1; RUN: opt < %s  -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
2
3target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
4target triple = "x86_64-apple-macosx10.9.0"
5
6define i32 @foo(i32* nocapture %A) nounwind uwtable readonly ssp {
7vector.ph:
8  br label %vector.body
9
10vector.body:                                      ; preds = %vector.body, %vector.ph
11  %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
12  %vec.phi = phi <2 x i32> [ zeroinitializer, %vector.ph ], [ %12, %vector.body ]
13  %0 = getelementptr inbounds i32* %A, i64 %index
14  %1 = bitcast i32* %0 to <2 x i32>*
15  %2 = load <2 x i32>* %1, align 4
16  %3 = sext <2 x i32> %2 to <2 x i64>
17  ;CHECK: cost of 1 {{.*}} extract
18  %4 = extractelement <2 x i64> %3, i32 0
19  %5 = getelementptr inbounds i32* %A, i64 %4
20  ;CHECK: cost of 1 {{.*}} extract
21  %6 = extractelement <2 x i64> %3, i32 1
22  %7 = getelementptr inbounds i32* %A, i64 %6
23  %8 = load i32* %5, align 4
24  ;CHECK: cost of 1 {{.*}} insert
25  %9 = insertelement <2 x i32> undef, i32 %8, i32 0
26  %10 = load i32* %7, align 4
27  ;CHECK: cost of 1 {{.*}} insert
28  %11 = insertelement <2 x i32> %9, i32 %10, i32 1
29  %12 = add nsw <2 x i32> %11, %vec.phi
30  %index.next = add i64 %index, 2
31  %13 = icmp eq i64 %index.next, 192
32  br i1 %13, label %for.end, label %vector.body
33
34for.end:                                          ; preds = %vector.body
35  %14 = extractelement <2 x i32> %12, i32 0
36  %15 = extractelement <2 x i32> %12, i32 1
37  %16 = add i32 %14, %15
38  ret i32 %16
39}
40