1 /*
2  *  Test program for MSA instruction ADDS_U.W
3  *
4  *  Copyright (C) 2019  Wave Computing, Inc.
5  *  Copyright (C) 2019  Aleksandar Markovic <amarkovic@wavecomp.com>
6  *  Copyright (C) 2019  RT-RK Computer Based Systems LLC
7  *  Copyright (C) 2019  Mateja Marjanovic <mateja.marjanovic@rt-rk.com>
8  *
9  *  This program is free software: you can redistribute it and/or modify
10  *  it under the terms of the GNU General Public License as published by
11  *  the Free Software Foundation, either version 2 of the License, or
12  *  (at your option) any later version.
13  *
14  *  This program is distributed in the hope that it will be useful,
15  *  but WITHOUT ANY WARRANTY; without even the implied warranty of
16  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
17  *  GNU General Public License for more details.
18  *
19  *  You should have received a copy of the GNU General Public License
20  *  along with this program.  If not, see <https://www.gnu.org/licenses/>.
21  *
22  */
23 
24 #include <sys/time.h>
25 #include <stdint.h>
26 
27 #include "../../../../include/wrappers_msa.h"
28 #include "../../../../include/test_inputs_128.h"
29 #include "../../../../include/test_utils_128.h"
30 
31 #define TEST_COUNT_TOTAL (                                                \
32             (PATTERN_INPUTS_SHORT_COUNT) * (PATTERN_INPUTS_SHORT_COUNT) + \
33             (RANDOM_INPUTS_SHORT_COUNT) * (RANDOM_INPUTS_SHORT_COUNT))
34 
35 
36 int32_t main(void)
37 {
38     char *instruction_name = "ADDS_U.W";
39     int32_t ret;
40     uint32_t i, j;
41     struct timeval start, end;
42     double elapsed_time;
43 
44     uint64_t b128_result[TEST_COUNT_TOTAL][2];
45     uint64_t b128_expect[TEST_COUNT_TOTAL][2] = {
46         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },    /*   0  */
47         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
48         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
49         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
50         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
51         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
52         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
53         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
54         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },    /*   8  */
55         { 0x0000000000000000ULL, 0x0000000000000000ULL, },
56         { 0xaaaaaaaaaaaaaaaaULL, 0xaaaaaaaaaaaaaaaaULL, },
57         { 0x5555555555555555ULL, 0x5555555555555555ULL, },
58         { 0xccccccccccccccccULL, 0xccccccccccccccccULL, },
59         { 0x3333333333333333ULL, 0x3333333333333333ULL, },
60         { 0xe38e38e38e38e38eULL, 0x38e38e38e38e38e3ULL, },
61         { 0x1c71c71c71c71c71ULL, 0xc71c71c71c71c71cULL, },
62         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },    /*  16  */
63         { 0xaaaaaaaaaaaaaaaaULL, 0xaaaaaaaaaaaaaaaaULL, },
64         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
65         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
66         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
67         { 0xddddddddddddddddULL, 0xddddddddddddddddULL, },
68         { 0xffffffffffffffffULL, 0xe38e38e2ffffffffULL, },
69         { 0xc71c71c6ffffffffULL, 0xffffffffc71c71c6ULL, },
70         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },    /*  24  */
71         { 0x5555555555555555ULL, 0x5555555555555555ULL, },
72         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
73         { 0xaaaaaaaaaaaaaaaaULL, 0xaaaaaaaaaaaaaaaaULL, },
74         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
75         { 0x8888888888888888ULL, 0x8888888888888888ULL, },
76         { 0xffffffffe38e38e3ULL, 0x8e38e38dffffffffULL, },
77         { 0x71c71c71c71c71c6ULL, 0xffffffff71c71c71ULL, },
78         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },    /*  32  */
79         { 0xccccccccccccccccULL, 0xccccccccccccccccULL, },
80         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
81         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
82         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
83         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
84         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
85         { 0xe93e93e8ffffffffULL, 0xffffffffe93e93e8ULL, },
86         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },    /*  40  */
87         { 0x3333333333333333ULL, 0x3333333333333333ULL, },
88         { 0xddddddddddddddddULL, 0xddddddddddddddddULL, },
89         { 0x8888888888888888ULL, 0x8888888888888888ULL, },
90         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
91         { 0x6666666666666666ULL, 0x6666666666666666ULL, },
92         { 0xffffffffc16c16c1ULL, 0x6c16c16bffffffffULL, },
93         { 0x4fa4fa4fa4fa4fa4ULL, 0xfa4fa4fa4fa4fa4fULL, },
94         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },    /*  48  */
95         { 0xe38e38e38e38e38eULL, 0x38e38e38e38e38e3ULL, },
96         { 0xffffffffffffffffULL, 0xe38e38e2ffffffffULL, },
97         { 0xffffffffe38e38e3ULL, 0x8e38e38dffffffffULL, },
98         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
99         { 0xffffffffc16c16c1ULL, 0x6c16c16bffffffffULL, },
100         { 0xffffffffffffffffULL, 0x71c71c70ffffffffULL, },
101         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
102         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },    /*  56  */
103         { 0x1c71c71c71c71c71ULL, 0xc71c71c71c71c71cULL, },
104         { 0xc71c71c6ffffffffULL, 0xffffffffc71c71c6ULL, },
105         { 0x71c71c71c71c71c6ULL, 0xffffffff71c71c71ULL, },
106         { 0xe93e93e8ffffffffULL, 0xffffffffe93e93e8ULL, },
107         { 0x4fa4fa4fa4fa4fa4ULL, 0xfa4fa4fa4fa4fa4fULL, },
108         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
109         { 0x38e38e38e38e38e2ULL, 0xffffffff38e38e38ULL, },
110         { 0xffffffff50c4aa80ULL, 0x96ce16bcffffffffULL, },    /*  64  */
111         { 0xffffffff75f61c48ULL, 0x5e5ec678ffffffffULL, },
112         { 0xffffffffe231e0c0ULL, 0x733fd25dffffffffULL, },
113         { 0xf8b9fd198694378eULL, 0xd9589436ffffffffULL, },
114         { 0xffffffff75f61c48ULL, 0x5e5ec678ffffffffULL, },
115         { 0xffffffff9b278e10ULL, 0x25ef76342a7ea5f8ULL, },
116         { 0xffffffffffffffffULL, 0x3ad08219c06a7810ULL, },
117         { 0xffffffffabc5a956ULL, 0xa0e943f2be82359cULL, },
118         { 0xffffffffe231e0c0ULL, 0x733fd25dffffffffULL, },    /*  72  */
119         { 0xffffffffffffffffULL, 0x3ad08219c06a7810ULL, },
120         { 0xffffffffffffffffULL, 0x4fb18dfeffffffffULL, },
121         { 0xffffffffffffffffULL, 0xb5ca4fd7ffffffffULL, },
122         { 0xf8b9fd198694378eULL, 0xd9589436ffffffffULL, },
123         { 0xffffffffabc5a956ULL, 0xa0e943f2be82359cULL, },
124         { 0xffffffffffffffffULL, 0xb5ca4fd7ffffffffULL, },
125         { 0xe09e2c9abc63c49cULL, 0xffffffffffffffffULL, },
126 };
127 
128     gettimeofday(&start, NULL);
129 
130     for (i = 0; i < PATTERN_INPUTS_SHORT_COUNT; i++) {
131         for (j = 0; j < PATTERN_INPUTS_SHORT_COUNT; j++) {
132             do_msa_ADDS_U_W(b128_pattern[i], b128_pattern[j],
133                            b128_result[PATTERN_INPUTS_SHORT_COUNT * i + j]);
134         }
135     }
136 
137     for (i = 0; i < RANDOM_INPUTS_SHORT_COUNT; i++) {
138         for (j = 0; j < RANDOM_INPUTS_SHORT_COUNT; j++) {
139             do_msa_ADDS_U_W(b128_random[i], b128_random[j],
140                            b128_result[((PATTERN_INPUTS_SHORT_COUNT) *
141                                         (PATTERN_INPUTS_SHORT_COUNT)) +
142                                        RANDOM_INPUTS_SHORT_COUNT * i + j]);
143         }
144     }
145 
146     gettimeofday(&end, NULL);
147 
148     elapsed_time = (end.tv_sec - start.tv_sec) * 1000.0;
149     elapsed_time += (end.tv_usec - start.tv_usec) / 1000.0;
150 
151     ret = check_results(instruction_name, TEST_COUNT_TOTAL, elapsed_time,
152                         &b128_result[0][0], &b128_expect[0][0]);
153 
154     return ret;
155 }
156