; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -mtriple=x86_64 -passes=slp-vectorizer -S -mcpu=skylake-avx512 | FileCheck %s

declare void @llvm.masked.scatter.v2f64.v2p0(<2 x double>, <2 x ptr>, i32 immarg, <2 x i1>)

define void @rdx_feeds_single_insert(<2 x double> %v, ptr nocapture readonly %arg, ptr nocapture readonly %arg1, ptr nocapture %arg2) {
; CHECK-LABEL: @rdx_feeds_single_insert(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    [[TMP0:%.*]] = load <8 x double>, ptr [[ARG1:%.*]], align 8
; CHECK-NEXT:    [[TMP1:%.*]] = fmul fast <8 x double> [[TMP0]], <double 1.000000e+01, double 1.100000e+01, double 1.200000e+01, double 1.300000e+01, double 1.400000e+01, double 1.500000e+01, double 1.600000e+01, double 1.700000e+01>
; CHECK-NEXT:    [[TMP2:%.*]] = call fast double @llvm.vector.reduce.fadd.v8f64(double 0.000000e+00, <8 x double> [[TMP1]])
; CHECK-NEXT:    [[I:%.*]] = insertelement <2 x double> [[V:%.*]], double [[TMP2]], i64 1
; CHECK-NEXT:    [[P:%.*]] = getelementptr inbounds double, ptr [[ARG2:%.*]], <2 x i64> <i64 0, i64 16>
; CHECK-NEXT:    call void @llvm.masked.scatter.v2f64.v2p0(<2 x double> [[I]], <2 x ptr> [[P]], i32 8, <2 x i1> splat (i1 true))
; CHECK-NEXT:    ret void
;
entry:
  %ld0.0 = load double, ptr %arg1, align 8
  %mul1.0 = fmul fast double %ld0.0, 10.0
  %gep0.1 = getelementptr inbounds double, ptr %arg1, i64 1
  %ld0.1 = load double, ptr %gep0.1, align 8
  %mul1.1 = fmul fast double %ld0.1, 11.0
  %rdx1.0 = fadd fast double %mul1.0, %mul1.1
  %gep0.2 = getelementptr inbounds double, ptr %arg1, i64 2
  %ld0.2 = load double, ptr %gep0.2, align 8
  %mul1.2 = fmul fast double %ld0.2, 12.0
  %rdx1.1 = fadd fast double %rdx1.0, %mul1.2
  %gep0.3 = getelementptr inbounds double, ptr %arg1, i64 3
  %ld0.3 = load double, ptr %gep0.3, align 8
  %mul1.3 = fmul fast double %ld0.3, 13.0
  %rdx1.2 = fadd fast double %rdx1.1, %mul1.3
  %gep0.4 = getelementptr inbounds double, ptr %arg1, i64 4
  %ld0.4 = load double, ptr %gep0.4, align 8
  %mul1.4 = fmul fast double %ld0.4, 14.0
  %rdx1.3 = fadd fast double %rdx1.2, %mul1.4
  %gep0.5 = getelementptr inbounds double, ptr %arg1, i64 5
  %ld0.5 = load double, ptr %gep0.5, align 8
  %mul1.5 = fmul fast double %ld0.5, 15.0
  %rdx1.4 = fadd fast double %rdx1.3, %mul1.5
  %gep0.6 = getelementptr inbounds double, ptr %arg1, i64 6
  %ld0.6 = load double, ptr %gep0.6, align 8
  %mul1.6 = fmul fast double %ld0.6, 16.0
  %rdx1.5 = fadd fast double %rdx1.4, %mul1.6
  %gep0.7 = getelementptr inbounds double, ptr %arg1, i64 7
  %ld0.7 = load double, ptr %gep0.7, align 8
  %mul1.7 = fmul fast double %ld0.7, 17.0
  %rdx1 = fadd fast double %rdx1.5, %mul1.7
  %i = insertelement <2 x double> %v, double %rdx1, i64 1
  %p = getelementptr inbounds double, ptr %arg2, <2 x i64> <i64 0, i64 16>
  call void @llvm.masked.scatter.v2f64.v2p0(<2 x double> %i, <2 x ptr> %p, i32 8, <2 x i1> <i1 true, i1 true>)
  ret void
}