; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: llc < %s -verify-machineinstrs -mtriple=aarch64-none-linux-gnu | FileCheck %s

define i8 @scalarize_v16i8(ptr %p) {
; CHECK-LABEL: scalarize_v16i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    umov w8, v0.b[0]
; CHECK-NEXT:    umov w9, v0.b[1]
; CHECK-NEXT:    umov w10, v0.b[2]
; CHECK-NEXT:    umov w11, v0.b[3]
; CHECK-NEXT:    umov w12, v0.b[4]
; CHECK-NEXT:    umov w13, v0.b[5]
; CHECK-NEXT:    umov w14, v0.b[6]
; CHECK-NEXT:    umov w15, v0.b[7]
; CHECK-NEXT:    umov w16, v0.b[8]
; CHECK-NEXT:    umov w17, v0.b[9]
; CHECK-NEXT:    umov w18, v0.b[10]
; CHECK-NEXT:    umov w0, v0.b[11]
; CHECK-NEXT:    umov w1, v0.b[12]
; CHECK-NEXT:    umov w2, v0.b[13]
; CHECK-NEXT:    add w8, w8, w9
; CHECK-NEXT:    umov w3, v0.b[14]
; CHECK-NEXT:    umov w4, v0.b[15]
; CHECK-NEXT:    add w9, w10, w11
; CHECK-NEXT:    add w10, w12, w13
; CHECK-NEXT:    add w11, w14, w15
; CHECK-NEXT:    add w8, w8, w9
; CHECK-NEXT:    add w12, w16, w17
; CHECK-NEXT:    add w13, w18, w0
; CHECK-NEXT:    add w9, w10, w11
; CHECK-NEXT:    add w14, w1, w2
; CHECK-NEXT:    add w10, w12, w13
; CHECK-NEXT:    add w8, w8, w9
; CHECK-NEXT:    add w15, w3, w4
; CHECK-NEXT:    add w11, w14, w15
; CHECK-NEXT:    add w9, w10, w11
; CHECK-NEXT:    add w0, w8, w9
; CHECK-NEXT:    ret
  %wide.load = load <16 x i8>, ptr %p, align 4
  %l0 = extractelement <16 x i8> %wide.load, i32 0
  %l1 = extractelement <16 x i8> %wide.load, i32 1
  %l2 = extractelement <16 x i8> %wide.load, i32 2
  %l3 = extractelement <16 x i8> %wide.load, i32 3
  %l4 = extractelement <16 x i8> %wide.load, i32 4
  %l5 = extractelement <16 x i8> %wide.load, i32 5
  %l6 = extractelement <16 x i8> %wide.load, i32 6
  %l7 = extractelement <16 x i8> %wide.load, i32 7
  %l8 = extractelement <16 x i8> %wide.load, i32 8
  %l9 = extractelement <16 x i8> %wide.load, i32 9
  %l10 = extractelement <16 x i8> %wide.load, i32 10
  %l11 = extractelement <16 x i8> %wide.load, i32 11
  %l12 = extractelement <16 x i8> %wide.load, i32 12
  %l13 = extractelement <16 x i8> %wide.load, i32 13
  %l14 = extractelement <16 x i8> %wide.load, i32 14
  %l15 = extractelement <16 x i8> %wide.load, i32 15
  %a0 = add i8 %l0, %l1
  %a1 = add i8 %l2, %l3
  %a2 = add i8 %l4, %l5
  %a3 = add i8 %l6, %l7
  %a4 = add i8 %l8, %l9
  %a5 = add i8 %l10, %l11
  %a6 = add i8 %l12, %l13
  %a7 = add i8 %l14, %l15
  %b0 = add i8 %a0, %a1
  %b1 = add i8 %a2, %a3
  %b2 = add i8 %a4, %a5
  %b3 = add i8 %a6, %a7
  %c0 = add i8 %b0, %b1
  %c1 = add i8 %b2, %b3
  %r = add i8 %c0, %c1
  ret i8 %r
}

define i8 @scalarize_v8i8(ptr %p) {
; CHECK-LABEL: scalarize_v8i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    umov w8, v0.b[0]
; CHECK-NEXT:    umov w9, v0.b[1]
; CHECK-NEXT:    umov w10, v0.b[2]
; CHECK-NEXT:    umov w11, v0.b[3]
; CHECK-NEXT:    umov w12, v0.b[4]
; CHECK-NEXT:    umov w13, v0.b[5]
; CHECK-NEXT:    umov w14, v0.b[6]
; CHECK-NEXT:    umov w15, v0.b[7]
; CHECK-NEXT:    add w8, w8, w9
; CHECK-NEXT:    add w9, w10, w11
; CHECK-NEXT:    add w10, w12, w13
; CHECK-NEXT:    add w11, w14, w15
; CHECK-NEXT:    add w8, w8, w9
; CHECK-NEXT:    add w9, w10, w11
; CHECK-NEXT:    add w0, w8, w9
; CHECK-NEXT:    ret
  %wide.load = load <8 x i8>, ptr %p, align 4
  %l0 = extractelement <8 x i8> %wide.load, i32 0
  %l1 = extractelement <8 x i8> %wide.load, i32 1
  %l2 = extractelement <8 x i8> %wide.load, i32 2
  %l3 = extractelement <8 x i8> %wide.load, i32 3
  %l4 = extractelement <8 x i8> %wide.load, i32 4
  %l5 = extractelement <8 x i8> %wide.load, i32 5
  %l6 = extractelement <8 x i8> %wide.load, i32 6
  %l7 = extractelement <8 x i8> %wide.load, i32 7
  %a0 = add i8 %l0, %l1
  %a1 = add i8 %l2, %l3
  %a2 = add i8 %l4, %l5
  %a3 = add i8 %l6, %l7
  %b0 = add i8 %a0, %a1
  %b1 = add i8 %a2, %a3
  %r = add i8 %b0, %b1
  ret i8 %r
}

define i16 @scalarize_v8i16(ptr %p) {
; CHECK-LABEL: scalarize_v8i16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    umov w8, v0.h[0]
; CHECK-NEXT:    umov w9, v0.h[1]
; CHECK-NEXT:    umov w10, v0.h[2]
; CHECK-NEXT:    umov w11, v0.h[3]
; CHECK-NEXT:    umov w12, v0.h[4]
; CHECK-NEXT:    umov w13, v0.h[5]
; CHECK-NEXT:    umov w14, v0.h[6]
; CHECK-NEXT:    umov w15, v0.h[7]
; CHECK-NEXT:    add w8, w8, w9
; CHECK-NEXT:    add w9, w10, w11
; CHECK-NEXT:    add w10, w12, w13
; CHECK-NEXT:    add w11, w14, w15
; CHECK-NEXT:    add w8, w8, w9
; CHECK-NEXT:    add w9, w10, w11
; CHECK-NEXT:    add w0, w8, w9
; CHECK-NEXT:    ret
  %wide.load = load <8 x i16>, ptr %p, align 4
  %l0 = extractelement <8 x i16> %wide.load, i32 0
  %l1 = extractelement <8 x i16> %wide.load, i32 1
  %l2 = extractelement <8 x i16> %wide.load, i32 2
  %l3 = extractelement <8 x i16> %wide.load, i32 3
  %l4 = extractelement <8 x i16> %wide.load, i32 4
  %l5 = extractelement <8 x i16> %wide.load, i32 5
  %l6 = extractelement <8 x i16> %wide.load, i32 6
  %l7 = extractelement <8 x i16> %wide.load, i32 7
  %a0 = add i16 %l0, %l1
  %a1 = add i16 %l2, %l3
  %a2 = add i16 %l4, %l5
  %a3 = add i16 %l6, %l7
  %b0 = add i16 %a0, %a1
  %b1 = add i16 %a2, %a3
  %r = add i16 %b0, %b1
  ret i16 %r
}

define i16 @scalarize_v4i16(ptr %p) {
; CHECK-LABEL: scalarize_v4i16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    umov w8, v0.h[0]
; CHECK-NEXT:    umov w9, v0.h[1]
; CHECK-NEXT:    umov w10, v0.h[2]
; CHECK-NEXT:    umov w11, v0.h[3]
; CHECK-NEXT:    add w8, w8, w9
; CHECK-NEXT:    add w9, w10, w11
; CHECK-NEXT:    add w0, w8, w9
; CHECK-NEXT:    ret
  %wide.load = load <4 x i16>, ptr %p, align 4
  %l0 = extractelement <4 x i16> %wide.load, i32 0
  %l1 = extractelement <4 x i16> %wide.load, i32 1
  %l2 = extractelement <4 x i16> %wide.load, i32 2
  %l3 = extractelement <4 x i16> %wide.load, i32 3
  %a0 = add i16 %l0, %l1
  %a1 = add i16 %l2, %l3
  %r = add i16 %a0, %a1
  ret i16 %r
}

define i32 @scalarize_v4i32(ptr %p) {
; CHECK-LABEL: scalarize_v4i32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    mov w8, v0.s[1]
; CHECK-NEXT:    mov w9, v0.s[2]
; CHECK-NEXT:    mov w10, v0.s[3]
; CHECK-NEXT:    fmov w11, s0
; CHECK-NEXT:    add w8, w11, w8
; CHECK-NEXT:    add w9, w9, w10
; CHECK-NEXT:    add w0, w8, w9
; CHECK-NEXT:    ret
  %wide.load = load <4 x i32>, ptr %p, align 4
  %l0 = extractelement <4 x i32> %wide.load, i32 0
  %l1 = extractelement <4 x i32> %wide.load, i32 1
  %l2 = extractelement <4 x i32> %wide.load, i32 2
  %l3 = extractelement <4 x i32> %wide.load, i32 3
  %a0 = add i32 %l0, %l1
  %a1 = add i32 %l2, %l3
  %r = add i32 %a0, %a1
  ret i32 %r
}

define i64 @scalarize_v4i64(ptr %p) {
; CHECK-LABEL: scalarize_v4i64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldp q1, q0, [x0]
; CHECK-NEXT:    addp d1, v1.2d
; CHECK-NEXT:    addp d0, v0.2d
; CHECK-NEXT:    fmov x8, d1
; CHECK-NEXT:    fmov x9, d0
; CHECK-NEXT:    add x0, x8, x9
; CHECK-NEXT:    ret
  %wide.load = load <4 x i64>, ptr %p, align 4
  %l0 = extractelement <4 x i64> %wide.load, i32 0
  %l1 = extractelement <4 x i64> %wide.load, i32 1
  %l2 = extractelement <4 x i64> %wide.load, i32 2
  %l3 = extractelement <4 x i64> %wide.load, i32 3
  %a0 = add i64 %l0, %l1
  %a1 = add i64 %l2, %l3
  %r = add i64 %a0, %a1
  ret i64 %r
}

define i64 @scalarize_v4i32_sext(ptr %p) {
; CHECK-LABEL: scalarize_v4i32_sext:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    sshll2 v1.2d, v0.4s, #0
; CHECK-NEXT:    sshll v0.2d, v0.2s, #0
; CHECK-NEXT:    addp d0, v0.2d
; CHECK-NEXT:    addp d1, v1.2d
; CHECK-NEXT:    fmov x8, d0
; CHECK-NEXT:    fmov x9, d1
; CHECK-NEXT:    add x0, x8, x9
; CHECK-NEXT:    ret
  %wide.load = load <4 x i32>, ptr %p, align 4
  %ext = sext <4 x i32> %wide.load to <4 x i64>
  %l0 = extractelement <4 x i64> %ext, i32 0
  %l1 = extractelement <4 x i64> %ext, i32 1
  %l2 = extractelement <4 x i64> %ext, i32 2
  %l3 = extractelement <4 x i64> %ext, i32 3
  %a0 = add i64 %l0, %l1
  %a1 = add i64 %l2, %l3
  %r = add i64 %a0, %a1
  ret i64 %r
}

define i64 @scalarize_v4i32_zext(ptr %p) {
; CHECK-LABEL: scalarize_v4i32_zext:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    ushll2 v1.2d, v0.4s, #0
; CHECK-NEXT:    ushll v0.2d, v0.2s, #0
; CHECK-NEXT:    addp d0, v0.2d
; CHECK-NEXT:    addp d1, v1.2d
; CHECK-NEXT:    fmov x8, d0
; CHECK-NEXT:    fmov x9, d1
; CHECK-NEXT:    add x0, x8, x9
; CHECK-NEXT:    ret
  %wide.load = load <4 x i32>, ptr %p, align 4
  %ext = zext <4 x i32> %wide.load to <4 x i64>
  %l0 = extractelement <4 x i64> %ext, i32 0
  %l1 = extractelement <4 x i64> %ext, i32 1
  %l2 = extractelement <4 x i64> %ext, i32 2
  %l3 = extractelement <4 x i64> %ext, i32 3
  %a0 = add i64 %l0, %l1
  %a1 = add i64 %l2, %l3
  %r = add i64 %a0, %a1
  ret i64 %r
}

define half @scalarize_v4f16(ptr %p) {
; CHECK-LABEL: scalarize_v4f16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    mov h1, v0.h[1]
; CHECK-NEXT:    mov h2, v0.h[2]
; CHECK-NEXT:    mov h3, v0.h[3]
; CHECK-NEXT:    fcvt s0, h0
; CHECK-NEXT:    fcvt s1, h1
; CHECK-NEXT:    fcvt s3, h3
; CHECK-NEXT:    fcvt s2, h2
; CHECK-NEXT:    fadd s0, s0, s1
; CHECK-NEXT:    fadd s1, s2, s3
; CHECK-NEXT:    fcvt h0, s0
; CHECK-NEXT:    fcvt h1, s1
; CHECK-NEXT:    fcvt s1, h1
; CHECK-NEXT:    fcvt s0, h0
; CHECK-NEXT:    fadd s0, s0, s1
; CHECK-NEXT:    fcvt h0, s0
; CHECK-NEXT:    ret
  %wide.load = load <4 x half>, ptr %p, align 4
  %l0 = extractelement <4 x half> %wide.load, i32 0
  %l1 = extractelement <4 x half> %wide.load, i32 1
  %l2 = extractelement <4 x half> %wide.load, i32 2
  %l3 = extractelement <4 x half> %wide.load, i32 3
  %a0 = fadd half %l0, %l1
  %a1 = fadd half %l2, %l3
  %r = fadd half %a0, %a1
  ret half %r
}

define float @scalarize_v4f32(ptr %p) {
; CHECK-LABEL: scalarize_v4f32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    mov s1, v0.s[2]
; CHECK-NEXT:    mov s2, v0.s[3]
; CHECK-NEXT:    faddp s0, v0.2s
; CHECK-NEXT:    fadd s1, s1, s2
; CHECK-NEXT:    fadd s0, s0, s1
; CHECK-NEXT:    ret
  %wide.load = load <4 x float>, ptr %p, align 4
  %l0 = extractelement <4 x float> %wide.load, i32 0
  %l1 = extractelement <4 x float> %wide.load, i32 1
  %l2 = extractelement <4 x float> %wide.load, i32 2
  %l3 = extractelement <4 x float> %wide.load, i32 3
  %a0 = fadd float %l0, %l1
  %a1 = fadd float %l2, %l3
  %r = fadd float %a0, %a1
  ret float %r
}

define double @scalarize_v4f64(ptr %p) {
; CHECK-LABEL: scalarize_v4f64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldp q1, q0, [x0]
; CHECK-NEXT:    faddp d1, v1.2d
; CHECK-NEXT:    faddp d0, v0.2d
; CHECK-NEXT:    fadd d0, d1, d0
; CHECK-NEXT:    ret
  %wide.load = load <4 x double>, ptr %p, align 4
  %l0 = extractelement <4 x double> %wide.load, i32 0
  %l1 = extractelement <4 x double> %wide.load, i32 1
  %l2 = extractelement <4 x double> %wide.load, i32 2
  %l3 = extractelement <4 x double> %wide.load, i32 3
  %a0 = fadd double %l0, %l1
  %a1 = fadd double %l2, %l3
  %r = fadd double %a0, %a1
  ret double %r
}


define float @scalarize_into_load(i64 %22, ptr %23, ptr %rawA, ptr %rawB) {
; CHECK-LABEL: scalarize_into_load:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ldp q1, q0, [x1]
; CHECK-NEXT:    ldp q3, q2, [x1, #96]
; CHECK-NEXT:    ldp q5, q4, [x1, #64]
; CHECK-NEXT:    ldp q7, q6, [x1, #32]
; CHECK-NEXT:    mov x8, v1.d[1]
; CHECK-NEXT:    mov x10, v0.d[1]
; CHECK-NEXT:    mov x1, v3.d[1]
; CHECK-NEXT:    mov x4, v2.d[1]
; CHECK-NEXT:    mov x16, v5.d[1]
; CHECK-NEXT:    mov x18, v4.d[1]
; CHECK-NEXT:    fmov x9, d1
; CHECK-NEXT:    mov x12, v7.d[1]
; CHECK-NEXT:    mov x14, v6.d[1]
; CHECK-NEXT:    fmov x11, d0
; CHECK-NEXT:    fmov x13, d7
; CHECK-NEXT:    fmov x15, d6
; CHECK-NEXT:    fmov x17, d5
; CHECK-NEXT:    fmov x0, d4
; CHECK-NEXT:    fmov x3, d3
; CHECK-NEXT:    fmov x5, d2
; CHECK-NEXT:    ldr s0, [x2, x9, lsl #2]
; CHECK-NEXT:    ldr s1, [x2, x8, lsl #2]
; CHECK-NEXT:    ldr s2, [x2, x11, lsl #2]
; CHECK-NEXT:    ldr s3, [x2, x10, lsl #2]
; CHECK-NEXT:    ldr s4, [x2, x13, lsl #2]
; CHECK-NEXT:    ldr s5, [x2, x12, lsl #2]
; CHECK-NEXT:    ldr s6, [x2, x15, lsl #2]
; CHECK-NEXT:    ldr s7, [x2, x14, lsl #2]
; CHECK-NEXT:    ldr s16, [x2, x17, lsl #2]
; CHECK-NEXT:    ldr s17, [x2, x16, lsl #2]
; CHECK-NEXT:    ldr s18, [x2, x0, lsl #2]
; CHECK-NEXT:    ldr s19, [x2, x18, lsl #2]
; CHECK-NEXT:    ldr s20, [x2, x3, lsl #2]
; CHECK-NEXT:    ldr s21, [x2, x1, lsl #2]
; CHECK-NEXT:    ldr s22, [x2, x5, lsl #2]
; CHECK-NEXT:    ldr s23, [x2, x4, lsl #2]
; CHECK-NEXT:    fadd s0, s0, s1
; CHECK-NEXT:    fadd s1, s2, s3
; CHECK-NEXT:    fadd s2, s4, s5
; CHECK-NEXT:    fadd s3, s6, s7
; CHECK-NEXT:    fadd s4, s16, s17
; CHECK-NEXT:    fadd s5, s18, s19
; CHECK-NEXT:    fadd s6, s20, s21
; CHECK-NEXT:    fadd s7, s22, s23
; CHECK-NEXT:    fadd s0, s0, s1
; CHECK-NEXT:    fadd s1, s2, s3
; CHECK-NEXT:    fadd s2, s4, s5
; CHECK-NEXT:    fadd s3, s6, s7
; CHECK-NEXT:    fadd s0, s0, s1
; CHECK-NEXT:    fadd s1, s2, s3
; CHECK-NEXT:    fadd s0, s0, s1
; CHECK-NEXT:    ret
entry:
  %wide.load = load <16 x i64>, ptr %23, align 4
  %25 = extractelement <16 x i64> %wide.load, i32 0
  %26 = getelementptr inbounds float, ptr %rawA, i64 %25
  %27 = extractelement <16 x i64> %wide.load, i32 1
  %28 = getelementptr inbounds float, ptr %rawA, i64 %27
  %29 = extractelement <16 x i64> %wide.load, i32 2
  %30 = getelementptr inbounds float, ptr %rawA, i64 %29
  %31 = extractelement <16 x i64> %wide.load, i32 3
  %32 = getelementptr inbounds float, ptr %rawA, i64 %31
  %33 = extractelement <16 x i64> %wide.load, i32 4
  %34 = getelementptr inbounds float, ptr %rawA, i64 %33
  %35 = extractelement <16 x i64> %wide.load, i32 5
  %36 = getelementptr inbounds float, ptr %rawA, i64 %35
  %37 = extractelement <16 x i64> %wide.load, i32 6
  %38 = getelementptr inbounds float, ptr %rawA, i64 %37
  %39 = extractelement <16 x i64> %wide.load, i32 7
  %40 = getelementptr inbounds float, ptr %rawA, i64 %39
  %41 = extractelement <16 x i64> %wide.load, i32 8
  %42 = getelementptr inbounds float, ptr %rawA, i64 %41
  %43 = extractelement <16 x i64> %wide.load, i32 9
  %44 = getelementptr inbounds float, ptr %rawA, i64 %43
  %45 = extractelement <16 x i64> %wide.load, i32 10
  %46 = getelementptr inbounds float, ptr %rawA, i64 %45
  %47 = extractelement <16 x i64> %wide.load, i32 11
  %48 = getelementptr inbounds float, ptr %rawA, i64 %47
  %49 = extractelement <16 x i64> %wide.load, i32 12
  %50 = getelementptr inbounds float, ptr %rawA, i64 %49
  %51 = extractelement <16 x i64> %wide.load, i32 13
  %52 = getelementptr inbounds float, ptr %rawA, i64 %51
  %53 = extractelement <16 x i64> %wide.load, i32 14
  %54 = getelementptr inbounds float, ptr %rawA, i64 %53
  %55 = extractelement <16 x i64> %wide.load, i32 15
  %56 = getelementptr inbounds float, ptr %rawA, i64 %55
  %59 = load float, ptr %26, align 4
  %60 = load float, ptr %28, align 4
  %61 = load float, ptr %30, align 4
  %62 = load float, ptr %32, align 4
  %63 = load float, ptr %34, align 4
  %64 = load float, ptr %36, align 4
  %65 = load float, ptr %38, align 4
  %66 = load float, ptr %40, align 4
  %67 = load float, ptr %42, align 4
  %68 = load float, ptr %44, align 4
  %69 = load float, ptr %46, align 4
  %70 = load float, ptr %48, align 4
  %71 = load float, ptr %50, align 4
  %72 = load float, ptr %52, align 4
  %73 = load float, ptr %54, align 4
  %74 = load float, ptr %56, align 4
  %a1 = fadd float %59, %60
  %a2 = fadd float %61, %62
  %a3 = fadd float %63, %64
  %a4 = fadd float %65, %66
  %a5 = fadd float %67, %68
  %a6 = fadd float %69, %70
  %a7 = fadd float %71, %72
  %a8 = fadd float %73, %74
  %a9 = fadd float %a1, %a2
  %a10 = fadd float %a3, %a4
  %a11 = fadd float %a5, %a6
  %a12 = fadd float %a7, %a8
  %a13 = fadd float %a9, %a10
  %a14 = fadd float %a11, %a12
  %a15 = fadd float %a13, %a14
  ret float %a15
}

define float @scalarize_into_load_sext(i64 %22, ptr %23, ptr %rawA, ptr %rawB) {
; CHECK-LABEL: scalarize_into_load_sext:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ldp q0, q2, [x1]
; CHECK-NEXT:    ldp q4, q1, [x1, #32]
; CHECK-NEXT:    sshll v3.2d, v0.2s, #0
; CHECK-NEXT:    sshll2 v0.2d, v0.4s, #0
; CHECK-NEXT:    sshll2 v6.2d, v2.4s, #0
; CHECK-NEXT:    sshll2 v5.2d, v1.4s, #0
; CHECK-NEXT:    sshll v1.2d, v1.2s, #0
; CHECK-NEXT:    sshll v2.2d, v2.2s, #0
; CHECK-NEXT:    sshll2 v7.2d, v4.4s, #0
; CHECK-NEXT:    sshll v4.2d, v4.2s, #0
; CHECK-NEXT:    mov x8, v3.d[1]
; CHECK-NEXT:    mov x10, v0.d[1]
; CHECK-NEXT:    mov x14, v6.d[1]
; CHECK-NEXT:    mov x12, v2.d[1]
; CHECK-NEXT:    mov x1, v1.d[1]
; CHECK-NEXT:    mov x4, v5.d[1]
; CHECK-NEXT:    mov x16, v4.d[1]
; CHECK-NEXT:    mov x18, v7.d[1]
; CHECK-NEXT:    fmov x9, d3
; CHECK-NEXT:    fmov x11, d0
; CHECK-NEXT:    fmov x13, d2
; CHECK-NEXT:    fmov x15, d6
; CHECK-NEXT:    fmov x17, d4
; CHECK-NEXT:    fmov x0, d7
; CHECK-NEXT:    ldr s2, [x2, x8, lsl #2]
; CHECK-NEXT:    fmov x3, d1
; CHECK-NEXT:    fmov x5, d5
; CHECK-NEXT:    ldr s0, [x2, x9, lsl #2]
; CHECK-NEXT:    ldr s1, [x2, x11, lsl #2]
; CHECK-NEXT:    ldr s3, [x2, x10, lsl #2]
; CHECK-NEXT:    ldr s4, [x2, x13, lsl #2]
; CHECK-NEXT:    ldr s5, [x2, x12, lsl #2]
; CHECK-NEXT:    ldr s6, [x2, x15, lsl #2]
; CHECK-NEXT:    ldr s7, [x2, x14, lsl #2]
; CHECK-NEXT:    ldr s16, [x2, x17, lsl #2]
; CHECK-NEXT:    ldr s17, [x2, x16, lsl #2]
; CHECK-NEXT:    ldr s18, [x2, x0, lsl #2]
; CHECK-NEXT:    ldr s19, [x2, x18, lsl #2]
; CHECK-NEXT:    ldr s20, [x2, x3, lsl #2]
; CHECK-NEXT:    ldr s21, [x2, x1, lsl #2]
; CHECK-NEXT:    ldr s22, [x2, x5, lsl #2]
; CHECK-NEXT:    ldr s23, [x2, x4, lsl #2]
; CHECK-NEXT:    fadd s0, s0, s2
; CHECK-NEXT:    fadd s1, s1, s3
; CHECK-NEXT:    fadd s2, s4, s5
; CHECK-NEXT:    fadd s3, s6, s7
; CHECK-NEXT:    fadd s4, s16, s17
; CHECK-NEXT:    fadd s5, s18, s19
; CHECK-NEXT:    fadd s6, s20, s21
; CHECK-NEXT:    fadd s7, s22, s23
; CHECK-NEXT:    fadd s0, s0, s1
; CHECK-NEXT:    fadd s1, s2, s3
; CHECK-NEXT:    fadd s2, s4, s5
; CHECK-NEXT:    fadd s3, s6, s7
; CHECK-NEXT:    fadd s0, s0, s1
; CHECK-NEXT:    fadd s1, s2, s3
; CHECK-NEXT:    fadd s0, s0, s1
; CHECK-NEXT:    ret
entry:
  %wide.load = load <16 x i32>, ptr %23, align 4
  %24 = sext <16 x i32> %wide.load to <16 x i64>
  %25 = extractelement <16 x i64> %24, i32 0
  %26 = getelementptr inbounds float, ptr %rawA, i64 %25
  %27 = extractelement <16 x i64> %24, i32 1
  %28 = getelementptr inbounds float, ptr %rawA, i64 %27
  %29 = extractelement <16 x i64> %24, i32 2
  %30 = getelementptr inbounds float, ptr %rawA, i64 %29
  %31 = extractelement <16 x i64> %24, i32 3
  %32 = getelementptr inbounds float, ptr %rawA, i64 %31
  %33 = extractelement <16 x i64> %24, i32 4
  %34 = getelementptr inbounds float, ptr %rawA, i64 %33
  %35 = extractelement <16 x i64> %24, i32 5
  %36 = getelementptr inbounds float, ptr %rawA, i64 %35
  %37 = extractelement <16 x i64> %24, i32 6
  %38 = getelementptr inbounds float, ptr %rawA, i64 %37
  %39 = extractelement <16 x i64> %24, i32 7
  %40 = getelementptr inbounds float, ptr %rawA, i64 %39
  %41 = extractelement <16 x i64> %24, i32 8
  %42 = getelementptr inbounds float, ptr %rawA, i64 %41
  %43 = extractelement <16 x i64> %24, i32 9
  %44 = getelementptr inbounds float, ptr %rawA, i64 %43
  %45 = extractelement <16 x i64> %24, i32 10
  %46 = getelementptr inbounds float, ptr %rawA, i64 %45
  %47 = extractelement <16 x i64> %24, i32 11
  %48 = getelementptr inbounds float, ptr %rawA, i64 %47
  %49 = extractelement <16 x i64> %24, i32 12
  %50 = getelementptr inbounds float, ptr %rawA, i64 %49
  %51 = extractelement <16 x i64> %24, i32 13
  %52 = getelementptr inbounds float, ptr %rawA, i64 %51
  %53 = extractelement <16 x i64> %24, i32 14
  %54 = getelementptr inbounds float, ptr %rawA, i64 %53
  %55 = extractelement <16 x i64> %24, i32 15
  %56 = getelementptr inbounds float, ptr %rawA, i64 %55
  %59 = load float, ptr %26, align 4
  %60 = load float, ptr %28, align 4
  %61 = load float, ptr %30, align 4
  %62 = load float, ptr %32, align 4
  %63 = load float, ptr %34, align 4
  %64 = load float, ptr %36, align 4
  %65 = load float, ptr %38, align 4
  %66 = load float, ptr %40, align 4
  %67 = load float, ptr %42, align 4
  %68 = load float, ptr %44, align 4
  %69 = load float, ptr %46, align 4
  %70 = load float, ptr %48, align 4
  %71 = load float, ptr %50, align 4
  %72 = load float, ptr %52, align 4
  %73 = load float, ptr %54, align 4
  %74 = load float, ptr %56, align 4
  %a1 = fadd float %59, %60
  %a2 = fadd float %61, %62
  %a3 = fadd float %63, %64
  %a4 = fadd float %65, %66
  %a5 = fadd float %67, %68
  %a6 = fadd float %69, %70
  %a7 = fadd float %71, %72
  %a8 = fadd float %73, %74
  %a9 = fadd float %a1, %a2
  %a10 = fadd float %a3, %a4
  %a11 = fadd float %a5, %a6
  %a12 = fadd float %a7, %a8
  %a13 = fadd float %a9, %a10
  %a14 = fadd float %a11, %a12
  %a15 = fadd float %a13, %a14
  ret float %a15
}

define float @scalarize_into_load_zext(i64 %22, ptr %23, ptr %rawA, ptr %rawB) {
; CHECK-LABEL: scalarize_into_load_zext:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ldp q0, q2, [x1]
; CHECK-NEXT:    ldp q4, q1, [x1, #32]
; CHECK-NEXT:    ushll v3.2d, v0.2s, #0
; CHECK-NEXT:    ushll2 v0.2d, v0.4s, #0
; CHECK-NEXT:    ushll2 v6.2d, v2.4s, #0
; CHECK-NEXT:    ushll2 v5.2d, v1.4s, #0
; CHECK-NEXT:    ushll v1.2d, v1.2s, #0
; CHECK-NEXT:    ushll v2.2d, v2.2s, #0
; CHECK-NEXT:    ushll2 v7.2d, v4.4s, #0
; CHECK-NEXT:    ushll v4.2d, v4.2s, #0
; CHECK-NEXT:    mov x8, v3.d[1]
; CHECK-NEXT:    mov x10, v0.d[1]
; CHECK-NEXT:    mov x14, v6.d[1]
; CHECK-NEXT:    mov x12, v2.d[1]
; CHECK-NEXT:    mov x1, v1.d[1]
; CHECK-NEXT:    mov x4, v5.d[1]
; CHECK-NEXT:    mov x16, v4.d[1]
; CHECK-NEXT:    mov x18, v7.d[1]
; CHECK-NEXT:    fmov x9, d3
; CHECK-NEXT:    fmov x11, d0
; CHECK-NEXT:    fmov x13, d2
; CHECK-NEXT:    fmov x15, d6
; CHECK-NEXT:    fmov x17, d4
; CHECK-NEXT:    fmov x0, d7
; CHECK-NEXT:    ldr s2, [x2, x8, lsl #2]
; CHECK-NEXT:    fmov x3, d1
; CHECK-NEXT:    fmov x5, d5
; CHECK-NEXT:    ldr s0, [x2, x9, lsl #2]
; CHECK-NEXT:    ldr s1, [x2, x11, lsl #2]
; CHECK-NEXT:    ldr s3, [x2, x10, lsl #2]
; CHECK-NEXT:    ldr s4, [x2, x13, lsl #2]
; CHECK-NEXT:    ldr s5, [x2, x12, lsl #2]
; CHECK-NEXT:    ldr s6, [x2, x15, lsl #2]
; CHECK-NEXT:    ldr s7, [x2, x14, lsl #2]
; CHECK-NEXT:    ldr s16, [x2, x17, lsl #2]
; CHECK-NEXT:    ldr s17, [x2, x16, lsl #2]
; CHECK-NEXT:    ldr s18, [x2, x0, lsl #2]
; CHECK-NEXT:    ldr s19, [x2, x18, lsl #2]
; CHECK-NEXT:    ldr s20, [x2, x3, lsl #2]
; CHECK-NEXT:    ldr s21, [x2, x1, lsl #2]
; CHECK-NEXT:    ldr s22, [x2, x5, lsl #2]
; CHECK-NEXT:    ldr s23, [x2, x4, lsl #2]
; CHECK-NEXT:    fadd s0, s0, s2
; CHECK-NEXT:    fadd s1, s1, s3
; CHECK-NEXT:    fadd s2, s4, s5
; CHECK-NEXT:    fadd s3, s6, s7
; CHECK-NEXT:    fadd s4, s16, s17
; CHECK-NEXT:    fadd s5, s18, s19
; CHECK-NEXT:    fadd s6, s20, s21
; CHECK-NEXT:    fadd s7, s22, s23
; CHECK-NEXT:    fadd s0, s0, s1
; CHECK-NEXT:    fadd s1, s2, s3
; CHECK-NEXT:    fadd s2, s4, s5
; CHECK-NEXT:    fadd s3, s6, s7
; CHECK-NEXT:    fadd s0, s0, s1
; CHECK-NEXT:    fadd s1, s2, s3
; CHECK-NEXT:    fadd s0, s0, s1
; CHECK-NEXT:    ret
entry:
  %wide.load = load <16 x i32>, ptr %23, align 4
  %24 = zext <16 x i32> %wide.load to <16 x i64>
  %25 = extractelement <16 x i64> %24, i32 0
  %26 = getelementptr inbounds float, ptr %rawA, i64 %25
  %27 = extractelement <16 x i64> %24, i32 1
  %28 = getelementptr inbounds float, ptr %rawA, i64 %27
  %29 = extractelement <16 x i64> %24, i32 2
  %30 = getelementptr inbounds float, ptr %rawA, i64 %29
  %31 = extractelement <16 x i64> %24, i32 3
  %32 = getelementptr inbounds float, ptr %rawA, i64 %31
  %33 = extractelement <16 x i64> %24, i32 4
  %34 = getelementptr inbounds float, ptr %rawA, i64 %33
  %35 = extractelement <16 x i64> %24, i32 5
  %36 = getelementptr inbounds float, ptr %rawA, i64 %35
  %37 = extractelement <16 x i64> %24, i32 6
  %38 = getelementptr inbounds float, ptr %rawA, i64 %37
  %39 = extractelement <16 x i64> %24, i32 7
  %40 = getelementptr inbounds float, ptr %rawA, i64 %39
  %41 = extractelement <16 x i64> %24, i32 8
  %42 = getelementptr inbounds float, ptr %rawA, i64 %41
  %43 = extractelement <16 x i64> %24, i32 9
  %44 = getelementptr inbounds float, ptr %rawA, i64 %43
  %45 = extractelement <16 x i64> %24, i32 10
  %46 = getelementptr inbounds float, ptr %rawA, i64 %45
  %47 = extractelement <16 x i64> %24, i32 11
  %48 = getelementptr inbounds float, ptr %rawA, i64 %47
  %49 = extractelement <16 x i64> %24, i32 12
  %50 = getelementptr inbounds float, ptr %rawA, i64 %49
  %51 = extractelement <16 x i64> %24, i32 13
  %52 = getelementptr inbounds float, ptr %rawA, i64 %51
  %53 = extractelement <16 x i64> %24, i32 14
  %54 = getelementptr inbounds float, ptr %rawA, i64 %53
  %55 = extractelement <16 x i64> %24, i32 15
  %56 = getelementptr inbounds float, ptr %rawA, i64 %55
  %59 = load float, ptr %26, align 4
  %60 = load float, ptr %28, align 4
  %61 = load float, ptr %30, align 4
  %62 = load float, ptr %32, align 4
  %63 = load float, ptr %34, align 4
  %64 = load float, ptr %36, align 4
  %65 = load float, ptr %38, align 4
  %66 = load float, ptr %40, align 4
  %67 = load float, ptr %42, align 4
  %68 = load float, ptr %44, align 4
  %69 = load float, ptr %46, align 4
  %70 = load float, ptr %48, align 4
  %71 = load float, ptr %50, align 4
  %72 = load float, ptr %52, align 4
  %73 = load float, ptr %54, align 4
  %74 = load float, ptr %56, align 4
  %a1 = fadd float %59, %60
  %a2 = fadd float %61, %62
  %a3 = fadd float %63, %64
  %a4 = fadd float %65, %66
  %a5 = fadd float %67, %68
  %a6 = fadd float %69, %70
  %a7 = fadd float %71, %72
  %a8 = fadd float %73, %74
  %a9 = fadd float %a1, %a2
  %a10 = fadd float %a3, %a4
  %a11 = fadd float %a5, %a6
  %a12 = fadd float %a7, %a8
  %a13 = fadd float %a9, %a10
  %a14 = fadd float %a11, %a12
  %a15 = fadd float %a13, %a14
  ret float %a15
}