/root/doris/contrib/openblas/interface/syr2.c
Line | Count | Source |
1 | | /*********************************************************************/ |
2 | | /* Copyright 2009, 2010 The University of Texas at Austin. */ |
3 | | /* All rights reserved. */ |
4 | | /* */ |
5 | | /* Redistribution and use in source and binary forms, with or */ |
6 | | /* without modification, are permitted provided that the following */ |
7 | | /* conditions are met: */ |
8 | | /* */ |
9 | | /* 1. Redistributions of source code must retain the above */ |
10 | | /* copyright notice, this list of conditions and the following */ |
11 | | /* disclaimer. */ |
12 | | /* */ |
13 | | /* 2. Redistributions in binary form must reproduce the above */ |
14 | | /* copyright notice, this list of conditions and the following */ |
15 | | /* disclaimer in the documentation and/or other materials */ |
16 | | /* provided with the distribution. */ |
17 | | /* */ |
18 | | /* THIS SOFTWARE IS PROVIDED BY THE UNIVERSITY OF TEXAS AT */ |
19 | | /* AUSTIN ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, */ |
20 | | /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */ |
21 | | /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */ |
22 | | /* DISCLAIMED. IN NO EVENT SHALL THE UNIVERSITY OF TEXAS AT */ |
23 | | /* AUSTIN OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, */ |
24 | | /* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES */ |
25 | | /* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE */ |
26 | | /* GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR */ |
27 | | /* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF */ |
28 | | /* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT */ |
29 | | /* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT */ |
30 | | /* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */ |
31 | | /* POSSIBILITY OF SUCH DAMAGE. */ |
32 | | /* */ |
33 | | /* The views and conclusions contained in the software and */ |
34 | | /* documentation are those of the authors and should not be */ |
35 | | /* interpreted as representing official policies, either expressed */ |
36 | | /* or implied, of The University of Texas at Austin. */ |
37 | | /*********************************************************************/ |
38 | | |
39 | | #include <stdio.h> |
40 | | #include <ctype.h> |
41 | | #include "common.h" |
42 | | #ifdef FUNCTION_PROFILE |
43 | | #include "functable.h" |
44 | | #endif |
45 | | |
46 | | #ifdef XDOUBLE |
47 | | #define ERROR_NAME "QSYR2 " |
48 | | #elif defined(DOUBLE) |
49 | 0 | #define ERROR_NAME "DSYR2 " |
50 | | #else |
51 | | #define ERROR_NAME "SSYR2 " |
52 | | #endif |
53 | | |
54 | | static int (*syr2[])(BLASLONG, FLOAT, FLOAT *, BLASLONG, FLOAT *, BLASLONG, FLOAT *, BLASLONG, FLOAT *) = { |
55 | | #ifdef XDOUBLE |
56 | | qsyr2_U, qsyr2_L, |
57 | | #elif defined(DOUBLE) |
58 | | dsyr2_U, dsyr2_L, |
59 | | #else |
60 | | ssyr2_U, ssyr2_L, |
61 | | #endif |
62 | | }; |
63 | | |
64 | | #ifdef SMP |
65 | | static int (*syr2_thread[])(BLASLONG, FLOAT, FLOAT *, BLASLONG, FLOAT *, BLASLONG, FLOAT *, BLASLONG, FLOAT *, int) = { |
66 | | #ifdef XDOUBLE |
67 | | qsyr2_thread_U, qsyr2_thread_L, |
68 | | #elif defined(DOUBLE) |
69 | | dsyr2_thread_U, dsyr2_thread_L, |
70 | | #else |
71 | | ssyr2_thread_U, ssyr2_thread_L, |
72 | | #endif |
73 | | }; |
74 | | #endif |
75 | | |
76 | | #ifndef CBLAS |
77 | | |
78 | | void NAME(char *UPLO, blasint *N, FLOAT *ALPHA, |
79 | 0 | FLOAT *x, blasint *INCX, FLOAT *y, blasint *INCY, FLOAT *a, blasint *LDA){ |
80 | |
|
81 | 0 | char uplo_arg = *UPLO; |
82 | 0 | blasint n = *N; |
83 | 0 | FLOAT alpha = *ALPHA; |
84 | 0 | blasint lda = *LDA; |
85 | 0 | blasint incx = *INCX; |
86 | 0 | blasint incy = *INCY; |
87 | |
|
88 | 0 | blasint info; |
89 | 0 | int uplo; |
90 | 0 | FLOAT *buffer; |
91 | 0 | #ifdef SMP |
92 | 0 | int nthreads; |
93 | 0 | #endif |
94 | |
|
95 | 0 | PRINT_DEBUG_NAME; |
96 | |
|
97 | 0 | TOUPPER(uplo_arg); |
98 | 0 | uplo = -1; |
99 | |
|
100 | 0 | if (uplo_arg == 'U') uplo = 0; |
101 | 0 | if (uplo_arg == 'L') uplo = 1; |
102 | |
|
103 | 0 | info = 0; |
104 | |
|
105 | 0 | if (lda < MAX(1, n)) info = 9; |
106 | 0 | if (incy == 0) info = 7; |
107 | 0 | if (incx == 0) info = 5; |
108 | 0 | if (n < 0) info = 2; |
109 | 0 | if (uplo < 0) info = 1; |
110 | |
|
111 | 0 | if (info != 0) { |
112 | 0 | BLASFUNC(xerbla)(ERROR_NAME, &info, sizeof(ERROR_NAME)); |
113 | 0 | return; |
114 | 0 | } |
115 | | |
116 | | #else |
117 | | |
118 | | void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint n, FLOAT alpha, FLOAT *x, blasint incx, FLOAT *y, blasint incy, FLOAT *a, blasint lda) { |
119 | | |
120 | | FLOAT *buffer; |
121 | | int uplo; |
122 | | blasint info; |
123 | | #ifdef SMP |
124 | | int nthreads; |
125 | | #endif |
126 | | |
127 | | PRINT_DEBUG_CNAME; |
128 | | |
129 | | uplo = -1; |
130 | | info = 0; |
131 | | |
132 | | if (order == CblasColMajor) { |
133 | | |
134 | | if (Uplo == CblasUpper) uplo = 0; |
135 | | if (Uplo == CblasLower) uplo = 1; |
136 | | |
137 | | info = -1; |
138 | | |
139 | | if (lda < MAX(1, n)) info = 9; |
140 | | if (incy == 0) info = 7; |
141 | | if (incx == 0) info = 5; |
142 | | if (n < 0) info = 2; |
143 | | if (uplo < 0) info = 1; |
144 | | } |
145 | | |
146 | | if (order == CblasRowMajor) { |
147 | | |
148 | | if (Uplo == CblasUpper) uplo = 1; |
149 | | if (Uplo == CblasLower) uplo = 0; |
150 | | |
151 | | info = -1; |
152 | | |
153 | | if (lda < MAX(1, n)) info = 9; |
154 | | if (incy == 0) info = 7; |
155 | | if (incx == 0) info = 5; |
156 | | if (n < 0) info = 2; |
157 | | if (uplo < 0) info = 1; |
158 | | } |
159 | | |
160 | | if (info >= 0) { |
161 | | BLASFUNC(xerbla)(ERROR_NAME, &info, sizeof(ERROR_NAME)); |
162 | | return; |
163 | | } |
164 | | |
165 | | #endif |
166 | | |
167 | 0 | if (n == 0) return; |
168 | | |
169 | 0 | if (alpha == ZERO) return; |
170 | | |
171 | 0 | IDEBUG_START; |
172 | |
|
173 | 0 | if (incx == 1 && incy == 1 && n < 100) { |
174 | 0 | blasint i; |
175 | 0 | if (!uplo) { |
176 | 0 | for (i = 0; i < n; i++){ |
177 | 0 | AXPYU_K(i + 1, 0, 0, alpha * x[i], y, 1, a, 1, NULL, 0); |
178 | 0 | AXPYU_K(i + 1, 0, 0, alpha * y[i], x, 1, a, 1, NULL, 0); |
179 | 0 | a += lda; |
180 | 0 | } |
181 | 0 | } else { |
182 | 0 | for (i = 0; i < n; i++){ |
183 | 0 | AXPYU_K(n - i, 0, 0, alpha * x[i], y + i, 1, a, 1, NULL, 0); |
184 | 0 | AXPYU_K(n - i, 0, 0, alpha * y[i], x + i, 1, a, 1, NULL, 0); |
185 | 0 | a += 1 + lda; |
186 | 0 | } |
187 | 0 | } |
188 | 0 | return; |
189 | 0 | } |
190 | | |
191 | | |
192 | 0 | FUNCTION_PROFILE_START(); |
193 | |
|
194 | 0 | if (incx < 0 ) x -= (n - 1) * incx; |
195 | 0 | if (incy < 0 ) y -= (n - 1) * incy; |
196 | |
|
197 | 0 | buffer = (FLOAT *)blas_memory_alloc(1); |
198 | |
|
199 | 0 | #ifdef SMP |
200 | 0 | nthreads = num_cpu_avail(2); |
201 | |
|
202 | 0 | if (nthreads == 1) { |
203 | 0 | #endif |
204 | |
|
205 | 0 | (syr2[uplo])(n, alpha, x, incx, y, incy, a, lda, buffer); |
206 | |
|
207 | 0 | #ifdef SMP |
208 | 0 | } else { |
209 | |
|
210 | 0 | (syr2_thread[uplo])(n, alpha, x, incx, y, incy, a, lda, buffer, nthreads); |
211 | |
|
212 | 0 | } |
213 | 0 | #endif |
214 | |
|
215 | 0 | blas_memory_free(buffer); |
216 | |
|
217 | 0 | FUNCTION_PROFILE_END(1, n * n / 2 + 2 * n, 2 * n * n); |
218 | |
|
219 | 0 | IDEBUG_END; |
220 | |
|
221 | 0 | return; |
222 | 0 | } |