %op1, <4 x i8> %op2) {
; CHECK-LABEL: shuffle_ext_byone_v4i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    adrp x8, .LCPI0_0
; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI0_0]
; CHECK-NEXT:    tbl z0.h, { z0.h }, z1.h
; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v4i8:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    str d0, [sp, #-16]!
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 16
; NONEON-NOSVE-NEXT:    ldrh w8, [sp, #6]
; NONEON-NOSVE-NEXT:    strh w8, [sp, #10]
; NONEON-NOSVE-NEXT:    ldrh w8, [sp]
; NONEON-NOSVE-NEXT:    strh w8, [sp, #8]
; NONEON-NOSVE-NEXT:    ldur w8, [sp, #2]
; NONEON-NOSVE-NEXT:    ror w8, w8, #16
; NONEON-NOSVE-NEXT:    str w8, [sp, #12]
; NONEON-NOSVE-NEXT:    ldr d0, [sp, #8]
; NONEON-NOSVE-NEXT:    add sp, sp, #16
; NONEON-NOSVE-NEXT:    ret
  %ret = shufflevector <4 x i8> %op1, <4 x i8> %op2, <4 x i32> <i32 0, i32 3, i32 2, i32 1>
  ret <4 x i8> %ret
}

define <8 x i8> @shuffle_ext_byone_v8i8(<8 x i8> %op1, <8 x i8> %op2) {
; CHECK-LABEL: shuffle_ext_byone_v8i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
; CHECK-NEXT:    mov z0.b, z0.b[7]
; CHECK-NEXT:    fmov w8, s0
; CHECK-NEXT:    insr z1.b, w8
; CHECK-NEXT:    fmov d0, d1
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v8i8:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    sub sp, sp, #32
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 32
; NONEON-NOSVE-NEXT:    stp d0, d1, [sp, #8]
; NONEON-NOSVE-NEXT:    ldrb w8, [sp, #22]
; NONEON-NOSVE-NEXT:    strb w8, [sp, #31]
; NONEON-NOSVE-NEXT:    ldrh w8, [sp, #20]
; NONEON-NOSVE-NEXT:    sturh w8, [sp, #29]
; NONEON-NOSVE-NEXT:    ldr w8, [sp, #16]
; NONEON-NOSVE-NEXT:    stur w8, [sp, #25]
; NONEON-NOSVE-NEXT:    ldrb w8, [sp, #15]
; NONEON-NOSVE-NEXT:    strb w8, [sp, #24]
; NONEON-NOSVE-NEXT:    ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT:    add sp, sp, #32
; NONEON-NOSVE-NEXT:    ret
  %ret = shufflevector <8 x i8> %op1, <8 x i8> %op2, <8 x i32> <i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14>
  ret <8 x i8> %ret
}

define <16 x i8> @shuffle_ext_byone_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
; CHECK-LABEL: shuffle_ext_byone_v16i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
; CHECK-NEXT:    mov z0.b, z0.b[15]
; CHECK-NEXT:    fmov w8, s0
; CHECK-NEXT:    insr z1.b, w8
; CHECK-NEXT:    mov z0.d, z1.d
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v16i8:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    stp q0, q1, [sp, #-48]!
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 48
; NONEON-NOSVE-NEXT:    ldrb w8, [sp, #30]
; NONEON-NOSVE-NEXT:    strb w8, [sp, #47]
; NONEON-NOSVE-NEXT:    ldrh w8, [sp, #28]
; NONEON-NOSVE-NEXT:    sturh w8, [sp, #45]
; NONEON-NOSVE-NEXT:    ldr w8, [sp, #24]
; NONEON-NOSVE-NEXT:    stur w8, [sp, #41]
; NONEON-NOSVE-NEXT:    ldr x8, [sp, #16]
; NONEON-NOSVE-NEXT:    stur x8, [sp, #33]
; NONEON-NOSVE-NEXT:    ldrb w8, [sp, #15]
; NONEON-NOSVE-NEXT:    strb w8, [sp, #32]
; NONEON-NOSVE-NEXT:    ldr q0, [sp, #32]
; NONEON-NOSVE-NEXT:    add sp, sp, #48
; NONEON-NOSVE-NEXT:    ret
  %ret = shufflevector <16 x i8> %op1, <16 x i8> %op2, <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22,
                                                                   i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
  ret <16 x i8> %ret
}

define void @shuffle_ext_byone_v32i8(ptr %a, ptr %b) {
; CHECK-LABEL: shuffle_ext_byone_v32i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldp q0, q2, [x1]
; CHECK-NEXT:    ldrb w8, [x0, #31]
; CHECK-NEXT:    mov z1.b, z0.b[15]
; CHECK-NEXT:    insr z0.b, w8
; CHECK-NEXT:    fmov w8, s1
; CHECK-NEXT:    insr z2.b, w8
; CHECK-NEXT:    stp q0, q2, [x0]
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v32i8:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    ldp q1, q0, [x1]
; NONEON-NOSVE-NEXT:    ldr q2, [x0, #16]
; NONEON-NOSVE-NEXT:    str q0, [sp, #-80]!
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 80
; NONEON-NOSVE-NEXT:    ldrb w8, [sp, #14]
; NONEON-NOSVE-NEXT:    stp q2, q1, [sp, #32]
; NONEON-NOSVE-NEXT:    strb w8, [sp, #31]
; NONEON-NOSVE-NEXT:    ldrh w8, [sp, #12]
; NONEON-NOSVE-NEXT:    sturh w8, [sp, #29]
; NONEON-NOSVE-NEXT:    ldr w8, [sp, #8]
; NONEON-NOSVE-NEXT:    stur w8, [sp, #25]
; NONEON-NOSVE-NEXT:    ldr x8, [sp]
; NONEON-NOSVE-NEXT:    stur x8, [sp, #17]
; NONEON-NOSVE-NEXT:    ldrb w8, [sp, #63]
; NONEON-NOSVE-NEXT:    strb w8, [sp, #16]
; NONEON-NOSVE-NEXT:    ldrb w8, [sp, #62]
; NONEON-NOSVE-NEXT:    ldr q1, [sp, #16]
; NONEON-NOSVE-NEXT:    strb w8, [sp, #79]
; NONEON-NOSVE-NEXT:    ldrh w8, [sp, #60]
; NONEON-NOSVE-NEXT:    sturh w8, [sp, #77]
; NONEON-NOSVE-NEXT:    ldr w8, [sp, #56]
; NONEON-NOSVE-NEXT:    stur w8, [sp, #73]
; NONEON-NOSVE-NEXT:    ldr x8, [sp, #48]
; NONEON-NOSVE-NEXT:    stur x8, [sp, #65]
; NONEON-NOSVE-NEXT:    ldrb w8, [sp, #47]
; NONEON-NOSVE-NEXT:    strb w8, [sp, #64]
; NONEON-NOSVE-NEXT:    ldr q0, [sp, #64]
; NONEON-NOSVE-NEXT:    stp q0, q1, [x0]
; NONEON-NOSVE-NEXT:    add sp, sp, #80
; NONEON-NOSVE-NEXT:    ret
  %op1 = load <32 x i8>, ptr %a
  %op2 = load <32 x i8>, ptr %b
  %ret = shufflevector <32 x i8> %op1, <32 x i8> %op2, <32 x i32> <i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38,
                                                                   i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46,
                                                                   i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54,
                                                                   i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62>
  store <32 x i8> %ret, ptr %a
  ret void
}

define <2 x i16> @shuffle_ext_byone_v2i16(<2 x i16> %op1, <2 x i16> %op2) {
; CHECK-LABEL: shuffle_ext_byone_v2i16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ptrue p0.d
; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT:    revw z0.d, p0/m, z0.d
; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v2i16:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    str d0, [sp, #-16]!
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 16
; NONEON-NOSVE-NEXT:    ldp w9, w8, [sp]
; NONEON-NOSVE-NEXT:    stp w8, w9, [sp, #8]
; NONEON-NOSVE-NEXT:    ldr d0, [sp, #8]
; NONEON-NOSVE-NEXT:    add sp, sp, #16
; NONEON-NOSVE-NEXT:    ret
  %ret = shufflevector <2 x i16> %op1, <2 x i16> %op2, <2 x i32> <i32 1, i32 0>
  ret <2 x i16> %ret
}

define <4 x i16> @shuffle_ext_byone_v4i16(<4 x i16> %op1, <4 x i16> %op2) {
; CHECK-LABEL: shuffle_ext_byone_v4i16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
; CHECK-NEXT:    mov z0.h, z0.h[3]
; CHECK-NEXT:    fmov w8, s0
; CHECK-NEXT:    insr z1.h, w8
; CHECK-NEXT:    fmov d0, d1
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v4i16:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    sub sp, sp, #32
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 32
; NONEON-NOSVE-NEXT:    stp d0, d1, [sp, #8]
; NONEON-NOSVE-NEXT:    ldrh w8, [sp, #20]
; NONEON-NOSVE-NEXT:    strh w8, [sp, #30]
; NONEON-NOSVE-NEXT:    ldr w8, [sp, #16]
; NONEON-NOSVE-NEXT:    stur w8, [sp, #26]
; NONEON-NOSVE-NEXT:    ldrh w8, [sp, #14]
; NONEON-NOSVE-NEXT:    strh w8, [sp, #24]
; NONEON-NOSVE-NEXT:    ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT:    add sp, sp, #32
; NONEON-NOSVE-NEXT:    ret
  %ret = shufflevector <4 x i16> %op1, <4 x i16> %op2, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
  ret <4 x i16> %ret
}

define <8 x i16> @shuffle_ext_byone_v8i16(<8 x i16> %op1, <8 x i16> %op2) {
; CHECK-LABEL: shuffle_ext_byone_v8i16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
; CHECK-NEXT:    mov z0.h, z0.h[7]
; CHECK-NEXT:    fmov w8, s0
; CHECK-NEXT:    insr z1.h, w8
; CHECK-NEXT:    mov z0.d, z1.d
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v8i16:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    stp q0, q1, [sp, #-48]!
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 48
; NONEON-NOSVE-NEXT:    ldrh w8, [sp, #28]
; NONEON-NOSVE-NEXT:    strh w8, [sp, #46]
; NONEON-NOSVE-NEXT:    ldr w8, [sp, #24]
; NONEON-NOSVE-NEXT:    stur w8, [sp, #42]
; NONEON-NOSVE-NEXT:    ldr x8, [sp, #16]
; NONEON-NOSVE-NEXT:    stur x8, [sp, #34]
; NONEON-NOSVE-NEXT:    ldrh w8, [sp, #14]
; NONEON-NOSVE-NEXT:    strh w8, [sp, #32]
; NONEON-NOSVE-NEXT:    ldr q0, [sp, #32]
; NONEON-NOSVE-NEXT:    add sp, sp, #48
; NONEON-NOSVE-NEXT:    ret
  %ret = shufflevector <8 x i16> %op1, <8 x i16> %op2, <8 x i32> <i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14>
  ret <8 x i16> %ret
}

define void @shuffle_ext_byone_v16i16(ptr %a, ptr %b) {
; CHECK-LABEL: shuffle_ext_byone_v16i16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldp q0, q2, [x1]
; CHECK-NEXT:    ldrh w8, [x0, #30]
; CHECK-NEXT:    mov z1.h, z0.h[7]
; CHECK-NEXT:    insr z0.h, w8
; CHECK-NEXT:    fmov w8, s1
; CHECK-NEXT:    insr z2.h, w8
; CHECK-NEXT:    stp q0, q2, [x0]
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v16i16:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    ldp q1, q0, [x1]
; NONEON-NOSVE-NEXT:    ldr q2, [x0, #16]
; NONEON-NOSVE-NEXT:    str q0, [sp, #-80]!
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 80
; NONEON-NOSVE-NEXT:    ldrh w8, [sp, #12]
; NONEON-NOSVE-NEXT:    stp q2, q1, [sp, #32]
; NONEON-NOSVE-NEXT:    strh w8, [sp, #30]
; NONEON-NOSVE-NEXT:    ldr w8, [sp, #8]
; NONEON-NOSVE-NEXT:    stur w8, [sp, #26]
; NONEON-NOSVE-NEXT:    ldr x8, [sp]
; NONEON-NOSVE-NEXT:    stur x8, [sp, #18]
; NONEON-NOSVE-NEXT:    ldrh w8, [sp, #62]
; NONEON-NOSVE-NEXT:    strh w8, [sp, #16]
; NONEON-NOSVE-NEXT:    ldrh w8, [sp, #60]
; NONEON-NOSVE-NEXT:    ldr q1, [sp, #16]
; NONEON-NOSVE-NEXT:    strh w8, [sp, #78]
; NONEON-NOSVE-NEXT:    ldr w8, [sp, #56]
; NONEON-NOSVE-NEXT:    stur w8, [sp, #74]
; NONEON-NOSVE-NEXT:    ldr x8, [sp, #48]
; NONEON-NOSVE-NEXT:    stur x8, [sp, #66]
; NONEON-NOSVE-NEXT:    ldrh w8, [sp, #46]
; NONEON-NOSVE-NEXT:    strh w8, [sp, #64]
; NONEON-NOSVE-NEXT:    ldr q0, [sp, #64]
; NONEON-NOSVE-NEXT:    stp q0, q1, [x0]
; NONEON-NOSVE-NEXT:    add sp, sp, #80
; NONEON-NOSVE-NEXT:    ret
  %op1 = load <16 x i16>, ptr %a
  %op2 = load <16 x i16>, ptr %b
  %ret = shufflevector <16 x i16> %op1, <16 x i16> %op2, <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22,
                                                                     i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
  store <16 x i16> %ret, ptr %a
  ret void
}

define <2 x i32> @shuffle_ext_byone_v2i32(<2 x i32> %op1, <2 x i32> %op2) {
; CHECK-LABEL: shuffle_ext_byone_v2i32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
; CHECK-NEXT:    mov z0.s, z0.s[1]
; CHECK-NEXT:    fmov w8, s0
; CHECK-NEXT:    insr z1.s, w8
; CHECK-NEXT:    fmov d0, d1
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v2i32:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    sub sp, sp, #32
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 32
; NONEON-NOSVE-NEXT:    stp d0, d1, [sp, #8]
; NONEON-NOSVE-NEXT:    ldp w8, w9, [sp, #12]
; NONEON-NOSVE-NEXT:    stp w8, w9, [sp, #24]
; NONEON-NOSVE-NEXT:    ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT:    add sp, sp, #32
; NONEON-NOSVE-NEXT:    ret
  %ret = shufflevector <2 x i32> %op1, <2 x i32> %op2, <2 x i32> <i32 1, i32 2>
  ret <2 x i32> %ret
}

define <4 x i32> @shuffle_ext_byone_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
; CHECK-LABEL: shuffle_ext_byone_v4i32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
; CHECK-NEXT:    mov z0.s, z0.s[3]
; CHECK-NEXT:    fmov w8, s0
; CHECK-NEXT:    insr z1.s, w8
; CHECK-NEXT:    mov z0.d, z1.d
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v4i32:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    stp q0, q1, [sp, #-48]!
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 48
; NONEON-NOSVE-NEXT:    ldr w8, [sp, #24]
; NONEON-NOSVE-NEXT:    str w8, [sp, #44]
; NONEON-NOSVE-NEXT:    ldr x8, [sp, #16]
; NONEON-NOSVE-NEXT:    stur x8, [sp, #36]
; NONEON-NOSVE-NEXT:    ldr w8, [sp, #12]
; NONEON-NOSVE-NEXT:    str w8, [sp, #32]
; NONEON-NOSVE-NEXT:    ldr q0, [sp, #32]
; NONEON-NOSVE-NEXT:    add sp, sp, #48
; NONEON-NOSVE-NEXT:    ret
  %ret = shufflevector <4 x i32> %op1, <4 x i32> %op2, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
  ret <4 x i32> %ret
}

define void @shuffle_ext_byone_v8i32(ptr %a, ptr %b) {
; CHECK-LABEL: shuffle_ext_byone_v8i32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldp q0, q2, [x1]
; CHECK-NEXT:    ldr w8, [x0, #28]
; CHECK-NEXT:    mov z1.s, z0.s[3]
; CHECK-NEXT:    insr z0.s, w8
; CHECK-NEXT:    fmov w8, s1
; CHECK-NEXT:    insr z2.s, w8
; CHECK-NEXT:    stp q0, q2, [x0]
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v8i32:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    ldp q1, q0, [x1]
; NONEON-NOSVE-NEXT:    ldr q2, [x0, #16]
; NONEON-NOSVE-NEXT:    str q0, [sp, #-80]!
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 80
; NONEON-NOSVE-NEXT:    ldr w8, [sp, #8]
; NONEON-NOSVE-NEXT:    stp q2, q1, [sp, #32]
; NONEON-NOSVE-NEXT:    str w8, [sp, #28]
; NONEON-NOSVE-NEXT:    ldr x8, [sp]
; NONEON-NOSVE-NEXT:    stur x8, [sp, #20]
; NONEON-NOSVE-NEXT:    ldp w8, w9, [sp, #56]
; NONEON-NOSVE-NEXT:    str w8, [sp, #76]
; NONEON-NOSVE-NEXT:    ldr x8, [sp, #48]
; NONEON-NOSVE-NEXT:    str w9, [sp, #16]
; NONEON-NOSVE-NEXT:    stur x8, [sp, #68]
; NONEON-NOSVE-NEXT:    ldr w8, [sp, #44]
; NONEON-NOSVE-NEXT:    ldr q1, [sp, #16]
; NONEON-NOSVE-NEXT:    str w8, [sp, #64]
; NONEON-NOSVE-NEXT:    ldr q0, [sp, #64]
; NONEON-NOSVE-NEXT:    stp q0, q1, [x0]
; NONEON-NOSVE-NEXT:    add sp, sp, #80
; NONEON-NOSVE-NEXT:    ret
  %op1 = load <8 x i32>, ptr %a
  %op2 = load <8 x i32>, ptr %b
  %ret = shufflevector <8 x i32> %op1, <8 x i32> %op2, <8 x i32> <i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14>
  store <8 x i32> %ret, ptr %a
  ret void
}

define <2 x i64> @shuffle_ext_byone_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
; CHECK-LABEL: shuffle_ext_byone_v2i64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
; CHECK-NEXT:    mov z0.d, z0.d[1]
; CHECK-NEXT:    fmov x8, d0
; CHECK-NEXT:    insr z1.d, x8
; CHECK-NEXT:    mov z0.d, z1.d
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v2i64:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    stp q0, q1, [sp, #-48]!
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 48
; NONEON-NOSVE-NEXT:    ldp x8, x9, [sp, #8]
; NONEON-NOSVE-NEXT:    stp x8, x9, [sp, #32]
; NONEON-NOSVE-NEXT:    ldr q0, [sp, #32]
; NONEON-NOSVE-NEXT:    add sp, sp, #48
; NONEON-NOSVE-NEXT:    ret
  %ret = shufflevector <2 x i64> %op1, <2 x i64> %op2, <2 x i32> <i32 1, i32 2>
  ret <2 x i64> %ret
}

define void @shuffle_ext_byone_v4i64(ptr %a, ptr %b) {
; CHECK-LABEL: shuffle_ext_byone_v4i64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldp q0, q2, [x1]
; CHECK-NEXT:    ldr x8, [x0, #24]
; CHECK-NEXT:    mov z1.d, z0.d[1]
; CHECK-NEXT:    insr z0.d, x8
; CHECK-NEXT:    fmov x8, d1
; CHECK-NEXT:    insr z2.d, x8
; CHECK-NEXT:    stp q0, q2, [x0]
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v4i64:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    ldp q1, q0, [x1]
; NONEON-NOSVE-NEXT:    ldr q2, [x0, #16]
; NONEON-NOSVE-NEXT:    str q0, [sp, #-80]!
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 80
; NONEON-NOSVE-NEXT:    stp q2, q1, [sp, #32]
; NONEON-NOSVE-NEXT:    ldr x9, [sp]
; NONEON-NOSVE-NEXT:    ldp x11, x10, [sp, #48]
; NONEON-NOSVE-NEXT:    ldr x8, [sp, #40]
; NONEON-NOSVE-NEXT:    stp x10, x9, [sp, #16]
; NONEON-NOSVE-NEXT:    stp x8, x11, [sp, #64]
; NONEON-NOSVE-NEXT:    ldr q1, [sp, #16]
; NONEON-NOSVE-NEXT:    ldr q0, [sp, #64]
; NONEON-NOSVE-NEXT:    stp q0, q1, [x0]
; NONEON-NOSVE-NEXT:    add sp, sp, #80
; NONEON-NOSVE-NEXT:    ret
  %op1 = load <4 x i64>, ptr %a
  %op2 = load <4 x i64>, ptr %b
  %ret = shufflevector <4 x i64> %op1, <4 x i64> %op2, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
  store <4 x i64> %ret, ptr %a
  ret void
}


define <4 x half> @shuffle_ext_byone_v4f16(<4 x half> %op1, <4 x half> %op2) {
; CHECK-LABEL: shuffle_ext_byone_v4f16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT:    mov z2.h, z0.h[3]
; CHECK-NEXT:    fmov d0, d1
; CHECK-NEXT:    insr z0.h, h2
; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v4f16:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    sub sp, sp, #32
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 32
; NONEON-NOSVE-NEXT:    stp d0, d1, [sp, #8]
; NONEON-NOSVE-NEXT:    ldr h0, [sp, #20]
; NONEON-NOSVE-NEXT:    ldr w8, [sp, #16]
; NONEON-NOSVE-NEXT:    str h0, [sp, #30]
; NONEON-NOSVE-NEXT:    ldr h0, [sp, #14]
; NONEON-NOSVE-NEXT:    stur w8, [sp, #26]
; NONEON-NOSVE-NEXT:    str h0, [sp, #24]
; NONEON-NOSVE-NEXT:    ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT:    add sp, sp, #32
; NONEON-NOSVE-NEXT:    ret
  %ret = shufflevector <4 x half> %op1, <4 x half> %op2, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
  ret <4 x half> %ret
}

define <8 x half> @shuffle_ext_byone_v8f16(<8 x half> %op1, <8 x half> %op2) {
; CHECK-LABEL: shuffle_ext_byone_v8f16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
; CHECK-NEXT:    mov z2.h, z0.h[7]
; CHECK-NEXT:    mov z0.d, z1.d
; CHECK-NEXT:    insr z0.h, h2
; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v8f16:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    stp q0, q1, [sp, #-48]!
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 48
; NONEON-NOSVE-NEXT:    ldr h0, [sp, #28]
; NONEON-NOSVE-NEXT:    ldr w8, [sp, #24]
; NONEON-NOSVE-NEXT:    str h0, [sp, #46]
; NONEON-NOSVE-NEXT:    ldr h0, [sp, #14]
; NONEON-NOSVE-NEXT:    stur w8, [sp, #42]
; NONEON-NOSVE-NEXT:    ldr x8, [sp, #16]
; NONEON-NOSVE-NEXT:    str h0, [sp, #32]
; NONEON-NOSVE-NEXT:    stur x8, [sp, #34]
; NONEON-NOSVE-NEXT:    ldr q0, [sp, #32]
; NONEON-NOSVE-NEXT:    add sp, sp, #48
; NONEON-NOSVE-NEXT:    ret
  %ret = shufflevector <8 x half> %op1, <8 x half> %op2, <8 x i32> <i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14>
  ret <8 x half> %ret
}

define void @shuffle_ext_byone_v16f16(ptr %a, ptr %b) {
; CHECK-LABEL: shuffle_ext_byone_v16f16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldp q1, q3, [x1]
; CHECK-NEXT:    ldr q0, [x0, #16]
; CHECK-NEXT:    mov z0.h, z0.h[7]
; CHECK-NEXT:    mov z2.h, z1.h[7]
; CHECK-NEXT:    insr z1.h, h0
; CHECK-NEXT:    insr z3.h, h2
; CHECK-NEXT:    stp q1, q3, [x0]
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v16f16:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    ldp q1, q0, [x1]
; NONEON-NOSVE-NEXT:    ldr q2, [x0, #16]
; NONEON-NOSVE-NEXT:    str q0, [sp, #-80]!
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 80
; NONEON-NOSVE-NEXT:    ldr h0, [sp, #12]
; NONEON-NOSVE-NEXT:    ldr w8, [sp, #8]
; NONEON-NOSVE-NEXT:    stp q2, q1, [sp, #32]
; NONEON-NOSVE-NEXT:    str h0, [sp, #30]
; NONEON-NOSVE-NEXT:    ldr h0, [sp, #62]
; NONEON-NOSVE-NEXT:    stur w8, [sp, #26]
; NONEON-NOSVE-NEXT:    ldr x8, [sp]
; NONEON-NOSVE-NEXT:    str h0, [sp, #16]
; NONEON-NOSVE-NEXT:    ldr h0, [sp, #60]
; NONEON-NOSVE-NEXT:    stur x8, [sp, #18]
; NONEON-NOSVE-NEXT:    ldr w8, [sp, #56]
; NONEON-NOSVE-NEXT:    str h0, [sp, #78]
; NONEON-NOSVE-NEXT:    ldr h0, [sp, #46]
; NONEON-NOSVE-NEXT:    ldr q1, [sp, #16]
; NONEON-NOSVE-NEXT:    stur w8, [sp, #74]
; NONEON-NOSVE-NEXT:    ldr x8, [sp, #48]
; NONEON-NOSVE-NEXT:    str h0, [sp, #64]
; NONEON-NOSVE-NEXT:    stur x8, [sp, #66]
; NONEON-NOSVE-NEXT:    ldr q0, [sp, #64]
; NONEON-NOSVE-NEXT:    stp q0, q1, [x0]
; NONEON-NOSVE-NEXT:    add sp, sp, #80
; NONEON-NOSVE-NEXT:    ret
  %op1 = load <16 x half>, ptr %a
  %op2 = load <16 x half>, ptr %b
  %ret = shufflevector <16 x half> %op1, <16 x half> %op2, <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22,
                                                                       i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
  store <16 x half> %ret, ptr %a
  ret void
}

define <2 x float> @shuffle_ext_byone_v2f32(<2 x float> %op1, <2 x float> %op2) {
; CHECK-LABEL: shuffle_ext_byone_v2f32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT:    mov z2.s, z0.s[1]
; CHECK-NEXT:    fmov d0, d1
; CHECK-NEXT:    insr z0.s, s2
; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v2f32:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    sub sp, sp, #32
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 32
; NONEON-NOSVE-NEXT:    stp d0, d1, [sp, #8]
; NONEON-NOSVE-NEXT:    ldp w8, w9, [sp, #12]
; NONEON-NOSVE-NEXT:    stp w8, w9, [sp, #24]
; NONEON-NOSVE-NEXT:    ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT:    add sp, sp, #32
; NONEON-NOSVE-NEXT:    ret
  %ret = shufflevector <2 x float> %op1, <2 x float> %op2, <2 x i32> <i32 1, i32 2>
  ret <2 x float> %ret
}

define <4 x float> @shuffle_ext_byone_v4f32(<4 x float> %op1, <4 x float> %op2) {
; CHECK-LABEL: shuffle_ext_byone_v4f32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
; CHECK-NEXT:    mov z2.s, z0.s[3]
; CHECK-NEXT:    mov z0.d, z1.d
; CHECK-NEXT:    insr z0.s, s2
; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v4f32:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    stp q0, q1, [sp, #-48]!
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 48
; NONEON-NOSVE-NEXT:    ldr s0, [sp, #24]
; NONEON-NOSVE-NEXT:    ldr x8, [sp, #16]
; NONEON-NOSVE-NEXT:    str s0, [sp, #44]
; NONEON-NOSVE-NEXT:    ldr s0, [sp, #12]
; NONEON-NOSVE-NEXT:    stur x8, [sp, #36]
; NONEON-NOSVE-NEXT:    str s0, [sp, #32]
; NONEON-NOSVE-NEXT:    ldr q0, [sp, #32]
; NONEON-NOSVE-NEXT:    add sp, sp, #48
; NONEON-NOSVE-NEXT:    ret
  %ret = shufflevector <4 x float> %op1, <4 x float> %op2, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
  ret <4 x float> %ret
}

define void @shuffle_ext_byone_v8f32(ptr %a, ptr %b) {
; CHECK-LABEL: shuffle_ext_byone_v8f32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldp q1, q3, [x1]
; CHECK-NEXT:    ldr q0, [x0, #16]
; CHECK-NEXT:    mov z0.s, z0.s[3]
; CHECK-NEXT:    mov z2.s, z1.s[3]
; CHECK-NEXT:    insr z1.s, s0
; CHECK-NEXT:    insr z3.s, s2
; CHECK-NEXT:    stp q1, q3, [x0]
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v8f32:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    ldp q1, q0, [x1]
; NONEON-NOSVE-NEXT:    ldr q2, [x0, #16]
; NONEON-NOSVE-NEXT:    str q0, [sp, #-80]!
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 80
; NONEON-NOSVE-NEXT:    ldr s0, [sp, #8]
; NONEON-NOSVE-NEXT:    stp q2, q1, [sp, #32]
; NONEON-NOSVE-NEXT:    ldr x8, [sp]
; NONEON-NOSVE-NEXT:    str s0, [sp, #28]
; NONEON-NOSVE-NEXT:    ldp s0, s1, [sp, #56]
; NONEON-NOSVE-NEXT:    stur x8, [sp, #20]
; NONEON-NOSVE-NEXT:    ldr x8, [sp, #48]
; NONEON-NOSVE-NEXT:    str s0, [sp, #76]
; NONEON-NOSVE-NEXT:    ldr s0, [sp, #44]
; NONEON-NOSVE-NEXT:    str s1, [sp, #16]
; NONEON-NOSVE-NEXT:    stur x8, [sp, #68]
; NONEON-NOSVE-NEXT:    ldr q1, [sp, #16]
; NONEON-NOSVE-NEXT:    str s0, [sp, #64]
; NONEON-NOSVE-NEXT:    ldr q0, [sp, #64]
; NONEON-NOSVE-NEXT:    stp q0, q1, [x0]
; NONEON-NOSVE-NEXT:    add sp, sp, #80
; NONEON-NOSVE-NEXT:    ret
  %op1 = load <8 x float>, ptr %a
  %op2 = load <8 x float>, ptr %b
  %ret = shufflevector <8 x float> %op1, <8 x float> %op2, <8 x i32> <i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14>
  store <8 x float> %ret, ptr %a
  ret void
}

define <2 x double> @shuffle_ext_byone_v2f64(<2 x double> %op1, <2 x double> %op2) {
; CHECK-LABEL: shuffle_ext_byone_v2f64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
; CHECK-NEXT:    mov z2.d, z0.d[1]
; CHECK-NEXT:    mov z0.d, z1.d
; CHECK-NEXT:    insr z0.d, d2
; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v2f64:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    stp q0, q1, [sp, #-48]!
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 48
; NONEON-NOSVE-NEXT:    ldp x8, x9, [sp, #8]
; NONEON-NOSVE-NEXT:    stp x8, x9, [sp, #32]
; NONEON-NOSVE-NEXT:    ldr q0, [sp, #32]
; NONEON-NOSVE-NEXT:    add sp, sp, #48
; NONEON-NOSVE-NEXT:    ret
  %ret = shufflevector <2 x double> %op1, <2 x double> %op2, <2 x i32> <i32 1, i32 2>
  ret <2 x double> %ret
}

define void @shuffle_ext_byone_v4f64(ptr %a, ptr %b) {
; CHECK-LABEL: shuffle_ext_byone_v4f64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldp q1, q3, [x1]
; CHECK-NEXT:    ldr q0, [x0, #16]
; CHECK-NEXT:    mov z0.d, z0.d[1]
; CHECK-NEXT:    mov z2.d, z1.d[1]
; CHECK-NEXT:    insr z1.d, d0
; CHECK-NEXT:    insr z3.d, d2
; CHECK-NEXT:    stp q1, q3, [x0]
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_v4f64:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    ldp q1, q0, [x1]
; NONEON-NOSVE-NEXT:    ldr q2, [x0, #16]
; NONEON-NOSVE-NEXT:    str q0, [sp, #-80]!
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 80
; NONEON-NOSVE-NEXT:    stp q2, q1, [sp, #32]
; NONEON-NOSVE-NEXT:    ldr d1, [sp]
; NONEON-NOSVE-NEXT:    ldp d3, d2, [sp, #48]
; NONEON-NOSVE-NEXT:    ldr d0, [sp, #40]
; NONEON-NOSVE-NEXT:    stp d2, d1, [sp, #16]
; NONEON-NOSVE-NEXT:    stp d0, d3, [sp, #64]
; NONEON-NOSVE-NEXT:    ldr q1, [sp, #16]
; NONEON-NOSVE-NEXT:    ldr q0, [sp, #64]
; NONEON-NOSVE-NEXT:    stp q0, q1, [x0]
; NONEON-NOSVE-NEXT:    add sp, sp, #80
; NONEON-NOSVE-NEXT:    ret
  %op1 = load <4 x double>, ptr %a
  %op2 = load <4 x double>, ptr %b
  %ret = shufflevector <4 x double> %op1, <4 x double> %op2, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
  store <4 x double> %ret, ptr %a
  ret void
}

define void @shuffle_ext_byone_reverse(ptr %a, ptr %b) {
; CHECK-LABEL: shuffle_ext_byone_reverse:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldp q1, q3, [x0]
; CHECK-NEXT:    ldr q0, [x1, #16]
; CHECK-NEXT:    mov z0.d, z0.d[1]
; CHECK-NEXT:    mov z2.d, z1.d[1]
; CHECK-NEXT:    insr z1.d, d0
; CHECK-NEXT:    insr z3.d, d2
; CHECK-NEXT:    stp q1, q3, [x0]
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_byone_reverse:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    sub sp, sp, #80
; NONEON-NOSVE-NEXT:    .cfi_def_cfa_offset 80
; NONEON-NOSVE-NEXT:    ldp q1, q0, [x0]
; NONEON-NOSVE-NEXT:    ldr q2, [x1, #16]
; NONEON-NOSVE-NEXT:    str q2, [sp]
; NONEON-NOSVE-NEXT:    stp q1, q0, [sp, #32]
; NONEON-NOSVE-NEXT:    ldp d0, d1, [sp, #40]
; NONEON-NOSVE-NEXT:    stp d0, d1, [sp, #64]
; NONEON-NOSVE-NEXT:    ldr d1, [sp, #32]
; NONEON-NOSVE-NEXT:    ldr d0, [sp, #8]
; NONEON-NOSVE-NEXT:    stp d0, d1, [sp, #16]
; NONEON-NOSVE-NEXT:    ldr q1, [sp, #64]
; NONEON-NOSVE-NEXT:    ldr q0, [sp, #16]
; NONEON-NOSVE-NEXT:    stp q0, q1, [x0]
; NONEON-NOSVE-NEXT:    add sp, sp, #80
; NONEON-NOSVE-NEXT:    ret
  %op1 = load <4 x double>, ptr %a
  %op2 = load <4 x double>, ptr %b
  %ret = shufflevector <4 x double> %op1, <4 x double> %op2, <4 x i32> <i32 7, i32 0, i32 1, i32 2>
  store <4 x double> %ret, ptr %a
  ret void
}

define void @shuffle_ext_invalid(ptr %a, ptr %b) {
; CHECK-LABEL: shuffle_ext_invalid:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0, #16]
; CHECK-NEXT:    ldr q1, [x1]
; CHECK-NEXT:    stp q0, q1, [x0]
; CHECK-NEXT:    ret
;
; NONEON-NOSVE-LABEL: shuffle_ext_invalid:
; NONEON-NOSVE:       // %bb.0:
; NONEON-NOSVE-NEXT:    ldr q0, [x0, #16]
; NONEON-NOSVE-NEXT:    ldr q1, [x1]
; NONEON-NOSVE-NEXT:    stp q0, q1, [x0]
; NONEON-NOSVE-NEXT:    ret
  %op1 = load <4 x double>, ptr %a
  %op2 = load <4 x double>, ptr %b
  %ret = shufflevector <4 x double> %op1, <4 x double> %op2, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
  store <4 x double> %ret, ptr %a
  ret void
}