1
; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
3
target 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"
4
target triple = "x86_64-apple-macosx10.8.0"
7
define i32 @rollable(i32* noalias nocapture %in, i32* noalias nocapture %out, i64 %n) {
9
br i1 %1, label %._crit_edge, label %.lr.ph
11
.lr.ph: ; preds = %0, %.lr.ph
12
%i.019 = phi i64 [ %26, %.lr.ph ], [ 0, %0 ]
13
%2 = shl i64 %i.019, 2
14
%3 = getelementptr inbounds i32, i32* %in, i64 %2
16
%4 = load i32, i32* %3, align 4
18
%6 = getelementptr inbounds i32, i32* %in, i64 %5
19
%7 = load i32, i32* %6, align 4
21
%9 = getelementptr inbounds i32, i32* %in, i64 %8
22
%10 = load i32, i32* %9, align 4
24
%12 = getelementptr inbounds i32, i32* %in, i64 %11
25
%13 = load i32, i32* %12, align 4
36
%22 = getelementptr inbounds i32, i32* %out, i64 %2
37
;CHECK:store <4 x i32>
38
store i32 %15, i32* %22, align 4
39
%23 = getelementptr inbounds i32, i32* %out, i64 %5
40
store i32 %17, i32* %23, align 4
41
%24 = getelementptr inbounds i32, i32* %out, i64 %8
42
store i32 %19, i32* %24, align 4
43
%25 = getelementptr inbounds i32, i32* %out, i64 %11
44
store i32 %21, i32* %25, align 4
45
%26 = add i64 %i.019, 1
46
%exitcond = icmp eq i64 %26, %n
47
br i1 %exitcond, label %._crit_edge, label %.lr.ph
49
._crit_edge: ; preds = %.lr.ph, %0
57
define i32 @unrollable(i32* %in, i32* %out, i64 %n) nounwind ssp uwtable {
58
%1 = icmp eq i64 %n, 0
59
br i1 %1, label %._crit_edge, label %.lr.ph
61
.lr.ph: ; preds = %0, %.lr.ph
62
%i.019 = phi i64 [ %26, %.lr.ph ], [ 0, %0 ]
63
%2 = shl i64 %i.019, 2
64
%3 = getelementptr inbounds i32, i32* %in, i64 %2
65
%4 = load i32, i32* %3, align 4
67
%6 = getelementptr inbounds i32, i32* %in, i64 %5
68
%7 = load i32, i32* %6, align 4
70
%9 = getelementptr inbounds i32, i32* %in, i64 %8
71
%10 = load i32, i32* %9, align 4
73
%12 = getelementptr inbounds i32, i32* %in, i64 %11
74
%13 = load i32, i32* %12, align 4
83
%22 = getelementptr inbounds i32, i32* %out, i64 %2
84
store i32 %15, i32* %22, align 4
85
%23 = getelementptr inbounds i32, i32* %out, i64 %5
86
store i32 %17, i32* %23, align 4
87
%barrier = call i32 @goo(i32 0) ; <---------------- memory barrier.
88
%24 = getelementptr inbounds i32, i32* %out, i64 %8
89
store i32 %19, i32* %24, align 4
90
%25 = getelementptr inbounds i32, i32* %out, i64 %11
91
store i32 %21, i32* %25, align 4
92
%26 = add i64 %i.019, 1
93
%exitcond = icmp eq i64 %26, %n
94
br i1 %exitcond, label %._crit_edge, label %.lr.ph
96
._crit_edge: ; preds = %.lr.ph, %0
100
declare i32 @goo(i32)