; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=thumbv8.1m.main-none-none-eabi -mattr=+mve.fp,+fp64 -verify-machineinstrs %s -o - | FileCheck %s

; i32

define void @vld3_v2i32(<6 x i32> *%src, <2 x i32> *%dst) {
; CHECK-LABEL: vld3_v2i32:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrw.u32 q0, [r0]
; CHECK-NEXT:    ldrd r12, r3, [r0, #16]
; CHECK-NEXT:    vmov r0, s1
; CHECK-NEXT:    vmov r2, s0
; CHECK-NEXT:    vmov.f64 d2, d0
; CHECK-NEXT:    vmov.f32 s6, s3
; CHECK-NEXT:    add r0, r2
; CHECK-NEXT:    vmov r2, s2
; CHECK-NEXT:    add r0, r2
; CHECK-NEXT:    vmov r2, s6
; CHECK-NEXT:    add r2, r12
; CHECK-NEXT:    add r2, r3
; CHECK-NEXT:    strd r0, r2, [r1]
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <6 x i32>, <6 x i32>* %src, align 4
  %s1 = shufflevector <6 x i32> %l1, <6 x i32> undef, <2 x i32> <i32 0, i32 3>
  %s2 = shufflevector <6 x i32> %l1, <6 x i32> undef, <2 x i32> <i32 1, i32 4>
  %s3 = shufflevector <6 x i32> %l1, <6 x i32> undef, <2 x i32> <i32 2, i32 5>
  %a1 = add <2 x i32> %s1, %s2
  %a = add <2 x i32> %a1, %s3
  store <2 x i32> %a, <2 x i32> *%dst
  ret void
}

define void @vld3_v4i32(<12 x i32> *%src, <4 x i32> *%dst) {
; CHECK-LABEL: vld3_v4i32:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .vsave {d8, d9}
; CHECK-NEXT:    vpush {d8, d9}
; CHECK-NEXT:    vldrw.u32 q1, [r0]
; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
; CHECK-NEXT:    vmov.f64 d6, d2
; CHECK-NEXT:    vmov.f32 s16, s5
; CHECK-NEXT:    vmov.f32 s13, s7
; CHECK-NEXT:    vmov.f32 s17, s0
; CHECK-NEXT:    vmov.f32 s14, s2
; CHECK-NEXT:    vmov.f32 s18, s3
; CHECK-NEXT:    vmov.f32 s0, s6
; CHECK-NEXT:    vmov.f32 s2, s8
; CHECK-NEXT:    vmov.f32 s19, s10
; CHECK-NEXT:    vmov.f32 s15, s9
; CHECK-NEXT:    vadd.i32 q3, q3, q4
; CHECK-NEXT:    vmov.f32 s3, s11
; CHECK-NEXT:    vadd.i32 q0, q3, q0
; CHECK-NEXT:    vstrw.32 q0, [r1]
; CHECK-NEXT:    vpop {d8, d9}
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <12 x i32>, <12 x i32>* %src, align 4
  %s1 = shufflevector <12 x i32> %l1, <12 x i32> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
  %s2 = shufflevector <12 x i32> %l1, <12 x i32> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
  %s3 = shufflevector <12 x i32> %l1, <12 x i32> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
  %a1 = add <4 x i32> %s1, %s2
  %a = add <4 x i32> %a1, %s3
  store <4 x i32> %a, <4 x i32> *%dst
  ret void
}

define void @vld3_v8i32(<24 x i32> *%src, <8 x i32> *%dst) {
; CHECK-LABEL: vld3_v8i32:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
; CHECK-NEXT:    vpush {d8, d9, d10, d11}
; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
; CHECK-NEXT:    vldrw.u32 q0, [r0, #64]
; CHECK-NEXT:    vldrw.u32 q2, [r0, #80]
; CHECK-NEXT:    vmov.f64 d6, d2
; CHECK-NEXT:    vmov.f32 s16, s5
; CHECK-NEXT:    vmov.f32 s13, s7
; CHECK-NEXT:    vmov.f32 s17, s0
; CHECK-NEXT:    vmov.f32 s14, s2
; CHECK-NEXT:    vmov.f32 s18, s3
; CHECK-NEXT:    vmov.f32 s0, s6
; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
; CHECK-NEXT:    vmov.f32 s2, s8
; CHECK-NEXT:    vmov.f32 s19, s10
; CHECK-NEXT:    vmov.f32 s15, s9
; CHECK-NEXT:    vmov.f32 s3, s11
; CHECK-NEXT:    vadd.i32 q3, q3, q4
; CHECK-NEXT:    vadd.i32 q0, q3, q0
; CHECK-NEXT:    vldrw.u32 q3, [r0]
; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
; CHECK-NEXT:    vmov.f32 s16, s13
; CHECK-NEXT:    vmov.f64 d10, d6
; CHECK-NEXT:    vmov.f32 s17, s4
; CHECK-NEXT:    vmov.f32 s21, s15
; CHECK-NEXT:    vmov.f32 s18, s7
; CHECK-NEXT:    vmov.f32 s22, s6
; CHECK-NEXT:    vmov.f32 s4, s14
; CHECK-NEXT:    vmov.f32 s6, s8
; CHECK-NEXT:    vmov.f32 s19, s10
; CHECK-NEXT:    vmov.f32 s23, s9
; CHECK-NEXT:    vadd.i32 q4, q5, q4
; CHECK-NEXT:    vmov.f32 s7, s11
; CHECK-NEXT:    vadd.i32 q1, q4, q1
; CHECK-NEXT:    vstrw.32 q1, [r1]
; CHECK-NEXT:    vpop {d8, d9, d10, d11}
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <24 x i32>, <24 x i32>* %src, align 4
  %s1 = shufflevector <24 x i32> %l1, <24 x i32> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
  %s2 = shufflevector <24 x i32> %l1, <24 x i32> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
  %s3 = shufflevector <24 x i32> %l1, <24 x i32> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
  %a1 = add <8 x i32> %s1, %s2
  %a = add <8 x i32> %a1, %s3
  store <8 x i32> %a, <8 x i32> *%dst
  ret void
}

define void @vld3_v16i32(<48 x i32> *%src, <16 x i32> *%dst) {
; CHECK-LABEL: vld3_v16i32:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
; CHECK-NEXT:    vldrw.u32 q0, [r0, #64]
; CHECK-NEXT:    vldrw.u32 q2, [r0, #80]
; CHECK-NEXT:    vmov.f64 d6, d2
; CHECK-NEXT:    vmov.f32 s16, s5
; CHECK-NEXT:    vmov.f32 s13, s7
; CHECK-NEXT:    vmov.f32 s17, s0
; CHECK-NEXT:    vmov.f32 s14, s2
; CHECK-NEXT:    vmov.f32 s18, s3
; CHECK-NEXT:    vmov.f32 s0, s6
; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
; CHECK-NEXT:    vmov.f32 s2, s8
; CHECK-NEXT:    vmov.f32 s19, s10
; CHECK-NEXT:    vmov.f32 s15, s9
; CHECK-NEXT:    vmov.f32 s3, s11
; CHECK-NEXT:    vadd.i32 q3, q3, q4
; CHECK-NEXT:    vadd.i32 q0, q3, q0
; CHECK-NEXT:    vldrw.u32 q3, [r0]
; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
; CHECK-NEXT:    vmov.f32 s16, s13
; CHECK-NEXT:    vmov.f64 d10, d6
; CHECK-NEXT:    vmov.f32 s17, s4
; CHECK-NEXT:    vmov.f32 s21, s15
; CHECK-NEXT:    vmov.f32 s18, s7
; CHECK-NEXT:    vmov.f32 s22, s6
; CHECK-NEXT:    vmov.f32 s4, s14
; CHECK-NEXT:    vldrw.u32 q3, [r0, #144]
; CHECK-NEXT:    vmov.f32 s6, s8
; CHECK-NEXT:    vmov.f32 s19, s10
; CHECK-NEXT:    vmov.f32 s23, s9
; CHECK-NEXT:    vmov.f32 s7, s11
; CHECK-NEXT:    vldrw.u32 q2, [r0, #160]
; CHECK-NEXT:    vadd.i32 q4, q5, q4
; CHECK-NEXT:    vmov.f64 d10, d6
; CHECK-NEXT:    vadd.i32 q1, q4, q1
; CHECK-NEXT:    vldrw.u32 q4, [r0, #176]
; CHECK-NEXT:    vmov.f32 s24, s13
; CHECK-NEXT:    vmov.f32 s21, s15
; CHECK-NEXT:    vmov.f32 s25, s8
; CHECK-NEXT:    vmov.f32 s22, s10
; CHECK-NEXT:    vmov.f32 s26, s11
; CHECK-NEXT:    vmov.f32 s8, s14
; CHECK-NEXT:    vldrw.u32 q3, [r0, #112]
; CHECK-NEXT:    vmov.f32 s10, s16
; CHECK-NEXT:    vmov.f32 s27, s18
; CHECK-NEXT:    vmov.f32 s23, s17
; CHECK-NEXT:    vmov.f32 s11, s19
; CHECK-NEXT:    vadd.i32 q5, q5, q6
; CHECK-NEXT:    vadd.i32 q2, q5, q2
; CHECK-NEXT:    vldrw.u32 q5, [r0, #96]
; CHECK-NEXT:    vldrw.u32 q4, [r0, #128]
; CHECK-NEXT:    vstrw.32 q2, [r1, #48]
; CHECK-NEXT:    vmov.f32 s24, s21
; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
; CHECK-NEXT:    vmov.f64 d14, d10
; CHECK-NEXT:    vstrw.32 q1, [r1]
; CHECK-NEXT:    vmov.f32 s25, s12
; CHECK-NEXT:    vmov.f32 s29, s23
; CHECK-NEXT:    vmov.f32 s26, s15
; CHECK-NEXT:    vmov.f32 s30, s14
; CHECK-NEXT:    vmov.f32 s12, s22
; CHECK-NEXT:    vmov.f32 s14, s16
; CHECK-NEXT:    vmov.f32 s27, s18
; CHECK-NEXT:    vmov.f32 s31, s17
; CHECK-NEXT:    vadd.i32 q6, q7, q6
; CHECK-NEXT:    vmov.f32 s15, s19
; CHECK-NEXT:    vadd.i32 q3, q6, q3
; CHECK-NEXT:    vstrw.32 q3, [r1, #32]
; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <48 x i32>, <48 x i32>* %src, align 4
  %s1 = shufflevector <48 x i32> %l1, <48 x i32> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
  %s2 = shufflevector <48 x i32> %l1, <48 x i32> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
  %s3 = shufflevector <48 x i32> %l1, <48 x i32> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
  %a1 = add <16 x i32> %s1, %s2
  %a = add <16 x i32> %a1, %s3
  store <16 x i32> %a, <16 x i32> *%dst
  ret void
}

; i16

define void @vld3_v2i16(<6 x i16> *%src, <2 x i16> *%dst) {
; CHECK-LABEL: vld3_v2i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .pad #8
; CHECK-NEXT:    sub sp, #8
; CHECK-NEXT:    vldrh.u32 q0, [r0]
; CHECK-NEXT:    ldr r2, [r0, #8]
; CHECK-NEXT:    mov r3, sp
; CHECK-NEXT:    str r2, [sp]
; CHECK-NEXT:    vmov.f64 d2, d0
; CHECK-NEXT:    vmov.f32 s6, s3
; CHECK-NEXT:    vmov.f32 s8, s1
; CHECK-NEXT:    vmov.f64 d6, d1
; CHECK-NEXT:    vmov r0, s6
; CHECK-NEXT:    vldrh.u32 q1, [r3]
; CHECK-NEXT:    vmov.f32 s10, s4
; CHECK-NEXT:    vmov.f32 s14, s5
; CHECK-NEXT:    vmov r2, s10
; CHECK-NEXT:    add r0, r2
; CHECK-NEXT:    vmov r2, s14
; CHECK-NEXT:    add r0, r2
; CHECK-NEXT:    strh r0, [r1, #2]
; CHECK-NEXT:    vmov r0, s8
; CHECK-NEXT:    vmov r2, s0
; CHECK-NEXT:    add r0, r2
; CHECK-NEXT:    vmov r2, s12
; CHECK-NEXT:    add r0, r2
; CHECK-NEXT:    strh r0, [r1]
; CHECK-NEXT:    add sp, #8
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <6 x i16>, <6 x i16>* %src, align 4
  %s1 = shufflevector <6 x i16> %l1, <6 x i16> undef, <2 x i32> <i32 0, i32 3>
  %s2 = shufflevector <6 x i16> %l1, <6 x i16> undef, <2 x i32> <i32 1, i32 4>
  %s3 = shufflevector <6 x i16> %l1, <6 x i16> undef, <2 x i32> <i32 2, i32 5>
  %a1 = add <2 x i16> %s1, %s2
  %a = add <2 x i16> %a1, %s3
  store <2 x i16> %a, <2 x i16> *%dst
  ret void
}

define void @vld3_v4i16(<12 x i16> *%src, <4 x i16> *%dst) {
; CHECK-LABEL: vld3_v4i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrw.u32 q0, [r0]
; CHECK-NEXT:    vldrh.u32 q3, [r0, #16]
; CHECK-NEXT:    vmov.u16 r2, q0[6]
; CHECK-NEXT:    vmov.u16 r3, q0[0]
; CHECK-NEXT:    vmov q1[2], q1[0], r3, r2
; CHECK-NEXT:    vmov.u16 r2, q0[7]
; CHECK-NEXT:    vmov.u16 r3, q0[1]
; CHECK-NEXT:    vmov r0, s14
; CHECK-NEXT:    vmov q2[2], q2[0], r3, r2
; CHECK-NEXT:    vmov.u16 r2, q0[4]
; CHECK-NEXT:    vmov q2[3], q2[1], r2, r0
; CHECK-NEXT:    vmov r0, s13
; CHECK-NEXT:    vmov.u16 r2, q0[3]
; CHECK-NEXT:    vmov q1[3], q1[1], r2, r0
; CHECK-NEXT:    vmov r0, s12
; CHECK-NEXT:    vmov.u16 r2, q0[2]
; CHECK-NEXT:    vadd.i32 q1, q1, q2
; CHECK-NEXT:    vmov q2[2], q2[0], r2, r0
; CHECK-NEXT:    vmov r0, s15
; CHECK-NEXT:    vmov.u16 r2, q0[5]
; CHECK-NEXT:    vmov q2[3], q2[1], r2, r0
; CHECK-NEXT:    vadd.i32 q0, q1, q2
; CHECK-NEXT:    vstrh.32 q0, [r1]
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <12 x i16>, <12 x i16>* %src, align 4
  %s1 = shufflevector <12 x i16> %l1, <12 x i16> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
  %s2 = shufflevector <12 x i16> %l1, <12 x i16> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
  %s3 = shufflevector <12 x i16> %l1, <12 x i16> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
  %a1 = add <4 x i16> %s1, %s2
  %a = add <4 x i16> %a1, %s3
  store <4 x i16> %a, <4 x i16> *%dst
  ret void
}

define void @vld3_v8i16(<24 x i16> *%src, <8 x i16> *%dst) {
; CHECK-LABEL: vld3_v8i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
; CHECK-NEXT:    vpush {d8, d9, d10, d11}
; CHECK-NEXT:    vldrw.u32 q1, [r0]
; CHECK-NEXT:    vmovx.f16 s8, s6
; CHECK-NEXT:    vmov.f32 s0, s5
; CHECK-NEXT:    vins.f16 s0, s8
; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
; CHECK-NEXT:    vmovx.f16 s12, s9
; CHECK-NEXT:    vmov.f32 s1, s8
; CHECK-NEXT:    vins.f16 s1, s12
; CHECK-NEXT:    vldrw.u32 q3, [r0, #32]
; CHECK-NEXT:    vmov.f32 s2, s11
; CHECK-NEXT:    vmov.u16 r0, q2[5]
; CHECK-NEXT:    vmovx.f16 s20, s15
; CHECK-NEXT:    vmov.f32 s19, s14
; CHECK-NEXT:    vins.f16 s19, s20
; CHECK-NEXT:    vmov.f32 s18, s12
; CHECK-NEXT:    vmov q5, q4
; CHECK-NEXT:    vmovnb.i32 q5, q0
; CHECK-NEXT:    vmov.f32 s2, s22
; CHECK-NEXT:    vmovx.f16 s20, s5
; CHECK-NEXT:    vmov.f32 s3, s19
; CHECK-NEXT:    vmov.f64 d8, d2
; CHECK-NEXT:    vins.f16 s16, s20
; CHECK-NEXT:    vmovx.f16 s20, s8
; CHECK-NEXT:    vmov.f32 s17, s7
; CHECK-NEXT:    vins.f16 s17, s20
; CHECK-NEXT:    vmovx.f16 s20, s11
; CHECK-NEXT:    vmov.f32 s18, s10
; CHECK-NEXT:    vins.f16 s18, s20
; CHECK-NEXT:    vmovx.f16 s20, s14
; CHECK-NEXT:    vmov.f32 s19, s13
; CHECK-NEXT:    vins.f16 s19, s20
; CHECK-NEXT:    vmovx.f16 s20, s4
; CHECK-NEXT:    vins.f16 s20, s6
; CHECK-NEXT:    vmovx.f16 s21, s7
; CHECK-NEXT:    vins.f16 s6, s12
; CHECK-NEXT:    vmovx.f16 s7, s13
; CHECK-NEXT:    vins.f16 s21, s9
; CHECK-NEXT:    vins.f16 s7, s15
; CHECK-NEXT:    vmov.16 q5[4], r0
; CHECK-NEXT:    vmov q2, q1
; CHECK-NEXT:    vmovnb.i32 q2, q5
; CHECK-NEXT:    vmov.f32 s22, s10
; CHECK-NEXT:    vmov.f32 s23, s7
; CHECK-NEXT:    vadd.i16 q1, q4, q5
; CHECK-NEXT:    vadd.i16 q0, q1, q0
; CHECK-NEXT:    vstrw.32 q0, [r1]
; CHECK-NEXT:    vpop {d8, d9, d10, d11}
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <24 x i16>, <24 x i16>* %src, align 4
  %s1 = shufflevector <24 x i16> %l1, <24 x i16> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
  %s2 = shufflevector <24 x i16> %l1, <24 x i16> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
  %s3 = shufflevector <24 x i16> %l1, <24 x i16> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
  %a1 = add <8 x i16> %s1, %s2
  %a = add <8 x i16> %a1, %s3
  store <8 x i16> %a, <8 x i16> *%dst
  ret void
}

define void @vld3_v16i16(<48 x i16> *%src, <16 x i16> *%dst) {
; CHECK-LABEL: vld3_v16i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT:    .pad #16
; CHECK-NEXT:    sub sp, #16
; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
; CHECK-NEXT:    vmovx.f16 s8, s6
; CHECK-NEXT:    vmov.f32 s0, s5
; CHECK-NEXT:    vins.f16 s0, s8
; CHECK-NEXT:    vldrw.u32 q2, [r0, #64]
; CHECK-NEXT:    vmovx.f16 s12, s9
; CHECK-NEXT:    vmov.f32 s1, s8
; CHECK-NEXT:    vins.f16 s1, s12
; CHECK-NEXT:    vldrw.u32 q3, [r0, #80]
; CHECK-NEXT:    vmov.f32 s2, s11
; CHECK-NEXT:    vmov.u16 r2, q2[5]
; CHECK-NEXT:    vmovx.f16 s20, s15
; CHECK-NEXT:    vmov.f32 s19, s14
; CHECK-NEXT:    vins.f16 s19, s20
; CHECK-NEXT:    vmov.f32 s18, s12
; CHECK-NEXT:    vmov q5, q4
; CHECK-NEXT:    vmovnb.i32 q5, q0
; CHECK-NEXT:    vmov.f32 s2, s22
; CHECK-NEXT:    vmovx.f16 s20, s5
; CHECK-NEXT:    vmov.f32 s3, s19
; CHECK-NEXT:    vmov.f64 d8, d2
; CHECK-NEXT:    vins.f16 s16, s20
; CHECK-NEXT:    vmovx.f16 s20, s8
; CHECK-NEXT:    vmov.f32 s17, s7
; CHECK-NEXT:    vins.f16 s17, s20
; CHECK-NEXT:    vmovx.f16 s20, s11
; CHECK-NEXT:    vmov.f32 s18, s10
; CHECK-NEXT:    vins.f16 s18, s20
; CHECK-NEXT:    vmovx.f16 s20, s14
; CHECK-NEXT:    vmov.f32 s19, s13
; CHECK-NEXT:    vins.f16 s19, s20
; CHECK-NEXT:    vmovx.f16 s20, s4
; CHECK-NEXT:    vins.f16 s20, s6
; CHECK-NEXT:    vmovx.f16 s21, s7
; CHECK-NEXT:    vins.f16 s6, s12
; CHECK-NEXT:    vmovx.f16 s7, s13
; CHECK-NEXT:    vins.f16 s21, s9
; CHECK-NEXT:    vins.f16 s7, s15
; CHECK-NEXT:    vmov.16 q5[4], r2
; CHECK-NEXT:    vmov q2, q1
; CHECK-NEXT:    vmovnb.i32 q2, q5
; CHECK-NEXT:    vmov.f32 s22, s10
; CHECK-NEXT:    vldrw.u32 q2, [r0]
; CHECK-NEXT:    vmov.f32 s23, s7
; CHECK-NEXT:    vadd.i16 q1, q4, q5
; CHECK-NEXT:    vmovx.f16 s12, s10
; CHECK-NEXT:    vadd.i16 q0, q1, q0
; CHECK-NEXT:    vmov.f32 s4, s9
; CHECK-NEXT:    vins.f16 s4, s12
; CHECK-NEXT:    vldrw.u32 q3, [r0, #16]
; CHECK-NEXT:    vstrw.32 q0, [sp] @ 16-byte Spill
; CHECK-NEXT:    vmovx.f16 s0, s9
; CHECK-NEXT:    vmovx.f16 s16, s13
; CHECK-NEXT:    vmov.f32 s5, s12
; CHECK-NEXT:    vins.f16 s5, s16
; CHECK-NEXT:    vldrw.u32 q4, [r0, #32]
; CHECK-NEXT:    vmov.f32 s6, s15
; CHECK-NEXT:    vmov.u16 r0, q3[5]
; CHECK-NEXT:    vmovx.f16 s20, s19
; CHECK-NEXT:    vmov.f32 s27, s18
; CHECK-NEXT:    vins.f16 s27, s20
; CHECK-NEXT:    vmov.f64 d10, d4
; CHECK-NEXT:    vins.f16 s20, s0
; CHECK-NEXT:    vmov.f32 s26, s16
; CHECK-NEXT:    vmovx.f16 s0, s12
; CHECK-NEXT:    vmov.f32 s21, s11
; CHECK-NEXT:    vins.f16 s21, s0
; CHECK-NEXT:    vmov q7, q6
; CHECK-NEXT:    vmovnb.i32 q7, q1
; CHECK-NEXT:    vmovx.f16 s0, s15
; CHECK-NEXT:    vmov.f32 s22, s14
; CHECK-NEXT:    vins.f16 s22, s0
; CHECK-NEXT:    vmov.f32 s6, s30
; CHECK-NEXT:    vmov.f32 s7, s27
; CHECK-NEXT:    vmovx.f16 s24, s8
; CHECK-NEXT:    vmovx.f16 s0, s18
; CHECK-NEXT:    vmov.f32 s23, s17
; CHECK-NEXT:    vins.f16 s24, s10
; CHECK-NEXT:    vins.f16 s23, s0
; CHECK-NEXT:    vins.f16 s2, s16
; CHECK-NEXT:    vmovx.f16 s25, s11
; CHECK-NEXT:    vmovx.f16 s3, s17
; CHECK-NEXT:    vins.f16 s25, s13
; CHECK-NEXT:    vins.f16 s3, s19
; CHECK-NEXT:    vmov.16 q6[4], r0
; CHECK-NEXT:    vmov q2, q0
; CHECK-NEXT:    vmovnb.i32 q2, q6
; CHECK-NEXT:    vmov.f32 s26, s10
; CHECK-NEXT:    vmov.f32 s27, s3
; CHECK-NEXT:    vadd.i16 q0, q5, q6
; CHECK-NEXT:    vadd.i16 q0, q0, q1
; CHECK-NEXT:    vldrw.u32 q1, [sp] @ 16-byte Reload
; CHECK-NEXT:    vstrw.32 q0, [r1]
; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
; CHECK-NEXT:    add sp, #16
; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <48 x i16>, <48 x i16>* %src, align 4
  %s1 = shufflevector <48 x i16> %l1, <48 x i16> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
  %s2 = shufflevector <48 x i16> %l1, <48 x i16> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
  %s3 = shufflevector <48 x i16> %l1, <48 x i16> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
  %a1 = add <16 x i16> %s1, %s2
  %a = add <16 x i16> %a1, %s3
  store <16 x i16> %a, <16 x i16> *%dst
  ret void
}

; i8

define void @vld3_v2i8(<6 x i8> *%src, <2 x i8> *%dst) {
; CHECK-LABEL: vld3_v2i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .pad #8
; CHECK-NEXT:    sub sp, #8
; CHECK-NEXT:    ldrd r2, r0, [r0]
; CHECK-NEXT:    strd r2, r0, [sp]
; CHECK-NEXT:    mov r0, sp
; CHECK-NEXT:    vldrb.u16 q0, [r0]
; CHECK-NEXT:    vmov.u16 r0, q0[4]
; CHECK-NEXT:    vmov.u16 r2, q0[3]
; CHECK-NEXT:    add r0, r2
; CHECK-NEXT:    vmov.u16 r2, q0[5]
; CHECK-NEXT:    add r0, r2
; CHECK-NEXT:    strb r0, [r1, #1]
; CHECK-NEXT:    vmov.u16 r0, q0[1]
; CHECK-NEXT:    vmov.u16 r2, q0[0]
; CHECK-NEXT:    add r0, r2
; CHECK-NEXT:    vmov.u16 r2, q0[2]
; CHECK-NEXT:    add r0, r2
; CHECK-NEXT:    strb r0, [r1]
; CHECK-NEXT:    add sp, #8
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <6 x i8>, <6 x i8>* %src, align 4
  %s1 = shufflevector <6 x i8> %l1, <6 x i8> undef, <2 x i32> <i32 0, i32 3>
  %s2 = shufflevector <6 x i8> %l1, <6 x i8> undef, <2 x i32> <i32 1, i32 4>
  %s3 = shufflevector <6 x i8> %l1, <6 x i8> undef, <2 x i32> <i32 2, i32 5>
  %a1 = add <2 x i8> %s1, %s2
  %a = add <2 x i8> %a1, %s3
  store <2 x i8> %a, <2 x i8> *%dst
  ret void
}

define void @vld3_v4i8(<12 x i8> *%src, <4 x i8> *%dst) {
; CHECK-LABEL: vld3_v4i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .save {r4, lr}
; CHECK-NEXT:    push {r4, lr}
; CHECK-NEXT:    .pad #8
; CHECK-NEXT:    sub sp, #8
; CHECK-NEXT:    vldrb.u16 q0, [r0]
; CHECK-NEXT:    ldr r0, [r0, #8]
; CHECK-NEXT:    str r0, [sp]
; CHECK-NEXT:    vmov.u16 r3, q0[6]
; CHECK-NEXT:    vmov.u16 r4, q0[0]
; CHECK-NEXT:    vmov q1[2], q1[0], r4, r3
; CHECK-NEXT:    vmov.u16 r3, q0[7]
; CHECK-NEXT:    vmov.u16 r4, q0[1]
; CHECK-NEXT:    vmov.u16 r12, q0[5]
; CHECK-NEXT:    vmov q2[2], q2[0], r4, r3
; CHECK-NEXT:    mov r3, sp
; CHECK-NEXT:    vmov.u16 lr, q0[2]
; CHECK-NEXT:    vmov.u16 r2, q0[3]
; CHECK-NEXT:    vmov.u16 r0, q0[4]
; CHECK-NEXT:    vldrb.u16 q0, [r3]
; CHECK-NEXT:    vmov.u16 r3, q0[2]
; CHECK-NEXT:    vmov q2[3], q2[1], r0, r3
; CHECK-NEXT:    vmov.u16 r0, q0[1]
; CHECK-NEXT:    vmov q1[3], q1[1], r2, r0
; CHECK-NEXT:    vmov.u16 r0, q0[0]
; CHECK-NEXT:    vadd.i32 q1, q1, q2
; CHECK-NEXT:    vmov q2[2], q2[0], lr, r0
; CHECK-NEXT:    vmov.u16 r0, q0[3]
; CHECK-NEXT:    vmov q2[3], q2[1], r12, r0
; CHECK-NEXT:    vadd.i32 q0, q1, q2
; CHECK-NEXT:    vstrb.32 q0, [r1]
; CHECK-NEXT:    add sp, #8
; CHECK-NEXT:    pop {r4, pc}
entry:
  %l1 = load <12 x i8>, <12 x i8>* %src, align 4
  %s1 = shufflevector <12 x i8> %l1, <12 x i8> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
  %s2 = shufflevector <12 x i8> %l1, <12 x i8> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
  %s3 = shufflevector <12 x i8> %l1, <12 x i8> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
  %a1 = add <4 x i8> %s1, %s2
  %a = add <4 x i8> %a1, %s3
  store <4 x i8> %a, <4 x i8> *%dst
  ret void
}

define void @vld3_v8i8(<24 x i8> *%src, <8 x i8> *%dst) {
; CHECK-LABEL: vld3_v8i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .vsave {d8, d9}
; CHECK-NEXT:    vpush {d8, d9}
; CHECK-NEXT:    vldrw.u32 q0, [r0]
; CHECK-NEXT:    vldrb.u16 q1, [r0, #16]
; CHECK-NEXT:    vmov.u8 r2, q0[1]
; CHECK-NEXT:    vmov.u8 r0, q0[0]
; CHECK-NEXT:    vmov.16 q2[0], r2
; CHECK-NEXT:    vmov.u8 r2, q0[4]
; CHECK-NEXT:    vmov.16 q2[1], r2
; CHECK-NEXT:    vmov.u8 r2, q0[7]
; CHECK-NEXT:    vmov.16 q3[0], r0
; CHECK-NEXT:    vmov.u8 r0, q0[3]
; CHECK-NEXT:    vmov.16 q2[2], r2
; CHECK-NEXT:    vmov.u8 r2, q0[10]
; CHECK-NEXT:    vmov.16 q3[1], r0
; CHECK-NEXT:    vmov.u8 r0, q0[6]
; CHECK-NEXT:    vmov.16 q2[3], r2
; CHECK-NEXT:    vmov.u8 r2, q0[13]
; CHECK-NEXT:    vmov.16 q3[2], r0
; CHECK-NEXT:    vmov.u8 r0, q0[9]
; CHECK-NEXT:    vmov.16 q2[4], r2
; CHECK-NEXT:    vmov.16 q3[3], r0
; CHECK-NEXT:    vmov.u8 r0, q0[12]
; CHECK-NEXT:    vins.f16 s10, s4
; CHECK-NEXT:    vmov.16 q3[4], r0
; CHECK-NEXT:    vmov.u8 r0, q0[15]
; CHECK-NEXT:    vmovx.f16 s16, s6
; CHECK-NEXT:    vmov.f32 s18, s5
; CHECK-NEXT:    vmovx.f16 s11, s5
; CHECK-NEXT:    vmov.16 q3[5], r0
; CHECK-NEXT:    vins.f16 s18, s16
; CHECK-NEXT:    vins.f16 s11, s7
; CHECK-NEXT:    vmov.f32 s15, s18
; CHECK-NEXT:    vmov.u8 r0, q0[2]
; CHECK-NEXT:    vadd.i16 q2, q3, q2
; CHECK-NEXT:    vmov.16 q3[0], r0
; CHECK-NEXT:    vmov.u8 r0, q0[5]
; CHECK-NEXT:    vmov.16 q3[1], r0
; CHECK-NEXT:    vmov.u8 r0, q0[8]
; CHECK-NEXT:    vmov.16 q3[2], r0
; CHECK-NEXT:    vmov.u8 r0, q0[11]
; CHECK-NEXT:    vmov.16 q3[3], r0
; CHECK-NEXT:    vmov.u8 r0, q0[14]
; CHECK-NEXT:    vmov.16 q3[4], r0
; CHECK-NEXT:    vmov.u16 r0, q1[1]
; CHECK-NEXT:    vmovx.f16 s0, s7
; CHECK-NEXT:    vmov.f32 s2, s6
; CHECK-NEXT:    vins.f16 s2, s0
; CHECK-NEXT:    vmov.16 q3[5], r0
; CHECK-NEXT:    vmov.f32 s15, s2
; CHECK-NEXT:    vadd.i16 q0, q2, q3
; CHECK-NEXT:    vstrb.16 q0, [r1]
; CHECK-NEXT:    vpop {d8, d9}
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <24 x i8>, <24 x i8>* %src, align 4
  %s1 = shufflevector <24 x i8> %l1, <24 x i8> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
  %s2 = shufflevector <24 x i8> %l1, <24 x i8> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
  %s3 = shufflevector <24 x i8> %l1, <24 x i8> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
  %a1 = add <8 x i8> %s1, %s2
  %a = add <8 x i8> %a1, %s3
  store <8 x i8> %a, <8 x i8> *%dst
  ret void
}

define void @vld3_v16i8(<48 x i8> *%src, <16 x i8> *%dst) {
; CHECK-LABEL: vld3_v16i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
; CHECK-NEXT:    vldrw.u32 q1, [r0]
; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
; CHECK-NEXT:    vmov.u8 r2, q1[1]
; CHECK-NEXT:    vmov.8 q3[0], r2
; CHECK-NEXT:    vmov.u8 r2, q1[4]
; CHECK-NEXT:    vmov.8 q3[1], r2
; CHECK-NEXT:    vmov.u8 r2, q1[7]
; CHECK-NEXT:    vmov.8 q3[2], r2
; CHECK-NEXT:    vmov.u8 r2, q1[10]
; CHECK-NEXT:    vmov.8 q3[3], r2
; CHECK-NEXT:    vmov.u8 r2, q1[13]
; CHECK-NEXT:    vmov.8 q3[4], r2
; CHECK-NEXT:    vmov.u8 r2, q0[0]
; CHECK-NEXT:    vmov.8 q3[5], r2
; CHECK-NEXT:    vmov.u8 r2, q0[3]
; CHECK-NEXT:    vmov.8 q3[6], r2
; CHECK-NEXT:    vmov.u8 r2, q0[6]
; CHECK-NEXT:    vmov.8 q3[7], r2
; CHECK-NEXT:    vmov.u8 r2, q0[9]
; CHECK-NEXT:    vmov.u8 r0, q2[5]
; CHECK-NEXT:    vmov.8 q3[8], r2
; CHECK-NEXT:    vmov.u8 r2, q0[12]
; CHECK-NEXT:    vmov.8 q4[12], r0
; CHECK-NEXT:    vmov.u8 r0, q2[8]
; CHECK-NEXT:    vmov.8 q3[9], r2
; CHECK-NEXT:    vmov.u8 r2, q0[15]
; CHECK-NEXT:    vmov.8 q4[13], r0
; CHECK-NEXT:    vmov.u8 r0, q2[11]
; CHECK-NEXT:    vmov.8 q3[10], r2
; CHECK-NEXT:    vmov.8 q4[14], r0
; CHECK-NEXT:    vmov.u8 r0, q2[14]
; CHECK-NEXT:    vmov.8 q4[15], r0
; CHECK-NEXT:    vmov.u8 r0, q2[2]
; CHECK-NEXT:    vmov q5, q3
; CHECK-NEXT:    vmov.8 q5[11], r0
; CHECK-NEXT:    vmov.u8 r0, q1[0]
; CHECK-NEXT:    vmov.f32 s14, s22
; CHECK-NEXT:    vmov.f32 s15, s19
; CHECK-NEXT:    vmov.8 q4[0], r0
; CHECK-NEXT:    vmov.u8 r0, q1[3]
; CHECK-NEXT:    vmov.8 q4[1], r0
; CHECK-NEXT:    vmov.u8 r0, q1[6]
; CHECK-NEXT:    vmov.8 q4[2], r0
; CHECK-NEXT:    vmov.u8 r0, q1[9]
; CHECK-NEXT:    vmov.8 q4[3], r0
; CHECK-NEXT:    vmov.u8 r0, q1[12]
; CHECK-NEXT:    vmov.8 q4[4], r0
; CHECK-NEXT:    vmov.u8 r0, q1[15]
; CHECK-NEXT:    vmov.8 q4[5], r0
; CHECK-NEXT:    vmov.u8 r0, q0[2]
; CHECK-NEXT:    vmov.8 q4[6], r0
; CHECK-NEXT:    vmov.u8 r0, q0[5]
; CHECK-NEXT:    vmov.8 q4[7], r0
; CHECK-NEXT:    vmov.u8 r0, q0[8]
; CHECK-NEXT:    vmov.8 q4[8], r0
; CHECK-NEXT:    vmov.u8 r0, q0[11]
; CHECK-NEXT:    vmov.8 q4[9], r0
; CHECK-NEXT:    vmov.u8 r0, q0[14]
; CHECK-NEXT:    vmov.8 q4[10], r0
; CHECK-NEXT:    vmov.u8 r0, q2[4]
; CHECK-NEXT:    vmov.8 q5[12], r0
; CHECK-NEXT:    vmov.u8 r0, q2[7]
; CHECK-NEXT:    vmov.8 q5[13], r0
; CHECK-NEXT:    vmov.u8 r0, q2[10]
; CHECK-NEXT:    vmov.8 q5[14], r0
; CHECK-NEXT:    vmov.u8 r0, q2[13]
; CHECK-NEXT:    vmov.8 q5[15], r0
; CHECK-NEXT:    vmov.u8 r0, q2[1]
; CHECK-NEXT:    vmov q6, q4
; CHECK-NEXT:    vmov.8 q6[11], r0
; CHECK-NEXT:    vmov.u8 r0, q1[2]
; CHECK-NEXT:    vmov.f32 s18, s26
; CHECK-NEXT:    vmov.f32 s19, s23
; CHECK-NEXT:    vadd.i8 q3, q4, q3
; CHECK-NEXT:    vmov.8 q4[0], r0
; CHECK-NEXT:    vmov.u8 r0, q1[5]
; CHECK-NEXT:    vmov.8 q4[1], r0
; CHECK-NEXT:    vmov.u8 r0, q1[8]
; CHECK-NEXT:    vmov.8 q4[2], r0
; CHECK-NEXT:    vmov.u8 r0, q1[11]
; CHECK-NEXT:    vmov.8 q4[3], r0
; CHECK-NEXT:    vmov.u8 r0, q1[14]
; CHECK-NEXT:    vmov.8 q4[4], r0
; CHECK-NEXT:    vmov.u8 r0, q0[1]
; CHECK-NEXT:    vmov.8 q4[5], r0
; CHECK-NEXT:    vmov.u8 r0, q0[4]
; CHECK-NEXT:    vmov.8 q4[6], r0
; CHECK-NEXT:    vmov.u8 r0, q2[6]
; CHECK-NEXT:    vmov.8 q1[12], r0
; CHECK-NEXT:    vmov.u8 r0, q2[9]
; CHECK-NEXT:    vmov.8 q1[13], r0
; CHECK-NEXT:    vmov.u8 r0, q2[12]
; CHECK-NEXT:    vmov.8 q1[14], r0
; CHECK-NEXT:    vmov.u8 r0, q2[15]
; CHECK-NEXT:    vmov.8 q1[15], r0
; CHECK-NEXT:    vmov.u8 r0, q0[10]
; CHECK-NEXT:    vmov.8 q5[8], r0
; CHECK-NEXT:    vmov.u8 r0, q0[13]
; CHECK-NEXT:    vmov.8 q5[9], r0
; CHECK-NEXT:    vmov.u8 r0, q2[0]
; CHECK-NEXT:    vmov.8 q5[10], r0
; CHECK-NEXT:    vmov.u8 r0, q2[3]
; CHECK-NEXT:    vmov.8 q5[11], r0
; CHECK-NEXT:    vmov.u8 r0, q0[7]
; CHECK-NEXT:    vmov.8 q4[7], r0
; CHECK-NEXT:    vmov.f32 s18, s22
; CHECK-NEXT:    vmov.f32 s19, s7
; CHECK-NEXT:    vadd.i8 q0, q3, q4
; CHECK-NEXT:    vstrw.32 q0, [r1]
; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <48 x i8>, <48 x i8>* %src, align 4
  %s1 = shufflevector <48 x i8> %l1, <48 x i8> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
  %s2 = shufflevector <48 x i8> %l1, <48 x i8> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
  %s3 = shufflevector <48 x i8> %l1, <48 x i8> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
  %a1 = add <16 x i8> %s1, %s2
  %a = add <16 x i8> %a1, %s3
  store <16 x i8> %a, <16 x i8> *%dst
  ret void
}

; i64

define void @vld3_v2i64(<6 x i64> *%src, <2 x i64> *%dst) {
; CHECK-LABEL: vld3_v2i64:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .save {r4, lr}
; CHECK-NEXT:    push {r4, lr}
; CHECK-NEXT:    .vsave {d8, d9}
; CHECK-NEXT:    vpush {d8, d9}
; CHECK-NEXT:    vldrw.u32 q1, [r0]
; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
; CHECK-NEXT:    vldrw.u32 q4, [r0, #32]
; CHECK-NEXT:    vmov.f64 d6, d3
; CHECK-NEXT:    vmov.f32 s13, s7
; CHECK-NEXT:    vmov.f32 s14, s16
; CHECK-NEXT:    vmov.f32 s6, s10
; CHECK-NEXT:    vmov.f32 s7, s11
; CHECK-NEXT:    vmov.f32 s15, s17
; CHECK-NEXT:    vmov r3, s14
; CHECK-NEXT:    vmov r0, s6
; CHECK-NEXT:    vmov.f64 d0, d4
; CHECK-NEXT:    vmov.f32 s1, s9
; CHECK-NEXT:    vmov.f32 s2, s18
; CHECK-NEXT:    vmov.f32 s3, s19
; CHECK-NEXT:    vmov r12, s15
; CHECK-NEXT:    vmov r2, s7
; CHECK-NEXT:    vmov r4, s4
; CHECK-NEXT:    adds.w lr, r0, r3
; CHECK-NEXT:    vmov r0, s2
; CHECK-NEXT:    vmov r3, s3
; CHECK-NEXT:    adc.w r2, r2, r12
; CHECK-NEXT:    adds.w lr, lr, r0
; CHECK-NEXT:    vmov r0, s12
; CHECK-NEXT:    adc.w r12, r2, r3
; CHECK-NEXT:    vmov r3, s13
; CHECK-NEXT:    vmov r2, s5
; CHECK-NEXT:    adds r0, r0, r4
; CHECK-NEXT:    vmov r4, s0
; CHECK-NEXT:    adcs r2, r3
; CHECK-NEXT:    vmov r3, s1
; CHECK-NEXT:    adds r0, r0, r4
; CHECK-NEXT:    vmov q0[2], q0[0], r0, lr
; CHECK-NEXT:    adcs r2, r3
; CHECK-NEXT:    vmov q0[3], q0[1], r2, r12
; CHECK-NEXT:    vstrw.32 q0, [r1]
; CHECK-NEXT:    vpop {d8, d9}
; CHECK-NEXT:    pop {r4, pc}
entry:
  %l1 = load <6 x i64>, <6 x i64>* %src, align 4
  %s1 = shufflevector <6 x i64> %l1, <6 x i64> undef, <2 x i32> <i32 0, i32 3>
  %s2 = shufflevector <6 x i64> %l1, <6 x i64> undef, <2 x i32> <i32 1, i32 4>
  %s3 = shufflevector <6 x i64> %l1, <6 x i64> undef, <2 x i32> <i32 2, i32 5>
  %a1 = add <2 x i64> %s1, %s2
  %a = add <2 x i64> %a1, %s3
  store <2 x i64> %a, <2 x i64> *%dst
  ret void
}

define void @vld3_v4i64(<12 x i64> *%src, <4 x i64> *%dst) {
; CHECK-LABEL: vld3_v4i64:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .save {r4, r5, r6, r7, r8, lr}
; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, lr}
; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT:    .pad #24
; CHECK-NEXT:    sub sp, #24
; CHECK-NEXT:    vldrw.u32 q2, [r0, #64]
; CHECK-NEXT:    vldrw.u32 q1, [r0, #80]
; CHECK-NEXT:    vldrw.u32 q5, [r0]
; CHECK-NEXT:    vldrw.u32 q3, [r0, #16]
; CHECK-NEXT:    vmov.f64 d0, d4
; CHECK-NEXT:    vstrw.32 q2, [sp] @ 16-byte Spill
; CHECK-NEXT:    vmov.f32 s1, s9
; CHECK-NEXT:    vldrw.u32 q2, [r0, #48]
; CHECK-NEXT:    vmov.f32 s2, s6
; CHECK-NEXT:    vmov.f64 d8, d5
; CHECK-NEXT:    vmov.f32 s17, s11
; CHECK-NEXT:    vmov.f32 s18, s4
; CHECK-NEXT:    vmov.f32 s19, s5
; CHECK-NEXT:    vmov.f64 d12, d11
; CHECK-NEXT:    vmov.f32 s3, s7
; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
; CHECK-NEXT:    vmov.f32 s25, s23
; CHECK-NEXT:    vmov.f32 s26, s4
; CHECK-NEXT:    vmov.f32 s22, s14
; CHECK-NEXT:    vmov.f32 s27, s5
; CHECK-NEXT:    vmov.f32 s23, s15
; CHECK-NEXT:    vmov r3, s26
; CHECK-NEXT:    vmov r0, s22
; CHECK-NEXT:    vmov.f64 d14, d6
; CHECK-NEXT:    vmov r12, s27
; CHECK-NEXT:    vmov r2, s23
; CHECK-NEXT:    vmov.f32 s29, s13
; CHECK-NEXT:    vmov.f32 s30, s6
; CHECK-NEXT:    vmov.f32 s31, s7
; CHECK-NEXT:    vldrw.u32 q1, [sp] @ 16-byte Reload
; CHECK-NEXT:    vmov r5, s16
; CHECK-NEXT:    vmov.f32 s10, s6
; CHECK-NEXT:    vmov.f32 s11, s7
; CHECK-NEXT:    vmov r4, s10
; CHECK-NEXT:    vmov r6, s8
; CHECK-NEXT:    vmov r7, s24
; CHECK-NEXT:    adds.w lr, r0, r3
; CHECK-NEXT:    vmov r0, s30
; CHECK-NEXT:    adc.w r3, r2, r12
; CHECK-NEXT:    vmov r2, s31
; CHECK-NEXT:    adds.w lr, lr, r0
; CHECK-NEXT:    vmov r0, s18
; CHECK-NEXT:    adc.w r12, r3, r2
; CHECK-NEXT:    vmov r3, s19
; CHECK-NEXT:    vmov r2, s11
; CHECK-NEXT:    adds r0, r0, r4
; CHECK-NEXT:    vmov r4, s2
; CHECK-NEXT:    adcs r2, r3
; CHECK-NEXT:    vmov r3, s3
; CHECK-NEXT:    adds r0, r0, r4
; CHECK-NEXT:    vmov r4, s9
; CHECK-NEXT:    adc.w r8, r2, r3
; CHECK-NEXT:    vmov r3, s17
; CHECK-NEXT:    adds r5, r5, r6
; CHECK-NEXT:    vmov r6, s0
; CHECK-NEXT:    vmov r2, s20
; CHECK-NEXT:    adcs r3, r4
; CHECK-NEXT:    vmov r4, s1
; CHECK-NEXT:    adds r5, r5, r6
; CHECK-NEXT:    vmov r6, s21
; CHECK-NEXT:    vmov q1[2], q1[0], r5, r0
; CHECK-NEXT:    vmov r0, s29
; CHECK-NEXT:    adcs r3, r4
; CHECK-NEXT:    vmov r4, s25
; CHECK-NEXT:    adds r2, r2, r7
; CHECK-NEXT:    vmov q1[3], q1[1], r3, r8
; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
; CHECK-NEXT:    adcs r4, r6
; CHECK-NEXT:    vmov r6, s28
; CHECK-NEXT:    adds r2, r2, r6
; CHECK-NEXT:    vmov q0[2], q0[0], r2, lr
; CHECK-NEXT:    adcs r0, r4
; CHECK-NEXT:    vmov q0[3], q0[1], r0, r12
; CHECK-NEXT:    vstrw.32 q0, [r1]
; CHECK-NEXT:    add sp, #24
; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, pc}
entry:
  %l1 = load <12 x i64>, <12 x i64>* %src, align 4
  %s1 = shufflevector <12 x i64> %l1, <12 x i64> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
  %s2 = shufflevector <12 x i64> %l1, <12 x i64> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
  %s3 = shufflevector <12 x i64> %l1, <12 x i64> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
  %a1 = add <4 x i64> %s1, %s2
  %a = add <4 x i64> %a1, %s3
  store <4 x i64> %a, <4 x i64> *%dst
  ret void
}

; f32

define void @vld3_v2f32(<6 x float> *%src, <2 x float> *%dst) {
; CHECK-LABEL: vld3_v2f32:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrw.u32 q2, [r0]
; CHECK-NEXT:    vldr s1, [r0, #16]
; CHECK-NEXT:    vldr s5, [r0, #20]
; CHECK-NEXT:    vmov.f64 d6, d4
; CHECK-NEXT:    vmov.f32 s13, s11
; CHECK-NEXT:    vmov.f32 s0, s9
; CHECK-NEXT:    vadd.f32 q0, q3, q0
; CHECK-NEXT:    vmov.f32 s4, s10
; CHECK-NEXT:    vadd.f32 q0, q0, q1
; CHECK-NEXT:    vstmia r1, {s0, s1}
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <6 x float>, <6 x float>* %src, align 4
  %s1 = shufflevector <6 x float> %l1, <6 x float> undef, <2 x i32> <i32 0, i32 3>
  %s2 = shufflevector <6 x float> %l1, <6 x float> undef, <2 x i32> <i32 1, i32 4>
  %s3 = shufflevector <6 x float> %l1, <6 x float> undef, <2 x i32> <i32 2, i32 5>
  %a1 = fadd <2 x float> %s1, %s2
  %a = fadd <2 x float> %a1, %s3
  store <2 x float> %a, <2 x float> *%dst
  ret void
}

define void @vld3_v4f32(<12 x float> *%src, <4 x float> *%dst) {
; CHECK-LABEL: vld3_v4f32:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .vsave {d8, d9}
; CHECK-NEXT:    vpush {d8, d9}
; CHECK-NEXT:    vldrw.u32 q1, [r0]
; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
; CHECK-NEXT:    vmov.f64 d6, d2
; CHECK-NEXT:    vmov.f32 s16, s5
; CHECK-NEXT:    vmov.f32 s13, s7
; CHECK-NEXT:    vmov.f32 s17, s0
; CHECK-NEXT:    vmov.f32 s14, s2
; CHECK-NEXT:    vmov.f32 s18, s3
; CHECK-NEXT:    vmov.f32 s0, s6
; CHECK-NEXT:    vmov.f32 s2, s8
; CHECK-NEXT:    vmov.f32 s19, s10
; CHECK-NEXT:    vmov.f32 s15, s9
; CHECK-NEXT:    vadd.f32 q3, q3, q4
; CHECK-NEXT:    vmov.f32 s3, s11
; CHECK-NEXT:    vadd.f32 q0, q3, q0
; CHECK-NEXT:    vstrw.32 q0, [r1]
; CHECK-NEXT:    vpop {d8, d9}
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <12 x float>, <12 x float>* %src, align 4
  %s1 = shufflevector <12 x float> %l1, <12 x float> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
  %s2 = shufflevector <12 x float> %l1, <12 x float> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
  %s3 = shufflevector <12 x float> %l1, <12 x float> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
  %a1 = fadd <4 x float> %s1, %s2
  %a = fadd <4 x float> %a1, %s3
  store <4 x float> %a, <4 x float> *%dst
  ret void
}

define void @vld3_v8f32(<24 x float> *%src, <8 x float> *%dst) {
; CHECK-LABEL: vld3_v8f32:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
; CHECK-NEXT:    vpush {d8, d9, d10, d11}
; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
; CHECK-NEXT:    vldrw.u32 q0, [r0, #64]
; CHECK-NEXT:    vldrw.u32 q2, [r0, #80]
; CHECK-NEXT:    vmov.f64 d6, d2
; CHECK-NEXT:    vmov.f32 s16, s5
; CHECK-NEXT:    vmov.f32 s13, s7
; CHECK-NEXT:    vmov.f32 s17, s0
; CHECK-NEXT:    vmov.f32 s14, s2
; CHECK-NEXT:    vmov.f32 s18, s3
; CHECK-NEXT:    vmov.f32 s0, s6
; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
; CHECK-NEXT:    vmov.f32 s2, s8
; CHECK-NEXT:    vmov.f32 s19, s10
; CHECK-NEXT:    vmov.f32 s15, s9
; CHECK-NEXT:    vmov.f32 s3, s11
; CHECK-NEXT:    vadd.f32 q3, q3, q4
; CHECK-NEXT:    vadd.f32 q0, q3, q0
; CHECK-NEXT:    vldrw.u32 q3, [r0]
; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
; CHECK-NEXT:    vmov.f32 s16, s13
; CHECK-NEXT:    vmov.f64 d10, d6
; CHECK-NEXT:    vmov.f32 s17, s4
; CHECK-NEXT:    vmov.f32 s21, s15
; CHECK-NEXT:    vmov.f32 s18, s7
; CHECK-NEXT:    vmov.f32 s22, s6
; CHECK-NEXT:    vmov.f32 s4, s14
; CHECK-NEXT:    vmov.f32 s6, s8
; CHECK-NEXT:    vmov.f32 s19, s10
; CHECK-NEXT:    vmov.f32 s23, s9
; CHECK-NEXT:    vadd.f32 q4, q5, q4
; CHECK-NEXT:    vmov.f32 s7, s11
; CHECK-NEXT:    vadd.f32 q1, q4, q1
; CHECK-NEXT:    vstrw.32 q1, [r1]
; CHECK-NEXT:    vpop {d8, d9, d10, d11}
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <24 x float>, <24 x float>* %src, align 4
  %s1 = shufflevector <24 x float> %l1, <24 x float> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
  %s2 = shufflevector <24 x float> %l1, <24 x float> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
  %s3 = shufflevector <24 x float> %l1, <24 x float> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
  %a1 = fadd <8 x float> %s1, %s2
  %a = fadd <8 x float> %a1, %s3
  store <8 x float> %a, <8 x float> *%dst
  ret void
}

define void @vld3_v16f32(<48 x float> *%src, <16 x float> *%dst) {
; CHECK-LABEL: vld3_v16f32:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
; CHECK-NEXT:    vldrw.u32 q0, [r0, #64]
; CHECK-NEXT:    vldrw.u32 q2, [r0, #80]
; CHECK-NEXT:    vmov.f64 d6, d2
; CHECK-NEXT:    vmov.f32 s16, s5
; CHECK-NEXT:    vmov.f32 s13, s7
; CHECK-NEXT:    vmov.f32 s17, s0
; CHECK-NEXT:    vmov.f32 s14, s2
; CHECK-NEXT:    vmov.f32 s18, s3
; CHECK-NEXT:    vmov.f32 s0, s6
; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
; CHECK-NEXT:    vmov.f32 s2, s8
; CHECK-NEXT:    vmov.f32 s19, s10
; CHECK-NEXT:    vmov.f32 s15, s9
; CHECK-NEXT:    vmov.f32 s3, s11
; CHECK-NEXT:    vadd.f32 q3, q3, q4
; CHECK-NEXT:    vadd.f32 q0, q3, q0
; CHECK-NEXT:    vldrw.u32 q3, [r0]
; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
; CHECK-NEXT:    vmov.f32 s16, s13
; CHECK-NEXT:    vmov.f64 d10, d6
; CHECK-NEXT:    vmov.f32 s17, s4
; CHECK-NEXT:    vmov.f32 s21, s15
; CHECK-NEXT:    vmov.f32 s18, s7
; CHECK-NEXT:    vmov.f32 s22, s6
; CHECK-NEXT:    vmov.f32 s4, s14
; CHECK-NEXT:    vldrw.u32 q3, [r0, #144]
; CHECK-NEXT:    vmov.f32 s6, s8
; CHECK-NEXT:    vmov.f32 s19, s10
; CHECK-NEXT:    vmov.f32 s23, s9
; CHECK-NEXT:    vmov.f32 s7, s11
; CHECK-NEXT:    vldrw.u32 q2, [r0, #160]
; CHECK-NEXT:    vadd.f32 q4, q5, q4
; CHECK-NEXT:    vmov.f64 d10, d6
; CHECK-NEXT:    vadd.f32 q1, q4, q1
; CHECK-NEXT:    vldrw.u32 q4, [r0, #176]
; CHECK-NEXT:    vmov.f32 s24, s13
; CHECK-NEXT:    vmov.f32 s21, s15
; CHECK-NEXT:    vmov.f32 s25, s8
; CHECK-NEXT:    vmov.f32 s22, s10
; CHECK-NEXT:    vmov.f32 s26, s11
; CHECK-NEXT:    vmov.f32 s8, s14
; CHECK-NEXT:    vldrw.u32 q3, [r0, #112]
; CHECK-NEXT:    vmov.f32 s10, s16
; CHECK-NEXT:    vmov.f32 s27, s18
; CHECK-NEXT:    vmov.f32 s23, s17
; CHECK-NEXT:    vmov.f32 s11, s19
; CHECK-NEXT:    vadd.f32 q5, q5, q6
; CHECK-NEXT:    vadd.f32 q2, q5, q2
; CHECK-NEXT:    vldrw.u32 q5, [r0, #96]
; CHECK-NEXT:    vldrw.u32 q4, [r0, #128]
; CHECK-NEXT:    vstrw.32 q2, [r1, #48]
; CHECK-NEXT:    vmov.f32 s24, s21
; CHECK-NEXT:    vstrw.32 q0, [r1, #16]
; CHECK-NEXT:    vmov.f64 d14, d10
; CHECK-NEXT:    vstrw.32 q1, [r1]
; CHECK-NEXT:    vmov.f32 s25, s12
; CHECK-NEXT:    vmov.f32 s29, s23
; CHECK-NEXT:    vmov.f32 s26, s15
; CHECK-NEXT:    vmov.f32 s30, s14
; CHECK-NEXT:    vmov.f32 s12, s22
; CHECK-NEXT:    vmov.f32 s14, s16
; CHECK-NEXT:    vmov.f32 s27, s18
; CHECK-NEXT:    vmov.f32 s31, s17
; CHECK-NEXT:    vadd.f32 q6, q7, q6
; CHECK-NEXT:    vmov.f32 s15, s19
; CHECK-NEXT:    vadd.f32 q3, q6, q3
; CHECK-NEXT:    vstrw.32 q3, [r1, #32]
; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <48 x float>, <48 x float>* %src, align 4
  %s1 = shufflevector <48 x float> %l1, <48 x float> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
  %s2 = shufflevector <48 x float> %l1, <48 x float> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
  %s3 = shufflevector <48 x float> %l1, <48 x float> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
  %a1 = fadd <16 x float> %s1, %s2
  %a = fadd <16 x float> %a1, %s3
  store <16 x float> %a, <16 x float> *%dst
  ret void
}

; f16

define void @vld3_v2f16(<6 x half> *%src, <2 x half> *%dst) {
; CHECK-LABEL: vld3_v2f16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    ldrd r2, r3, [r0]
; CHECK-NEXT:    ldr r0, [r0, #8]
; CHECK-NEXT:    vmov.32 q0[0], r2
; CHECK-NEXT:    vmov.32 q0[1], r3
; CHECK-NEXT:    vmov.32 q0[2], r0
; CHECK-NEXT:    vmovx.f16 s8, s0
; CHECK-NEXT:    vmovx.f16 s4, s2
; CHECK-NEXT:    vins.f16 s8, s2
; CHECK-NEXT:    vmovx.f16 s6, s1
; CHECK-NEXT:    vins.f16 s1, s4
; CHECK-NEXT:    vins.f16 s0, s6
; CHECK-NEXT:    vadd.f16 q1, q0, q2
; CHECK-NEXT:    vmov.f32 s0, s1
; CHECK-NEXT:    vadd.f16 q0, q1, q0
; CHECK-NEXT:    vmov r0, s0
; CHECK-NEXT:    str r0, [r1]
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <6 x half>, <6 x half>* %src, align 4
  %s1 = shufflevector <6 x half> %l1, <6 x half> undef, <2 x i32> <i32 0, i32 3>
  %s2 = shufflevector <6 x half> %l1, <6 x half> undef, <2 x i32> <i32 1, i32 4>
  %s3 = shufflevector <6 x half> %l1, <6 x half> undef, <2 x i32> <i32 2, i32 5>
  %a1 = fadd <2 x half> %s1, %s2
  %a = fadd <2 x half> %a1, %s3
  store <2 x half> %a, <2 x half> *%dst
  ret void
}

define void @vld3_v4f16(<12 x half> *%src, <4 x half> *%dst) {
; CHECK-LABEL: vld3_v4f16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .vsave {d8, d9}
; CHECK-NEXT:    vpush {d8, d9}
; CHECK-NEXT:    ldrd r2, r3, [r0, #16]
; CHECK-NEXT:    vmov.32 q2[0], r2
; CHECK-NEXT:    vmov.32 q2[1], r3
; CHECK-NEXT:    vmov.f32 s1, s8
; CHECK-NEXT:    vmovx.f16 s4, s9
; CHECK-NEXT:    vins.f16 s1, s4
; CHECK-NEXT:    vldrw.u32 q1, [r0]
; CHECK-NEXT:    vmovx.f16 s8, s8
; CHECK-NEXT:    vmovx.f16 s12, s4
; CHECK-NEXT:    vmovx.f16 s16, s6
; CHECK-NEXT:    vins.f16 s12, s6
; CHECK-NEXT:    vmovx.f16 s18, s5
; CHECK-NEXT:    vins.f16 s5, s16
; CHECK-NEXT:    vins.f16 s4, s18
; CHECK-NEXT:    vmovx.f16 s13, s7
; CHECK-NEXT:    vins.f16 s7, s8
; CHECK-NEXT:    vmov.f32 s0, s5
; CHECK-NEXT:    vins.f16 s13, s9
; CHECK-NEXT:    vmov.f32 s5, s7
; CHECK-NEXT:    vadd.f16 q1, q1, q3
; CHECK-NEXT:    vadd.f16 q0, q1, q0
; CHECK-NEXT:    vmov r2, s1
; CHECK-NEXT:    vmov r0, s0
; CHECK-NEXT:    strd r0, r2, [r1]
; CHECK-NEXT:    vpop {d8, d9}
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <12 x half>, <12 x half>* %src, align 4
  %s1 = shufflevector <12 x half> %l1, <12 x half> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
  %s2 = shufflevector <12 x half> %l1, <12 x half> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
  %s3 = shufflevector <12 x half> %l1, <12 x half> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
  %a1 = fadd <4 x half> %s1, %s2
  %a = fadd <4 x half> %a1, %s3
  store <4 x half> %a, <4 x half> *%dst
  ret void
}

define void @vld3_v8f16(<24 x half> *%src, <8 x half> *%dst) {
; CHECK-LABEL: vld3_v8f16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT:    vldrw.u32 q0, [r0]
; CHECK-NEXT:    vldrw.u32 q4, [r0, #16]
; CHECK-NEXT:    vmovx.f16 s8, s2
; CHECK-NEXT:    vmov.f32 s4, s1
; CHECK-NEXT:    vins.f16 s4, s8
; CHECK-NEXT:    vmovx.f16 s8, s17
; CHECK-NEXT:    vmov.f32 s5, s16
; CHECK-NEXT:    vmovx.f16 s24, s1
; CHECK-NEXT:    vins.f16 s5, s8
; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
; CHECK-NEXT:    vmov.f32 s6, s19
; CHECK-NEXT:    vmovx.f16 s26, s16
; CHECK-NEXT:    vmovx.f16 s20, s11
; CHECK-NEXT:    vmov.f32 s15, s10
; CHECK-NEXT:    vins.f16 s15, s20
; CHECK-NEXT:    vmovx.f16 s20, s8
; CHECK-NEXT:    vins.f16 s6, s20
; CHECK-NEXT:    vmovx.f16 s20, s19
; CHECK-NEXT:    vmov.f32 s28, s18
; CHECK-NEXT:    vmovx.f16 s30, s10
; CHECK-NEXT:    vins.f16 s28, s20
; CHECK-NEXT:    vmovx.f16 s20, s0
; CHECK-NEXT:    vins.f16 s0, s24
; CHECK-NEXT:    vins.f16 s20, s2
; CHECK-NEXT:    vmovx.f16 s21, s3
; CHECK-NEXT:    vins.f16 s3, s26
; CHECK-NEXT:    vins.f16 s21, s17
; CHECK-NEXT:    vmov.f32 s14, s8
; CHECK-NEXT:    vmovx.f16 s23, s9
; CHECK-NEXT:    vmov.f32 s1, s3
; CHECK-NEXT:    vins.f16 s9, s30
; CHECK-NEXT:    vins.f16 s23, s11
; CHECK-NEXT:    vmovx.f16 s22, s18
; CHECK-NEXT:    vmov.f32 s2, s28
; CHECK-NEXT:    vins.f16 s22, s8
; CHECK-NEXT:    vmov.f32 s3, s9
; CHECK-NEXT:    vmov.f32 s7, s15
; CHECK-NEXT:    vadd.f16 q0, q0, q5
; CHECK-NEXT:    vadd.f16 q0, q0, q1
; CHECK-NEXT:    vstrw.32 q0, [r1]
; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <24 x half>, <24 x half>* %src, align 4
  %s1 = shufflevector <24 x half> %l1, <24 x half> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
  %s2 = shufflevector <24 x half> %l1, <24 x half> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
  %s3 = shufflevector <24 x half> %l1, <24 x half> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
  %a1 = fadd <8 x half> %s1, %s2
  %a = fadd <8 x half> %a1, %s3
  store <8 x half> %a, <8 x half> *%dst
  ret void
}

define void @vld3_v16f16(<48 x half> *%src, <16 x half> *%dst) {
; CHECK-LABEL: vld3_v16f16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT:    vldrw.u32 q0, [r0, #48]
; CHECK-NEXT:    vldrw.u32 q4, [r0, #64]
; CHECK-NEXT:    vmovx.f16 s8, s2
; CHECK-NEXT:    vmov.f32 s4, s1
; CHECK-NEXT:    vins.f16 s4, s8
; CHECK-NEXT:    vmovx.f16 s8, s17
; CHECK-NEXT:    vmov.f32 s5, s16
; CHECK-NEXT:    vmovx.f16 s24, s1
; CHECK-NEXT:    vins.f16 s5, s8
; CHECK-NEXT:    vldrw.u32 q2, [r0, #80]
; CHECK-NEXT:    vmov.f32 s6, s19
; CHECK-NEXT:    vmovx.f16 s26, s16
; CHECK-NEXT:    vmovx.f16 s20, s11
; CHECK-NEXT:    vmov.f32 s15, s10
; CHECK-NEXT:    vins.f16 s15, s20
; CHECK-NEXT:    vmovx.f16 s20, s8
; CHECK-NEXT:    vins.f16 s6, s20
; CHECK-NEXT:    vmovx.f16 s20, s19
; CHECK-NEXT:    vmov.f32 s28, s18
; CHECK-NEXT:    vmovx.f16 s30, s10
; CHECK-NEXT:    vins.f16 s28, s20
; CHECK-NEXT:    vmovx.f16 s20, s0
; CHECK-NEXT:    vins.f16 s0, s24
; CHECK-NEXT:    vins.f16 s20, s2
; CHECK-NEXT:    vmovx.f16 s21, s3
; CHECK-NEXT:    vins.f16 s3, s26
; CHECK-NEXT:    vins.f16 s21, s17
; CHECK-NEXT:    vmov.f32 s14, s8
; CHECK-NEXT:    vmovx.f16 s23, s9
; CHECK-NEXT:    vmov.f32 s1, s3
; CHECK-NEXT:    vins.f16 s9, s30
; CHECK-NEXT:    vins.f16 s23, s11
; CHECK-NEXT:    vmovx.f16 s22, s18
; CHECK-NEXT:    vmov.f32 s2, s28
; CHECK-NEXT:    vins.f16 s22, s8
; CHECK-NEXT:    vmov.f32 s3, s9
; CHECK-NEXT:    vmov.f32 s7, s15
; CHECK-NEXT:    vadd.f16 q0, q0, q5
; CHECK-NEXT:    vadd.f16 q1, q0, q1
; CHECK-NEXT:    vldrw.u32 q0, [r0]
; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
; CHECK-NEXT:    vldrw.u32 q3, [r0, #16]
; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
; CHECK-NEXT:    vmovx.f16 s16, s2
; CHECK-NEXT:    vmov.f32 s4, s1
; CHECK-NEXT:    vmovx.f16 s20, s11
; CHECK-NEXT:    vins.f16 s4, s16
; CHECK-NEXT:    vmovx.f16 s16, s13
; CHECK-NEXT:    vmov.f32 s5, s12
; CHECK-NEXT:    vmovx.f16 s24, s1
; CHECK-NEXT:    vins.f16 s5, s16
; CHECK-NEXT:    vmov.f32 s19, s10
; CHECK-NEXT:    vins.f16 s19, s20
; CHECK-NEXT:    vmov.f32 s6, s15
; CHECK-NEXT:    vmovx.f16 s20, s8
; CHECK-NEXT:    vmov.f32 s28, s14
; CHECK-NEXT:    vins.f16 s6, s20
; CHECK-NEXT:    vmovx.f16 s20, s15
; CHECK-NEXT:    vins.f16 s28, s20
; CHECK-NEXT:    vmovx.f16 s20, s0
; CHECK-NEXT:    vins.f16 s0, s24
; CHECK-NEXT:    vins.f16 s20, s2
; CHECK-NEXT:    vmovx.f16 s21, s3
; CHECK-NEXT:    vmovx.f16 s26, s12
; CHECK-NEXT:    vins.f16 s21, s13
; CHECK-NEXT:    vins.f16 s3, s26
; CHECK-NEXT:    vmovx.f16 s30, s10
; CHECK-NEXT:    vmovx.f16 s23, s9
; CHECK-NEXT:    vmov.f32 s18, s8
; CHECK-NEXT:    vins.f16 s9, s30
; CHECK-NEXT:    vins.f16 s23, s11
; CHECK-NEXT:    vmov.f32 s1, s3
; CHECK-NEXT:    vmovx.f16 s22, s14
; CHECK-NEXT:    vmov.f32 s2, s28
; CHECK-NEXT:    vins.f16 s22, s8
; CHECK-NEXT:    vmov.f32 s3, s9
; CHECK-NEXT:    vmov.f32 s7, s19
; CHECK-NEXT:    vadd.f16 q0, q0, q5
; CHECK-NEXT:    vadd.f16 q0, q0, q1
; CHECK-NEXT:    vstrw.32 q0, [r1]
; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <48 x half>, <48 x half>* %src, align 4
  %s1 = shufflevector <48 x half> %l1, <48 x half> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
  %s2 = shufflevector <48 x half> %l1, <48 x half> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
  %s3 = shufflevector <48 x half> %l1, <48 x half> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
  %a1 = fadd <16 x half> %s1, %s2
  %a = fadd <16 x half> %a1, %s3
  store <16 x half> %a, <16 x half> *%dst
  ret void
}

; f64

define void @vld3_v2f64(<6 x double> *%src, <2 x double> *%dst) {
; CHECK-LABEL: vld3_v2f64:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrw.u32 q0, [r0, #32]
; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
; CHECK-NEXT:    vldrw.u32 q3, [r0]
; CHECK-NEXT:    vadd.f64 d4, d3, d0
; CHECK-NEXT:    vadd.f64 d5, d6, d7
; CHECK-NEXT:    vadd.f64 d1, d4, d1
; CHECK-NEXT:    vadd.f64 d0, d5, d2
; CHECK-NEXT:    vstrw.32 q0, [r1]
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <6 x double>, <6 x double>* %src, align 4
  %s1 = shufflevector <6 x double> %l1, <6 x double> undef, <2 x i32> <i32 0, i32 3>
  %s2 = shufflevector <6 x double> %l1, <6 x double> undef, <2 x i32> <i32 1, i32 4>
  %s3 = shufflevector <6 x double> %l1, <6 x double> undef, <2 x i32> <i32 2, i32 5>
  %a1 = fadd <2 x double> %s1, %s2
  %a = fadd <2 x double> %a1, %s3
  store <2 x double> %a, <2 x double> *%dst
  ret void
}

define void @vld3_v4f64(<12 x double> *%src, <4 x double> *%dst) {
; CHECK-LABEL: vld3_v4f64:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
; CHECK-NEXT:    vldrw.u32 q3, [r0, #48]
; CHECK-NEXT:    vldrw.u32 q1, [r0, #80]
; CHECK-NEXT:    vldrw.u32 q0, [r0, #64]
; CHECK-NEXT:    vldrw.u32 q4, [r0, #16]
; CHECK-NEXT:    vadd.f64 d5, d6, d7
; CHECK-NEXT:    vldrw.u32 q3, [r0, #32]
; CHECK-NEXT:    vldrw.u32 q6, [r0]
; CHECK-NEXT:    vadd.f64 d4, d1, d2
; CHECK-NEXT:    vadd.f64 d10, d9, d6
; CHECK-NEXT:    vadd.f64 d11, d12, d13
; CHECK-NEXT:    vadd.f64 d3, d4, d3
; CHECK-NEXT:    vadd.f64 d2, d5, d0
; CHECK-NEXT:    vadd.f64 d1, d10, d7
; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
; CHECK-NEXT:    vadd.f64 d0, d11, d8
; CHECK-NEXT:    vstrw.32 q0, [r1]
; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
; CHECK-NEXT:    bx lr
entry:
  %l1 = load <12 x double>, <12 x double>* %src, align 4
  %s1 = shufflevector <12 x double> %l1, <12 x double> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
  %s2 = shufflevector <12 x double> %l1, <12 x double> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
  %s3 = shufflevector <12 x double> %l1, <12 x double> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
  %a1 = fadd <4 x double> %s1, %s2
  %a = fadd <4 x double> %a1, %s3
  store <4 x double> %a, <4 x double> *%dst
  ret void
}