1 | ; RUN: opt < %s -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck -check-prefix=CHECK -check-prefix=AVX1 %s |
---|
2 | ; RUN: opt < %s -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mcpu=core-avx2 | FileCheck -check-prefix=CHECK -check-prefix=AVX2 %s |
---|
3 | ; RUN: opt < %s -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mcpu=knl | FileCheck -check-prefix=CHECK -check-prefix=AVX512 %s |
---|
4 | |
---|
5 | 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" |
---|
6 | target triple = "x86_64-apple-macosx10.8.0" |
---|
7 | |
---|
8 | define i32 @cmp(i32 %arg) { |
---|
9 | ; -- floats -- |
---|
10 | ;AVX1: cost of 1 {{.*}} fcmp |
---|
11 | ;AVX2: cost of 1 {{.*}} fcmp |
---|
12 | %A = fcmp olt <2 x float> undef, undef |
---|
13 | ;AVX1: cost of 1 {{.*}} fcmp |
---|
14 | ;AVX2: cost of 1 {{.*}} fcmp |
---|
15 | %B = fcmp olt <4 x float> undef, undef |
---|
16 | ;AVX1: cost of 1 {{.*}} fcmp |
---|
17 | ;AVX2: cost of 1 {{.*}} fcmp |
---|
18 | %C = fcmp olt <8 x float> undef, undef |
---|
19 | ;AVX1: cost of 1 {{.*}} fcmp |
---|
20 | ;AVX2: cost of 1 {{.*}} fcmp |
---|
21 | %D = fcmp olt <2 x double> undef, undef |
---|
22 | ;AVX1: cost of 1 {{.*}} fcmp |
---|
23 | ;AVX2: cost of 1 {{.*}} fcmp |
---|
24 | %E = fcmp olt <4 x double> undef, undef |
---|
25 | |
---|
26 | ; AVX512: cost of 1 {{.*}} %E1 = fcmp |
---|
27 | %E1 = fcmp olt <16 x float> undef, undef |
---|
28 | ; AVX512: cost of 2 {{.*}} %E2 = fcmp |
---|
29 | %E2 = fcmp olt <16 x double> undef, undef |
---|
30 | |
---|
31 | ; -- integers -- |
---|
32 | |
---|
33 | ;AVX1: cost of 1 {{.*}} icmp |
---|
34 | ;AVX2: cost of 1 {{.*}} icmp |
---|
35 | %F = icmp eq <16 x i8> undef, undef |
---|
36 | ;AVX1: cost of 1 {{.*}} icmp |
---|
37 | ;AVX2: cost of 1 {{.*}} icmp |
---|
38 | %G = icmp eq <8 x i16> undef, undef |
---|
39 | ;AVX1: cost of 1 {{.*}} icmp |
---|
40 | ;AVX2: cost of 1 {{.*}} icmp |
---|
41 | %H = icmp eq <4 x i32> undef, undef |
---|
42 | ;AVX1: cost of 1 {{.*}} icmp |
---|
43 | ;AVX2: cost of 1 {{.*}} icmp |
---|
44 | %I = icmp eq <2 x i64> undef, undef |
---|
45 | ;AVX1: cost of 4 {{.*}} icmp |
---|
46 | ;AVX2: cost of 1 {{.*}} icmp |
---|
47 | %J = icmp eq <4 x i64> undef, undef |
---|
48 | ;AVX1: cost of 4 {{.*}} icmp |
---|
49 | ;AVX2: cost of 1 {{.*}} icmp |
---|
50 | %K = icmp eq <8 x i32> undef, undef |
---|
51 | ;AVX1: cost of 4 {{.*}} icmp |
---|
52 | ;AVX2: cost of 1 {{.*}} icmp |
---|
53 | %L = icmp eq <16 x i16> undef, undef |
---|
54 | ;AVX1: cost of 4 {{.*}} icmp |
---|
55 | ;AVX2: cost of 1 {{.*}} icmp |
---|
56 | %M = icmp eq <32 x i8> undef, undef |
---|
57 | |
---|
58 | ; AVX512: cost of 1 {{.*}} %M1 = icmp |
---|
59 | %M1 = icmp eq <16 x i32> undef, undef |
---|
60 | ; AVX512: cost of 2 {{.*}} %M2 = icmp |
---|
61 | %M2 = icmp eq <16 x i64> undef, undef |
---|
62 | |
---|
63 | ;CHECK: cost of 0 {{.*}} ret |
---|
64 | ret i32 undef |
---|
65 | } |
---|
66 | |
---|
67 | |
---|