/root/doris/contrib/openblas/interface/trmv.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 "QTRMV " |
48 | | #elif defined(DOUBLE) |
49 | 0 | #define ERROR_NAME "DTRMV " |
50 | | #else |
51 | 0 | #define ERROR_NAME "STRMV " |
52 | | #endif |
53 | | |
54 | | static int (*trmv[])(BLASLONG, FLOAT *, BLASLONG, FLOAT *, BLASLONG, FLOAT *) = { |
55 | | #ifdef XDOUBLE |
56 | | qtrmv_NUU, qtrmv_NUN, qtrmv_NLU, qtrmv_NLN, |
57 | | qtrmv_TUU, qtrmv_TUN, qtrmv_TLU, qtrmv_TLN, |
58 | | #elif defined(DOUBLE) |
59 | | dtrmv_NUU, dtrmv_NUN, dtrmv_NLU, dtrmv_NLN, |
60 | | dtrmv_TUU, dtrmv_TUN, dtrmv_TLU, dtrmv_TLN, |
61 | | #else |
62 | | strmv_NUU, strmv_NUN, strmv_NLU, strmv_NLN, |
63 | | strmv_TUU, strmv_TUN, strmv_TLU, strmv_TLN, |
64 | | #endif |
65 | | }; |
66 | | |
67 | | #ifdef SMP |
68 | | static int (*trmv_thread[])(BLASLONG, FLOAT *, BLASLONG, FLOAT *, BLASLONG, FLOAT *, int) = { |
69 | | #ifdef XDOUBLE |
70 | | qtrmv_thread_NUU, qtrmv_thread_NUN, qtrmv_thread_NLU, qtrmv_thread_NLN, |
71 | | qtrmv_thread_TUU, qtrmv_thread_TUN, qtrmv_thread_TLU, qtrmv_thread_TLN, |
72 | | #elif defined(DOUBLE) |
73 | | dtrmv_thread_NUU, dtrmv_thread_NUN, dtrmv_thread_NLU, dtrmv_thread_NLN, |
74 | | dtrmv_thread_TUU, dtrmv_thread_TUN, dtrmv_thread_TLU, dtrmv_thread_TLN, |
75 | | #else |
76 | | strmv_thread_NUU, strmv_thread_NUN, strmv_thread_NLU, strmv_thread_NLN, |
77 | | strmv_thread_TUU, strmv_thread_TUN, strmv_thread_TLU, strmv_thread_TLN, |
78 | | #endif |
79 | | }; |
80 | | #endif |
81 | | |
82 | | #ifndef CBLAS |
83 | | |
84 | | void NAME(char *UPLO, char *TRANS, char *DIAG, |
85 | 0 | blasint *N, FLOAT *a, blasint *LDA, FLOAT *x, blasint *INCX){ |
86 | |
|
87 | 0 | char uplo_arg = *UPLO; |
88 | 0 | char trans_arg = *TRANS; |
89 | 0 | char diag_arg = *DIAG; |
90 | |
|
91 | 0 | blasint n = *N; |
92 | 0 | blasint lda = *LDA; |
93 | 0 | blasint incx = *INCX; |
94 | |
|
95 | 0 | blasint info; |
96 | 0 | int uplo; |
97 | 0 | int unit; |
98 | 0 | int trans; |
99 | 0 | FLOAT *buffer; |
100 | 0 | #ifdef SMP |
101 | 0 | int nthreads; |
102 | 0 | #endif |
103 | |
|
104 | 0 | PRINT_DEBUG_NAME; |
105 | |
|
106 | 0 | TOUPPER(uplo_arg); |
107 | 0 | TOUPPER(trans_arg); |
108 | 0 | TOUPPER(diag_arg); |
109 | |
|
110 | 0 | trans = -1; |
111 | 0 | unit = -1; |
112 | 0 | uplo = -1; |
113 | |
|
114 | 0 | if (trans_arg == 'N') trans = 0; |
115 | 0 | if (trans_arg == 'T') trans = 1; |
116 | 0 | if (trans_arg == 'R') trans = 0; |
117 | 0 | if (trans_arg == 'C') trans = 1; |
118 | |
|
119 | 0 | if (diag_arg == 'U') unit = 0; |
120 | 0 | if (diag_arg == 'N') unit = 1; |
121 | |
|
122 | 0 | if (uplo_arg == 'U') uplo = 0; |
123 | 0 | if (uplo_arg == 'L') uplo = 1; |
124 | |
|
125 | 0 | info = 0; |
126 | |
|
127 | 0 | if (incx == 0) info = 8; |
128 | 0 | if (lda < MAX(1, n)) info = 6; |
129 | 0 | if (n < 0) info = 4; |
130 | 0 | if (unit < 0) info = 3; |
131 | 0 | if (trans < 0) info = 2; |
132 | 0 | if (uplo < 0) info = 1; |
133 | |
|
134 | 0 | if (info != 0) { |
135 | 0 | BLASFUNC(xerbla)(ERROR_NAME, &info, sizeof(ERROR_NAME)); |
136 | 0 | return; |
137 | 0 | } |
138 | | |
139 | | #else |
140 | | |
141 | | void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, |
142 | | enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag, |
143 | | blasint n, FLOAT *a, blasint lda, FLOAT *x, blasint incx) { |
144 | | |
145 | | int trans, uplo, unit; |
146 | | blasint info; |
147 | | FLOAT *buffer; |
148 | | #ifdef SMP |
149 | | int nthreads; |
150 | | #endif |
151 | | |
152 | | PRINT_DEBUG_CNAME; |
153 | | |
154 | | unit = -1; |
155 | | uplo = -1; |
156 | | trans = -1; |
157 | | info = 0; |
158 | | |
159 | | if (order == CblasColMajor) { |
160 | | if (Uplo == CblasUpper) uplo = 0; |
161 | | if (Uplo == CblasLower) uplo = 1; |
162 | | |
163 | | if (TransA == CblasNoTrans) trans = 0; |
164 | | if (TransA == CblasTrans) trans = 1; |
165 | | if (TransA == CblasConjNoTrans) trans = 0; |
166 | | if (TransA == CblasConjTrans) trans = 1; |
167 | | |
168 | | if (Diag == CblasUnit) unit = 0; |
169 | | if (Diag == CblasNonUnit) unit = 1; |
170 | | |
171 | | info = -1; |
172 | | |
173 | | if (incx == 0) info = 8; |
174 | | if (lda < MAX(1, n)) info = 6; |
175 | | if (n < 0) info = 4; |
176 | | if (unit < 0) info = 3; |
177 | | if (trans < 0) info = 2; |
178 | | if (uplo < 0) info = 1; |
179 | | } |
180 | | |
181 | | if (order == CblasRowMajor) { |
182 | | if (Uplo == CblasUpper) uplo = 1; |
183 | | if (Uplo == CblasLower) uplo = 0; |
184 | | |
185 | | if (TransA == CblasNoTrans) trans = 1; |
186 | | if (TransA == CblasTrans) trans = 0; |
187 | | if (TransA == CblasConjNoTrans) trans = 1; |
188 | | if (TransA == CblasConjTrans) trans = 0; |
189 | | |
190 | | if (Diag == CblasUnit) unit = 0; |
191 | | if (Diag == CblasNonUnit) unit = 1; |
192 | | |
193 | | info = -1; |
194 | | |
195 | | if (incx == 0) info = 8; |
196 | | if (lda < MAX(1, n)) info = 6; |
197 | | if (n < 0) info = 4; |
198 | | if (unit < 0) info = 3; |
199 | | if (trans < 0) info = 2; |
200 | | if (uplo < 0) info = 1; |
201 | | } |
202 | | |
203 | | if (info >= 0) { |
204 | | BLASFUNC(xerbla)(ERROR_NAME, &info, sizeof(ERROR_NAME)); |
205 | | return; |
206 | | } |
207 | | |
208 | | #endif |
209 | | |
210 | 0 | if (n == 0) return; |
211 | | |
212 | 0 | IDEBUG_START; |
213 | |
|
214 | 0 | FUNCTION_PROFILE_START(); |
215 | |
|
216 | 0 | if (incx < 0 ) x -= (n - 1) * incx; |
217 | |
|
218 | 0 | buffer = (FLOAT *)blas_memory_alloc(1); |
219 | |
|
220 | 0 | #ifdef SMP |
221 | 0 | nthreads = num_cpu_avail(2); |
222 | |
|
223 | 0 | if (nthreads == 1) { |
224 | 0 | #endif |
225 | |
|
226 | 0 | (trmv[(trans<<2) | (uplo<<1) | unit])(n, a, lda, x, incx, buffer); |
227 | |
|
228 | 0 | #ifdef SMP |
229 | 0 | } else { |
230 | |
|
231 | 0 | (trmv_thread[(trans<<2) | (uplo<<1) | unit])(n, a, lda, x, incx, buffer, nthreads); |
232 | |
|
233 | 0 | } |
234 | 0 | #endif |
235 | |
|
236 | 0 | blas_memory_free(buffer); |
237 | |
|
238 | 0 | FUNCTION_PROFILE_END(1, n * n / 2 + n, n * n); |
239 | |
|
240 | 0 | IDEBUG_END; |
241 | |
|
242 | 0 | return; |
243 | 0 | } Unexecuted instantiation: strmv_ Unexecuted instantiation: dtrmv_ |