; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=thumbv8.1m.main -mattr=+mve.fp,+fp-armv8d16sp,+fp16,+fullfp16 %s -o - | FileCheck %s

define arm_aapcs_vfpcc void @float_float_mul(float* nocapture readonly %a, float* nocapture readonly %b, float* nocapture %c, i32 %N) {
; CHECK-LABEL: float_float_mul:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    push {r4, r5, r6, r7, lr}
; CHECK-NEXT:    cmp r3, #0
; CHECK-NEXT:    beq .LBB0_10
; CHECK-NEXT:  @ %bb.1: @ %for.body.preheader
; CHECK-NEXT:    cmp r3, #3
; CHECK-NEXT:    bhi .LBB0_3
; CHECK-NEXT:  @ %bb.2:
; CHECK-NEXT:    mov.w r12, #0
; CHECK-NEXT:    b .LBB0_4
; CHECK-NEXT:  .LBB0_3: @ %vector.memcheck
; CHECK-NEXT:    add.w r5, r2, r3, lsl #2
; CHECK-NEXT:    add.w r6, r1, r3, lsl #2
; CHECK-NEXT:    cmp r5, r1
; CHECK-NEXT:    add.w r4, r0, r3, lsl #2
; CHECK-NEXT:    cset r7, hi
; CHECK-NEXT:    cmp r6, r2
; CHECK-NEXT:    cset r6, hi
; CHECK-NEXT:    cmp r5, r0
; CHECK-NEXT:    cset r5, hi
; CHECK-NEXT:    cmp r4, r2
; CHECK-NEXT:    cset r4, hi
; CHECK-NEXT:    mov.w r12, #0
; CHECK-NEXT:    ands r5, r4
; CHECK-NEXT:    lsls r5, r5, #31
; CHECK-NEXT:    itt eq
; CHECK-NEXT:    andeq r7, r6
; CHECK-NEXT:    lslseq.w r7, r7, #31
; CHECK-NEXT:    beq .LBB0_11
; CHECK-NEXT:  .LBB0_4: @ %for.body.preheader22
; CHECK-NEXT:    mvn.w r7, r12
; CHECK-NEXT:    adds r4, r7, r3
; CHECK-NEXT:    and lr, r3, #3
; CHECK-NEXT:    wls lr, lr, .LBB0_7
; CHECK-NEXT:  @ %bb.5: @ %for.body.prol.preheader
; CHECK-NEXT:    add.w r5, r0, r12, lsl #2
; CHECK-NEXT:    add.w r6, r1, r12, lsl #2
; CHECK-NEXT:    add.w r7, r2, r12, lsl #2
; CHECK-NEXT:  .LBB0_6: @ %for.body.prol
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldr s0, [r6]
; CHECK-NEXT:    adds r6, #4
; CHECK-NEXT:    vldr s2, [r5]
; CHECK-NEXT:    adds r5, #4
; CHECK-NEXT:    add.w r12, r12, #1
; CHECK-NEXT:    vmul.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r7]
; CHECK-NEXT:    adds r7, #4
; CHECK-NEXT:    le lr, .LBB0_6
; CHECK-NEXT:  .LBB0_7: @ %for.body.prol.loopexit
; CHECK-NEXT:    cmp r4, #3
; CHECK-NEXT:    blo .LBB0_10
; CHECK-NEXT:  @ %bb.8: @ %for.body.preheader1
; CHECK-NEXT:    sub.w r3, r3, r12
; CHECK-NEXT:    lsl.w r12, r12, #2
; CHECK-NEXT:  .LBB0_9: @ %for.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    add.w r7, r1, r12
; CHECK-NEXT:    add.w r6, r0, r12
; CHECK-NEXT:    add.w r5, r2, r12
; CHECK-NEXT:    adds r0, #16
; CHECK-NEXT:    vldr s0, [r7]
; CHECK-NEXT:    adds r1, #16
; CHECK-NEXT:    vldr s2, [r6]
; CHECK-NEXT:    adds r2, #16
; CHECK-NEXT:    subs r3, #4
; CHECK-NEXT:    vmul.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r5]
; CHECK-NEXT:    vldr s0, [r7, #4]
; CHECK-NEXT:    vldr s2, [r6, #4]
; CHECK-NEXT:    vmul.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r5, #4]
; CHECK-NEXT:    vldr s0, [r7, #8]
; CHECK-NEXT:    vldr s2, [r6, #8]
; CHECK-NEXT:    vmul.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r5, #8]
; CHECK-NEXT:    vldr s0, [r7, #12]
; CHECK-NEXT:    vldr s2, [r6, #12]
; CHECK-NEXT:    vmul.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r5, #12]
; CHECK-NEXT:    bne .LBB0_9
; CHECK-NEXT:  .LBB0_10: @ %for.cond.cleanup
; CHECK-NEXT:    pop {r4, r5, r6, r7, pc}
; CHECK-NEXT:  .LBB0_11: @ %vector.ph
; CHECK-NEXT:    bic r12, r3, #3
; CHECK-NEXT:    movs r6, #1
; CHECK-NEXT:    sub.w r7, r12, #4
; CHECK-NEXT:    mov r4, r0
; CHECK-NEXT:    mov r5, r1
; CHECK-NEXT:    add.w lr, r6, r7, lsr #2
; CHECK-NEXT:    mov r6, r2
; CHECK-NEXT:    dls lr, lr
; CHECK-NEXT:  .LBB0_12: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldrw.u32 q0, [r5], #16
; CHECK-NEXT:    vldrw.u32 q1, [r4], #16
; CHECK-NEXT:    vmul.f32 q0, q1, q0
; CHECK-NEXT:    vstrb.8 q0, [r6], #16
; CHECK-NEXT:    le lr, .LBB0_12
; CHECK-NEXT:  @ %bb.13: @ %middle.block
; CHECK-NEXT:    cmp r12, r3
; CHECK-NEXT:    bne .LBB0_4
; CHECK-NEXT:    b .LBB0_10
entry:
  %cmp8 = icmp eq i32 %N, 0
  br i1 %cmp8, label %for.cond.cleanup, label %for.body.preheader

for.body.preheader:                               ; preds = %entry
  %min.iters.check = icmp ult i32 %N, 4
  br i1 %min.iters.check, label %for.body.preheader22, label %vector.memcheck

for.body.preheader22:                             ; preds = %middle.block, %vector.memcheck, %for.body.preheader
  %i.09.ph = phi i32 [ 0, %vector.memcheck ], [ 0, %for.body.preheader ], [ %n.vec, %middle.block ]
  %0 = xor i32 %i.09.ph, -1
  %1 = add i32 %0, %N
  %xtraiter = and i32 %N, 3
  %lcmp.mod = icmp eq i32 %xtraiter, 0
  br i1 %lcmp.mod, label %for.body.prol.loopexit, label %for.body.prol

for.body.prol:                                    ; preds = %for.body.preheader22, %for.body.prol
  %i.09.prol = phi i32 [ %inc.prol, %for.body.prol ], [ %i.09.ph, %for.body.preheader22 ]
  %prol.iter = phi i32 [ %prol.iter.sub, %for.body.prol ], [ %xtraiter, %for.body.preheader22 ]
  %arrayidx.prol = getelementptr inbounds float, float* %a, i32 %i.09.prol
  %2 = load float, float* %arrayidx.prol, align 4
  %arrayidx1.prol = getelementptr inbounds float, float* %b, i32 %i.09.prol
  %3 = load float, float* %arrayidx1.prol, align 4
  %mul.prol = fmul float %2, %3
  %arrayidx2.prol = getelementptr inbounds float, float* %c, i32 %i.09.prol
  store float %mul.prol, float* %arrayidx2.prol, align 4
  %inc.prol = add nuw i32 %i.09.prol, 1
  %prol.iter.sub = add i32 %prol.iter, -1
  %prol.iter.cmp = icmp eq i32 %prol.iter.sub, 0
  br i1 %prol.iter.cmp, label %for.body.prol.loopexit, label %for.body.prol

for.body.prol.loopexit:                           ; preds = %for.body.prol, %for.body.preheader22
  %i.09.unr = phi i32 [ %i.09.ph, %for.body.preheader22 ], [ %inc.prol, %for.body.prol ]
  %4 = icmp ult i32 %1, 3
  br i1 %4, label %for.cond.cleanup, label %for.body

vector.memcheck:                                  ; preds = %for.body.preheader
  %scevgep = getelementptr float, float* %c, i32 %N
  %scevgep13 = getelementptr float, float* %a, i32 %N
  %scevgep16 = getelementptr float, float* %b, i32 %N
  %bound0 = icmp ugt float* %scevgep13, %c
  %bound1 = icmp ugt float* %scevgep, %a
  %found.conflict = and i1 %bound0, %bound1
  %bound018 = icmp ugt float* %scevgep16, %c
  %bound119 = icmp ugt float* %scevgep, %b
  %found.conflict20 = and i1 %bound018, %bound119
  %conflict.rdx = or i1 %found.conflict, %found.conflict20
  br i1 %conflict.rdx, label %for.body.preheader22, label %vector.ph

vector.ph:                                        ; preds = %vector.memcheck
  %n.vec = and i32 %N, -4
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %vector.ph
  %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
  %5 = getelementptr inbounds float, float* %a, i32 %index
  %6 = bitcast float* %5 to <4 x float>*
  %wide.load = load <4 x float>, <4 x float>* %6, align 4
  %7 = getelementptr inbounds float, float* %b, i32 %index
  %8 = bitcast float* %7 to <4 x float>*
  %wide.load21 = load <4 x float>, <4 x float>* %8, align 4
  %9 = fmul <4 x float> %wide.load, %wide.load21
  %10 = getelementptr inbounds float, float* %c, i32 %index
  %11 = bitcast float* %10 to <4 x float>*
  store <4 x float> %9, <4 x float>* %11, align 4
  %index.next = add i32 %index, 4
  %12 = icmp eq i32 %index.next, %n.vec
  br i1 %12, label %middle.block, label %vector.body

middle.block:                                     ; preds = %vector.body
  %cmp.n = icmp eq i32 %n.vec, %N
  br i1 %cmp.n, label %for.cond.cleanup, label %for.body.preheader22

for.cond.cleanup:                                 ; preds = %for.body.prol.loopexit, %for.body, %middle.block, %entry
  ret void

for.body:                                         ; preds = %for.body.prol.loopexit, %for.body
  %i.09 = phi i32 [ %inc.3, %for.body ], [ %i.09.unr, %for.body.prol.loopexit ]
  %arrayidx = getelementptr inbounds float, float* %a, i32 %i.09
  %13 = load float, float* %arrayidx, align 4
  %arrayidx1 = getelementptr inbounds float, float* %b, i32 %i.09
  %14 = load float, float* %arrayidx1, align 4
  %mul = fmul float %13, %14
  %arrayidx2 = getelementptr inbounds float, float* %c, i32 %i.09
  store float %mul, float* %arrayidx2, align 4
  %inc = add nuw i32 %i.09, 1
  %arrayidx.1 = getelementptr inbounds float, float* %a, i32 %inc
  %15 = load float, float* %arrayidx.1, align 4
  %arrayidx1.1 = getelementptr inbounds float, float* %b, i32 %inc
  %16 = load float, float* %arrayidx1.1, align 4
  %mul.1 = fmul float %15, %16
  %arrayidx2.1 = getelementptr inbounds float, float* %c, i32 %inc
  store float %mul.1, float* %arrayidx2.1, align 4
  %inc.1 = add nuw i32 %i.09, 2
  %arrayidx.2 = getelementptr inbounds float, float* %a, i32 %inc.1
  %17 = load float, float* %arrayidx.2, align 4
  %arrayidx1.2 = getelementptr inbounds float, float* %b, i32 %inc.1
  %18 = load float, float* %arrayidx1.2, align 4
  %mul.2 = fmul float %17, %18
  %arrayidx2.2 = getelementptr inbounds float, float* %c, i32 %inc.1
  store float %mul.2, float* %arrayidx2.2, align 4
  %inc.2 = add nuw i32 %i.09, 3
  %arrayidx.3 = getelementptr inbounds float, float* %a, i32 %inc.2
  %19 = load float, float* %arrayidx.3, align 4
  %arrayidx1.3 = getelementptr inbounds float, float* %b, i32 %inc.2
  %20 = load float, float* %arrayidx1.3, align 4
  %mul.3 = fmul float %19, %20
  %arrayidx2.3 = getelementptr inbounds float, float* %c, i32 %inc.2
  store float %mul.3, float* %arrayidx2.3, align 4
  %inc.3 = add nuw i32 %i.09, 4
  %exitcond.3 = icmp eq i32 %inc.3, %N
  br i1 %exitcond.3, label %for.cond.cleanup, label %for.body
}

define arm_aapcs_vfpcc void @float_float_add(float* nocapture readonly %a, float* nocapture readonly %b, float* nocapture %c, i32 %N) {
; CHECK-LABEL: float_float_add:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    push {r4, r5, r6, r7, lr}
; CHECK-NEXT:    cmp r3, #0
; CHECK-NEXT:    beq .LBB1_10
; CHECK-NEXT:  @ %bb.1: @ %for.body.preheader
; CHECK-NEXT:    cmp r3, #3
; CHECK-NEXT:    bhi .LBB1_3
; CHECK-NEXT:  @ %bb.2:
; CHECK-NEXT:    mov.w r12, #0
; CHECK-NEXT:    b .LBB1_4
; CHECK-NEXT:  .LBB1_3: @ %vector.memcheck
; CHECK-NEXT:    add.w r5, r2, r3, lsl #2
; CHECK-NEXT:    add.w r6, r1, r3, lsl #2
; CHECK-NEXT:    cmp r5, r1
; CHECK-NEXT:    add.w r4, r0, r3, lsl #2
; CHECK-NEXT:    cset r7, hi
; CHECK-NEXT:    cmp r6, r2
; CHECK-NEXT:    cset r6, hi
; CHECK-NEXT:    cmp r5, r0
; CHECK-NEXT:    cset r5, hi
; CHECK-NEXT:    cmp r4, r2
; CHECK-NEXT:    cset r4, hi
; CHECK-NEXT:    mov.w r12, #0
; CHECK-NEXT:    ands r5, r4
; CHECK-NEXT:    lsls r5, r5, #31
; CHECK-NEXT:    itt eq
; CHECK-NEXT:    andeq r7, r6
; CHECK-NEXT:    lslseq.w r7, r7, #31
; CHECK-NEXT:    beq .LBB1_11
; CHECK-NEXT:  .LBB1_4: @ %for.body.preheader22
; CHECK-NEXT:    mvn.w r7, r12
; CHECK-NEXT:    adds r4, r7, r3
; CHECK-NEXT:    and lr, r3, #3
; CHECK-NEXT:    wls lr, lr, .LBB1_7
; CHECK-NEXT:  @ %bb.5: @ %for.body.prol.preheader
; CHECK-NEXT:    add.w r5, r0, r12, lsl #2
; CHECK-NEXT:    add.w r6, r1, r12, lsl #2
; CHECK-NEXT:    add.w r7, r2, r12, lsl #2
; CHECK-NEXT:  .LBB1_6: @ %for.body.prol
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldr s0, [r6]
; CHECK-NEXT:    adds r6, #4
; CHECK-NEXT:    vldr s2, [r5]
; CHECK-NEXT:    adds r5, #4
; CHECK-NEXT:    add.w r12, r12, #1
; CHECK-NEXT:    vadd.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r7]
; CHECK-NEXT:    adds r7, #4
; CHECK-NEXT:    le lr, .LBB1_6
; CHECK-NEXT:  .LBB1_7: @ %for.body.prol.loopexit
; CHECK-NEXT:    cmp r4, #3
; CHECK-NEXT:    blo .LBB1_10
; CHECK-NEXT:  @ %bb.8: @ %for.body.preheader1
; CHECK-NEXT:    sub.w r3, r3, r12
; CHECK-NEXT:    lsl.w r12, r12, #2
; CHECK-NEXT:  .LBB1_9: @ %for.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    add.w r7, r1, r12
; CHECK-NEXT:    add.w r6, r0, r12
; CHECK-NEXT:    add.w r5, r2, r12
; CHECK-NEXT:    adds r0, #16
; CHECK-NEXT:    vldr s0, [r7]
; CHECK-NEXT:    adds r1, #16
; CHECK-NEXT:    vldr s2, [r6]
; CHECK-NEXT:    adds r2, #16
; CHECK-NEXT:    subs r3, #4
; CHECK-NEXT:    vadd.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r5]
; CHECK-NEXT:    vldr s0, [r7, #4]
; CHECK-NEXT:    vldr s2, [r6, #4]
; CHECK-NEXT:    vadd.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r5, #4]
; CHECK-NEXT:    vldr s0, [r7, #8]
; CHECK-NEXT:    vldr s2, [r6, #8]
; CHECK-NEXT:    vadd.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r5, #8]
; CHECK-NEXT:    vldr s0, [r7, #12]
; CHECK-NEXT:    vldr s2, [r6, #12]
; CHECK-NEXT:    vadd.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r5, #12]
; CHECK-NEXT:    bne .LBB1_9
; CHECK-NEXT:  .LBB1_10: @ %for.cond.cleanup
; CHECK-NEXT:    pop {r4, r5, r6, r7, pc}
; CHECK-NEXT:  .LBB1_11: @ %vector.ph
; CHECK-NEXT:    bic r12, r3, #3
; CHECK-NEXT:    movs r6, #1
; CHECK-NEXT:    sub.w r7, r12, #4
; CHECK-NEXT:    mov r4, r0
; CHECK-NEXT:    mov r5, r1
; CHECK-NEXT:    add.w lr, r6, r7, lsr #2
; CHECK-NEXT:    mov r6, r2
; CHECK-NEXT:    dls lr, lr
; CHECK-NEXT:  .LBB1_12: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldrw.u32 q0, [r5], #16
; CHECK-NEXT:    vldrw.u32 q1, [r4], #16
; CHECK-NEXT:    vadd.f32 q0, q1, q0
; CHECK-NEXT:    vstrb.8 q0, [r6], #16
; CHECK-NEXT:    le lr, .LBB1_12
; CHECK-NEXT:  @ %bb.13: @ %middle.block
; CHECK-NEXT:    cmp r12, r3
; CHECK-NEXT:    bne .LBB1_4
; CHECK-NEXT:    b .LBB1_10
entry:
  %cmp8 = icmp eq i32 %N, 0
  br i1 %cmp8, label %for.cond.cleanup, label %for.body.preheader

for.body.preheader:                               ; preds = %entry
  %min.iters.check = icmp ult i32 %N, 4
  br i1 %min.iters.check, label %for.body.preheader22, label %vector.memcheck

for.body.preheader22:                             ; preds = %middle.block, %vector.memcheck, %for.body.preheader
  %i.09.ph = phi i32 [ 0, %vector.memcheck ], [ 0, %for.body.preheader ], [ %n.vec, %middle.block ]
  %0 = xor i32 %i.09.ph, -1
  %1 = add i32 %0, %N
  %xtraiter = and i32 %N, 3
  %lcmp.mod = icmp eq i32 %xtraiter, 0
  br i1 %lcmp.mod, label %for.body.prol.loopexit, label %for.body.prol

for.body.prol:                                    ; preds = %for.body.preheader22, %for.body.prol
  %i.09.prol = phi i32 [ %inc.prol, %for.body.prol ], [ %i.09.ph, %for.body.preheader22 ]
  %prol.iter = phi i32 [ %prol.iter.sub, %for.body.prol ], [ %xtraiter, %for.body.preheader22 ]
  %arrayidx.prol = getelementptr inbounds float, float* %a, i32 %i.09.prol
  %2 = load float, float* %arrayidx.prol, align 4
  %arrayidx1.prol = getelementptr inbounds float, float* %b, i32 %i.09.prol
  %3 = load float, float* %arrayidx1.prol, align 4
  %add.prol = fadd float %2, %3
  %arrayidx2.prol = getelementptr inbounds float, float* %c, i32 %i.09.prol
  store float %add.prol, float* %arrayidx2.prol, align 4
  %inc.prol = add nuw i32 %i.09.prol, 1
  %prol.iter.sub = add i32 %prol.iter, -1
  %prol.iter.cmp = icmp eq i32 %prol.iter.sub, 0
  br i1 %prol.iter.cmp, label %for.body.prol.loopexit, label %for.body.prol

for.body.prol.loopexit:                           ; preds = %for.body.prol, %for.body.preheader22
  %i.09.unr = phi i32 [ %i.09.ph, %for.body.preheader22 ], [ %inc.prol, %for.body.prol ]
  %4 = icmp ult i32 %1, 3
  br i1 %4, label %for.cond.cleanup, label %for.body

vector.memcheck:                                  ; preds = %for.body.preheader
  %scevgep = getelementptr float, float* %c, i32 %N
  %scevgep13 = getelementptr float, float* %a, i32 %N
  %scevgep16 = getelementptr float, float* %b, i32 %N
  %bound0 = icmp ugt float* %scevgep13, %c
  %bound1 = icmp ugt float* %scevgep, %a
  %found.conflict = and i1 %bound0, %bound1
  %bound018 = icmp ugt float* %scevgep16, %c
  %bound119 = icmp ugt float* %scevgep, %b
  %found.conflict20 = and i1 %bound018, %bound119
  %conflict.rdx = or i1 %found.conflict, %found.conflict20
  br i1 %conflict.rdx, label %for.body.preheader22, label %vector.ph

vector.ph:                                        ; preds = %vector.memcheck
  %n.vec = and i32 %N, -4
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %vector.ph
  %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
  %5 = getelementptr inbounds float, float* %a, i32 %index
  %6 = bitcast float* %5 to <4 x float>*
  %wide.load = load <4 x float>, <4 x float>* %6, align 4
  %7 = getelementptr inbounds float, float* %b, i32 %index
  %8 = bitcast float* %7 to <4 x float>*
  %wide.load21 = load <4 x float>, <4 x float>* %8, align 4
  %9 = fadd <4 x float> %wide.load, %wide.load21
  %10 = getelementptr inbounds float, float* %c, i32 %index
  %11 = bitcast float* %10 to <4 x float>*
  store <4 x float> %9, <4 x float>* %11, align 4
  %index.next = add i32 %index, 4
  %12 = icmp eq i32 %index.next, %n.vec
  br i1 %12, label %middle.block, label %vector.body

middle.block:                                     ; preds = %vector.body
  %cmp.n = icmp eq i32 %n.vec, %N
  br i1 %cmp.n, label %for.cond.cleanup, label %for.body.preheader22

for.cond.cleanup:                                 ; preds = %for.body.prol.loopexit, %for.body, %middle.block, %entry
  ret void

for.body:                                         ; preds = %for.body.prol.loopexit, %for.body
  %i.09 = phi i32 [ %inc.3, %for.body ], [ %i.09.unr, %for.body.prol.loopexit ]
  %arrayidx = getelementptr inbounds float, float* %a, i32 %i.09
  %13 = load float, float* %arrayidx, align 4
  %arrayidx1 = getelementptr inbounds float, float* %b, i32 %i.09
  %14 = load float, float* %arrayidx1, align 4
  %add = fadd float %13, %14
  %arrayidx2 = getelementptr inbounds float, float* %c, i32 %i.09
  store float %add, float* %arrayidx2, align 4
  %inc = add nuw i32 %i.09, 1
  %arrayidx.1 = getelementptr inbounds float, float* %a, i32 %inc
  %15 = load float, float* %arrayidx.1, align 4
  %arrayidx1.1 = getelementptr inbounds float, float* %b, i32 %inc
  %16 = load float, float* %arrayidx1.1, align 4
  %add.1 = fadd float %15, %16
  %arrayidx2.1 = getelementptr inbounds float, float* %c, i32 %inc
  store float %add.1, float* %arrayidx2.1, align 4
  %inc.1 = add nuw i32 %i.09, 2
  %arrayidx.2 = getelementptr inbounds float, float* %a, i32 %inc.1
  %17 = load float, float* %arrayidx.2, align 4
  %arrayidx1.2 = getelementptr inbounds float, float* %b, i32 %inc.1
  %18 = load float, float* %arrayidx1.2, align 4
  %add.2 = fadd float %17, %18
  %arrayidx2.2 = getelementptr inbounds float, float* %c, i32 %inc.1
  store float %add.2, float* %arrayidx2.2, align 4
  %inc.2 = add nuw i32 %i.09, 3
  %arrayidx.3 = getelementptr inbounds float, float* %a, i32 %inc.2
  %19 = load float, float* %arrayidx.3, align 4
  %arrayidx1.3 = getelementptr inbounds float, float* %b, i32 %inc.2
  %20 = load float, float* %arrayidx1.3, align 4
  %add.3 = fadd float %19, %20
  %arrayidx2.3 = getelementptr inbounds float, float* %c, i32 %inc.2
  store float %add.3, float* %arrayidx2.3, align 4
  %inc.3 = add nuw i32 %i.09, 4
  %exitcond.3 = icmp eq i32 %inc.3, %N
  br i1 %exitcond.3, label %for.cond.cleanup, label %for.body
}

define arm_aapcs_vfpcc void @float_float_sub(float* nocapture readonly %a, float* nocapture readonly %b, float* nocapture %c, i32 %N) {
; CHECK-LABEL: float_float_sub:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    push {r4, r5, r6, r7, lr}
; CHECK-NEXT:    cmp r3, #0
; CHECK-NEXT:    beq .LBB2_10
; CHECK-NEXT:  @ %bb.1: @ %for.body.preheader
; CHECK-NEXT:    cmp r3, #3
; CHECK-NEXT:    bhi .LBB2_3
; CHECK-NEXT:  @ %bb.2:
; CHECK-NEXT:    mov.w r12, #0
; CHECK-NEXT:    b .LBB2_4
; CHECK-NEXT:  .LBB2_3: @ %vector.memcheck
; CHECK-NEXT:    add.w r5, r2, r3, lsl #2
; CHECK-NEXT:    add.w r6, r1, r3, lsl #2
; CHECK-NEXT:    cmp r5, r1
; CHECK-NEXT:    add.w r4, r0, r3, lsl #2
; CHECK-NEXT:    cset r7, hi
; CHECK-NEXT:    cmp r6, r2
; CHECK-NEXT:    cset r6, hi
; CHECK-NEXT:    cmp r5, r0
; CHECK-NEXT:    cset r5, hi
; CHECK-NEXT:    cmp r4, r2
; CHECK-NEXT:    cset r4, hi
; CHECK-NEXT:    mov.w r12, #0
; CHECK-NEXT:    ands r5, r4
; CHECK-NEXT:    lsls r5, r5, #31
; CHECK-NEXT:    itt eq
; CHECK-NEXT:    andeq r7, r6
; CHECK-NEXT:    lslseq.w r7, r7, #31
; CHECK-NEXT:    beq .LBB2_11
; CHECK-NEXT:  .LBB2_4: @ %for.body.preheader22
; CHECK-NEXT:    mvn.w r7, r12
; CHECK-NEXT:    adds r4, r7, r3
; CHECK-NEXT:    and lr, r3, #3
; CHECK-NEXT:    wls lr, lr, .LBB2_7
; CHECK-NEXT:  @ %bb.5: @ %for.body.prol.preheader
; CHECK-NEXT:    add.w r5, r0, r12, lsl #2
; CHECK-NEXT:    add.w r6, r1, r12, lsl #2
; CHECK-NEXT:    add.w r7, r2, r12, lsl #2
; CHECK-NEXT:  .LBB2_6: @ %for.body.prol
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldr s0, [r6]
; CHECK-NEXT:    adds r6, #4
; CHECK-NEXT:    vldr s2, [r5]
; CHECK-NEXT:    adds r5, #4
; CHECK-NEXT:    add.w r12, r12, #1
; CHECK-NEXT:    vsub.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r7]
; CHECK-NEXT:    adds r7, #4
; CHECK-NEXT:    le lr, .LBB2_6
; CHECK-NEXT:  .LBB2_7: @ %for.body.prol.loopexit
; CHECK-NEXT:    cmp r4, #3
; CHECK-NEXT:    blo .LBB2_10
; CHECK-NEXT:  @ %bb.8: @ %for.body.preheader1
; CHECK-NEXT:    sub.w r3, r3, r12
; CHECK-NEXT:    lsl.w r12, r12, #2
; CHECK-NEXT:  .LBB2_9: @ %for.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    add.w r7, r1, r12
; CHECK-NEXT:    add.w r6, r0, r12
; CHECK-NEXT:    add.w r5, r2, r12
; CHECK-NEXT:    adds r0, #16
; CHECK-NEXT:    vldr s0, [r7]
; CHECK-NEXT:    adds r1, #16
; CHECK-NEXT:    vldr s2, [r6]
; CHECK-NEXT:    adds r2, #16
; CHECK-NEXT:    subs r3, #4
; CHECK-NEXT:    vsub.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r5]
; CHECK-NEXT:    vldr s0, [r7, #4]
; CHECK-NEXT:    vldr s2, [r6, #4]
; CHECK-NEXT:    vsub.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r5, #4]
; CHECK-NEXT:    vldr s0, [r7, #8]
; CHECK-NEXT:    vldr s2, [r6, #8]
; CHECK-NEXT:    vsub.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r5, #8]
; CHECK-NEXT:    vldr s0, [r7, #12]
; CHECK-NEXT:    vldr s2, [r6, #12]
; CHECK-NEXT:    vsub.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r5, #12]
; CHECK-NEXT:    bne .LBB2_9
; CHECK-NEXT:  .LBB2_10: @ %for.cond.cleanup
; CHECK-NEXT:    pop {r4, r5, r6, r7, pc}
; CHECK-NEXT:  .LBB2_11: @ %vector.ph
; CHECK-NEXT:    bic r12, r3, #3
; CHECK-NEXT:    movs r6, #1
; CHECK-NEXT:    sub.w r7, r12, #4
; CHECK-NEXT:    mov r4, r0
; CHECK-NEXT:    mov r5, r1
; CHECK-NEXT:    add.w lr, r6, r7, lsr #2
; CHECK-NEXT:    mov r6, r2
; CHECK-NEXT:    dls lr, lr
; CHECK-NEXT:  .LBB2_12: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldrw.u32 q0, [r5], #16
; CHECK-NEXT:    vldrw.u32 q1, [r4], #16
; CHECK-NEXT:    vsub.f32 q0, q1, q0
; CHECK-NEXT:    vstrb.8 q0, [r6], #16
; CHECK-NEXT:    le lr, .LBB2_12
; CHECK-NEXT:  @ %bb.13: @ %middle.block
; CHECK-NEXT:    cmp r12, r3
; CHECK-NEXT:    bne .LBB2_4
; CHECK-NEXT:    b .LBB2_10
entry:
  %cmp8 = icmp eq i32 %N, 0
  br i1 %cmp8, label %for.cond.cleanup, label %for.body.preheader

for.body.preheader:                               ; preds = %entry
  %min.iters.check = icmp ult i32 %N, 4
  br i1 %min.iters.check, label %for.body.preheader22, label %vector.memcheck

for.body.preheader22:                             ; preds = %middle.block, %vector.memcheck, %for.body.preheader
  %i.09.ph = phi i32 [ 0, %vector.memcheck ], [ 0, %for.body.preheader ], [ %n.vec, %middle.block ]
  %0 = xor i32 %i.09.ph, -1
  %1 = add i32 %0, %N
  %xtraiter = and i32 %N, 3
  %lcmp.mod = icmp eq i32 %xtraiter, 0
  br i1 %lcmp.mod, label %for.body.prol.loopexit, label %for.body.prol

for.body.prol:                                    ; preds = %for.body.preheader22, %for.body.prol
  %i.09.prol = phi i32 [ %inc.prol, %for.body.prol ], [ %i.09.ph, %for.body.preheader22 ]
  %prol.iter = phi i32 [ %prol.iter.sub, %for.body.prol ], [ %xtraiter, %for.body.preheader22 ]
  %arrayidx.prol = getelementptr inbounds float, float* %a, i32 %i.09.prol
  %2 = load float, float* %arrayidx.prol, align 4
  %arrayidx1.prol = getelementptr inbounds float, float* %b, i32 %i.09.prol
  %3 = load float, float* %arrayidx1.prol, align 4
  %sub.prol = fsub float %2, %3
  %arrayidx2.prol = getelementptr inbounds float, float* %c, i32 %i.09.prol
  store float %sub.prol, float* %arrayidx2.prol, align 4
  %inc.prol = add nuw i32 %i.09.prol, 1
  %prol.iter.sub = add i32 %prol.iter, -1
  %prol.iter.cmp = icmp eq i32 %prol.iter.sub, 0
  br i1 %prol.iter.cmp, label %for.body.prol.loopexit, label %for.body.prol

for.body.prol.loopexit:                           ; preds = %for.body.prol, %for.body.preheader22
  %i.09.unr = phi i32 [ %i.09.ph, %for.body.preheader22 ], [ %inc.prol, %for.body.prol ]
  %4 = icmp ult i32 %1, 3
  br i1 %4, label %for.cond.cleanup, label %for.body

vector.memcheck:                                  ; preds = %for.body.preheader
  %scevgep = getelementptr float, float* %c, i32 %N
  %scevgep13 = getelementptr float, float* %a, i32 %N
  %scevgep16 = getelementptr float, float* %b, i32 %N
  %bound0 = icmp ugt float* %scevgep13, %c
  %bound1 = icmp ugt float* %scevgep, %a
  %found.conflict = and i1 %bound0, %bound1
  %bound018 = icmp ugt float* %scevgep16, %c
  %bound119 = icmp ugt float* %scevgep, %b
  %found.conflict20 = and i1 %bound018, %bound119
  %conflict.rdx = or i1 %found.conflict, %found.conflict20
  br i1 %conflict.rdx, label %for.body.preheader22, label %vector.ph

vector.ph:                                        ; preds = %vector.memcheck
  %n.vec = and i32 %N, -4
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %vector.ph
  %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
  %5 = getelementptr inbounds float, float* %a, i32 %index
  %6 = bitcast float* %5 to <4 x float>*
  %wide.load = load <4 x float>, <4 x float>* %6, align 4
  %7 = getelementptr inbounds float, float* %b, i32 %index
  %8 = bitcast float* %7 to <4 x float>*
  %wide.load21 = load <4 x float>, <4 x float>* %8, align 4
  %9 = fsub <4 x float> %wide.load, %wide.load21
  %10 = getelementptr inbounds float, float* %c, i32 %index
  %11 = bitcast float* %10 to <4 x float>*
  store <4 x float> %9, <4 x float>* %11, align 4
  %index.next = add i32 %index, 4
  %12 = icmp eq i32 %index.next, %n.vec
  br i1 %12, label %middle.block, label %vector.body

middle.block:                                     ; preds = %vector.body
  %cmp.n = icmp eq i32 %n.vec, %N
  br i1 %cmp.n, label %for.cond.cleanup, label %for.body.preheader22

for.cond.cleanup:                                 ; preds = %for.body.prol.loopexit, %for.body, %middle.block, %entry
  ret void

for.body:                                         ; preds = %for.body.prol.loopexit, %for.body
  %i.09 = phi i32 [ %inc.3, %for.body ], [ %i.09.unr, %for.body.prol.loopexit ]
  %arrayidx = getelementptr inbounds float, float* %a, i32 %i.09
  %13 = load float, float* %arrayidx, align 4
  %arrayidx1 = getelementptr inbounds float, float* %b, i32 %i.09
  %14 = load float, float* %arrayidx1, align 4
  %sub = fsub float %13, %14
  %arrayidx2 = getelementptr inbounds float, float* %c, i32 %i.09
  store float %sub, float* %arrayidx2, align 4
  %inc = add nuw i32 %i.09, 1
  %arrayidx.1 = getelementptr inbounds float, float* %a, i32 %inc
  %15 = load float, float* %arrayidx.1, align 4
  %arrayidx1.1 = getelementptr inbounds float, float* %b, i32 %inc
  %16 = load float, float* %arrayidx1.1, align 4
  %sub.1 = fsub float %15, %16
  %arrayidx2.1 = getelementptr inbounds float, float* %c, i32 %inc
  store float %sub.1, float* %arrayidx2.1, align 4
  %inc.1 = add nuw i32 %i.09, 2
  %arrayidx.2 = getelementptr inbounds float, float* %a, i32 %inc.1
  %17 = load float, float* %arrayidx.2, align 4
  %arrayidx1.2 = getelementptr inbounds float, float* %b, i32 %inc.1
  %18 = load float, float* %arrayidx1.2, align 4
  %sub.2 = fsub float %17, %18
  %arrayidx2.2 = getelementptr inbounds float, float* %c, i32 %inc.1
  store float %sub.2, float* %arrayidx2.2, align 4
  %inc.2 = add nuw i32 %i.09, 3
  %arrayidx.3 = getelementptr inbounds float, float* %a, i32 %inc.2
  %19 = load float, float* %arrayidx.3, align 4
  %arrayidx1.3 = getelementptr inbounds float, float* %b, i32 %inc.2
  %20 = load float, float* %arrayidx1.3, align 4
  %sub.3 = fsub float %19, %20
  %arrayidx2.3 = getelementptr inbounds float, float* %c, i32 %inc.2
  store float %sub.3, float* %arrayidx2.3, align 4
  %inc.3 = add nuw i32 %i.09, 4
  %exitcond.3 = icmp eq i32 %inc.3, %N
  br i1 %exitcond.3, label %for.cond.cleanup, label %for.body
}

define arm_aapcs_vfpcc void @float_int_mul(float* nocapture readonly %a, i32* nocapture readonly %b, float* nocapture %c, i32 %N) {
; CHECK-LABEL: float_int_mul:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, lr}
; CHECK-NEXT:    cmp r3, #0
; CHECK-NEXT:    beq.w .LBB3_13
; CHECK-NEXT:  @ %bb.1: @ %for.body.preheader
; CHECK-NEXT:    cmp r3, #3
; CHECK-NEXT:    bls .LBB3_6
; CHECK-NEXT:  @ %bb.2: @ %vector.memcheck
; CHECK-NEXT:    add.w r7, r0, r3, lsl #2
; CHECK-NEXT:    cmp r7, r2
; CHECK-NEXT:    itt hi
; CHECK-NEXT:    addhi.w r7, r2, r3, lsl #2
; CHECK-NEXT:    cmphi r7, r0
; CHECK-NEXT:    bhi .LBB3_6
; CHECK-NEXT:  @ %bb.3: @ %vector.ph
; CHECK-NEXT:    bic r12, r3, #3
; CHECK-NEXT:    movs r6, #1
; CHECK-NEXT:    sub.w r7, r12, #4
; CHECK-NEXT:    mov r4, r0
; CHECK-NEXT:    mov r5, r1
; CHECK-NEXT:    add.w lr, r6, r7, lsr #2
; CHECK-NEXT:    mov r6, r2
; CHECK-NEXT:    dls lr, lr
; CHECK-NEXT:  .LBB3_4: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldrw.u32 q0, [r5], #16
; CHECK-NEXT:    vldrw.u32 q1, [r4], #16
; CHECK-NEXT:    vcvt.f32.s32 q0, q0
; CHECK-NEXT:    vmul.f32 q0, q1, q0
; CHECK-NEXT:    vstrb.8 q0, [r6], #16
; CHECK-NEXT:    le lr, .LBB3_4
; CHECK-NEXT:  @ %bb.5: @ %middle.block
; CHECK-NEXT:    cmp r12, r3
; CHECK-NEXT:    bne .LBB3_7
; CHECK-NEXT:    b .LBB3_13
; CHECK-NEXT:  .LBB3_6:
; CHECK-NEXT:    mov.w r12, #0
; CHECK-NEXT:  .LBB3_7: @ %for.body.preheader16
; CHECK-NEXT:    mvn.w r7, r12
; CHECK-NEXT:    add.w r8, r7, r3
; CHECK-NEXT:    and lr, r3, #3
; CHECK-NEXT:    wls lr, lr, .LBB3_10
; CHECK-NEXT:  @ %bb.8: @ %for.body.prol.preheader
; CHECK-NEXT:    add.w r5, r0, r12, lsl #2
; CHECK-NEXT:    add.w r6, r1, r12, lsl #2
; CHECK-NEXT:    add.w r7, r2, r12, lsl #2
; CHECK-NEXT:  .LBB3_9: @ %for.body.prol
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    ldr r4, [r6], #4
; CHECK-NEXT:    add.w r12, r12, #1
; CHECK-NEXT:    vldr s2, [r5]
; CHECK-NEXT:    adds r5, #4
; CHECK-NEXT:    vmov s0, r4
; CHECK-NEXT:    vcvt.f32.s32 s0, s0
; CHECK-NEXT:    vmul.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r7]
; CHECK-NEXT:    adds r7, #4
; CHECK-NEXT:    le lr, .LBB3_9
; CHECK-NEXT:  .LBB3_10: @ %for.body.prol.loopexit
; CHECK-NEXT:    cmp.w r8, #3
; CHECK-NEXT:    blo .LBB3_13
; CHECK-NEXT:  @ %bb.11: @ %for.body.preheader1
; CHECK-NEXT:    add.w r1, r1, r12, lsl #2
; CHECK-NEXT:    sub.w r3, r3, r12
; CHECK-NEXT:    adds r1, #8
; CHECK-NEXT:    lsl.w r12, r12, #2
; CHECK-NEXT:  .LBB3_12: @ %for.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldr s0, [r1, #-8]
; CHECK-NEXT:    add.w r7, r0, r12
; CHECK-NEXT:    add.w r6, r2, r12
; CHECK-NEXT:    adds r0, #16
; CHECK-NEXT:    vcvt.f32.s32 s0, s0
; CHECK-NEXT:    vldr s2, [r7]
; CHECK-NEXT:    adds r2, #16
; CHECK-NEXT:    subs r3, #4
; CHECK-NEXT:    vmul.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r6]
; CHECK-NEXT:    vldr s0, [r1, #-4]
; CHECK-NEXT:    vldr s2, [r7, #4]
; CHECK-NEXT:    vcvt.f32.s32 s0, s0
; CHECK-NEXT:    vmul.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r6, #4]
; CHECK-NEXT:    vldr s0, [r1]
; CHECK-NEXT:    vldr s2, [r7, #8]
; CHECK-NEXT:    vcvt.f32.s32 s0, s0
; CHECK-NEXT:    vmul.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r6, #8]
; CHECK-NEXT:    vldr s0, [r1, #4]
; CHECK-NEXT:    add.w r1, r1, #16
; CHECK-NEXT:    vldr s2, [r7, #12]
; CHECK-NEXT:    vcvt.f32.s32 s0, s0
; CHECK-NEXT:    vmul.f32 s0, s2, s0
; CHECK-NEXT:    vstr s0, [r6, #12]
; CHECK-NEXT:    bne .LBB3_12
; CHECK-NEXT:  .LBB3_13: @ %for.cond.cleanup
; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, pc}
entry:
  %cmp8 = icmp eq i32 %N, 0
  br i1 %cmp8, label %for.cond.cleanup, label %for.body.preheader

for.body.preheader:                               ; preds = %entry
  %min.iters.check = icmp ult i32 %N, 4
  br i1 %min.iters.check, label %for.body.preheader16, label %vector.memcheck

for.body.preheader16:                             ; preds = %middle.block, %vector.memcheck, %for.body.preheader
  %i.09.ph = phi i32 [ 0, %vector.memcheck ], [ 0, %for.body.preheader ], [ %n.vec, %middle.block ]
  %0 = xor i32 %i.09.ph, -1
  %1 = add i32 %0, %N
  %xtraiter = and i32 %N, 3
  %lcmp.mod = icmp eq i32 %xtraiter, 0
  br i1 %lcmp.mod, label %for.body.prol.loopexit, label %for.body.prol

for.body.prol:                                    ; preds = %for.body.preheader16, %for.body.prol
  %i.09.prol = phi i32 [ %inc.prol, %for.body.prol ], [ %i.09.ph, %for.body.preheader16 ]
  %prol.iter = phi i32 [ %prol.iter.sub, %for.body.prol ], [ %xtraiter, %for.body.preheader16 ]
  %arrayidx.prol = getelementptr inbounds float, float* %a, i32 %i.09.prol
  %2 = load float, float* %arrayidx.prol, align 4
  %arrayidx1.prol = getelementptr inbounds i32, i32* %b, i32 %i.09.prol
  %3 = load i32, i32* %arrayidx1.prol, align 4
  %conv.prol = sitofp i32 %3 to float
  %mul.prol = fmul float %2, %conv.prol
  %arrayidx2.prol = getelementptr inbounds float, float* %c, i32 %i.09.prol
  store float %mul.prol, float* %arrayidx2.prol, align 4
  %inc.prol = add nuw i32 %i.09.prol, 1
  %prol.iter.sub = add i32 %prol.iter, -1
  %prol.iter.cmp = icmp eq i32 %prol.iter.sub, 0
  br i1 %prol.iter.cmp, label %for.body.prol.loopexit, label %for.body.prol

for.body.prol.loopexit:                           ; preds = %for.body.prol, %for.body.preheader16
  %i.09.unr = phi i32 [ %i.09.ph, %for.body.preheader16 ], [ %inc.prol, %for.body.prol ]
  %4 = icmp ult i32 %1, 3
  br i1 %4, label %for.cond.cleanup, label %for.body

vector.memcheck:                                  ; preds = %for.body.preheader
  %scevgep = getelementptr float, float* %c, i32 %N
  %scevgep13 = getelementptr float, float* %a, i32 %N
  %bound0 = icmp ugt float* %scevgep13, %c
  %bound1 = icmp ugt float* %scevgep, %a
  %found.conflict = and i1 %bound0, %bound1
  br i1 %found.conflict, label %for.body.preheader16, label %vector.ph

vector.ph:                                        ; preds = %vector.memcheck
  %n.vec = and i32 %N, -4
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %vector.ph
  %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
  %5 = getelementptr inbounds float, float* %a, i32 %index
  %6 = bitcast float* %5 to <4 x float>*
  %wide.load = load <4 x float>, <4 x float>* %6, align 4
  %7 = getelementptr inbounds i32, i32* %b, i32 %index
  %8 = bitcast i32* %7 to <4 x i32>*
  %wide.load15 = load <4 x i32>, <4 x i32>* %8, align 4
  %9 = sitofp <4 x i32> %wide.load15 to <4 x float>
  %10 = fmul <4 x float> %wide.load, %9
  %11 = getelementptr inbounds float, float* %c, i32 %index
  %12 = bitcast float* %11 to <4 x float>*
  store <4 x float> %10, <4 x float>* %12, align 4
  %index.next = add i32 %index, 4
  %13 = icmp eq i32 %index.next, %n.vec
  br i1 %13, label %middle.block, label %vector.body

middle.block:                                     ; preds = %vector.body
  %cmp.n = icmp eq i32 %n.vec, %N
  br i1 %cmp.n, label %for.cond.cleanup, label %for.body.preheader16

for.cond.cleanup:                                 ; preds = %for.body.prol.loopexit, %for.body, %middle.block, %entry
  ret void

for.body:                                         ; preds = %for.body.prol.loopexit, %for.body
  %i.09 = phi i32 [ %inc.3, %for.body ], [ %i.09.unr, %for.body.prol.loopexit ]
  %arrayidx = getelementptr inbounds float, float* %a, i32 %i.09
  %14 = load float, float* %arrayidx, align 4
  %arrayidx1 = getelementptr inbounds i32, i32* %b, i32 %i.09
  %15 = load i32, i32* %arrayidx1, align 4
  %conv = sitofp i32 %15 to float
  %mul = fmul float %14, %conv
  %arrayidx2 = getelementptr inbounds float, float* %c, i32 %i.09
  store float %mul, float* %arrayidx2, align 4
  %inc = add nuw i32 %i.09, 1
  %arrayidx.1 = getelementptr inbounds float, float* %a, i32 %inc
  %16 = load float, float* %arrayidx.1, align 4
  %arrayidx1.1 = getelementptr inbounds i32, i32* %b, i32 %inc
  %17 = load i32, i32* %arrayidx1.1, align 4
  %conv.1 = sitofp i32 %17 to float
  %mul.1 = fmul float %16, %conv.1
  %arrayidx2.1 = getelementptr inbounds float, float* %c, i32 %inc
  store float %mul.1, float* %arrayidx2.1, align 4
  %inc.1 = add nuw i32 %i.09, 2
  %arrayidx.2 = getelementptr inbounds float, float* %a, i32 %inc.1
  %18 = load float, float* %arrayidx.2, align 4
  %arrayidx1.2 = getelementptr inbounds i32, i32* %b, i32 %inc.1
  %19 = load i32, i32* %arrayidx1.2, align 4
  %conv.2 = sitofp i32 %19 to float
  %mul.2 = fmul float %18, %conv.2
  %arrayidx2.2 = getelementptr inbounds float, float* %c, i32 %inc.1
  store float %mul.2, float* %arrayidx2.2, align 4
  %inc.2 = add nuw i32 %i.09, 3
  %arrayidx.3 = getelementptr inbounds float, float* %a, i32 %inc.2
  %20 = load float, float* %arrayidx.3, align 4
  %arrayidx1.3 = getelementptr inbounds i32, i32* %b, i32 %inc.2
  %21 = load i32, i32* %arrayidx1.3, align 4
  %conv.3 = sitofp i32 %21 to float
  %mul.3 = fmul float %20, %conv.3
  %arrayidx2.3 = getelementptr inbounds float, float* %c, i32 %inc.2
  store float %mul.3, float* %arrayidx2.3, align 4
  %inc.3 = add nuw i32 %i.09, 4
  %exitcond.3 = icmp eq i32 %inc.3, %N
  br i1 %exitcond.3, label %for.cond.cleanup, label %for.body
}

define arm_aapcs_vfpcc void @float_int_int_mul(i32* nocapture readonly %a, i32* nocapture readonly %b, float* nocapture %c, i32 %N) {
; CHECK-LABEL: float_int_int_mul:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    push {r4, r5, r6, lr}
; CHECK-NEXT:    cmp r3, #0
; CHECK-NEXT:    beq .LBB4_8
; CHECK-NEXT:  @ %bb.1: @ %for.body.preheader
; CHECK-NEXT:    cmp r3, #3
; CHECK-NEXT:    bhi .LBB4_3
; CHECK-NEXT:  @ %bb.2:
; CHECK-NEXT:    mov.w r12, #0
; CHECK-NEXT:    b .LBB4_6
; CHECK-NEXT:  .LBB4_3: @ %vector.ph
; CHECK-NEXT:    bic r12, r3, #3
; CHECK-NEXT:    movs r5, #1
; CHECK-NEXT:    sub.w r6, r12, #4
; CHECK-NEXT:    mov r4, r0
; CHECK-NEXT:    add.w lr, r5, r6, lsr #2
; CHECK-NEXT:    mov r5, r1
; CHECK-NEXT:    mov r6, r2
; CHECK-NEXT:    dls lr, lr
; CHECK-NEXT:  .LBB4_4: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldrw.u32 q0, [r4], #16
; CHECK-NEXT:    vldrw.u32 q1, [r5], #16
; CHECK-NEXT:    vmul.i32 q0, q1, q0
; CHECK-NEXT:    vcvt.f32.s32 q0, q0
; CHECK-NEXT:    vstrb.8 q0, [r6], #16
; CHECK-NEXT:    le lr, .LBB4_4
; CHECK-NEXT:  @ %bb.5: @ %middle.block
; CHECK-NEXT:    cmp r12, r3
; CHECK-NEXT:    it eq
; CHECK-NEXT:    popeq {r4, r5, r6, pc}
; CHECK-NEXT:  .LBB4_6: @ %for.body.preheader11
; CHECK-NEXT:    sub.w lr, r3, r12
; CHECK-NEXT:    add.w r0, r0, r12, lsl #2
; CHECK-NEXT:    add.w r1, r1, r12, lsl #2
; CHECK-NEXT:    add.w r2, r2, r12, lsl #2
; CHECK-NEXT:    dls lr, lr
; CHECK-NEXT:  .LBB4_7: @ %for.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    ldr r3, [r0], #4
; CHECK-NEXT:    ldr r6, [r1], #4
; CHECK-NEXT:    muls r3, r6, r3
; CHECK-NEXT:    vmov s0, r3
; CHECK-NEXT:    vcvt.f32.s32 s0, s0
; CHECK-NEXT:    vstmia r2!, {s0}
; CHECK-NEXT:    le lr, .LBB4_7
; CHECK-NEXT:  .LBB4_8: @ %for.cond.cleanup
; CHECK-NEXT:    pop {r4, r5, r6, pc}
entry:
  %cmp8 = icmp eq i32 %N, 0
  br i1 %cmp8, label %for.cond.cleanup, label %for.body.preheader

for.body.preheader:                               ; preds = %entry
  %min.iters.check = icmp ult i32 %N, 4
  br i1 %min.iters.check, label %for.body.preheader11, label %vector.ph

for.body.preheader11:                             ; preds = %middle.block, %for.body.preheader
  %i.09.ph = phi i32 [ 0, %for.body.preheader ], [ %n.vec, %middle.block ]
  br label %for.body

vector.ph:                                        ; preds = %for.body.preheader
  %n.vec = and i32 %N, -4
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %vector.ph
  %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
  %0 = getelementptr inbounds i32, i32* %a, i32 %index
  %1 = bitcast i32* %0 to <4 x i32>*
  %wide.load = load <4 x i32>, <4 x i32>* %1, align 4
  %2 = getelementptr inbounds i32, i32* %b, i32 %index
  %3 = bitcast i32* %2 to <4 x i32>*
  %wide.load10 = load <4 x i32>, <4 x i32>* %3, align 4
  %4 = mul nsw <4 x i32> %wide.load10, %wide.load
  %5 = sitofp <4 x i32> %4 to <4 x float>
  %6 = getelementptr inbounds float, float* %c, i32 %index
  %7 = bitcast float* %6 to <4 x float>*
  store <4 x float> %5, <4 x float>* %7, align 4
  %index.next = add i32 %index, 4
  %8 = icmp eq i32 %index.next, %n.vec
  br i1 %8, label %middle.block, label %vector.body

middle.block:                                     ; preds = %vector.body
  %cmp.n = icmp eq i32 %n.vec, %N
  br i1 %cmp.n, label %for.cond.cleanup, label %for.body.preheader11

for.cond.cleanup:                                 ; preds = %for.body, %middle.block, %entry
  ret void

for.body:                                         ; preds = %for.body.preheader11, %for.body
  %i.09 = phi i32 [ %inc, %for.body ], [ %i.09.ph, %for.body.preheader11 ]
  %arrayidx = getelementptr inbounds i32, i32* %a, i32 %i.09
  %9 = load i32, i32* %arrayidx, align 4
  %arrayidx1 = getelementptr inbounds i32, i32* %b, i32 %i.09
  %10 = load i32, i32* %arrayidx1, align 4
  %mul = mul nsw i32 %10, %9
  %conv = sitofp i32 %mul to float
  %arrayidx2 = getelementptr inbounds float, float* %c, i32 %i.09
  store float %conv, float* %arrayidx2, align 4
  %inc = add nuw i32 %i.09, 1
  %exitcond = icmp eq i32 %inc, %N
  br i1 %exitcond, label %for.cond.cleanup, label %for.body
}

define arm_aapcs_vfpcc void @half_half_mul(half* nocapture readonly %a, half* nocapture readonly %b, float* nocapture %c, i32 %N) {
; CHECK-LABEL: half_half_mul:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, r9, r10, lr}
; CHECK-NEXT:    cmp r3, #0
; CHECK-NEXT:    beq .LBB5_8
; CHECK-NEXT:  @ %bb.1: @ %for.body.preheader
; CHECK-NEXT:    cmp r3, #3
; CHECK-NEXT:    bhi .LBB5_3
; CHECK-NEXT:  @ %bb.2:
; CHECK-NEXT:    mov.w r12, #0
; CHECK-NEXT:    b .LBB5_6
; CHECK-NEXT:  .LBB5_3: @ %vector.ph
; CHECK-NEXT:    bic r12, r3, #3
; CHECK-NEXT:    movs r5, #1
; CHECK-NEXT:    sub.w r6, r12, #4
; CHECK-NEXT:    mov r4, r0
; CHECK-NEXT:    add.w lr, r5, r6, lsr #2
; CHECK-NEXT:    mov r5, r1
; CHECK-NEXT:    mov r6, r2
; CHECK-NEXT:    dls lr, lr
; CHECK-NEXT:  .LBB5_4: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    ldr.w r9, [r4]
; CHECK-NEXT:    ldr r7, [r5]
; CHECK-NEXT:    ldr.w r8, [r4, #4]
; CHECK-NEXT:    vmov.32 q0[0], r9
; CHECK-NEXT:    ldr.w r10, [r5, #4]
; CHECK-NEXT:    vmov.32 q1[0], r7
; CHECK-NEXT:    vmov.32 q0[1], r8
; CHECK-NEXT:    adds r4, #8
; CHECK-NEXT:    vmov.32 q1[1], r10
; CHECK-NEXT:    adds r5, #8
; CHECK-NEXT:    vmul.f16 q0, q0, q1
; CHECK-NEXT:    vcvtt.f32.f16 s7, s1
; CHECK-NEXT:    vcvtb.f32.f16 s6, s1
; CHECK-NEXT:    vcvtt.f32.f16 s5, s0
; CHECK-NEXT:    vcvtb.f32.f16 s4, s0
; CHECK-NEXT:    vstrb.8 q1, [r6], #16
; CHECK-NEXT:    le lr, .LBB5_4
; CHECK-NEXT:  @ %bb.5: @ %middle.block
; CHECK-NEXT:    cmp r12, r3
; CHECK-NEXT:    beq .LBB5_8
; CHECK-NEXT:  .LBB5_6: @ %for.body.preheader11
; CHECK-NEXT:    sub.w lr, r3, r12
; CHECK-NEXT:    add.w r0, r0, r12, lsl #1
; CHECK-NEXT:    add.w r1, r1, r12, lsl #1
; CHECK-NEXT:    add.w r2, r2, r12, lsl #2
; CHECK-NEXT:    dls lr, lr
; CHECK-NEXT:  .LBB5_7: @ %for.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldr.16 s0, [r1]
; CHECK-NEXT:    vldr.16 s2, [r0]
; CHECK-NEXT:    adds r0, #2
; CHECK-NEXT:    adds r1, #2
; CHECK-NEXT:    vmul.f16 s0, s2, s0
; CHECK-NEXT:    vcvtb.f32.f16 s0, s0
; CHECK-NEXT:    vstr s0, [r2]
; CHECK-NEXT:    adds r2, #4
; CHECK-NEXT:    le lr, .LBB5_7
; CHECK-NEXT:  .LBB5_8: @ %for.cond.cleanup
; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, r9, r10, pc}
entry:
  %cmp8 = icmp eq i32 %N, 0
  br i1 %cmp8, label %for.cond.cleanup, label %for.body.preheader

for.body.preheader:                               ; preds = %entry
  %min.iters.check = icmp ult i32 %N, 4
  br i1 %min.iters.check, label %for.body.preheader11, label %vector.ph

for.body.preheader11:                             ; preds = %middle.block, %for.body.preheader
  %i.09.ph = phi i32 [ 0, %for.body.preheader ], [ %n.vec, %middle.block ]
  br label %for.body

vector.ph:                                        ; preds = %for.body.preheader
  %n.vec = and i32 %N, -4
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %vector.ph
  %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
  %0 = getelementptr inbounds half, half* %a, i32 %index
  %1 = bitcast half* %0 to <4 x half>*
  %wide.load = load <4 x half>, <4 x half>* %1, align 2
  %2 = getelementptr inbounds half, half* %b, i32 %index
  %3 = bitcast half* %2 to <4 x half>*
  %wide.load10 = load <4 x half>, <4 x half>* %3, align 2
  %4 = fmul <4 x half> %wide.load, %wide.load10
  %5 = fpext <4 x half> %4 to <4 x float>
  %6 = getelementptr inbounds float, float* %c, i32 %index
  %7 = bitcast float* %6 to <4 x float>*
  store <4 x float> %5, <4 x float>* %7, align 4
  %index.next = add i32 %index, 4
  %8 = icmp eq i32 %index.next, %n.vec
  br i1 %8, label %middle.block, label %vector.body

middle.block:                                     ; preds = %vector.body
  %cmp.n = icmp eq i32 %n.vec, %N
  br i1 %cmp.n, label %for.cond.cleanup, label %for.body.preheader11

for.cond.cleanup:                                 ; preds = %for.body, %middle.block, %entry
  ret void

for.body:                                         ; preds = %for.body.preheader11, %for.body
  %i.09 = phi i32 [ %inc, %for.body ], [ %i.09.ph, %for.body.preheader11 ]
  %arrayidx = getelementptr inbounds half, half* %a, i32 %i.09
  %9 = load half, half* %arrayidx, align 2
  %arrayidx1 = getelementptr inbounds half, half* %b, i32 %i.09
  %10 = load half, half* %arrayidx1, align 2
  %mul = fmul half %9, %10
  %conv = fpext half %mul to float
  %arrayidx2 = getelementptr inbounds float, float* %c, i32 %i.09
  store float %conv, float* %arrayidx2, align 4
  %inc = add nuw i32 %i.09, 1
  %exitcond = icmp eq i32 %inc, %N
  br i1 %exitcond, label %for.cond.cleanup, label %for.body
}

define arm_aapcs_vfpcc void @half_half_add(half* nocapture readonly %a, half* nocapture readonly %b, float* nocapture %c, i32 %N) {
; CHECK-LABEL: half_half_add:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, r9, r10, lr}
; CHECK-NEXT:    cmp r3, #0
; CHECK-NEXT:    beq .LBB6_8
; CHECK-NEXT:  @ %bb.1: @ %for.body.preheader
; CHECK-NEXT:    cmp r3, #3
; CHECK-NEXT:    bhi .LBB6_3
; CHECK-NEXT:  @ %bb.2:
; CHECK-NEXT:    mov.w r12, #0
; CHECK-NEXT:    b .LBB6_6
; CHECK-NEXT:  .LBB6_3: @ %vector.ph
; CHECK-NEXT:    bic r12, r3, #3
; CHECK-NEXT:    movs r5, #1
; CHECK-NEXT:    sub.w r6, r12, #4
; CHECK-NEXT:    mov r4, r0
; CHECK-NEXT:    add.w lr, r5, r6, lsr #2
; CHECK-NEXT:    mov r5, r1
; CHECK-NEXT:    mov r6, r2
; CHECK-NEXT:    dls lr, lr
; CHECK-NEXT:  .LBB6_4: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    ldr.w r9, [r4]
; CHECK-NEXT:    ldr r7, [r5]
; CHECK-NEXT:    ldr.w r8, [r4, #4]
; CHECK-NEXT:    vmov.32 q0[0], r9
; CHECK-NEXT:    ldr.w r10, [r5, #4]
; CHECK-NEXT:    vmov.32 q1[0], r7
; CHECK-NEXT:    vmov.32 q0[1], r8
; CHECK-NEXT:    adds r4, #8
; CHECK-NEXT:    vmov.32 q1[1], r10
; CHECK-NEXT:    adds r5, #8
; CHECK-NEXT:    vadd.f16 q0, q0, q1
; CHECK-NEXT:    vcvtt.f32.f16 s7, s1
; CHECK-NEXT:    vcvtb.f32.f16 s6, s1
; CHECK-NEXT:    vcvtt.f32.f16 s5, s0
; CHECK-NEXT:    vcvtb.f32.f16 s4, s0
; CHECK-NEXT:    vstrb.8 q1, [r6], #16
; CHECK-NEXT:    le lr, .LBB6_4
; CHECK-NEXT:  @ %bb.5: @ %middle.block
; CHECK-NEXT:    cmp r12, r3
; CHECK-NEXT:    beq .LBB6_8
; CHECK-NEXT:  .LBB6_6: @ %for.body.preheader11
; CHECK-NEXT:    sub.w lr, r3, r12
; CHECK-NEXT:    add.w r0, r0, r12, lsl #1
; CHECK-NEXT:    add.w r1, r1, r12, lsl #1
; CHECK-NEXT:    add.w r2, r2, r12, lsl #2
; CHECK-NEXT:    dls lr, lr
; CHECK-NEXT:  .LBB6_7: @ %for.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldr.16 s0, [r1]
; CHECK-NEXT:    vldr.16 s2, [r0]
; CHECK-NEXT:    adds r0, #2
; CHECK-NEXT:    adds r1, #2
; CHECK-NEXT:    vadd.f16 s0, s2, s0
; CHECK-NEXT:    vcvtb.f32.f16 s0, s0
; CHECK-NEXT:    vstr s0, [r2]
; CHECK-NEXT:    adds r2, #4
; CHECK-NEXT:    le lr, .LBB6_7
; CHECK-NEXT:  .LBB6_8: @ %for.cond.cleanup
; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, r9, r10, pc}
entry:
  %cmp8 = icmp eq i32 %N, 0
  br i1 %cmp8, label %for.cond.cleanup, label %for.body.preheader

for.body.preheader:                               ; preds = %entry
  %min.iters.check = icmp ult i32 %N, 4
  br i1 %min.iters.check, label %for.body.preheader11, label %vector.ph

for.body.preheader11:                             ; preds = %middle.block, %for.body.preheader
  %i.09.ph = phi i32 [ 0, %for.body.preheader ], [ %n.vec, %middle.block ]
  br label %for.body

vector.ph:                                        ; preds = %for.body.preheader
  %n.vec = and i32 %N, -4
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %vector.ph
  %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
  %0 = getelementptr inbounds half, half* %a, i32 %index
  %1 = bitcast half* %0 to <4 x half>*
  %wide.load = load <4 x half>, <4 x half>* %1, align 2
  %2 = getelementptr inbounds half, half* %b, i32 %index
  %3 = bitcast half* %2 to <4 x half>*
  %wide.load10 = load <4 x half>, <4 x half>* %3, align 2
  %4 = fadd <4 x half> %wide.load, %wide.load10
  %5 = fpext <4 x half> %4 to <4 x float>
  %6 = getelementptr inbounds float, float* %c, i32 %index
  %7 = bitcast float* %6 to <4 x float>*
  store <4 x float> %5, <4 x float>* %7, align 4
  %index.next = add i32 %index, 4
  %8 = icmp eq i32 %index.next, %n.vec
  br i1 %8, label %middle.block, label %vector.body

middle.block:                                     ; preds = %vector.body
  %cmp.n = icmp eq i32 %n.vec, %N
  br i1 %cmp.n, label %for.cond.cleanup, label %for.body.preheader11

for.cond.cleanup:                                 ; preds = %for.body, %middle.block, %entry
  ret void

for.body:                                         ; preds = %for.body.preheader11, %for.body
  %i.09 = phi i32 [ %inc, %for.body ], [ %i.09.ph, %for.body.preheader11 ]
  %arrayidx = getelementptr inbounds half, half* %a, i32 %i.09
  %9 = load half, half* %arrayidx, align 2
  %arrayidx1 = getelementptr inbounds half, half* %b, i32 %i.09
  %10 = load half, half* %arrayidx1, align 2
  %add = fadd half %9, %10
  %conv = fpext half %add to float
  %arrayidx2 = getelementptr inbounds float, float* %c, i32 %i.09
  store float %conv, float* %arrayidx2, align 4
  %inc = add nuw i32 %i.09, 1
  %exitcond = icmp eq i32 %inc, %N
  br i1 %exitcond, label %for.cond.cleanup, label %for.body
}

define arm_aapcs_vfpcc void @half_half_sub(half* nocapture readonly %a, half* nocapture readonly %b, float* nocapture %c, i32 %N) {
; CHECK-LABEL: half_half_sub:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, r9, r10, lr}
; CHECK-NEXT:    cmp r3, #0
; CHECK-NEXT:    beq .LBB7_8
; CHECK-NEXT:  @ %bb.1: @ %for.body.preheader
; CHECK-NEXT:    cmp r3, #3
; CHECK-NEXT:    bhi .LBB7_3
; CHECK-NEXT:  @ %bb.2:
; CHECK-NEXT:    mov.w r12, #0
; CHECK-NEXT:    b .LBB7_6
; CHECK-NEXT:  .LBB7_3: @ %vector.ph
; CHECK-NEXT:    bic r12, r3, #3
; CHECK-NEXT:    movs r5, #1
; CHECK-NEXT:    sub.w r6, r12, #4
; CHECK-NEXT:    mov r4, r0
; CHECK-NEXT:    add.w lr, r5, r6, lsr #2
; CHECK-NEXT:    mov r5, r1
; CHECK-NEXT:    mov r6, r2
; CHECK-NEXT:    dls lr, lr
; CHECK-NEXT:  .LBB7_4: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    ldr.w r9, [r4]
; CHECK-NEXT:    ldr r7, [r5]
; CHECK-NEXT:    ldr.w r8, [r4, #4]
; CHECK-NEXT:    vmov.32 q0[0], r9
; CHECK-NEXT:    ldr.w r10, [r5, #4]
; CHECK-NEXT:    vmov.32 q1[0], r7
; CHECK-NEXT:    vmov.32 q0[1], r8
; CHECK-NEXT:    adds r4, #8
; CHECK-NEXT:    vmov.32 q1[1], r10
; CHECK-NEXT:    adds r5, #8
; CHECK-NEXT:    vsub.f16 q0, q0, q1
; CHECK-NEXT:    vcvtt.f32.f16 s7, s1
; CHECK-NEXT:    vcvtb.f32.f16 s6, s1
; CHECK-NEXT:    vcvtt.f32.f16 s5, s0
; CHECK-NEXT:    vcvtb.f32.f16 s4, s0
; CHECK-NEXT:    vstrb.8 q1, [r6], #16
; CHECK-NEXT:    le lr, .LBB7_4
; CHECK-NEXT:  @ %bb.5: @ %middle.block
; CHECK-NEXT:    cmp r12, r3
; CHECK-NEXT:    beq .LBB7_8
; CHECK-NEXT:  .LBB7_6: @ %for.body.preheader11
; CHECK-NEXT:    sub.w lr, r3, r12
; CHECK-NEXT:    add.w r0, r0, r12, lsl #1
; CHECK-NEXT:    add.w r1, r1, r12, lsl #1
; CHECK-NEXT:    add.w r2, r2, r12, lsl #2
; CHECK-NEXT:    dls lr, lr
; CHECK-NEXT:  .LBB7_7: @ %for.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldr.16 s0, [r1]
; CHECK-NEXT:    vldr.16 s2, [r0]
; CHECK-NEXT:    adds r0, #2
; CHECK-NEXT:    adds r1, #2
; CHECK-NEXT:    vsub.f16 s0, s2, s0
; CHECK-NEXT:    vcvtb.f32.f16 s0, s0
; CHECK-NEXT:    vstr s0, [r2]
; CHECK-NEXT:    adds r2, #4
; CHECK-NEXT:    le lr, .LBB7_7
; CHECK-NEXT:  .LBB7_8: @ %for.cond.cleanup
; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, r9, r10, pc}
entry:
  %cmp8 = icmp eq i32 %N, 0
  br i1 %cmp8, label %for.cond.cleanup, label %for.body.preheader

for.body.preheader:                               ; preds = %entry
  %min.iters.check = icmp ult i32 %N, 4
  br i1 %min.iters.check, label %for.body.preheader11, label %vector.ph

for.body.preheader11:                             ; preds = %middle.block, %for.body.preheader
  %i.09.ph = phi i32 [ 0, %for.body.preheader ], [ %n.vec, %middle.block ]
  br label %for.body

vector.ph:                                        ; preds = %for.body.preheader
  %n.vec = and i32 %N, -4
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %vector.ph
  %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
  %0 = getelementptr inbounds half, half* %a, i32 %index
  %1 = bitcast half* %0 to <4 x half>*
  %wide.load = load <4 x half>, <4 x half>* %1, align 2
  %2 = getelementptr inbounds half, half* %b, i32 %index
  %3 = bitcast half* %2 to <4 x half>*
  %wide.load10 = load <4 x half>, <4 x half>* %3, align 2
  %4 = fsub <4 x half> %wide.load, %wide.load10
  %5 = fpext <4 x half> %4 to <4 x float>
  %6 = getelementptr inbounds float, float* %c, i32 %index
  %7 = bitcast float* %6 to <4 x float>*
  store <4 x float> %5, <4 x float>* %7, align 4
  %index.next = add i32 %index, 4
  %8 = icmp eq i32 %index.next, %n.vec
  br i1 %8, label %middle.block, label %vector.body

middle.block:                                     ; preds = %vector.body
  %cmp.n = icmp eq i32 %n.vec, %N
  br i1 %cmp.n, label %for.cond.cleanup, label %for.body.preheader11

for.cond.cleanup:                                 ; preds = %for.body, %middle.block, %entry
  ret void

for.body:                                         ; preds = %for.body.preheader11, %for.body
  %i.09 = phi i32 [ %inc, %for.body ], [ %i.09.ph, %for.body.preheader11 ]
  %arrayidx = getelementptr inbounds half, half* %a, i32 %i.09
  %9 = load half, half* %arrayidx, align 2
  %arrayidx1 = getelementptr inbounds half, half* %b, i32 %i.09
  %10 = load half, half* %arrayidx1, align 2
  %sub = fsub half %9, %10
  %conv = fpext half %sub to float
  %arrayidx2 = getelementptr inbounds float, float* %c, i32 %i.09
  store float %conv, float* %arrayidx2, align 4
  %inc = add nuw i32 %i.09, 1
  %exitcond = icmp eq i32 %inc, %N
  br i1 %exitcond, label %for.cond.cleanup, label %for.body
}

define arm_aapcs_vfpcc void @half_short_mul(half* nocapture readonly %a, i16* nocapture readonly %b, float* nocapture %c, i32 %N) {
; CHECK-LABEL: half_short_mul:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, r9, lr}
; CHECK-NEXT:    cmp r3, #0
; CHECK-NEXT:    beq .LBB8_8
; CHECK-NEXT:  @ %bb.1: @ %for.body.preheader
; CHECK-NEXT:    cmp r3, #3
; CHECK-NEXT:    bhi .LBB8_3
; CHECK-NEXT:  @ %bb.2:
; CHECK-NEXT:    mov.w r12, #0
; CHECK-NEXT:    b .LBB8_6
; CHECK-NEXT:  .LBB8_3: @ %vector.ph
; CHECK-NEXT:    bic r12, r3, #3
; CHECK-NEXT:    movs r5, #1
; CHECK-NEXT:    sub.w r6, r12, #4
; CHECK-NEXT:    mov r4, r0
; CHECK-NEXT:    add.w lr, r5, r6, lsr #2
; CHECK-NEXT:    mov r5, r1
; CHECK-NEXT:    mov r6, r2
; CHECK-NEXT:    dls lr, lr
; CHECK-NEXT:  .LBB8_4: @ %vector.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldrh.u32 q0, [r5], #8
; CHECK-NEXT:    ldr.w r9, [r4]
; CHECK-NEXT:    ldr.w r8, [r4, #4]
; CHECK-NEXT:    adds r4, #8
; CHECK-NEXT:    vmov r7, s0
; CHECK-NEXT:    vmov.16 q1[0], r7
; CHECK-NEXT:    vmov r7, s1
; CHECK-NEXT:    vmov.16 q1[1], r7
; CHECK-NEXT:    vmov r7, s2
; CHECK-NEXT:    vmov.16 q1[2], r7
; CHECK-NEXT:    vmov r7, s3
; CHECK-NEXT:    vmov.16 q1[3], r7
; CHECK-NEXT:    vcvt.f16.s16 q0, q1
; CHECK-NEXT:    vmov.32 q1[0], r9
; CHECK-NEXT:    vmov.32 q1[1], r8
; CHECK-NEXT:    vmul.f16 q0, q1, q0
; CHECK-NEXT:    vcvtt.f32.f16 s7, s1
; CHECK-NEXT:    vcvtb.f32.f16 s6, s1
; CHECK-NEXT:    vcvtt.f32.f16 s5, s0
; CHECK-NEXT:    vcvtb.f32.f16 s4, s0
; CHECK-NEXT:    vstrb.8 q1, [r6], #16
; CHECK-NEXT:    le lr, .LBB8_4
; CHECK-NEXT:  @ %bb.5: @ %middle.block
; CHECK-NEXT:    cmp r12, r3
; CHECK-NEXT:    beq .LBB8_8
; CHECK-NEXT:  .LBB8_6: @ %for.body.preheader13
; CHECK-NEXT:    sub.w lr, r3, r12
; CHECK-NEXT:    add.w r0, r0, r12, lsl #1
; CHECK-NEXT:    add.w r1, r1, r12, lsl #1
; CHECK-NEXT:    add.w r2, r2, r12, lsl #2
; CHECK-NEXT:    dls lr, lr
; CHECK-NEXT:  .LBB8_7: @ %for.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    ldrsh r3, [r1], #2
; CHECK-NEXT:    vldr.16 s0, [r0]
; CHECK-NEXT:    adds r0, #2
; CHECK-NEXT:    vmov s2, r3
; CHECK-NEXT:    vcvt.f16.s32 s2, s2
; CHECK-NEXT:    vmul.f16 s0, s0, s2
; CHECK-NEXT:    vcvtb.f32.f16 s0, s0
; CHECK-NEXT:    vstr s0, [r2]
; CHECK-NEXT:    adds r2, #4
; CHECK-NEXT:    le lr, .LBB8_7
; CHECK-NEXT:  .LBB8_8: @ %for.cond.cleanup
; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, r9, pc}
entry:
  %cmp10 = icmp eq i32 %N, 0
  br i1 %cmp10, label %for.cond.cleanup, label %for.body.preheader

for.body.preheader:                               ; preds = %entry
  %min.iters.check = icmp ult i32 %N, 4
  br i1 %min.iters.check, label %for.body.preheader13, label %vector.ph

for.body.preheader13:                             ; preds = %middle.block, %for.body.preheader
  %i.011.ph = phi i32 [ 0, %for.body.preheader ], [ %n.vec, %middle.block ]
  br label %for.body

vector.ph:                                        ; preds = %for.body.preheader
  %n.vec = and i32 %N, -4
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %vector.ph
  %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
  %0 = getelementptr inbounds half, half* %a, i32 %index
  %1 = bitcast half* %0 to <4 x half>*
  %wide.load = load <4 x half>, <4 x half>* %1, align 2
  %2 = getelementptr inbounds i16, i16* %b, i32 %index
  %3 = bitcast i16* %2 to <4 x i16>*
  %wide.load12 = load <4 x i16>, <4 x i16>* %3, align 2
  %4 = sitofp <4 x i16> %wide.load12 to <4 x half>
  %5 = fmul <4 x half> %wide.load, %4
  %6 = fpext <4 x half> %5 to <4 x float>
  %7 = getelementptr inbounds float, float* %c, i32 %index
  %8 = bitcast float* %7 to <4 x float>*
  store <4 x float> %6, <4 x float>* %8, align 4
  %index.next = add i32 %index, 4
  %9 = icmp eq i32 %index.next, %n.vec
  br i1 %9, label %middle.block, label %vector.body

middle.block:                                     ; preds = %vector.body
  %cmp.n = icmp eq i32 %n.vec, %N
  br i1 %cmp.n, label %for.cond.cleanup, label %for.body.preheader13

for.cond.cleanup:                                 ; preds = %for.body, %middle.block, %entry
  ret void

for.body:                                         ; preds = %for.body.preheader13, %for.body
  %i.011 = phi i32 [ %inc, %for.body ], [ %i.011.ph, %for.body.preheader13 ]
  %arrayidx = getelementptr inbounds half, half* %a, i32 %i.011
  %10 = load half, half* %arrayidx, align 2
  %arrayidx1 = getelementptr inbounds i16, i16* %b, i32 %i.011
  %11 = load i16, i16* %arrayidx1, align 2
  %conv2 = sitofp i16 %11 to half
  %mul = fmul half %10, %conv2
  %conv3 = fpext half %mul to float
  %arrayidx4 = getelementptr inbounds float, float* %c, i32 %i.011
  store float %conv3, float* %arrayidx4, align 4
  %inc = add nuw i32 %i.011, 1
  %exitcond = icmp eq i32 %inc, %N
  br i1 %exitcond, label %for.cond.cleanup, label %for.body
}

define arm_aapcs_vfpcc float @half_half_mac(half* nocapture readonly %a, half* nocapture readonly %b, i32 %N) {
; CHECK-LABEL: half_half_mac:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    push {r4, r5, r7, lr}
; CHECK-NEXT:    cbz r2, .LBB9_3
; CHECK-NEXT:  @ %bb.1: @ %for.body.preheader
; CHECK-NEXT:    subs r3, r2, #1
; CHECK-NEXT:    and r5, r2, #3
; CHECK-NEXT:    cmp r3, #3
; CHECK-NEXT:    bhs .LBB9_4
; CHECK-NEXT:  @ %bb.2:
; CHECK-NEXT:    vldr s0, .LCPI9_0
; CHECK-NEXT:    mov.w r12, #0
; CHECK-NEXT:    b .LBB9_6
; CHECK-NEXT:  .LBB9_3:
; CHECK-NEXT:    vldr s0, .LCPI9_0
; CHECK-NEXT:    b .LBB9_9
; CHECK-NEXT:  .LBB9_4: @ %for.body.preheader.new
; CHECK-NEXT:    bic r2, r2, #3
; CHECK-NEXT:    movs r3, #1
; CHECK-NEXT:    subs r2, #4
; CHECK-NEXT:    vldr s0, .LCPI9_0
; CHECK-NEXT:    mov.w r12, #0
; CHECK-NEXT:    add.w lr, r3, r2, lsr #2
; CHECK-NEXT:    movs r3, #0
; CHECK-NEXT:    dls lr, lr
; CHECK-NEXT:  .LBB9_5: @ %for.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    adds r4, r0, r3
; CHECK-NEXT:    adds r2, r1, r3
; CHECK-NEXT:    vldr.16 s2, [r2, #6]
; CHECK-NEXT:    vldr.16 s4, [r4, #6]
; CHECK-NEXT:    vldr.16 s6, [r4, #4]
; CHECK-NEXT:    vldr.16 s8, [r4, #2]
; CHECK-NEXT:    vmul.f16 s2, s4, s2
; CHECK-NEXT:    vldr.16 s4, [r2, #4]
; CHECK-NEXT:    vldr.16 s10, [r4]
; CHECK-NEXT:    vcvtb.f32.f16 s2, s2
; CHECK-NEXT:    vmul.f16 s4, s6, s4
; CHECK-NEXT:    vldr.16 s6, [r2, #2]
; CHECK-NEXT:    vcvtb.f32.f16 s4, s4
; CHECK-NEXT:    adds r3, #8
; CHECK-NEXT:    vmul.f16 s6, s8, s6
; CHECK-NEXT:    vldr.16 s8, [r2]
; CHECK-NEXT:    vcvtb.f32.f16 s6, s6
; CHECK-NEXT:    add.w r12, r12, #4
; CHECK-NEXT:    vmul.f16 s8, s10, s8
; CHECK-NEXT:    vcvtb.f32.f16 s8, s8
; CHECK-NEXT:    vadd.f32 s0, s0, s8
; CHECK-NEXT:    vadd.f32 s0, s0, s6
; CHECK-NEXT:    vadd.f32 s0, s0, s4
; CHECK-NEXT:    vadd.f32 s0, s0, s2
; CHECK-NEXT:    le lr, .LBB9_5
; CHECK-NEXT:  .LBB9_6: @ %for.cond.cleanup.loopexit.unr-lcssa
; CHECK-NEXT:    wls lr, r5, .LBB9_9
; CHECK-NEXT:  @ %bb.7: @ %for.body.epil.preheader
; CHECK-NEXT:    add.w r0, r0, r12, lsl #1
; CHECK-NEXT:    add.w r1, r1, r12, lsl #1
; CHECK-NEXT:    mov lr, r5
; CHECK-NEXT:  .LBB9_8: @ %for.body.epil
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldr.16 s2, [r1]
; CHECK-NEXT:    vldr.16 s4, [r0]
; CHECK-NEXT:    adds r0, #2
; CHECK-NEXT:    adds r1, #2
; CHECK-NEXT:    vmul.f16 s2, s4, s2
; CHECK-NEXT:    vcvtb.f32.f16 s2, s2
; CHECK-NEXT:    vadd.f32 s0, s0, s2
; CHECK-NEXT:    le lr, .LBB9_8
; CHECK-NEXT:  .LBB9_9: @ %for.cond.cleanup
; CHECK-NEXT:    pop {r4, r5, r7, pc}
; CHECK-NEXT:    .p2align 2
; CHECK-NEXT:  @ %bb.10:
; CHECK-NEXT:  .LCPI9_0:
; CHECK-NEXT:    .long 0x00000000 @ float 0
entry:
  %cmp8 = icmp eq i32 %N, 0
  br i1 %cmp8, label %for.cond.cleanup, label %for.body.preheader

for.body.preheader:                               ; preds = %entry
  %0 = add i32 %N, -1
  %xtraiter = and i32 %N, 3
  %1 = icmp ult i32 %0, 3
  br i1 %1, label %for.cond.cleanup.loopexit.unr-lcssa, label %for.body.preheader.new

for.body.preheader.new:                           ; preds = %for.body.preheader
  %unroll_iter = sub i32 %N, %xtraiter
  br label %for.body

for.cond.cleanup.loopexit.unr-lcssa:              ; preds = %for.body, %for.body.preheader
  %add.lcssa.ph = phi float [ undef, %for.body.preheader ], [ %add.3, %for.body ]
  %i.010.unr = phi i32 [ 0, %for.body.preheader ], [ %inc.3, %for.body ]
  %res.09.unr = phi float [ 0.000000e+00, %for.body.preheader ], [ %add.3, %for.body ]
  %lcmp.mod = icmp eq i32 %xtraiter, 0
  br i1 %lcmp.mod, label %for.cond.cleanup, label %for.body.epil

for.body.epil:                                    ; preds = %for.cond.cleanup.loopexit.unr-lcssa, %for.body.epil
  %i.010.epil = phi i32 [ %inc.epil, %for.body.epil ], [ %i.010.unr, %for.cond.cleanup.loopexit.unr-lcssa ]
  %res.09.epil = phi float [ %add.epil, %for.body.epil ], [ %res.09.unr, %for.cond.cleanup.loopexit.unr-lcssa ]
  %epil.iter = phi i32 [ %epil.iter.sub, %for.body.epil ], [ %xtraiter, %for.cond.cleanup.loopexit.unr-lcssa ]
  %arrayidx.epil = getelementptr inbounds half, half* %a, i32 %i.010.epil
  %2 = load half, half* %arrayidx.epil, align 2
  %arrayidx1.epil = getelementptr inbounds half, half* %b, i32 %i.010.epil
  %3 = load half, half* %arrayidx1.epil, align 2
  %mul.epil = fmul half %2, %3
  %conv.epil = fpext half %mul.epil to float
  %add.epil = fadd float %res.09.epil, %conv.epil
  %inc.epil = add nuw i32 %i.010.epil, 1
  %epil.iter.sub = add i32 %epil.iter, -1
  %epil.iter.cmp = icmp eq i32 %epil.iter.sub, 0
  br i1 %epil.iter.cmp, label %for.cond.cleanup, label %for.body.epil

for.cond.cleanup:                                 ; preds = %for.cond.cleanup.loopexit.unr-lcssa, %for.body.epil, %entry
  %res.0.lcssa = phi float [ 0.000000e+00, %entry ], [ %add.lcssa.ph, %for.cond.cleanup.loopexit.unr-lcssa ], [ %add.epil, %for.body.epil ]
  ret float %res.0.lcssa

for.body:                                         ; preds = %for.body, %for.body.preheader.new
  %i.010 = phi i32 [ 0, %for.body.preheader.new ], [ %inc.3, %for.body ]
  %res.09 = phi float [ 0.000000e+00, %for.body.preheader.new ], [ %add.3, %for.body ]
  %niter = phi i32 [ %unroll_iter, %for.body.preheader.new ], [ %niter.nsub.3, %for.body ]
  %arrayidx = getelementptr inbounds half, half* %a, i32 %i.010
  %4 = load half, half* %arrayidx, align 2
  %arrayidx1 = getelementptr inbounds half, half* %b, i32 %i.010
  %5 = load half, half* %arrayidx1, align 2
  %mul = fmul half %4, %5
  %conv = fpext half %mul to float
  %add = fadd float %res.09, %conv
  %inc = or i32 %i.010, 1
  %arrayidx.1 = getelementptr inbounds half, half* %a, i32 %inc
  %6 = load half, half* %arrayidx.1, align 2
  %arrayidx1.1 = getelementptr inbounds half, half* %b, i32 %inc
  %7 = load half, half* %arrayidx1.1, align 2
  %mul.1 = fmul half %6, %7
  %conv.1 = fpext half %mul.1 to float
  %add.1 = fadd float %add, %conv.1
  %inc.1 = or i32 %i.010, 2
  %arrayidx.2 = getelementptr inbounds half, half* %a, i32 %inc.1
  %8 = load half, half* %arrayidx.2, align 2
  %arrayidx1.2 = getelementptr inbounds half, half* %b, i32 %inc.1
  %9 = load half, half* %arrayidx1.2, align 2
  %mul.2 = fmul half %8, %9
  %conv.2 = fpext half %mul.2 to float
  %add.2 = fadd float %add.1, %conv.2
  %inc.2 = or i32 %i.010, 3
  %arrayidx.3 = getelementptr inbounds half, half* %a, i32 %inc.2
  %10 = load half, half* %arrayidx.3, align 2
  %arrayidx1.3 = getelementptr inbounds half, half* %b, i32 %inc.2
  %11 = load half, half* %arrayidx1.3, align 2
  %mul.3 = fmul half %10, %11
  %conv.3 = fpext half %mul.3 to float
  %add.3 = fadd float %add.2, %conv.3
  %inc.3 = add nuw i32 %i.010, 4
  %niter.nsub.3 = add i32 %niter, -4
  %niter.ncmp.3 = icmp eq i32 %niter.nsub.3, 0
  br i1 %niter.ncmp.3, label %for.cond.cleanup.loopexit.unr-lcssa, label %for.body
}

define arm_aapcs_vfpcc float @half_half_acc(half* nocapture readonly %a, half* nocapture readonly %b, i32 %N) {
; CHECK-LABEL: half_half_acc:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    push {r4, r5, r7, lr}
; CHECK-NEXT:    cbz r2, .LBB10_3
; CHECK-NEXT:  @ %bb.1: @ %for.body.preheader
; CHECK-NEXT:    subs r3, r2, #1
; CHECK-NEXT:    and r5, r2, #3
; CHECK-NEXT:    cmp r3, #3
; CHECK-NEXT:    bhs .LBB10_4
; CHECK-NEXT:  @ %bb.2:
; CHECK-NEXT:    vldr s0, .LCPI10_0
; CHECK-NEXT:    mov.w r12, #0
; CHECK-NEXT:    b .LBB10_6
; CHECK-NEXT:  .LBB10_3:
; CHECK-NEXT:    vldr s0, .LCPI10_0
; CHECK-NEXT:    b .LBB10_9
; CHECK-NEXT:  .LBB10_4: @ %for.body.preheader.new
; CHECK-NEXT:    bic r2, r2, #3
; CHECK-NEXT:    movs r3, #1
; CHECK-NEXT:    subs r2, #4
; CHECK-NEXT:    vldr s0, .LCPI10_0
; CHECK-NEXT:    mov.w r12, #0
; CHECK-NEXT:    add.w lr, r3, r2, lsr #2
; CHECK-NEXT:    movs r3, #0
; CHECK-NEXT:    dls lr, lr
; CHECK-NEXT:  .LBB10_5: @ %for.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    adds r4, r0, r3
; CHECK-NEXT:    adds r2, r1, r3
; CHECK-NEXT:    vldr.16 s2, [r2, #6]
; CHECK-NEXT:    vldr.16 s4, [r4, #6]
; CHECK-NEXT:    vldr.16 s6, [r4, #4]
; CHECK-NEXT:    vldr.16 s8, [r4, #2]
; CHECK-NEXT:    vadd.f16 s2, s4, s2
; CHECK-NEXT:    vldr.16 s4, [r2, #4]
; CHECK-NEXT:    vldr.16 s10, [r4]
; CHECK-NEXT:    vcvtb.f32.f16 s2, s2
; CHECK-NEXT:    vadd.f16 s4, s6, s4
; CHECK-NEXT:    vldr.16 s6, [r2, #2]
; CHECK-NEXT:    vcvtb.f32.f16 s4, s4
; CHECK-NEXT:    adds r3, #8
; CHECK-NEXT:    vadd.f16 s6, s8, s6
; CHECK-NEXT:    vldr.16 s8, [r2]
; CHECK-NEXT:    vcvtb.f32.f16 s6, s6
; CHECK-NEXT:    add.w r12, r12, #4
; CHECK-NEXT:    vadd.f16 s8, s10, s8
; CHECK-NEXT:    vcvtb.f32.f16 s8, s8
; CHECK-NEXT:    vadd.f32 s0, s0, s8
; CHECK-NEXT:    vadd.f32 s0, s0, s6
; CHECK-NEXT:    vadd.f32 s0, s0, s4
; CHECK-NEXT:    vadd.f32 s0, s0, s2
; CHECK-NEXT:    le lr, .LBB10_5
; CHECK-NEXT:  .LBB10_6: @ %for.cond.cleanup.loopexit.unr-lcssa
; CHECK-NEXT:    wls lr, r5, .LBB10_9
; CHECK-NEXT:  @ %bb.7: @ %for.body.epil.preheader
; CHECK-NEXT:    add.w r0, r0, r12, lsl #1
; CHECK-NEXT:    add.w r1, r1, r12, lsl #1
; CHECK-NEXT:    mov lr, r5
; CHECK-NEXT:  .LBB10_8: @ %for.body.epil
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    vldr.16 s2, [r1]
; CHECK-NEXT:    vldr.16 s4, [r0]
; CHECK-NEXT:    adds r0, #2
; CHECK-NEXT:    adds r1, #2
; CHECK-NEXT:    vadd.f16 s2, s4, s2
; CHECK-NEXT:    vcvtb.f32.f16 s2, s2
; CHECK-NEXT:    vadd.f32 s0, s0, s2
; CHECK-NEXT:    le lr, .LBB10_8
; CHECK-NEXT:  .LBB10_9: @ %for.cond.cleanup
; CHECK-NEXT:    pop {r4, r5, r7, pc}
; CHECK-NEXT:    .p2align 2
; CHECK-NEXT:  @ %bb.10:
; CHECK-NEXT:  .LCPI10_0:
; CHECK-NEXT:    .long 0x00000000 @ float 0
entry:
  %cmp9 = icmp eq i32 %N, 0
  br i1 %cmp9, label %for.cond.cleanup, label %for.body.preheader

for.body.preheader:                               ; preds = %entry
  %0 = add i32 %N, -1
  %xtraiter = and i32 %N, 3
  %1 = icmp ult i32 %0, 3
  br i1 %1, label %for.cond.cleanup.loopexit.unr-lcssa, label %for.body.preheader.new

for.body.preheader.new:                           ; preds = %for.body.preheader
  %unroll_iter = sub i32 %N, %xtraiter
  br label %for.body

for.cond.cleanup.loopexit.unr-lcssa:              ; preds = %for.body, %for.body.preheader
  %add2.lcssa.ph = phi float [ undef, %for.body.preheader ], [ %add2.3, %for.body ]
  %i.011.unr = phi i32 [ 0, %for.body.preheader ], [ %inc.3, %for.body ]
  %res.010.unr = phi float [ 0.000000e+00, %for.body.preheader ], [ %add2.3, %for.body ]
  %lcmp.mod = icmp eq i32 %xtraiter, 0
  br i1 %lcmp.mod, label %for.cond.cleanup, label %for.body.epil

for.body.epil:                                    ; preds = %for.cond.cleanup.loopexit.unr-lcssa, %for.body.epil
  %i.011.epil = phi i32 [ %inc.epil, %for.body.epil ], [ %i.011.unr, %for.cond.cleanup.loopexit.unr-lcssa ]
  %res.010.epil = phi float [ %add2.epil, %for.body.epil ], [ %res.010.unr, %for.cond.cleanup.loopexit.unr-lcssa ]
  %epil.iter = phi i32 [ %epil.iter.sub, %for.body.epil ], [ %xtraiter, %for.cond.cleanup.loopexit.unr-lcssa ]
  %arrayidx.epil = getelementptr inbounds half, half* %a, i32 %i.011.epil
  %2 = load half, half* %arrayidx.epil, align 2
  %arrayidx1.epil = getelementptr inbounds half, half* %b, i32 %i.011.epil
  %3 = load half, half* %arrayidx1.epil, align 2
  %add.epil = fadd half %2, %3
  %conv.epil = fpext half %add.epil to float
  %add2.epil = fadd float %res.010.epil, %conv.epil
  %inc.epil = add nuw i32 %i.011.epil, 1
  %epil.iter.sub = add i32 %epil.iter, -1
  %epil.iter.cmp = icmp eq i32 %epil.iter.sub, 0
  br i1 %epil.iter.cmp, label %for.cond.cleanup, label %for.body.epil

for.cond.cleanup:                                 ; preds = %for.cond.cleanup.loopexit.unr-lcssa, %for.body.epil, %entry
  %res.0.lcssa = phi float [ 0.000000e+00, %entry ], [ %add2.lcssa.ph, %for.cond.cleanup.loopexit.unr-lcssa ], [ %add2.epil, %for.body.epil ]
  ret float %res.0.lcssa

for.body:                                         ; preds = %for.body, %for.body.preheader.new
  %i.011 = phi i32 [ 0, %for.body.preheader.new ], [ %inc.3, %for.body ]
  %res.010 = phi float [ 0.000000e+00, %for.body.preheader.new ], [ %add2.3, %for.body ]
  %niter = phi i32 [ %unroll_iter, %for.body.preheader.new ], [ %niter.nsub.3, %for.body ]
  %arrayidx = getelementptr inbounds half, half* %a, i32 %i.011
  %4 = load half, half* %arrayidx, align 2
  %arrayidx1 = getelementptr inbounds half, half* %b, i32 %i.011
  %5 = load half, half* %arrayidx1, align 2
  %add = fadd half %4, %5
  %conv = fpext half %add to float
  %add2 = fadd float %res.010, %conv
  %inc = or i32 %i.011, 1
  %arrayidx.1 = getelementptr inbounds half, half* %a, i32 %inc
  %6 = load half, half* %arrayidx.1, align 2
  %arrayidx1.1 = getelementptr inbounds half, half* %b, i32 %inc
  %7 = load half, half* %arrayidx1.1, align 2
  %add.1 = fadd half %6, %7
  %conv.1 = fpext half %add.1 to float
  %add2.1 = fadd float %add2, %conv.1
  %inc.1 = or i32 %i.011, 2
  %arrayidx.2 = getelementptr inbounds half, half* %a, i32 %inc.1
  %8 = load half, half* %arrayidx.2, align 2
  %arrayidx1.2 = getelementptr inbounds half, half* %b, i32 %inc.1
  %9 = load half, half* %arrayidx1.2, align 2
  %add.2 = fadd half %8, %9
  %conv.2 = fpext half %add.2 to float
  %add2.2 = fadd float %add2.1, %conv.2
  %inc.2 = or i32 %i.011, 3
  %arrayidx.3 = getelementptr inbounds half, half* %a, i32 %inc.2
  %10 = load half, half* %arrayidx.3, align 2
  %arrayidx1.3 = getelementptr inbounds half, half* %b, i32 %inc.2
  %11 = load half, half* %arrayidx1.3, align 2
  %add.3 = fadd half %10, %11
  %conv.3 = fpext half %add.3 to float
  %add2.3 = fadd float %add2.2, %conv.3
  %inc.3 = add nuw i32 %i.011, 4
  %niter.nsub.3 = add i32 %niter, -4
  %niter.ncmp.3 = icmp eq i32 %niter.nsub.3, 0
  br i1 %niter.ncmp.3, label %for.cond.cleanup.loopexit.unr-lcssa, label %for.body
}

define arm_aapcs_vfpcc float @half_short_mac(half* nocapture readonly %a, i16* nocapture readonly %b, i32 %N) {
; CHECK-LABEL: half_short_mac:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    push {r4, r5, r6, lr}
; CHECK-NEXT:    cbz r2, .LBB11_3
; CHECK-NEXT:  @ %bb.1: @ %for.body.preheader
; CHECK-NEXT:    subs r3, r2, #1
; CHECK-NEXT:    and r6, r2, #3
; CHECK-NEXT:    cmp r3, #3
; CHECK-NEXT:    bhs .LBB11_4
; CHECK-NEXT:  @ %bb.2:
; CHECK-NEXT:    vldr s0, .LCPI11_0
; CHECK-NEXT:    mov.w r12, #0
; CHECK-NEXT:    b .LBB11_6
; CHECK-NEXT:  .LBB11_3:
; CHECK-NEXT:    vldr s0, .LCPI11_0
; CHECK-NEXT:    b .LBB11_9
; CHECK-NEXT:  .LBB11_4: @ %for.body.preheader.new
; CHECK-NEXT:    bic r2, r2, #3
; CHECK-NEXT:    movs r3, #1
; CHECK-NEXT:    subs r2, #4
; CHECK-NEXT:    vldr s0, .LCPI11_0
; CHECK-NEXT:    mov.w r12, #0
; CHECK-NEXT:    add.w lr, r3, r2, lsr #2
; CHECK-NEXT:    adds r3, r1, #4
; CHECK-NEXT:    adds r2, r0, #4
; CHECK-NEXT:    dls lr, lr
; CHECK-NEXT:  .LBB11_5: @ %for.body
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    ldrsh.w r4, [r3, #2]
; CHECK-NEXT:    vldr.16 s2, [r2, #2]
; CHECK-NEXT:    ldrsh r5, [r3, #-2]
; CHECK-NEXT:    add.w r12, r12, #4
; CHECK-NEXT:    vmov s4, r4
; CHECK-NEXT:    ldrsh.w r4, [r3]
; CHECK-NEXT:    vcvt.f16.s32 s4, s4
; CHECK-NEXT:    vmov s8, r5
; CHECK-NEXT:    vmul.f16 s2, s2, s4
; CHECK-NEXT:    vldr.16 s4, [r2]
; CHECK-NEXT:    vmov s6, r4
; CHECK-NEXT:    ldrsh r4, [r3, #-4]
; CHECK-NEXT:    vcvt.f16.s32 s6, s6
; CHECK-NEXT:    vcvt.f16.s32 s8, s8
; CHECK-NEXT:    vmul.f16 s4, s4, s6
; CHECK-NEXT:    vldr.16 s6, [r2, #-2]
; CHECK-NEXT:    vmov s10, r4
; CHECK-NEXT:    vcvtb.f32.f16 s4, s4
; CHECK-NEXT:    vmul.f16 s6, s6, s8
; CHECK-NEXT:    vldr.16 s8, [r2, #-4]
; CHECK-NEXT:    vcvt.f16.s32 s10, s10
; CHECK-NEXT:    vcvtb.f32.f16 s6, s6
; CHECK-NEXT:    vmul.f16 s8, s8, s10
; CHECK-NEXT:    vcvtb.f32.f16 s2, s2
; CHECK-NEXT:    vcvtb.f32.f16 s8, s8
; CHECK-NEXT:    adds r3, #8
; CHECK-NEXT:    vadd.f32 s0, s0, s8
; CHECK-NEXT:    adds r2, #8
; CHECK-NEXT:    vadd.f32 s0, s0, s6
; CHECK-NEXT:    vadd.f32 s0, s0, s4
; CHECK-NEXT:    vadd.f32 s0, s0, s2
; CHECK-NEXT:    le lr, .LBB11_5
; CHECK-NEXT:  .LBB11_6: @ %for.cond.cleanup.loopexit.unr-lcssa
; CHECK-NEXT:    wls lr, r6, .LBB11_9
; CHECK-NEXT:  @ %bb.7: @ %for.body.epil.preheader
; CHECK-NEXT:    add.w r0, r0, r12, lsl #1
; CHECK-NEXT:    add.w r1, r1, r12, lsl #1
; CHECK-NEXT:    mov lr, r6
; CHECK-NEXT:  .LBB11_8: @ %for.body.epil
; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    ldrsh r2, [r1], #2
; CHECK-NEXT:    vldr.16 s2, [r0]
; CHECK-NEXT:    adds r0, #2
; CHECK-NEXT:    vmov s4, r2
; CHECK-NEXT:    vcvt.f16.s32 s4, s4
; CHECK-NEXT:    vmul.f16 s2, s2, s4
; CHECK-NEXT:    vcvtb.f32.f16 s2, s2
; CHECK-NEXT:    vadd.f32 s0, s0, s2
; CHECK-NEXT:    le lr, .LBB11_8
; CHECK-NEXT:  .LBB11_9: @ %for.cond.cleanup
; CHECK-NEXT:    pop {r4, r5, r6, pc}
; CHECK-NEXT:    .p2align 2
; CHECK-NEXT:  @ %bb.10:
; CHECK-NEXT:  .LCPI11_0:
; CHECK-NEXT:    .long 0x00000000 @ float 0
entry:
  %cmp10 = icmp eq i32 %N, 0
  br i1 %cmp10, label %for.cond.cleanup, label %for.body.preheader

for.body.preheader:                               ; preds = %entry
  %0 = add i32 %N, -1
  %xtraiter = and i32 %N, 3
  %1 = icmp ult i32 %0, 3
  br i1 %1, label %for.cond.cleanup.loopexit.unr-lcssa, label %for.body.preheader.new

for.body.preheader.new:                           ; preds = %for.body.preheader
  %unroll_iter = sub i32 %N, %xtraiter
  br label %for.body

for.cond.cleanup.loopexit.unr-lcssa:              ; preds = %for.body, %for.body.preheader
  %add.lcssa.ph = phi float [ undef, %for.body.preheader ], [ %add.3, %for.body ]
  %i.012.unr = phi i32 [ 0, %for.body.preheader ], [ %inc.3, %for.body ]
  %res.011.unr = phi float [ 0.000000e+00, %for.body.preheader ], [ %add.3, %for.body ]
  %lcmp.mod = icmp eq i32 %xtraiter, 0
  br i1 %lcmp.mod, label %for.cond.cleanup, label %for.body.epil

for.body.epil:                                    ; preds = %for.cond.cleanup.loopexit.unr-lcssa, %for.body.epil
  %i.012.epil = phi i32 [ %inc.epil, %for.body.epil ], [ %i.012.unr, %for.cond.cleanup.loopexit.unr-lcssa ]
  %res.011.epil = phi float [ %add.epil, %for.body.epil ], [ %res.011.unr, %for.cond.cleanup.loopexit.unr-lcssa ]
  %epil.iter = phi i32 [ %epil.iter.sub, %for.body.epil ], [ %xtraiter, %for.cond.cleanup.loopexit.unr-lcssa ]
  %arrayidx.epil = getelementptr inbounds half, half* %a, i32 %i.012.epil
  %2 = load half, half* %arrayidx.epil, align 2
  %arrayidx1.epil = getelementptr inbounds i16, i16* %b, i32 %i.012.epil
  %3 = load i16, i16* %arrayidx1.epil, align 2
  %conv2.epil = sitofp i16 %3 to half
  %mul.epil = fmul half %2, %conv2.epil
  %conv3.epil = fpext half %mul.epil to float
  %add.epil = fadd float %res.011.epil, %conv3.epil
  %inc.epil = add nuw i32 %i.012.epil, 1
  %epil.iter.sub = add i32 %epil.iter, -1
  %epil.iter.cmp = icmp eq i32 %epil.iter.sub, 0
  br i1 %epil.iter.cmp, label %for.cond.cleanup, label %for.body.epil

for.cond.cleanup:                                 ; preds = %for.cond.cleanup.loopexit.unr-lcssa, %for.body.epil, %entry
  %res.0.lcssa = phi float [ 0.000000e+00, %entry ], [ %add.lcssa.ph, %for.cond.cleanup.loopexit.unr-lcssa ], [ %add.epil, %for.body.epil ]
  ret float %res.0.lcssa

for.body:                                         ; preds = %for.body, %for.body.preheader.new
  %i.012 = phi i32 [ 0, %for.body.preheader.new ], [ %inc.3, %for.body ]
  %res.011 = phi float [ 0.000000e+00, %for.body.preheader.new ], [ %add.3, %for.body ]
  %niter = phi i32 [ %unroll_iter, %for.body.preheader.new ], [ %niter.nsub.3, %for.body ]
  %arrayidx = getelementptr inbounds half, half* %a, i32 %i.012
  %4 = load half, half* %arrayidx, align 2
  %arrayidx1 = getelementptr inbounds i16, i16* %b, i32 %i.012
  %5 = load i16, i16* %arrayidx1, align 2
  %conv2 = sitofp i16 %5 to half
  %mul = fmul half %4, %conv2
  %conv3 = fpext half %mul to float
  %add = fadd float %res.011, %conv3
  %inc = or i32 %i.012, 1
  %arrayidx.1 = getelementptr inbounds half, half* %a, i32 %inc
  %6 = load half, half* %arrayidx.1, align 2
  %arrayidx1.1 = getelementptr inbounds i16, i16* %b, i32 %inc
  %7 = load i16, i16* %arrayidx1.1, align 2
  %conv2.1 = sitofp i16 %7 to half
  %mul.1 = fmul half %6, %conv2.1
  %conv3.1 = fpext half %mul.1 to float
  %add.1 = fadd float %add, %conv3.1
  %inc.1 = or i32 %i.012, 2
  %arrayidx.2 = getelementptr inbounds half, half* %a, i32 %inc.1
  %8 = load half, half* %arrayidx.2, align 2
  %arrayidx1.2 = getelementptr inbounds i16, i16* %b, i32 %inc.1
  %9 = load i16, i16* %arrayidx1.2, align 2
  %conv2.2 = sitofp i16 %9 to half
  %mul.2 = fmul half %8, %conv2.2
  %conv3.2 = fpext half %mul.2 to float
  %add.2 = fadd float %add.1, %conv3.2
  %inc.2 = or i32 %i.012, 3
  %arrayidx.3 = getelementptr inbounds half, half* %a, i32 %inc.2
  %10 = load half, half* %arrayidx.3, align 2
  %arrayidx1.3 = getelementptr inbounds i16, i16* %b, i32 %inc.2
  %11 = load i16, i16* %arrayidx1.3, align 2
  %conv2.3 = sitofp i16 %11 to half
  %mul.3 = fmul half %10, %conv2.3
  %conv3.3 = fpext half %mul.3 to float
  %add.3 = fadd float %add.2, %conv3.3
  %inc.3 = add nuw i32 %i.012, 4
  %niter.nsub.3 = add i32 %niter, -4
  %niter.ncmp.3 = icmp eq i32 %niter.nsub.3, 0
  br i1 %niter.ncmp.3, label %for.cond.cleanup.loopexit.unr-lcssa, label %for.body
}