; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
; RUN: llc < %s -mtriple=x86_64- -mcpu=x86-64-v4 | FileCheck %s

define void @PR93000(ptr %a0, ptr %a1, ptr %a2, <32 x i16> %a3) {
; CHECK-LABEL: PR93000:
; CHECK:       # %bb.0: # %Entry
; CHECK-NEXT:    movl (%rdi), %eax
; CHECK-NEXT:    addq $4, %rdi
; CHECK-NEXT:    .p2align 4
; CHECK-NEXT:  .LBB0_1: # %Loop
; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
; CHECK-NEXT:    kmovd %eax, %k1
; CHECK-NEXT:    knotd %k1, %k2
; CHECK-NEXT:    vpblendmw (%rsi), %zmm0, %zmm1 {%k1}
; CHECK-NEXT:    vmovdqu16 (%rdx), %zmm1 {%k2}
; CHECK-NEXT:    vmovdqu64 %zmm1, (%rsi)
; CHECK-NEXT:    movl (%rdi), %eax
; CHECK-NEXT:    addq $4, %rdi
; CHECK-NEXT:    testl %eax, %eax
; CHECK-NEXT:    jne .LBB0_1
; CHECK-NEXT:  # %bb.2: # %Then
; CHECK-NEXT:    vzeroupper
; CHECK-NEXT:    retq
Entry:
  %pre = load i32, ptr %a0, align 4
  br label %Loop

Loop:                                             ; preds = %Loop, %Entry
  %p = phi i32 [ %limit, %Loop ], [ %pre, %Entry ]
  %lsr.iv.pn = phi ptr [ %lsr.iv, %Loop ], [ %a0, %Entry ]
  %lsr.iv = getelementptr i8, ptr %lsr.iv.pn, i64 4
  %pn = xor i32 %p, -1
  %m = bitcast i32 %p to <32 x i1>
  %mn = bitcast i32 %pn to <32 x i1>
  %mload0 = tail call <32 x i16> @llvm.masked.load.v32i16.p0(ptr %a1, i32 2, <32 x i1> %m, <32 x i16> %a3)
  %mload1 = tail call <32 x i16> @llvm.masked.load.v32i16.p0(ptr %a2, i32 2, <32 x i1> %mn, <32 x i16> %mload0)
  store <32 x i16> %mload1, ptr %a1, align 2
  %limit = load i32, ptr %lsr.iv, align 4
  %icmp = icmp eq i32 %limit, 0
  br i1 %icmp, label %Then, label %Loop

Then:                                             ; preds = %Loop
  ret void
}