1 /*
2  *  Test program for MSA instruction SUBSUS_U.H
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 *isa_ase_name = "MSA";
39     char *group_name = "Int Subtract";
40     char *instruction_name =  "SUBSUS_U.H";
41     int32_t ret;
42     uint32_t i, j;
43     struct timeval start, end;
44     double elapsed_time;
45 
46     uint64_t b128_result[TEST_COUNT_TOTAL][2];
47     uint64_t b128_expect[TEST_COUNT_TOTAL][2] = {
48         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },    /*   0  */
49         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
50         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
51         { 0xaaaaaaaaaaaaaaaaULL, 0xaaaaaaaaaaaaaaaaULL, },
52         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
53         { 0xccccccccccccccccULL, 0xccccccccccccccccULL, },
54         { 0xffffc71cffffffffULL, 0xc71cffffffffc71cULL, },
55         { 0xe38effff8e38e38eULL, 0xffff8e38e38effffULL, },
56         { 0x0001000100010001ULL, 0x0001000100010001ULL, },    /*   8  */
57         { 0x0000000000000000ULL, 0x0000000000000000ULL, },
58         { 0x5556555655565556ULL, 0x5556555655565556ULL, },
59         { 0x0000000000000000ULL, 0x0000000000000000ULL, },
60         { 0x3334333433343334ULL, 0x3334333433343334ULL, },
61         { 0x0000000000000000ULL, 0x0000000000000000ULL, },
62         { 0x1c72000071c81c72ULL, 0x000071c81c720000ULL, },
63         { 0x000038e400000000ULL, 0x38e40000000038e4ULL, },
64         { 0xaaabaaabaaabaaabULL, 0xaaabaaabaaabaaabULL, },    /*  16  */
65         { 0xaaaaaaaaaaaaaaaaULL, 0xaaaaaaaaaaaaaaaaULL, },
66         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
67         { 0x5555555555555555ULL, 0x5555555555555555ULL, },
68         { 0xdddedddedddedddeULL, 0xdddedddedddedddeULL, },
69         { 0x7777777777777777ULL, 0x7777777777777777ULL, },
70         { 0xc71c71c7ffffc71cULL, 0x71c7ffffc71c71c7ULL, },
71         { 0x8e39e38e38e38e39ULL, 0xe38e38e38e39e38eULL, },
72         { 0x5556555655565556ULL, 0x5556555655565556ULL, },    /*  24  */
73         { 0x5555555555555555ULL, 0x5555555555555555ULL, },
74         { 0xaaabaaabaaabaaabULL, 0xaaabaaabaaabaaabULL, },
75         { 0x0000000000000000ULL, 0x0000000000000000ULL, },
76         { 0x8889888988898889ULL, 0x8889888988898889ULL, },
77         { 0x2222222222222222ULL, 0x2222222222222222ULL, },
78         { 0x71c71c72c71d71c7ULL, 0x1c72c71d71c71c72ULL, },
79         { 0x38e48e39000038e4ULL, 0x8e39000038e48e39ULL, },
80         { 0xcccdcccdcccdcccdULL, 0xcccdcccdcccdcccdULL, },    /*  32  */
81         { 0xccccccccccccccccULL, 0xccccccccccccccccULL, },
82         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
83         { 0x7777777777777777ULL, 0x7777777777777777ULL, },
84         { 0xffffffffffffffffULL, 0xffffffffffffffffULL, },
85         { 0x9999999999999999ULL, 0x9999999999999999ULL, },
86         { 0xe93e93e9ffffe93eULL, 0x93e9ffffe93e93e9ULL, },
87         { 0xb05bffff5b05b05bULL, 0xffff5b05b05bffffULL, },
88         { 0x3334333433343334ULL, 0x3334333433343334ULL, },    /*  40  */
89         { 0x3333333333333333ULL, 0x3333333333333333ULL, },
90         { 0x8889888988898889ULL, 0x8889888988898889ULL, },
91         { 0x0000000000000000ULL, 0x0000000000000000ULL, },
92         { 0x6667666766676667ULL, 0x6667666766676667ULL, },
93         { 0x0000000000000000ULL, 0x0000000000000000ULL, },
94         { 0x4fa50000a4fb4fa5ULL, 0x0000a4fb4fa50000ULL, },
95         { 0x16c26c17000016c2ULL, 0x6c17000016c26c17ULL, },
96         { 0xe38f38e48e39e38fULL, 0x38e48e39e38f38e4ULL, },    /*  48  */
97         { 0xe38e38e38e38e38eULL, 0x38e38e38e38e38e3ULL, },
98         { 0xffff8e39e38effffULL, 0x8e39e38effff8e39ULL, },
99         { 0x8e39000038e38e39ULL, 0x000038e38e390000ULL, },
100         { 0xffff6c17c16cffffULL, 0x6c17c16cffff6c17ULL, },
101         { 0xb05b05b05b05b05bULL, 0x05b05b05b05b05b0ULL, },
102         { 0xffff0000ffffffffULL, 0x0000ffffffff0000ULL, },
103         { 0xc71d71c71c71c71dULL, 0x71c71c71c71d71c7ULL, },
104         { 0x1c72c71d71c81c72ULL, 0xc71d71c81c72c71dULL, },    /*  56  */
105         { 0x1c71c71c71c71c71ULL, 0xc71c71c71c71c71cULL, },
106         { 0x71c7ffffc71d71c7ULL, 0xffffc71d71c7ffffULL, },
107         { 0x000071c71c720000ULL, 0x71c71c72000071c7ULL, },
108         { 0x4fa5fa50a4fb4fa5ULL, 0xfa50a4fb4fa5fa50ULL, },
109         { 0x000093e93e940000ULL, 0x93e93e94000093e9ULL, },
110         { 0x38e38e39e38f38e3ULL, 0x8e39e38f38e38e39ULL, },
111         { 0x0000ffff00000000ULL, 0xffff00000000ffffULL, },
112         { 0xffffffff00000000ULL, 0x00000000ffffffffULL, },    /*  64  */
113         { 0x8cace66900008e38ULL, 0x38705044e93c5d10ULL, },
114         { 0xdc10ffff6e93c9c0ULL, 0x238f445fffff8af8ULL, },
115         { 0x181bd07f000072f2ULL, 0xbd768286ffffcd6cULL, },
116         { 0xffff1997253171c8ULL, 0x0000afbc16c4a2f0ULL, },
117         { 0xffff00000000ffffULL, 0x0000ffff00000000ULL, },
118         { 0xffff51b993c4ffffULL, 0x0000f41b6a142de8ULL, },
119         { 0x8b6f00000000e4baULL, 0x8506ffff6bfd705cULL, },
120         { 0xffffc7de916d3640ULL, 0x0000bba1acb07508ULL, },    /*  72  */
121         { 0xb09cae476c3cc478ULL, 0x14e1ffff95ec0000ULL, },
122         { 0xffffffffffffffffULL, 0x0000ffffffff0000ULL, },
123         { 0x3c0b985d5b9ea932ULL, 0x99e7ffffffff4274ULL, },
124         { 0xe7e52f8135cf8d0eULL, 0x428a7d7aaac7ffffULL, },
125         { 0x749115ea109effffULL, 0x7afacdbe94038fa4ULL, },
126         { 0xc3f567a3a462ffffULL, 0x6619c1d9fe17bd8cULL, },
127         { 0x000000000000ffffULL, 0xffffffffffffffffULL, },
128 };
129 
130     reset_msa_registers();
131 
132     gettimeofday(&start, NULL);
133 
134     for (i = 0; i < PATTERN_INPUTS_SHORT_COUNT; i++) {
135         for (j = 0; j < PATTERN_INPUTS_SHORT_COUNT; j++) {
136             do_msa_SUBSUS_U_H(b128_pattern[i], b128_pattern[j],
137                            b128_result[PATTERN_INPUTS_SHORT_COUNT * i + j]);
138         }
139     }
140 
141     for (i = 0; i < RANDOM_INPUTS_SHORT_COUNT; i++) {
142         for (j = 0; j < RANDOM_INPUTS_SHORT_COUNT; j++) {
143             do_msa_SUBSUS_U_H(b128_random[i], b128_random[j],
144                            b128_result[((PATTERN_INPUTS_SHORT_COUNT) *
145                                         (PATTERN_INPUTS_SHORT_COUNT)) +
146                                        RANDOM_INPUTS_SHORT_COUNT * i + j]);
147         }
148     }
149 
150     gettimeofday(&end, NULL);
151 
152     elapsed_time = (end.tv_sec - start.tv_sec) * 1000.0;
153     elapsed_time += (end.tv_usec - start.tv_usec) / 1000.0;
154 
155     ret = check_results_128(isa_ase_name, group_name, instruction_name,
156                             TEST_COUNT_TOTAL, elapsed_time,
157                             &b128_result[0][0], &b128_expect[0][0]);
158 
159     return ret;
160 }
161