1 /*
2 * Test program for MSA instruction HADD_U.D
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
main(void)36 int32_t main(void)
37 {
38 char *isa_ase_name = "MSA";
39 char *group_name = "Int Add";
40 char *instruction_name = "HADD_U.D";
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 { 0x00000001fffffffeULL, 0x00000001fffffffeULL, }, /* 0 */
49 { 0x00000000ffffffffULL, 0x00000000ffffffffULL, },
50 { 0x00000001aaaaaaa9ULL, 0x00000001aaaaaaa9ULL, },
51 { 0x0000000155555554ULL, 0x0000000155555554ULL, },
52 { 0x00000001cccccccbULL, 0x00000001cccccccbULL, },
53 { 0x0000000133333332ULL, 0x0000000133333332ULL, },
54 { 0x000000018e38e38dULL, 0x00000001e38e38e2ULL, },
55 { 0x0000000171c71c70ULL, 0x000000011c71c71bULL, },
56 { 0x00000000ffffffffULL, 0x00000000ffffffffULL, }, /* 8 */
57 { 0x0000000000000000ULL, 0x0000000000000000ULL, },
58 { 0x00000000aaaaaaaaULL, 0x00000000aaaaaaaaULL, },
59 { 0x0000000055555555ULL, 0x0000000055555555ULL, },
60 { 0x00000000ccccccccULL, 0x00000000ccccccccULL, },
61 { 0x0000000033333333ULL, 0x0000000033333333ULL, },
62 { 0x000000008e38e38eULL, 0x00000000e38e38e3ULL, },
63 { 0x0000000071c71c71ULL, 0x000000001c71c71cULL, },
64 { 0x00000001aaaaaaa9ULL, 0x00000001aaaaaaa9ULL, }, /* 16 */
65 { 0x00000000aaaaaaaaULL, 0x00000000aaaaaaaaULL, },
66 { 0x0000000155555554ULL, 0x0000000155555554ULL, },
67 { 0x00000000ffffffffULL, 0x00000000ffffffffULL, },
68 { 0x0000000177777776ULL, 0x0000000177777776ULL, },
69 { 0x00000000ddddddddULL, 0x00000000ddddddddULL, },
70 { 0x0000000138e38e38ULL, 0x000000018e38e38dULL, },
71 { 0x000000011c71c71bULL, 0x00000000c71c71c6ULL, },
72 { 0x0000000155555554ULL, 0x0000000155555554ULL, }, /* 24 */
73 { 0x0000000055555555ULL, 0x0000000055555555ULL, },
74 { 0x00000000ffffffffULL, 0x00000000ffffffffULL, },
75 { 0x00000000aaaaaaaaULL, 0x00000000aaaaaaaaULL, },
76 { 0x0000000122222221ULL, 0x0000000122222221ULL, },
77 { 0x0000000088888888ULL, 0x0000000088888888ULL, },
78 { 0x00000000e38e38e3ULL, 0x0000000138e38e38ULL, },
79 { 0x00000000c71c71c6ULL, 0x0000000071c71c71ULL, },
80 { 0x00000001cccccccbULL, 0x00000001cccccccbULL, }, /* 32 */
81 { 0x00000000ccccccccULL, 0x00000000ccccccccULL, },
82 { 0x0000000177777776ULL, 0x0000000177777776ULL, },
83 { 0x0000000122222221ULL, 0x0000000122222221ULL, },
84 { 0x0000000199999998ULL, 0x0000000199999998ULL, },
85 { 0x00000000ffffffffULL, 0x00000000ffffffffULL, },
86 { 0x000000015b05b05aULL, 0x00000001b05b05afULL, },
87 { 0x000000013e93e93dULL, 0x00000000e93e93e8ULL, },
88 { 0x0000000133333332ULL, 0x0000000133333332ULL, }, /* 40 */
89 { 0x0000000033333333ULL, 0x0000000033333333ULL, },
90 { 0x00000000ddddddddULL, 0x00000000ddddddddULL, },
91 { 0x0000000088888888ULL, 0x0000000088888888ULL, },
92 { 0x00000000ffffffffULL, 0x00000000ffffffffULL, },
93 { 0x0000000066666666ULL, 0x0000000066666666ULL, },
94 { 0x00000000c16c16c1ULL, 0x0000000116c16c16ULL, },
95 { 0x00000000a4fa4fa4ULL, 0x000000004fa4fa4fULL, },
96 { 0x00000001e38e38e2ULL, 0x0000000138e38e37ULL, }, /* 48 */
97 { 0x00000000e38e38e3ULL, 0x0000000038e38e38ULL, },
98 { 0x000000018e38e38dULL, 0x00000000e38e38e2ULL, },
99 { 0x0000000138e38e38ULL, 0x000000008e38e38dULL, },
100 { 0x00000001b05b05afULL, 0x0000000105b05b04ULL, },
101 { 0x0000000116c16c16ULL, 0x000000006c16c16bULL, },
102 { 0x0000000171c71c71ULL, 0x000000011c71c71bULL, },
103 { 0x0000000155555554ULL, 0x0000000055555554ULL, },
104 { 0x000000011c71c71bULL, 0x00000001c71c71c6ULL, }, /* 56 */
105 { 0x000000001c71c71cULL, 0x00000000c71c71c7ULL, },
106 { 0x00000000c71c71c6ULL, 0x0000000171c71c71ULL, },
107 { 0x0000000071c71c71ULL, 0x000000011c71c71cULL, },
108 { 0x00000000e93e93e8ULL, 0x0000000193e93e93ULL, },
109 { 0x000000004fa4fa4fULL, 0x00000000fa4fa4faULL, },
110 { 0x00000000aaaaaaaaULL, 0x00000001aaaaaaaaULL, },
111 { 0x000000008e38e38dULL, 0x00000000e38e38e3ULL, },
112 { 0x00000000b0cd3c0cULL, 0x0000000149e2bb6aULL, }, /* 64 */
113 { 0x00000000d5feadd4ULL, 0x0000000060a65e5aULL, },
114 { 0x00000001423a724cULL, 0x00000000f6923072ULL, },
115 { 0x00000000e69cc91aULL, 0x00000000f4a9edfeULL, },
116 { 0x00000001242055a3ULL, 0x0000000111736b26ULL, },
117 { 0x000000014951c76bULL, 0x0000000028370e16ULL, },
118 { 0x00000001b58d8be3ULL, 0x00000000be22e02eULL, },
119 { 0x0000000159efe2b1ULL, 0x00000000bc3a9dbaULL, },
120 { 0x00000000d4bd03eaULL, 0x000000012654770bULL, }, /* 72 */
121 { 0x00000000f9ee75b2ULL, 0x000000003d1819fbULL, },
122 { 0x00000001662a3a2aULL, 0x00000000d303ec13ULL, },
123 { 0x000000010a8c90f8ULL, 0x00000000d11ba99fULL, },
124 { 0x0000000098b16b8dULL, 0x000000018c6d38e4ULL, },
125 { 0x00000000bde2dd55ULL, 0x00000000a330dbd4ULL, },
126 { 0x000000012a1ea1cdULL, 0x00000001391cadecULL, },
127 { 0x00000000ce80f89bULL, 0x0000000137346b78ULL, },
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_HADD_U_D(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_HADD_U_D(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