1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
|
/* Test the `vmaxnmf32' ARM Neon intrinsic. */
/* { dg-do run } */
/* { dg-require-effective-target arm_v8_neon_hw } */
/* { dg-options "-save-temps -O3 -march=armv8-a" } */
/* { dg-add-options arm_v8_neon } */
#include "arm_neon.h"
extern void abort ();
void __attribute__ ((noinline))
test_vmaxnm_f32__regular_input1 ()
{
float32_t a1[] = {1,2};
float32_t b1[] = {3,4};
float32x2_t a = vld1_f32 (a1);
float32x2_t b = vld1_f32 (b1);
float32x2_t c = vmaxnm_f32 (a, b);
float32_t actual[2];
vst1_f32 (actual, c);
for (int i = 0; i < 2; ++i)
if (actual[i] != b1[i])
abort ();
}
void __attribute__ ((noinline))
test_vmaxnm_f32__regular_input2 ()
{
float32_t a1[] = {3,2};
float32_t b1[] = {1,4};
float32_t e[] = {3,4};
float32x2_t a = vld1_f32 (a1);
float32x2_t b = vld1_f32 (b1);
float32x2_t c = vmaxnm_f32 (a, b);
float32_t actual[2];
vst1_f32 (actual, c);
for (int i = 0; i < 2; ++i)
if (actual[i] != e[i])
abort ();
}
void __attribute__ ((noinline))
test_vmaxnm_f32__quiet_NaN_one_arg ()
{
/* When given a quiet NaN, vmaxnm returns the other operand.
In this test case we have NaNs in only one operand. */
float32_t n = __builtin_nanf ("");
float32_t a1[] = {1,2};
float32_t b1[] = {n,n};
float32_t e[] = {1,2};
float32x2_t a = vld1_f32 (a1);
float32x2_t b = vld1_f32 (b1);
float32x2_t c = vmaxnm_f32 (a, b);
float32_t actual[2];
vst1_f32 (actual, c);
for (int i = 0; i < 2; ++i)
if (actual[i] != e[i])
abort ();
}
void __attribute__ ((noinline))
test_vmaxnm_f32__quiet_NaN_both_args ()
{
/* When given a quiet NaN, vmaxnm returns the other operand.
In this test case we have NaNs in both operands. */
float32_t n = __builtin_nanf ("");
float32_t a1[] = {n,2};
float32_t b1[] = {1,n};
float32_t e[] = {1,2};
float32x2_t a = vld1_f32 (a1);
float32x2_t b = vld1_f32 (b1);
float32x2_t c = vmaxnm_f32 (a, b);
float32_t actual[2];
vst1_f32 (actual, c);
for (int i = 0; i < 2; ++i)
if (actual[i] != e[i])
abort ();
}
void __attribute__ ((noinline))
test_vmaxnm_f32__zero_both_args ()
{
/* For 0 and -0, vmaxnm returns 0. Since 0 == -0, check sign bit. */
float32_t a1[] = {0.0, 0.0};
float32_t b1[] = {-0.0, -0.0};
float32_t e[] = {0.0, 0.0};
float32x2_t a = vld1_f32 (a1);
float32x2_t b = vld1_f32 (b1);
float32x2_t c = vmaxnm_f32 (a, b);
float32_t actual1[2];
vst1_f32 (actual1, c);
for (int i = 0; i < 2; ++i)
if (actual1[i] != e[i] || __builtin_signbit (actual1[i]) != 0)
abort ();
}
void __attribute__ ((noinline))
test_vmaxnm_f32__inf_both_args ()
{
/* The max of inf and inf is inf. The max of -inf and -inf is -inf. */
float32_t inf = __builtin_huge_valf ();
float32_t a1[] = {inf, -inf};
float32_t b1[] = {inf, -inf};
float32_t e[] = {inf, -inf};
float32x2_t a = vld1_f32 (a1);
float32x2_t b = vld1_f32 (b1);
float32x2_t c = vmaxnm_f32 (a, b);
float32_t actual1[2];
vst1_f32 (actual1, c);
for (int i = 0; i < 2; ++i)
if (actual1[i] != e[i])
abort ();
}
void __attribute__ ((noinline))
test_vmaxnm_f32__two_quiet_NaNs_both_args ()
{
/* When given 2 NaNs, return a NaN. Since a NaN is not equal to anything,
not even another NaN, use __builtin_isnan () to check. */
float32_t n = __builtin_nanf ("");
float32_t a1[] = {n,n};
float32_t b1[] = {n,n};
float32_t e[] = {n,n};
float32x2_t a = vld1_f32 (a1);
float32x2_t b = vld1_f32 (b1);
float32x2_t c = vmaxnm_f32 (a, b);
float32_t actual[2];
vst1_f32 (actual, c);
for (int i = 0; i < 2; ++i)
if (!__builtin_isnan (actual[i]))
abort ();
}
int
main ()
{
test_vmaxnm_f32__regular_input1 ();
test_vmaxnm_f32__regular_input2 ();
test_vmaxnm_f32__quiet_NaN_one_arg ();
test_vmaxnm_f32__quiet_NaN_both_args ();
test_vmaxnm_f32__zero_both_args ();
test_vmaxnm_f32__inf_both_args ();
test_vmaxnm_f32__two_quiet_NaNs_both_args ();
return 0;
}
/* { dg-final { scan-assembler-times "vmaxnm\.f32\t\[dD\]\[0-9\]+, ?\[dD\]\[0-9\]+, ?\[dD\]\[0-9\]+\n" 7 } } */
|
