1; RUN: opt < %s -O3 | \
2; RUN:   llc -mtriple=thumbv7-apple-darwin10 -mattr=+neon | FileCheck %s
3
4target datalayout = "e-p:32:32:32-i1:8:32-i8:8:32-i16:16:32-i32:32:32-i64:32:32-f32:32:32-f64:32:32-v64:64:64-v128:128:128-a0:0:32"
5
6define void @fred(i32 %three_by_three, i8* %in, double %dt1, i32 %x_size, i32 %y_size, i8* %bp) nounwind {
7entry:
8; -- The loop following the load should only use a single add-literation
9;    instruction.
10; CHECK: vldr
11; CHECK: adds r{{[0-9]+.*}}#1
12; CHECK-NOT: adds
13; CHECK: subsections_via_symbols
14
15
16  %three_by_three_addr = alloca i32               ; <i32*> [#uses=2]
17  %in_addr = alloca i8*                           ; <i8**> [#uses=2]
18  %dt_addr = alloca float                         ; <float*> [#uses=4]
19  %x_size_addr = alloca i32                       ; <i32*> [#uses=2]
20  %y_size_addr = alloca i32                       ; <i32*> [#uses=1]
21  %bp_addr = alloca i8*                           ; <i8**> [#uses=1]
22  %tmp_image = alloca i8*                         ; <i8**> [#uses=0]
23  %out = alloca i8*                               ; <i8**> [#uses=1]
24  %cp = alloca i8*                                ; <i8**> [#uses=0]
25  %dpt = alloca i8*                               ; <i8**> [#uses=4]
26  %dp = alloca i8*                                ; <i8**> [#uses=2]
27  %ip = alloca i8*                                ; <i8**> [#uses=0]
28  %centre = alloca i32                            ; <i32*> [#uses=0]
29  %tmp = alloca i32                               ; <i32*> [#uses=0]
30  %brightness = alloca i32                        ; <i32*> [#uses=0]
31  %area = alloca i32                              ; <i32*> [#uses=0]
32  %y = alloca i32                                 ; <i32*> [#uses=0]
33  %x = alloca i32                                 ; <i32*> [#uses=2]
34  %j = alloca i32                                 ; <i32*> [#uses=6]
35  %i = alloca i32                                 ; <i32*> [#uses=1]
36  %mask_size = alloca i32                         ; <i32*> [#uses=5]
37  %increment = alloca i32                         ; <i32*> [#uses=1]
38  %n_max = alloca i32                             ; <i32*> [#uses=4]
39  %temp = alloca float                            ; <float*> [#uses=1]
40  %"alloca point" = bitcast i32 0 to i32          ; <i32> [#uses=0]
41  store i32 %three_by_three, i32* %three_by_three_addr
42  store i8* %in, i8** %in_addr
43  %dt = fptrunc double %dt1 to float              ; <float> [#uses=1]
44  store float %dt, float* %dt_addr
45  store i32 %x_size, i32* %x_size_addr
46  store i32 %y_size, i32* %y_size_addr
47  store i8* %bp, i8** %bp_addr
48  %0 = load i8*, i8** %in_addr, align 4                ; <i8*> [#uses=1]
49  store i8* %0, i8** %out, align 4
50  %1 = call  i32 (...) @foo() nounwind ; <i32> [#uses=1]
51  store i32 %1, i32* %i, align 4
52  %2 = load i32, i32* %three_by_three_addr, align 4    ; <i32> [#uses=1]
53  %3 = icmp eq i32 %2, 0                          ; <i1> [#uses=1]
54  br i1 %3, label %bb, label %bb2
55
56bb:                                               ; preds = %entry
57  %4 = load float, float* %dt_addr, align 4              ; <float> [#uses=1]
58  %5 = fpext float %4 to double                   ; <double> [#uses=1]
59  %6 = fmul double %5, 1.500000e+00               ; <double> [#uses=1]
60  %7 = fptosi double %6 to i32                    ; <i32> [#uses=1]
61  %8 = add nsw i32 %7, 1                          ; <i32> [#uses=1]
62  store i32 %8, i32* %mask_size, align 4
63  br label %bb3
64
65bb2:                                              ; preds = %entry
66  store i32 1, i32* %mask_size, align 4
67  br label %bb3
68
69bb3:                                              ; preds = %bb2, %bb
70  %9 = load i32, i32* %mask_size, align 4              ; <i32> [#uses=1]
71  %10 = mul i32 %9, 2                             ; <i32> [#uses=1]
72  %11 = add nsw i32 %10, 1                        ; <i32> [#uses=1]
73  store i32 %11, i32* %n_max, align 4
74  %12 = load i32, i32* %x_size_addr, align 4           ; <i32> [#uses=1]
75  %13 = load i32, i32* %n_max, align 4                 ; <i32> [#uses=1]
76  %14 = sub i32 %12, %13                          ; <i32> [#uses=1]
77  store i32 %14, i32* %increment, align 4
78  %15 = load i32, i32* %n_max, align 4                 ; <i32> [#uses=1]
79  %16 = load i32, i32* %n_max, align 4                 ; <i32> [#uses=1]
80  %17 = mul i32 %15, %16                          ; <i32> [#uses=1]
81  %18 = call  noalias i8* @malloc(i32 %17) nounwind ; <i8*> [#uses=1]
82  store i8* %18, i8** %dp, align 4
83  %19 = load i8*, i8** %dp, align 4                    ; <i8*> [#uses=1]
84  store i8* %19, i8** %dpt, align 4
85  %20 = load float, float* %dt_addr, align 4             ; <float> [#uses=1]
86  %21 = load float, float* %dt_addr, align 4             ; <float> [#uses=1]
87  %22 = fmul float %20, %21                       ; <float> [#uses=1]
88  %23 = fsub float -0.000000e+00, %22             ; <float> [#uses=1]
89  store float %23, float* %temp, align 4
90  %24 = load i32, i32* %mask_size, align 4             ; <i32> [#uses=1]
91  %25 = sub i32 0, %24                            ; <i32> [#uses=1]
92  store i32 %25, i32* %j, align 4
93  br label %bb5
94
95bb4:                                              ; preds = %bb5
96  %26 = load i32, i32* %j, align 4                     ; <i32> [#uses=1]
97  %27 = load i32, i32* %j, align 4                     ; <i32> [#uses=1]
98  %28 = mul i32 %26, %27                          ; <i32> [#uses=1]
99  %29 = sitofp i32 %28 to double                  ; <double> [#uses=1]
100  %30 = fmul double %29, 1.234000e+00             ; <double> [#uses=1]
101  %31 = fptosi double %30 to i32                  ; <i32> [#uses=1]
102  store i32 %31, i32* %x, align 4
103  %32 = load i32, i32* %x, align 4                     ; <i32> [#uses=1]
104  %33 = trunc i32 %32 to i8                       ; <i8> [#uses=1]
105  %34 = load i8*, i8** %dpt, align 4                   ; <i8*> [#uses=1]
106  store i8 %33, i8* %34, align 1
107  %35 = load i8*, i8** %dpt, align 4                   ; <i8*> [#uses=1]
108  %36 = getelementptr inbounds i8, i8* %35, i64 1     ; <i8*> [#uses=1]
109  store i8* %36, i8** %dpt, align 4
110  %37 = load i32, i32* %j, align 4                     ; <i32> [#uses=1]
111  %38 = add nsw i32 %37, 1                        ; <i32> [#uses=1]
112  store i32 %38, i32* %j, align 4
113  br label %bb5
114
115bb5:                                              ; preds = %bb4, %bb3
116  %39 = load i32, i32* %j, align 4                     ; <i32> [#uses=1]
117  %40 = load i32, i32* %mask_size, align 4             ; <i32> [#uses=1]
118  %41 = icmp sle i32 %39, %40                     ; <i1> [#uses=1]
119  br i1 %41, label %bb4, label %bb6
120
121bb6:                                              ; preds = %bb5
122  br label %return
123
124return:                                           ; preds = %bb6
125  ret void
126}
127
128declare i32 @foo(...)
129
130declare noalias i8* @malloc(i32) nounwind
131