1 /*
2  *  Test program for MSA instruction PCNT.B
3  *
4  *  Copyright (C) 2019  Wave Computing, Inc.
5  *  Copyright (C) 2019  Aleksandar Markovic <amarkovic@wavecomp.com>
6  *
7  *  This program is free software: you can redistribute it and/or modify
8  *  it under the terms of the GNU General Public License as published by
9  *  the Free Software Foundation, either version 2 of the License, or
10  *  (at your option) any later version.
11  *
12  *  This program is distributed in the hope that it will be useful,
13  *  but WITHOUT ANY WARRANTY; without even the implied warranty of
14  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
15  *  GNU General Public License for more details.
16  *
17  *  You should have received a copy of the GNU General Public License
18  *  along with this program.  If not, see <https://www.gnu.org/licenses/>.
19  *
20  */
21 
22 #include <sys/time.h>
23 #include <stdint.h>
24 
25 #include "../../../../include/wrappers_msa.h"
26 #include "../../../../include/test_inputs_128.h"
27 #include "../../../../include/test_utils_128.h"
28 
29 #define TEST_COUNT_TOTAL (PATTERN_INPUTS_COUNT + RANDOM_INPUTS_COUNT)
30 
31 
32 int32_t main(void)
33 {
34     char *isa_ase_name = "MSA";
35     char *group_name = "Bit Count";
36     char *instruction_name =  "PCNT.B";
37     int32_t ret;
38     uint32_t i;
39     struct timeval start, end;
40     double elapsed_time;
41 
42     uint64_t b128_result[TEST_COUNT_TOTAL][2];
43     uint64_t b128_expect[TEST_COUNT_TOTAL][2] = {
44         { 0x0808080808080808ULL, 0x0808080808080808ULL, },    /*   0  */
45         { 0x0000000000000000ULL, 0x0000000000000000ULL, },
46         { 0x0404040404040404ULL, 0x0404040404040404ULL, },
47         { 0x0404040404040404ULL, 0x0404040404040404ULL, },
48         { 0x0404040404040404ULL, 0x0404040404040404ULL, },
49         { 0x0404040404040404ULL, 0x0404040404040404ULL, },
50         { 0x0504030504030504ULL, 0x0305040305040305ULL, },
51         { 0x0304050304050304ULL, 0x0503040503040503ULL, },
52         { 0x0404040404040404ULL, 0x0404040404040404ULL, },    /*   8  */
53         { 0x0404040404040404ULL, 0x0404040404040404ULL, },
54         { 0x0505040303050504ULL, 0x0303050504030305ULL, },
55         { 0x0303040505030304ULL, 0x0505030304050503ULL, },
56         { 0x0604020604020604ULL, 0x0206040206040206ULL, },
57         { 0x0204060204060204ULL, 0x0602040602040602ULL, },
58         { 0x0702050403060107ULL, 0x0205040306010702ULL, },
59         { 0x0106030405020701ULL, 0x0603040502070106ULL, },
60         { 0x0800080008000800ULL, 0x0800080008000800ULL, },    /*  16  */
61         { 0x0008000800080008ULL, 0x0008000800080008ULL, },
62         { 0x0801060304050207ULL, 0x0008010603040502ULL, },
63         { 0x0007020504030601ULL, 0x0800070205040306ULL, },
64         { 0x0802040600080204ULL, 0x0600080204060008ULL, },
65         { 0x0006040208000604ULL, 0x0208000604020800ULL, },
66         { 0x0803020801040700ULL, 0x0605000803020801ULL, },
67         { 0x0005060007040108ULL, 0x0203080005060007ULL, },
68         { 0x0804000804000804ULL, 0x0008040008040008ULL, },    /*  24  */
69         { 0x0004080004080004ULL, 0x0800040800040800ULL, },
70         { 0x0805000607000408ULL, 0x0102080300080500ULL, },
71         { 0x0003080201080400ULL, 0x0706000508000308ULL, },
72         { 0x0806000408020008ULL, 0x0600040802000806ULL, },
73         { 0x0002080400060800ULL, 0x0208040006080002ULL, },
74         { 0x0807000208050004ULL, 0x0803000608010008ULL, },
75         { 0x0001080600030804ULL, 0x0005080200070800ULL, },
76         { 0x0808000008080000ULL, 0x0808000008080000ULL, },    /*  32  */
77         { 0x0000080800000808ULL, 0x0000080800000808ULL, },
78         { 0x0808010006080300ULL, 0x0408050002080700ULL, },
79         { 0x0000070802000508ULL, 0x0400030806000108ULL, },
80         { 0x0808020004080600ULL, 0x0008080200040806ULL, },
81         { 0x0000060804000208ULL, 0x0800000608040002ULL, },
82         { 0x0808030002080801ULL, 0x0004080700000608ULL, },
83         { 0x0000050806000007ULL, 0x0804000108080200ULL, },
84         { 0x0808040000080804ULL, 0x0000080804000008ULL, },    /*  40  */
85         { 0x0000040808000004ULL, 0x0808000004080800ULL, },
86         { 0x0808050000060807ULL, 0x0000040808010002ULL, },
87         { 0x0000030808020001ULL, 0x0808040000070806ULL, },
88         { 0x0808060000040808ULL, 0x0200000808060000ULL, },
89         { 0x0000020808040000ULL, 0x0608080000020808ULL, },
90         { 0x0808070000020808ULL, 0x0500000408080300ULL, },
91         { 0x0000010808060000ULL, 0x0308080400000508ULL, },
92         { 0x0808080000000808ULL, 0x0800000008080800ULL, },    /*  48  */
93         { 0x0000000808080000ULL, 0x0008080800000008ULL, },
94         { 0x0808080100000608ULL, 0x0803000004080805ULL, },
95         { 0x0000000708080200ULL, 0x0005080804000003ULL, },
96         { 0x0808080200000408ULL, 0x0806000000080808ULL, },
97         { 0x0000000608080400ULL, 0x0002080808000000ULL, },
98         { 0x0808080300000208ULL, 0x0808010000040808ULL, },
99         { 0x0000000508080600ULL, 0x0000070808040000ULL, },
100         { 0x0808080400000008ULL, 0x0808040000000808ULL, },    /*  56  */
101         { 0x0000000408080800ULL, 0x0000040808080000ULL, },
102         { 0x0808080500000006ULL, 0x0808070000000408ULL, },
103         { 0x0000000308080802ULL, 0x0000010808080400ULL, },
104         { 0x0808080600000004ULL, 0x0808080200000008ULL, },
105         { 0x0000000208080804ULL, 0x0000000608080800ULL, },
106         { 0x0808080700000002ULL, 0x0808080500000004ULL, },
107         { 0x0000000108080806ULL, 0x0000000308080804ULL, },
108         { 0x0204050402030401ULL, 0x0405030507060302ULL, },    /*  64  */
109         { 0x0706000404040501ULL, 0x0207060303060306ULL, },
110         { 0x0404050405060401ULL, 0x0404040805040302ULL, },
111         { 0x0305030405030404ULL, 0x0405020404020402ULL, },
112         { 0x0503050506060203ULL, 0x0302050505050405ULL, },
113         { 0x0304060308020406ULL, 0x0403020207040204ULL, },
114         { 0x0405030204040503ULL, 0x0303010207040503ULL, },
115         { 0x0605030404040602ULL, 0x0407040502020505ULL, },
116         { 0x0104020504050503ULL, 0x0705030404040403ULL, },    /*  72  */
117         { 0x0405050304050506ULL, 0x0402050302050504ULL, },
118         { 0x0304060604050404ULL, 0x0403040305030502ULL, },
119         { 0x0304050503030305ULL, 0x0404040505050306ULL, },
120         { 0x0502030504030502ULL, 0x0303030504050403ULL, },
121         { 0x0303040703030303ULL, 0x0603040404030405ULL, },
122         { 0x0302020402060704ULL, 0x0604030705030204ULL, },
123         { 0x0404050305040605ULL, 0x0404050504030405ULL, },
124     };
125 
126     reset_msa_registers();
127 
128     gettimeofday(&start, NULL);
129 
130     for (i = 0; i < TEST_COUNT_TOTAL; i++) {
131         if (i < PATTERN_INPUTS_COUNT) {
132             do_msa_PCNT_B(b128_pattern[i], b128_result[i]);
133         } else {
134             do_msa_PCNT_B(b128_random[i - PATTERN_INPUTS_COUNT],
135                           b128_result[i]);
136         }
137     }
138 
139     gettimeofday(&end, NULL);
140 
141     elapsed_time = (end.tv_sec - start.tv_sec) * 1000.0;
142     elapsed_time += (end.tv_usec - start.tv_usec) / 1000.0;
143 
144     ret = check_results_128(isa_ase_name, group_name, instruction_name,
145                             TEST_COUNT_TOTAL, elapsed_time,
146                             &b128_result[0][0], &b128_expect[0][0]);
147 
148     return ret;
149 }
150