; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
; RUN: opt < %s -mtriple=armv8a-linux-gnueabihf -mattr=+fp64 -passes="print<cost-model>" 2>&1 -disable-output | FileCheck %s --check-prefix=CHECK-V8
; RUN: opt < %s -mtriple=thumbv8.1m.main-none-eabi -mattr=+mve.fp -passes="print<cost-model>" 2>&1 -disable-output | FileCheck %s --check-prefix=CHECK-MVEFP
; RUN: opt < %s -mtriple=thumbv8.1m.main-none-eabi -mattr=+mve -passes="print<cost-model>" 2>&1 -disable-output | FileCheck %s --check-prefix=CHECK-MVEI

target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"

define void @fmin_strict() {
; CHECK-V8-LABEL: 'fmin_strict'
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %fmin_v2f16 = call half @llvm.vector.reduce.fmin.v2f16(<2 x half> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %fmin_v4f16 = call half @llvm.vector.reduce.fmin.v4f16(<4 x half> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %fmin_v8f16 = call half @llvm.vector.reduce.fmin.v8f16(<8 x half> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 91 for instruction: %fmin_v16f16 = call half @llvm.vector.reduce.fmin.v16f16(<16 x half> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fmin_v2f32 = call float @llvm.vector.reduce.fmin.v2f32(<2 x float> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fmin_v4f32 = call float @llvm.vector.reduce.fmin.v4f32(<4 x float> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fmin_v8f32 = call float @llvm.vector.reduce.fmin.v8f32(<8 x float> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fmin_v2f64 = call double @llvm.vector.reduce.fmin.v2f64(<2 x double> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %fmin_v4f64 = call double @llvm.vector.reduce.fmin.v4f64(<4 x double> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 94 for instruction: %fmin_v4f128 = call fp128 @llvm.vector.reduce.fmin.v4f128(<4 x fp128> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; CHECK-MVEFP-LABEL: 'fmin_strict'
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fmin_v2f16 = call half @llvm.vector.reduce.fmin.v2f16(<2 x half> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %fmin_v4f16 = call half @llvm.vector.reduce.fmin.v4f16(<4 x half> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %fmin_v8f16 = call half @llvm.vector.reduce.fmin.v8f16(<8 x half> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %fmin_v16f16 = call half @llvm.vector.reduce.fmin.v16f16(<16 x half> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fmin_v2f32 = call float @llvm.vector.reduce.fmin.v2f32(<2 x float> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fmin_v4f32 = call float @llvm.vector.reduce.fmin.v4f32(<4 x float> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %fmin_v8f32 = call float @llvm.vector.reduce.fmin.v8f32(<8 x float> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %fmin_v2f64 = call double @llvm.vector.reduce.fmin.v2f64(<2 x double> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %fmin_v4f64 = call double @llvm.vector.reduce.fmin.v4f64(<4 x double> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %fmin_v4f128 = call fp128 @llvm.vector.reduce.fmin.v4f128(<4 x fp128> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; CHECK-MVEI-LABEL: 'fmin_strict'
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %fmin_v2f16 = call half @llvm.vector.reduce.fmin.v2f16(<2 x half> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 137 for instruction: %fmin_v4f16 = call half @llvm.vector.reduce.fmin.v4f16(<4 x half> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 409 for instruction: %fmin_v8f16 = call half @llvm.vector.reduce.fmin.v8f16(<8 x half> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 545 for instruction: %fmin_v16f16 = call half @llvm.vector.reduce.fmin.v16f16(<16 x half> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %fmin_v2f32 = call float @llvm.vector.reduce.fmin.v2f32(<2 x float> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 137 for instruction: %fmin_v4f32 = call float @llvm.vector.reduce.fmin.v4f32(<4 x float> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 205 for instruction: %fmin_v8f32 = call float @llvm.vector.reduce.fmin.v8f32(<8 x float> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %fmin_v2f64 = call double @llvm.vector.reduce.fmin.v2f64(<2 x double> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %fmin_v4f64 = call double @llvm.vector.reduce.fmin.v4f64(<4 x double> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %fmin_v4f128 = call fp128 @llvm.vector.reduce.fmin.v4f128(<4 x fp128> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
;
  %fmin_v2f16 = call half @llvm.vector.reduce.fmin.v2f16(<2 x half> undef)
  %fmin_v4f16 = call half @llvm.vector.reduce.fmin.v4f16(<4 x half> undef)
  %fmin_v8f16 = call half @llvm.vector.reduce.fmin.v8f16(<8 x half> undef)
  %fmin_v16f16 = call half @llvm.vector.reduce.fmin.v16f16(<16 x half> undef)
  %fmin_v2f32 = call float @llvm.vector.reduce.fmin.v2f32(<2 x float> undef)
  %fmin_v4f32 = call float @llvm.vector.reduce.fmin.v4f32(<4 x float> undef)
  %fmin_v8f32 = call float @llvm.vector.reduce.fmin.v8f32(<8 x float> undef)
  %fmin_v2f64 = call double @llvm.vector.reduce.fmin.v2f64(<2 x double> undef)
  %fmin_v4f64 = call double @llvm.vector.reduce.fmin.v4f64(<4 x double> undef)
  %fmin_v4f128 = call fp128 @llvm.vector.reduce.fmin.v4f128(<4 x fp128> undef)
  ret void
}


define void @fmin_unordered() {
; CHECK-V8-LABEL: 'fmin_unordered'
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %fmin_v2f16 = call reassoc half @llvm.vector.reduce.fmin.v2f16(<2 x half> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %fmin_v4f16 = call reassoc half @llvm.vector.reduce.fmin.v4f16(<4 x half> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %fmin_v8f16 = call reassoc half @llvm.vector.reduce.fmin.v8f16(<8 x half> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 91 for instruction: %fmin_v16f16 = call reassoc half @llvm.vector.reduce.fmin.v16f16(<16 x half> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fmin_v2f32 = call reassoc float @llvm.vector.reduce.fmin.v2f32(<2 x float> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fmin_v4f32 = call reassoc float @llvm.vector.reduce.fmin.v4f32(<4 x float> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fmin_v8f32 = call reassoc float @llvm.vector.reduce.fmin.v8f32(<8 x float> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fmin_v2f64 = call reassoc double @llvm.vector.reduce.fmin.v2f64(<2 x double> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %fmin_v4f64 = call reassoc double @llvm.vector.reduce.fmin.v4f64(<4 x double> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 94 for instruction: %fmin_v4f128 = call reassoc fp128 @llvm.vector.reduce.fmin.v4f128(<4 x fp128> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; CHECK-MVEFP-LABEL: 'fmin_unordered'
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fmin_v2f16 = call reassoc half @llvm.vector.reduce.fmin.v2f16(<2 x half> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %fmin_v4f16 = call reassoc half @llvm.vector.reduce.fmin.v4f16(<4 x half> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %fmin_v8f16 = call reassoc half @llvm.vector.reduce.fmin.v8f16(<8 x half> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %fmin_v16f16 = call reassoc half @llvm.vector.reduce.fmin.v16f16(<16 x half> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fmin_v2f32 = call reassoc float @llvm.vector.reduce.fmin.v2f32(<2 x float> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fmin_v4f32 = call reassoc float @llvm.vector.reduce.fmin.v4f32(<4 x float> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %fmin_v8f32 = call reassoc float @llvm.vector.reduce.fmin.v8f32(<8 x float> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %fmin_v2f64 = call reassoc double @llvm.vector.reduce.fmin.v2f64(<2 x double> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %fmin_v4f64 = call reassoc double @llvm.vector.reduce.fmin.v4f64(<4 x double> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %fmin_v4f128 = call reassoc fp128 @llvm.vector.reduce.fmin.v4f128(<4 x fp128> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; CHECK-MVEI-LABEL: 'fmin_unordered'
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %fmin_v2f16 = call reassoc half @llvm.vector.reduce.fmin.v2f16(<2 x half> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 137 for instruction: %fmin_v4f16 = call reassoc half @llvm.vector.reduce.fmin.v4f16(<4 x half> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 409 for instruction: %fmin_v8f16 = call reassoc half @llvm.vector.reduce.fmin.v8f16(<8 x half> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 545 for instruction: %fmin_v16f16 = call reassoc half @llvm.vector.reduce.fmin.v16f16(<16 x half> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %fmin_v2f32 = call reassoc float @llvm.vector.reduce.fmin.v2f32(<2 x float> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 137 for instruction: %fmin_v4f32 = call reassoc float @llvm.vector.reduce.fmin.v4f32(<4 x float> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 205 for instruction: %fmin_v8f32 = call reassoc float @llvm.vector.reduce.fmin.v8f32(<8 x float> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %fmin_v2f64 = call reassoc double @llvm.vector.reduce.fmin.v2f64(<2 x double> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %fmin_v4f64 = call reassoc double @llvm.vector.reduce.fmin.v4f64(<4 x double> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %fmin_v4f128 = call reassoc fp128 @llvm.vector.reduce.fmin.v4f128(<4 x fp128> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
;
  %fmin_v2f16 = call reassoc half @llvm.vector.reduce.fmin.v2f16(<2 x half> undef)
  %fmin_v4f16 = call reassoc half @llvm.vector.reduce.fmin.v4f16(<4 x half> undef)
  %fmin_v8f16 = call reassoc half @llvm.vector.reduce.fmin.v8f16(<8 x half> undef)
  %fmin_v16f16 = call reassoc half @llvm.vector.reduce.fmin.v16f16(<16 x half> undef)
  %fmin_v2f32 = call reassoc float @llvm.vector.reduce.fmin.v2f32(<2 x float> undef)
  %fmin_v4f32 = call reassoc float @llvm.vector.reduce.fmin.v4f32(<4 x float> undef)
  %fmin_v8f32 = call reassoc float @llvm.vector.reduce.fmin.v8f32(<8 x float> undef)
  %fmin_v2f64 = call reassoc double @llvm.vector.reduce.fmin.v2f64(<2 x double> undef)
  %fmin_v4f64 = call reassoc double @llvm.vector.reduce.fmin.v4f64(<4 x double> undef)
  %fmin_v4f128 = call reassoc fp128 @llvm.vector.reduce.fmin.v4f128(<4 x fp128> undef)
  ret void
}

define void @fmax_strict() {
; CHECK-V8-LABEL: 'fmax_strict'
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %fmax_v2f16 = call half @llvm.vector.reduce.fmax.v2f16(<2 x half> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %fmax_v4f16 = call half @llvm.vector.reduce.fmax.v4f16(<4 x half> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %fmax_v8f16 = call half @llvm.vector.reduce.fmax.v8f16(<8 x half> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 91 for instruction: %fmax_v16f16 = call half @llvm.vector.reduce.fmax.v16f16(<16 x half> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fmax_v2f32 = call float @llvm.vector.reduce.fmax.v2f32(<2 x float> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fmax_v4f32 = call float @llvm.vector.reduce.fmax.v4f32(<4 x float> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fmax_v8f32 = call float @llvm.vector.reduce.fmax.v8f32(<8 x float> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fmax_v2f64 = call double @llvm.vector.reduce.fmax.v2f64(<2 x double> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %fmax_v4f64 = call double @llvm.vector.reduce.fmax.v4f64(<4 x double> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 94 for instruction: %fmax_v4f128 = call fp128 @llvm.vector.reduce.fmax.v4f128(<4 x fp128> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; CHECK-MVEFP-LABEL: 'fmax_strict'
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fmax_v2f16 = call half @llvm.vector.reduce.fmax.v2f16(<2 x half> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %fmax_v4f16 = call half @llvm.vector.reduce.fmax.v4f16(<4 x half> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %fmax_v8f16 = call half @llvm.vector.reduce.fmax.v8f16(<8 x half> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %fmax_v16f16 = call half @llvm.vector.reduce.fmax.v16f16(<16 x half> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fmax_v2f32 = call float @llvm.vector.reduce.fmax.v2f32(<2 x float> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fmax_v4f32 = call float @llvm.vector.reduce.fmax.v4f32(<4 x float> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %fmax_v8f32 = call float @llvm.vector.reduce.fmax.v8f32(<8 x float> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %fmax_v2f64 = call double @llvm.vector.reduce.fmax.v2f64(<2 x double> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %fmax_v4f64 = call double @llvm.vector.reduce.fmax.v4f64(<4 x double> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %fmax_v4f128 = call fp128 @llvm.vector.reduce.fmax.v4f128(<4 x fp128> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; CHECK-MVEI-LABEL: 'fmax_strict'
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %fmax_v2f16 = call half @llvm.vector.reduce.fmax.v2f16(<2 x half> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 137 for instruction: %fmax_v4f16 = call half @llvm.vector.reduce.fmax.v4f16(<4 x half> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 409 for instruction: %fmax_v8f16 = call half @llvm.vector.reduce.fmax.v8f16(<8 x half> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 545 for instruction: %fmax_v16f16 = call half @llvm.vector.reduce.fmax.v16f16(<16 x half> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %fmax_v2f32 = call float @llvm.vector.reduce.fmax.v2f32(<2 x float> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 137 for instruction: %fmax_v4f32 = call float @llvm.vector.reduce.fmax.v4f32(<4 x float> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 205 for instruction: %fmax_v8f32 = call float @llvm.vector.reduce.fmax.v8f32(<8 x float> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %fmax_v2f64 = call double @llvm.vector.reduce.fmax.v2f64(<2 x double> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %fmax_v4f64 = call double @llvm.vector.reduce.fmax.v4f64(<4 x double> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %fmax_v4f128 = call fp128 @llvm.vector.reduce.fmax.v4f128(<4 x fp128> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
;
  %fmax_v2f16 = call half @llvm.vector.reduce.fmax.v2f16(<2 x half> undef)
  %fmax_v4f16 = call half @llvm.vector.reduce.fmax.v4f16(<4 x half> undef)
  %fmax_v8f16 = call half @llvm.vector.reduce.fmax.v8f16(<8 x half> undef)
  %fmax_v16f16 = call half @llvm.vector.reduce.fmax.v16f16(<16 x half> undef)
  %fmax_v2f32 = call float @llvm.vector.reduce.fmax.v2f32(<2 x float> undef)
  %fmax_v4f32 = call float @llvm.vector.reduce.fmax.v4f32(<4 x float> undef)
  %fmax_v8f32 = call float @llvm.vector.reduce.fmax.v8f32(<8 x float> undef)
  %fmax_v2f64 = call double @llvm.vector.reduce.fmax.v2f64(<2 x double> undef)
  %fmax_v4f64 = call double @llvm.vector.reduce.fmax.v4f64(<4 x double> undef)
  %fmax_v4f128 = call fp128 @llvm.vector.reduce.fmax.v4f128(<4 x fp128> undef)
  ret void
}


define void @fmax_unordered() {
; CHECK-V8-LABEL: 'fmax_unordered'
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %fmax_v2f16 = call reassoc half @llvm.vector.reduce.fmax.v2f16(<2 x half> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %fmax_v4f16 = call reassoc half @llvm.vector.reduce.fmax.v4f16(<4 x half> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %fmax_v8f16 = call reassoc half @llvm.vector.reduce.fmax.v8f16(<8 x half> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 91 for instruction: %fmax_v16f16 = call reassoc half @llvm.vector.reduce.fmax.v16f16(<16 x half> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fmax_v2f32 = call reassoc float @llvm.vector.reduce.fmax.v2f32(<2 x float> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fmax_v4f32 = call reassoc float @llvm.vector.reduce.fmax.v4f32(<4 x float> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fmax_v8f32 = call reassoc float @llvm.vector.reduce.fmax.v8f32(<8 x float> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fmax_v2f64 = call reassoc double @llvm.vector.reduce.fmax.v2f64(<2 x double> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %fmax_v4f64 = call reassoc double @llvm.vector.reduce.fmax.v4f64(<4 x double> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 94 for instruction: %fmax_v4f128 = call reassoc fp128 @llvm.vector.reduce.fmax.v4f128(<4 x fp128> undef)
; CHECK-V8-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; CHECK-MVEFP-LABEL: 'fmax_unordered'
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fmax_v2f16 = call reassoc half @llvm.vector.reduce.fmax.v2f16(<2 x half> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %fmax_v4f16 = call reassoc half @llvm.vector.reduce.fmax.v4f16(<4 x half> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %fmax_v8f16 = call reassoc half @llvm.vector.reduce.fmax.v8f16(<8 x half> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %fmax_v16f16 = call reassoc half @llvm.vector.reduce.fmax.v16f16(<16 x half> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fmax_v2f32 = call reassoc float @llvm.vector.reduce.fmax.v2f32(<2 x float> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fmax_v4f32 = call reassoc float @llvm.vector.reduce.fmax.v4f32(<4 x float> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %fmax_v8f32 = call reassoc float @llvm.vector.reduce.fmax.v8f32(<8 x float> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %fmax_v2f64 = call reassoc double @llvm.vector.reduce.fmax.v2f64(<2 x double> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %fmax_v4f64 = call reassoc double @llvm.vector.reduce.fmax.v4f64(<4 x double> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %fmax_v4f128 = call reassoc fp128 @llvm.vector.reduce.fmax.v4f128(<4 x fp128> undef)
; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; CHECK-MVEI-LABEL: 'fmax_unordered'
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %fmax_v2f16 = call reassoc half @llvm.vector.reduce.fmax.v2f16(<2 x half> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 137 for instruction: %fmax_v4f16 = call reassoc half @llvm.vector.reduce.fmax.v4f16(<4 x half> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 409 for instruction: %fmax_v8f16 = call reassoc half @llvm.vector.reduce.fmax.v8f16(<8 x half> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 545 for instruction: %fmax_v16f16 = call reassoc half @llvm.vector.reduce.fmax.v16f16(<16 x half> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %fmax_v2f32 = call reassoc float @llvm.vector.reduce.fmax.v2f32(<2 x float> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 137 for instruction: %fmax_v4f32 = call reassoc float @llvm.vector.reduce.fmax.v4f32(<4 x float> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 205 for instruction: %fmax_v8f32 = call reassoc float @llvm.vector.reduce.fmax.v8f32(<8 x float> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %fmax_v2f64 = call reassoc double @llvm.vector.reduce.fmax.v2f64(<2 x double> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %fmax_v4f64 = call reassoc double @llvm.vector.reduce.fmax.v4f64(<4 x double> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %fmax_v4f128 = call reassoc fp128 @llvm.vector.reduce.fmax.v4f128(<4 x fp128> undef)
; CHECK-MVEI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
;
  %fmax_v2f16 = call reassoc half @llvm.vector.reduce.fmax.v2f16(<2 x half> undef)
  %fmax_v4f16 = call reassoc half @llvm.vector.reduce.fmax.v4f16(<4 x half> undef)
  %fmax_v8f16 = call reassoc half @llvm.vector.reduce.fmax.v8f16(<8 x half> undef)
  %fmax_v16f16 = call reassoc half @llvm.vector.reduce.fmax.v16f16(<16 x half> undef)
  %fmax_v2f32 = call reassoc float @llvm.vector.reduce.fmax.v2f32(<2 x float> undef)
  %fmax_v4f32 = call reassoc float @llvm.vector.reduce.fmax.v4f32(<4 x float> undef)
  %fmax_v8f32 = call reassoc float @llvm.vector.reduce.fmax.v8f32(<8 x float> undef)
  %fmax_v2f64 = call reassoc double @llvm.vector.reduce.fmax.v2f64(<2 x double> undef)
  %fmax_v4f64 = call reassoc double @llvm.vector.reduce.fmax.v4f64(<4 x double> undef)
  %fmax_v4f128 = call reassoc fp128 @llvm.vector.reduce.fmax.v4f128(<4 x fp128> undef)
  ret void
}


declare half @llvm.vector.reduce.fmin.v2f16(<2 x half>)
declare half @llvm.vector.reduce.fmin.v4f16(<4 x half>)
declare half @llvm.vector.reduce.fmin.v8f16(<8 x half>)
declare half @llvm.vector.reduce.fmin.v16f16(<16 x half>)
declare float @llvm.vector.reduce.fmin.v2f32(<2 x float>)
declare float @llvm.vector.reduce.fmin.v4f32(<4 x float>)
declare float @llvm.vector.reduce.fmin.v8f32(<8 x float>)
declare double @llvm.vector.reduce.fmin.v2f64(<2 x double>)
declare double @llvm.vector.reduce.fmin.v4f64(<4 x double>)
declare fp128 @llvm.vector.reduce.fmin.v4f128(<4 x fp128>)


declare half @llvm.vector.reduce.fmax.v2f16(<2 x half>)
declare half @llvm.vector.reduce.fmax.v4f16(<4 x half>)
declare half @llvm.vector.reduce.fmax.v8f16(<8 x half>)
declare half @llvm.vector.reduce.fmax.v16f16(<16 x half>)
declare float @llvm.vector.reduce.fmax.v2f32(<2 x float>)
declare float @llvm.vector.reduce.fmax.v4f32(<4 x float>)
declare float @llvm.vector.reduce.fmax.v8f32(<8 x float>)
declare double @llvm.vector.reduce.fmax.v2f64(<2 x double>)
declare double @llvm.vector.reduce.fmax.v4f64(<4 x double>)
declare fp128 @llvm.vector.reduce.fmax.v4f128(<4 x fp128>)