• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1; RUN: opt -cost-model -analyze -mtriple=thumbv7-apple-ios6.0.0 -mcpu=swift < %s | FileCheck %s
2
3target datalayout = "e-p:32:32:32-i1:8:32-i8:8:32-i16:16:32-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:32:64-v128:32:128-a0:0:32-n32-S32"
4target triple = "thumbv7-apple-ios6.0.0"
5
6define void @test_geps(i32 %i) {
7  ; GEPs with index 0 are essentially NOOPs.
8;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds i8, i8*
9  %a0 = getelementptr inbounds i8, i8* undef, i32 0
10;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds i16, i16*
11  %a1 = getelementptr inbounds i16, i16* undef, i32 0
12;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds i32, i32*
13  %a2 = getelementptr inbounds i32, i32* undef, i32 0
14;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds i64, i64*
15  %a3 = getelementptr inbounds i64, i64* undef, i32 0
16;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds float, float*
17  %a4 = getelementptr inbounds float, float* undef, i32 0
18;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds double, double*
19  %a5 = getelementptr inbounds double, double* undef, i32 0
20;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds <4 x i8>, <4 x i8>*
21  %a7 = getelementptr inbounds <4 x i8>, <4 x i8>* undef, i32 0
22;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds <4 x i16>, <4 x i16>*
23  %a8 = getelementptr inbounds <4 x i16>, <4 x i16>* undef, i32 0
24;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds <4 x i32>, <4 x i32>*
25  %a9 = getelementptr inbounds <4 x i32>, <4 x i32>* undef, i32 0
26;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds <4 x i64>, <4 x i64>*
27  %a10 = getelementptr inbounds <4 x i64>, <4 x i64>* undef, i32 0
28;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds <4 x float>, <4 x float>*
29  %a11 = getelementptr inbounds <4 x float>, <4 x float>* undef, i32 0
30;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds <4 x double>, <4 x double>*
31  %a12 = getelementptr inbounds <4 x double>, <4 x double>* undef, i32 0
32
33  ; Cost of GEPs is one if we cannot fold the address computation.
34;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds i8, i8*
35  %b0 = getelementptr inbounds i8, i8* undef, i32 1024
36;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds i16, i16*
37  %b1 = getelementptr inbounds i16, i16* undef, i32 1024
38  ; Thumb-2 cannot fold offset >= 2^12 into address computation.
39;CHECK: cost of 1 for instruction: {{.*}} getelementptr inbounds i32, i32*
40  %b2 = getelementptr inbounds i32, i32* undef, i32 1024
41;CHECK: cost of 1 for instruction: {{.*}} getelementptr inbounds i64, i64*
42  %b3 = getelementptr inbounds i64, i64* undef, i32 1024
43;CHECK: cost of 1 for instruction: {{.*}} getelementptr inbounds float, float*
44  %b4 = getelementptr inbounds float, float* undef, i32 1024
45;CHECK: cost of 1 for instruction: {{.*}} getelementptr inbounds double, double*
46  %b5 = getelementptr inbounds double, double* undef, i32 1024
47;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds <4 x i8>, <4 x i8>*
48  %b7 = getelementptr inbounds <4 x i8>, <4 x i8>* undef, i32 1
49;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds <4 x i16>, <4 x i16>*
50  %b8 = getelementptr inbounds <4 x i16>, <4 x i16>* undef, i32 1
51;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds <4 x i32>, <4 x i32>*
52  %b9 = getelementptr inbounds <4 x i32>, <4 x i32>* undef, i32 1
53;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds <4 x i64>, <4 x i64>*
54  %b10 = getelementptr inbounds <4 x i64>, <4 x i64>* undef, i32 1
55;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds <4 x float>, <4 x float>*
56  %b11 = getelementptr inbounds <4 x float>, <4 x float>* undef, i32 1
57;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds <4 x double>, <4 x double>*
58  %b12 = getelementptr inbounds <4 x double>, <4 x double>* undef, i32 1
59
60;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds i8, i8*
61  %c0 = getelementptr inbounds i8, i8* undef, i32 %i
62;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds i16, i16*
63  %c1 = getelementptr inbounds i16, i16* undef, i32 %i
64;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds i32, i32*
65  %c2 = getelementptr inbounds i32, i32* undef, i32 %i
66;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds i64, i64*
67  %c3 = getelementptr inbounds i64, i64* undef, i32 %i
68;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds float, float*
69  %c4 = getelementptr inbounds float, float* undef, i32 %i
70;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds double, double*
71  %c5 = getelementptr inbounds double, double* undef, i32 %i
72;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds <4 x i8>, <4 x i8>*
73  %c7 = getelementptr inbounds <4 x i8>, <4 x i8>* undef, i32 %i
74;CHECK: cost of 0 for instruction: {{.*}} getelementptr inbounds <4 x i16>, <4 x i16>*
75  %c8 = getelementptr inbounds <4 x i16>, <4 x i16>* undef, i32 %i
76  ; Thumb-2 cannot fold scales larger than 8 to address computation.
77;CHECK: cost of 1 for instruction: {{.*}} getelementptr inbounds <4 x i32>, <4 x i32>*
78  %c9 = getelementptr inbounds <4 x i32>, <4 x i32>* undef, i32 %i
79;CHECK: cost of 1 for instruction: {{.*}} getelementptr inbounds <4 x i64>, <4 x i64>*
80  %c10 = getelementptr inbounds <4 x i64>, <4 x i64>* undef, i32 %i
81;CHECK: cost of 1 for instruction: {{.*}} getelementptr inbounds <4 x float>, <4 x float>*
82  %c11 = getelementptr inbounds <4 x float>, <4 x float>* undef, i32 %i
83;CHECK: cost of 1 for instruction: {{.*}} getelementptr inbounds <4 x double>, <4 x double>*
84  %c12 = getelementptr inbounds <4 x double>, <4 x double>* undef, i32 %i
85
86  ret void
87}
88