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

define void @to_4(ptr nocapture readonly %x, ptr noalias nocapture %y) {
; CHECK-LABEL: to_4:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .save {r7, lr}
; CHECK-NEXT:    push {r7, lr}
; CHECK-NEXT:    mov.w lr, #256
; CHECK-NEXT:    movw r2, #26214
; CHECK-NEXT:    movt r2, #16390
; CHECK-NEXT:  .LBB0_1: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldrw.u32 q0, [r0], #16
; CHECK-NEXT:    vmul.f32 q0, q0, r2
; CHECK-NEXT:    vcvtb.f16.f32 q0, q0
; CHECK-NEXT:    vstrh.32 q0, [r1], #8
; CHECK-NEXT:    le lr, .LBB0_1
; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
; CHECK-NEXT:    pop {r7, pc}
entry:
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %entry
  %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
  %0 = getelementptr inbounds float, ptr %x, i32 %index
  %wide.load = load <4 x float>, ptr %0, align 4
  %1 = fmul <4 x float> %wide.load, <float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000>
  %2 = fptrunc <4 x float> %1 to <4 x half>
  %3 = getelementptr inbounds half, ptr %y, i32 %index
  store <4 x half> %2, ptr %3, align 2
  %index.next = add i32 %index, 4
  %4 = icmp eq i32 %index.next, 1024
  br i1 %4, label %for.cond.cleanup, label %vector.body

for.cond.cleanup:                                 ; preds = %vector.body
  ret void
}

define void @to_8(ptr nocapture readonly %x, ptr noalias nocapture %y) {
; CHECK-LABEL: to_8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .save {r7, lr}
; CHECK-NEXT:    push {r7, lr}
; CHECK-NEXT:    mov.w lr, #128
; CHECK-NEXT:    movw r2, #26214
; CHECK-NEXT:    movt r2, #16390
; CHECK-NEXT:  .LBB1_1: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
; CHECK-NEXT:    vmul.f32 q0, q0, r2
; CHECK-NEXT:    vcvtb.f16.f32 q0, q0
; CHECK-NEXT:    vstrh.32 q0, [r1, #8]
; CHECK-NEXT:    vldrw.u32 q0, [r0], #32
; CHECK-NEXT:    vmul.f32 q0, q0, r2
; CHECK-NEXT:    vcvtb.f16.f32 q0, q0
; CHECK-NEXT:    vstrh.32 q0, [r1], #16
; CHECK-NEXT:    le lr, .LBB1_1
; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
; CHECK-NEXT:    pop {r7, pc}
entry:
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %entry
  %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
  %0 = getelementptr inbounds float, ptr %x, i32 %index
  %wide.load = load <8 x float>, ptr %0, align 4
  %1 = fmul <8 x float> %wide.load, <float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000>
  %2 = fptrunc <8 x float> %1 to <8 x half>
  %3 = getelementptr inbounds half, ptr %y, i32 %index
  store <8 x half> %2, ptr %3, align 2
  %index.next = add i32 %index, 8
  %4 = icmp eq i32 %index.next, 1024
  br i1 %4, label %for.cond.cleanup, label %vector.body

for.cond.cleanup:                                 ; preds = %vector.body
  ret void
}

define void @to_16(ptr nocapture readonly %x, ptr noalias nocapture %y) {
; CHECK-LABEL: to_16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .save {r7, lr}
; CHECK-NEXT:    push {r7, lr}
; CHECK-NEXT:    mov.w lr, #64
; CHECK-NEXT:    movw r2, #26214
; CHECK-NEXT:    movt r2, #16390
; CHECK-NEXT:  .LBB2_1: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldrw.u32 q0, [r0, #48]
; CHECK-NEXT:    vmul.f32 q0, q0, r2
; CHECK-NEXT:    vcvtb.f16.f32 q0, q0
; CHECK-NEXT:    vstrh.32 q0, [r1, #24]
; CHECK-NEXT:    vldrw.u32 q0, [r0, #32]
; CHECK-NEXT:    vmul.f32 q0, q0, r2
; CHECK-NEXT:    vcvtb.f16.f32 q0, q0
; CHECK-NEXT:    vstrh.32 q0, [r1, #16]
; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
; CHECK-NEXT:    vmul.f32 q0, q0, r2
; CHECK-NEXT:    vcvtb.f16.f32 q0, q0
; CHECK-NEXT:    vstrh.32 q0, [r1, #8]
; CHECK-NEXT:    vldrw.u32 q0, [r0], #64
; CHECK-NEXT:    vmul.f32 q0, q0, r2
; CHECK-NEXT:    vcvtb.f16.f32 q0, q0
; CHECK-NEXT:    vstrh.32 q0, [r1], #32
; CHECK-NEXT:    le lr, .LBB2_1
; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
; CHECK-NEXT:    pop {r7, pc}
entry:
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %entry
  %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
  %0 = getelementptr inbounds float, ptr %x, i32 %index
  %wide.load = load <16 x float>, ptr %0, align 4
  %1 = fmul <16 x float> %wide.load, <float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000>
  %2 = fptrunc <16 x float> %1 to <16 x half>
  %3 = getelementptr inbounds half, ptr %y, i32 %index
  store <16 x half> %2, ptr %3, align 2
  %index.next = add i32 %index, 16
  %4 = icmp eq i32 %index.next, 1024
  br i1 %4, label %for.cond.cleanup, label %vector.body

for.cond.cleanup:                                 ; preds = %vector.body
  ret void
}

define void @from_4(ptr nocapture readonly %x, ptr noalias nocapture %y) {
; CHECK-LABEL: from_4:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .save {r7, lr}
; CHECK-NEXT:    push {r7, lr}
; CHECK-NEXT:    mov.w lr, #256
; CHECK-NEXT:    movw r2, #26214
; CHECK-NEXT:    movt r2, #16390
; CHECK-NEXT:  .LBB3_1: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldrh.u32 q0, [r0], #8
; CHECK-NEXT:    vcvtb.f32.f16 q0, q0
; CHECK-NEXT:    vmul.f32 q0, q0, r2
; CHECK-NEXT:    vstrb.8 q0, [r1], #16
; CHECK-NEXT:    le lr, .LBB3_1
; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
; CHECK-NEXT:    pop {r7, pc}
entry:
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %entry
  %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
  %0 = getelementptr inbounds half, ptr %x, i32 %index
  %wide.load = load <4 x half>, ptr %0, align 2
  %1 = fpext <4 x half> %wide.load to <4 x float>
  %2 = fmul <4 x float> %1, <float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000>
  %3 = getelementptr inbounds float, ptr %y, i32 %index
  store <4 x float> %2, ptr %3, align 4
  %index.next = add i32 %index, 4
  %4 = icmp eq i32 %index.next, 1024
  br i1 %4, label %for.cond.cleanup, label %vector.body

for.cond.cleanup:                                 ; preds = %vector.body
  ret void
}

define void @from_8(ptr nocapture readonly %x, ptr noalias nocapture %y) {
; CHECK-LABEL: from_8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .save {r7, lr}
; CHECK-NEXT:    push {r7, lr}
; CHECK-NEXT:    mov.w lr, #128
; CHECK-NEXT:    movw r2, #26214
; CHECK-NEXT:    movt r2, #16390
; CHECK-NEXT:  .LBB4_1: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldrh.u32 q0, [r0], #16
; CHECK-NEXT:    vldrh.u32 q1, [r0, #-8]
; CHECK-NEXT:    vcvtb.f32.f16 q0, q0
; CHECK-NEXT:    vmul.f32 q0, q0, r2
; CHECK-NEXT:    vcvtb.f32.f16 q1, q1
; CHECK-NEXT:    vmul.f32 q1, q1, r2
; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
; CHECK-NEXT:    vstrw.32 q0, [r1], #32
; CHECK-NEXT:    le lr, .LBB4_1
; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
; CHECK-NEXT:    pop {r7, pc}
entry:
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %entry
  %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
  %0 = getelementptr inbounds half, ptr %x, i32 %index
  %wide.load = load <8 x half>, ptr %0, align 2
  %1 = fpext <8 x half> %wide.load to <8 x float>
  %2 = fmul <8 x float> %1, <float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000>
  %3 = getelementptr inbounds float, ptr %y, i32 %index
  store <8 x float> %2, ptr %3, align 4
  %index.next = add i32 %index, 8
  %4 = icmp eq i32 %index.next, 1024
  br i1 %4, label %for.cond.cleanup, label %vector.body

for.cond.cleanup:                                 ; preds = %vector.body
  ret void
}

define void @from_16(ptr nocapture readonly %x, ptr noalias nocapture %y) {
; CHECK-LABEL: from_16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .save {r7, lr}
; CHECK-NEXT:    push {r7, lr}
; CHECK-NEXT:    mov.w lr, #64
; CHECK-NEXT:    movw r2, #26214
; CHECK-NEXT:    movt r2, #16390
; CHECK-NEXT:  .LBB5_1: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldrh.u32 q0, [r0], #32
; CHECK-NEXT:    vldrh.u32 q1, [r0, #-24]
; CHECK-NEXT:    vldrh.u32 q2, [r0, #-16]
; CHECK-NEXT:    vldrh.u32 q3, [r0, #-8]
; CHECK-NEXT:    vcvtb.f32.f16 q0, q0
; CHECK-NEXT:    vcvtb.f32.f16 q1, q1
; CHECK-NEXT:    vcvtb.f32.f16 q2, q2
; CHECK-NEXT:    vcvtb.f32.f16 q3, q3
; CHECK-NEXT:    vmul.f32 q2, q2, r2
; CHECK-NEXT:    vmul.f32 q3, q3, r2
; CHECK-NEXT:    vmul.f32 q1, q1, r2
; CHECK-NEXT:    vmul.f32 q0, q0, r2
; CHECK-NEXT:    vstrw.32 q3, [r1, #48]
; CHECK-NEXT:    vstrw.32 q2, [r1, #32]
; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
; CHECK-NEXT:    vstrw.32 q0, [r1], #64
; CHECK-NEXT:    le lr, .LBB5_1
; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
; CHECK-NEXT:    pop {r7, pc}
entry:
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %entry
  %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
  %0 = getelementptr inbounds half, ptr %x, i32 %index
  %wide.load = load <16 x half>, ptr %0, align 2
  %1 = fpext <16 x half> %wide.load to <16 x float>
  %2 = fmul <16 x float> %1, <float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000>
  %3 = getelementptr inbounds float, ptr %y, i32 %index
  store <16 x float> %2, ptr %3, align 4
  %index.next = add i32 %index, 16
  %4 = icmp eq i32 %index.next, 1024
  br i1 %4, label %for.cond.cleanup, label %vector.body

for.cond.cleanup:                                 ; preds = %vector.body
  ret void
}

define void @both_4(ptr nocapture readonly %x, ptr noalias nocapture %y) {
; CHECK-LABEL: both_4:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .save {r7, lr}
; CHECK-NEXT:    push {r7, lr}
; CHECK-NEXT:    mov.w lr, #256
; CHECK-NEXT:    movw r2, #26214
; CHECK-NEXT:    movt r2, #16390
; CHECK-NEXT:  .LBB6_1: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldrh.u32 q0, [r0], #8
; CHECK-NEXT:    vcvtb.f32.f16 q0, q0
; CHECK-NEXT:    vmul.f32 q0, q0, r2
; CHECK-NEXT:    vcvtb.f16.f32 q0, q0
; CHECK-NEXT:    vstrh.32 q0, [r1], #8
; CHECK-NEXT:    le lr, .LBB6_1
; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
; CHECK-NEXT:    pop {r7, pc}
entry:
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %entry
  %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
  %0 = getelementptr inbounds half, ptr %x, i32 %index
  %wide.load = load <4 x half>, ptr %0, align 2
  %1 = fpext <4 x half> %wide.load to <4 x float>
  %2 = fmul <4 x float> %1, <float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000>
  %3 = fptrunc <4 x float> %2 to <4 x half>
  %4 = getelementptr inbounds half, ptr %y, i32 %index
  store <4 x half> %3, ptr %4, align 2
  %index.next = add i32 %index, 4
  %5 = icmp eq i32 %index.next, 1024
  br i1 %5, label %for.cond.cleanup, label %vector.body

for.cond.cleanup:                                 ; preds = %vector.body
  ret void
}

define void @both_8(ptr nocapture readonly %x, ptr noalias nocapture %y) {
; CHECK-LABEL: both_8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .save {r7, lr}
; CHECK-NEXT:    push {r7, lr}
; CHECK-NEXT:    mov.w lr, #128
; CHECK-NEXT:    movw r2, #26214
; CHECK-NEXT:    movt r2, #16390
; CHECK-NEXT:  .LBB7_1: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldrh.u16 q0, [r0], #16
; CHECK-NEXT:    vcvtb.f32.f16 q1, q0
; CHECK-NEXT:    vcvtt.f32.f16 q0, q0
; CHECK-NEXT:    vmul.f32 q1, q1, r2
; CHECK-NEXT:    vmul.f32 q0, q0, r2
; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
; CHECK-NEXT:    vcvtt.f16.f32 q1, q0
; CHECK-NEXT:    vstrb.8 q1, [r1], #16
; CHECK-NEXT:    le lr, .LBB7_1
; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
; CHECK-NEXT:    pop {r7, pc}
entry:
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %entry
  %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
  %0 = getelementptr inbounds half, ptr %x, i32 %index
  %wide.load = load <8 x half>, ptr %0, align 2
  %1 = fpext <8 x half> %wide.load to <8 x float>
  %2 = fmul <8 x float> %1, <float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000>
  %3 = fptrunc <8 x float> %2 to <8 x half>
  %4 = getelementptr inbounds half, ptr %y, i32 %index
  store <8 x half> %3, ptr %4, align 2
  %index.next = add i32 %index, 8
  %5 = icmp eq i32 %index.next, 1024
  br i1 %5, label %for.cond.cleanup, label %vector.body

for.cond.cleanup:                                 ; preds = %vector.body
  ret void
}

define void @both_16(ptr nocapture readonly %x, ptr noalias nocapture %y) {
; CHECK-LABEL: both_16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .save {r7, lr}
; CHECK-NEXT:    push {r7, lr}
; CHECK-NEXT:    mov.w lr, #64
; CHECK-NEXT:    movw r2, #26214
; CHECK-NEXT:    movt r2, #16390
; CHECK-NEXT:  .LBB8_1: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldrh.u16 q0, [r0, #16]
; CHECK-NEXT:    vcvtb.f32.f16 q1, q0
; CHECK-NEXT:    vcvtt.f32.f16 q0, q0
; CHECK-NEXT:    vmul.f32 q1, q1, r2
; CHECK-NEXT:    vmul.f32 q0, q0, r2
; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
; CHECK-NEXT:    vcvtt.f16.f32 q1, q0
; CHECK-NEXT:    vldrh.u16 q0, [r0], #32
; CHECK-NEXT:    vstrh.16 q1, [r1, #16]
; CHECK-NEXT:    vcvtb.f32.f16 q1, q0
; CHECK-NEXT:    vcvtt.f32.f16 q0, q0
; CHECK-NEXT:    vmul.f32 q1, q1, r2
; CHECK-NEXT:    vmul.f32 q0, q0, r2
; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
; CHECK-NEXT:    vcvtt.f16.f32 q1, q0
; CHECK-NEXT:    vstrh.16 q1, [r1], #32
; CHECK-NEXT:    le lr, .LBB8_1
; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
; CHECK-NEXT:    pop {r7, pc}
entry:
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %entry
  %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
  %0 = getelementptr inbounds half, ptr %x, i32 %index
  %wide.load = load <16 x half>, ptr %0, align 2
  %1 = fpext <16 x half> %wide.load to <16 x float>
  %2 = fmul <16 x float> %1, <float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000>
  %3 = fptrunc <16 x float> %2 to <16 x half>
  %4 = getelementptr inbounds half, ptr %y, i32 %index
  store <16 x half> %3, ptr %4, align 2
  %index.next = add i32 %index, 16
  %5 = icmp eq i32 %index.next, 1024
  br i1 %5, label %for.cond.cleanup, label %vector.body

for.cond.cleanup:                                 ; preds = %vector.body
  ret void
}

define void @both_8_I(ptr nocapture readonly %x, ptr noalias nocapture %y) {
; CHECK-LABEL: both_8_I:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .save {r7, lr}
; CHECK-NEXT:    push {r7, lr}
; CHECK-NEXT:    mov.w lr, #128
; CHECK-NEXT:    movw r2, #26214
; CHECK-NEXT:    movt r2, #16390
; CHECK-NEXT:  .LBB9_1: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldrh.u16 q0, [r0], #16
; CHECK-NEXT:    vcvtb.f32.f16 q1, q0
; CHECK-NEXT:    vcvtt.f32.f16 q0, q0
; CHECK-NEXT:    vmul.f32 q1, q1, r2
; CHECK-NEXT:    vmul.f32 q0, q0, r2
; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
; CHECK-NEXT:    vcvtt.f16.f32 q1, q0
; CHECK-NEXT:    vstrb.8 q1, [r1], #16
; CHECK-NEXT:    le lr, .LBB9_1
; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
; CHECK-NEXT:    pop {r7, pc}
entry:
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %entry
  %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
  %0 = getelementptr inbounds half, ptr %x, i32 %index
  %wide.load = load <8 x half>, ptr %0, align 2
  %1 = shufflevector <8 x half> %wide.load, <8 x half> %wide.load, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
  %2 = shufflevector <8 x half> %wide.load, <8 x half> %wide.load, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
  %3 = fpext <4 x half> %1 to <4 x float>
  %4 = fpext <4 x half> %2 to <4 x float>
  %5 = fmul <4 x float> %3, <float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000>
  %6 = fmul <4 x float> %4, <float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000>
  %7 = shufflevector <4 x float> %5, <4 x float> %6, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
  %8 = fptrunc <8 x float> %7 to <8 x half>
  %9 = getelementptr inbounds half, ptr %y, i32 %index
  store <8 x half> %8, ptr %9, align 2
  %index.next = add i32 %index, 8
  %10 = icmp eq i32 %index.next, 1024
  br i1 %10, label %for.cond.cleanup, label %vector.body

for.cond.cleanup:                                 ; preds = %vector.body
  ret void
}

define void @both_16_I(ptr nocapture readonly %x, ptr noalias nocapture %y) {
; CHECK-LABEL: both_16_I:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    .save {r7, lr}
; CHECK-NEXT:    push {r7, lr}
; CHECK-NEXT:    mov.w lr, #128
; CHECK-NEXT:    movw r2, #26214
; CHECK-NEXT:    movt r2, #16390
; CHECK-NEXT:  .LBB10_1: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldrh.u16 q0, [r0]
; CHECK-NEXT:    vcvtb.f32.f16 q1, q0
; CHECK-NEXT:    vcvtt.f32.f16 q0, q0
; CHECK-NEXT:    vmul.f32 q1, q1, r2
; CHECK-NEXT:    vmul.f32 q0, q0, r2
; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
; CHECK-NEXT:    vcvtt.f16.f32 q1, q0
; CHECK-NEXT:    vldrh.u16 q0, [r0, #16]!
; CHECK-NEXT:    vstrh.16 q1, [r1]
; CHECK-NEXT:    vcvtb.f32.f16 q1, q0
; CHECK-NEXT:    vcvtt.f32.f16 q0, q0
; CHECK-NEXT:    vmul.f32 q1, q1, r2
; CHECK-NEXT:    vmul.f32 q0, q0, r2
; CHECK-NEXT:    vcvtb.f16.f32 q1, q1
; CHECK-NEXT:    vcvtt.f16.f32 q1, q0
; CHECK-NEXT:    vstrb.8 q1, [r1, #16]!
; CHECK-NEXT:    le lr, .LBB10_1
; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
; CHECK-NEXT:    pop {r7, pc}
entry:
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %entry
  %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
  %0 = getelementptr inbounds half, ptr %x, i32 %index
  %wide.load = load <16 x half>, ptr %0, align 2
  %1 = shufflevector <16 x half> %wide.load, <16 x half> %wide.load, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
  %2 = shufflevector <16 x half> %wide.load, <16 x half> %wide.load, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
  %3 = fpext <8 x half> %1 to <8 x float>
  %4 = fpext <8 x half> %2 to <8 x float>
  %5 = fmul <8 x float> %3, <float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000>
  %6 = fmul <8 x float> %4, <float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000, float 0x4000CCCCC0000000>
  %7 = shufflevector <8 x float> %5, <8 x float> %6, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
  %8 = fptrunc <16 x float> %7 to <16 x half>
  %9 = getelementptr inbounds half, ptr %y, i32 %index
  store <16 x half> %8, ptr %9, align 2
  %index.next = add i32 %index, 8
  %10 = icmp eq i32 %index.next, 1024
  br i1 %10, label %for.cond.cleanup, label %vector.body

for.cond.cleanup:                                 ; preds = %vector.body
  ret void
}