2 * Copyright 1993-2014 NVIDIA Corporation. All rights reserved.
6 * This source code and/or documentation ("Licensed Deliverables") are
7 * subject to NVIDIA intellectual property rights under U.S. and
8 * international Copyright laws.
10 * These Licensed Deliverables contained herein is PROPRIETARY and
11 * CONFIDENTIAL to NVIDIA and is being provided under the terms and
12 * conditions of a form of NVIDIA software license agreement by and
13 * between NVIDIA and Licensee ("License Agreement") or electronically
14 * accepted by Licensee. Notwithstanding any terms or conditions to
15 * the contrary in the License Agreement, reproduction or disclosure
16 * of the Licensed Deliverables to any third party without the express
17 * written consent of NVIDIA is prohibited.
19 * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
20 * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
21 * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE. IT IS
22 * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
23 * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
24 * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
25 * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
26 * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
27 * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
28 * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
29 * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
30 * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
31 * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
32 * OF THESE LICENSED DELIVERABLES.
34 * U.S. Government End Users. These Licensed Deliverables are a
35 * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
36 * 1995), consisting of "commercial computer software" and "commercial
37 * computer software documentation" as such terms are used in 48
38 * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
39 * only as a commercial end item. Consistent with 48 C.F.R.12.212 and
40 * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
41 * U.S. Government End Users acquire the Licensed Deliverables with
42 * only those rights set forth herein.
44 * Any use of the Licensed Deliverables in individual and commercial
45 * software must include, in the user documentation and internal
46 * comments to the code, the above Disclaimer and U.S. Government End
50 #if !defined(NVBLAS_H_)
53 #include "driver_types.h"
54 #include "cuComplex.h" /* import complex data type */
56 #if defined(__cplusplus)
61 void sgemm_ (const char *transa, const char *transb, const int *m, const int *n, const int *k,
62 const float *alpha, const float *a, const int *lda, const float *b, const int *ldb,
63 const float *beta, float *c, const int *ldc);
65 void dgemm_ ( const char *transa, const char *transb, const int *m, const int *n, const int *k,
66 const double *alpha, const double *a, const int *lda, const double *b, const int *ldb,
67 const double *beta, double *c, const int *ldc);
69 void cgemm_ (const char *transa, const char *transb, const int *m, const int *n, const int *k,
70 const cuComplex *alpha, const cuComplex *a, const int *lda,
71 const cuComplex *b, const int *ldb, const cuComplex *beta,
72 cuComplex *c, const int *ldc);
74 void zgemm_ (const char *transa, const char *transb, const int *m, const int *n, const int *k,
75 const cuDoubleComplex *alpha, const cuDoubleComplex *a, const int *lda,
76 const cuDoubleComplex *b, const int *ldb, const cuDoubleComplex *beta,
77 cuDoubleComplex *c, const int *ldc);
79 void sgemm (const char *transa, const char *transb, const int *m, const int *n, const int *k,
80 const float *alpha, const float *a, const int *lda, const float *b, const int *ldb,
81 const float *beta, float *c, const int *ldc);
83 void dgemm ( const char *transa, const char *transb, const int *m, const int *n, const int *k,
84 const double *alpha, const double *a, const int *lda, const double *b, const int *ldb,
85 const double *beta, double *c, const int *ldc);
87 void cgemm (const char *transa, const char *transb, const int *m, const int *n, const int *k,
88 const cuComplex *alpha, const cuComplex *a, const int *lda,
89 const cuComplex *b, const int *ldb, const cuComplex *beta,
90 cuComplex *c, const int *ldc);
92 void zgemm (const char *transa, const char *transb, const int *m, const int *n, const int *k,
93 const cuDoubleComplex *alpha, const cuDoubleComplex *a, const int *lda,
94 const cuDoubleComplex *b, const int *ldb, const cuDoubleComplex *beta,
95 cuDoubleComplex *c, const int *ldc);
98 void ssyrk_(const char *uplo, const char *trans, const int *n, const int *k,
99 const float *alpha, const float *a, const int *lda,
100 const float *beta, float *c, const int *ldc);
102 void dsyrk_(const char *uplo, const char *trans, const int *n, const int *k,
103 const double *alpha, const double *a, const int *lda, const double *beta,
104 double *c, const int *ldc);
106 void csyrk_(const char *uplo, const char *trans, const int *n, const int *k,
107 const cuComplex *alpha, const cuComplex *a, const int *lda,
108 const cuComplex *beta, cuComplex *c, const int *ldc);
110 void zsyrk_(const char *uplo, const char *trans, const int *n, const int *k,
111 const cuDoubleComplex *alpha, const cuDoubleComplex *a, const int *lda,
112 const cuDoubleComplex *beta, cuDoubleComplex *c, const int *ldc);
114 void ssyrk (const char *uplo, const char *trans, const int *n, const int *k,
115 const float *alpha, const float *a, const int *lda,
116 const float *beta, float *c, const int *ldc);
118 void dsyrk (const char *uplo, const char *trans, const int *n, const int *k,
119 const double *alpha, const double *a, const int *lda, const double *beta,
120 double *c, const int *ldc);
122 void csyrk (const char *uplo, const char *trans, const int *n, const int *k,
123 const cuComplex *alpha, const cuComplex *a, const int *lda,
124 const cuComplex *beta, cuComplex *c, const int *ldc);
126 void zsyrk (const char *uplo, const char *trans, const int *n, const int *k,
127 const cuDoubleComplex *alpha, const cuDoubleComplex *a, const int *lda,
128 const cuDoubleComplex *beta, cuDoubleComplex *c, const int *ldc);
131 void cherk_(const char *uplo, const char *trans, const int *n, const int *k,
132 const float *alpha, const cuComplex *a, const int *lda, const float *beta,
133 cuComplex *c, const int *ldc);
135 void zherk_(const char *uplo, const char *trans, const int *n, const int *k,
136 const double *alpha, const cuDoubleComplex *a, const int *lda,
137 const double *beta, cuDoubleComplex *c, const int *ldc);
139 void cherk(const char *uplo, const char *trans, const int *n, const int *k,
140 const float *alpha, const cuComplex *a, const int *lda, const float *beta,
141 cuComplex *c, const int *ldc);
143 void zherk(const char *uplo, const char *trans, const int *n, const int *k,
144 const double *alpha, const cuDoubleComplex *a, const int *lda,
145 const double *beta, cuDoubleComplex *c, const int *ldc);
148 void strsm_(const char *side, const char *uplo, const char *transa, const char *diag,
149 const int *m, const int *n, const float *alpha, const float *a, const int *lda,
150 float *b, const int *ldb);
152 void dtrsm_(const char *side, const char *uplo, const char *transa, const char *diag,
153 const int *m, const int *n, const double *alpha, const double *a, const int *lda,
154 double *b, const int *ldb);
156 void ctrsm_(const char *side, const char *uplo, const char *transa, const char *diag,
157 const int *m, const int *n, const cuComplex *alpha,
158 const cuComplex *a, const int *lda, cuComplex *b, const int *ldb);
160 void ztrsm_(const char *side, const char *uplo, const char *transa, const char *diag,
161 const int *m, const int *n, const cuDoubleComplex *alpha,
162 const cuDoubleComplex *a, const int *lda, cuDoubleComplex *b, const int *ldb);
164 void strsm(const char *side, const char *uplo, const char *transa, const char *diag,
165 const int *m, const int *n, const float *alpha, const float *a, const int *lda,
166 float *b, const int *ldb);
168 void dtrsm(const char *side, const char *uplo, const char *transa, const char *diag,
169 const int *m, const int *n, const double *alpha, const double *a, const int *lda,
170 double *b, const int *ldb);
172 void ctrsm(const char *side, const char *uplo, const char *transa, const char *diag,
173 const int *m, const int *n, const cuComplex *alpha,
174 const cuComplex *a, const int *lda, cuComplex *b, const int *ldb);
176 void ztrsm(const char *side, const char *uplo, const char *transa, const char *diag,
177 const int *m, const int *n, const cuDoubleComplex *alpha,
178 const cuDoubleComplex *a, const int *lda, cuDoubleComplex *b, const int *ldb);
181 void ssymm_(const char *side, const char *uplo, const int *m, const int *n,
182 const float *alpha, const float *a, const int *lda, const float *b, const int *ldb,
183 const float *beta, float *c, const int *ldc);
185 void dsymm_(const char *side, const char *uplo, const int *m, const int *n,
186 const double *alpha, const double *a, const int *lda, const double *b, const int *ldb,
187 const double *beta, double *c, const int *ldc);
189 void csymm_(const char *side, const char *uplo, const int *m, const int *n,
190 const cuComplex *alpha, const cuComplex *a, const int *lda,
191 const cuComplex *b, const int *ldb, const cuComplex *beta,
192 cuComplex *c, const int *ldc);
194 void zsymm_(const char *side, const char *uplo, const int *m, const int *n,
195 const cuDoubleComplex *alpha, const cuDoubleComplex *a, const int *lda,
196 const cuDoubleComplex *b, const int *ldb, const cuDoubleComplex *beta,
197 cuDoubleComplex *c, const int *ldc);
199 void ssymm(const char *side, const char *uplo, const int *m, const int *n,
200 const float *alpha, const float *a, const int *lda, const float *b, const int *ldb,
201 const float *beta, float *c, const int *ldc);
203 void dsymm(const char *side, const char *uplo, const int *m, const int *n,
204 const double *alpha, const double *a, const int *lda, const double *b, const int *ldb,
205 const double *beta, double *c, const int *ldc);
207 void csymm(const char *side, const char *uplo, const int *m, const int *n,
208 const cuComplex *alpha, const cuComplex *a, const int *lda,
209 const cuComplex *b, const int *ldb, const cuComplex *beta,
210 cuComplex *c, const int *ldc);
212 void zsymm(const char *side, const char *uplo, const int *m, const int *n,
213 const cuDoubleComplex *alpha, const cuDoubleComplex *a, const int *lda,
214 const cuDoubleComplex *b, const int *ldb, const cuDoubleComplex *beta,
215 cuDoubleComplex *c, const int *ldc);
218 void chemm_(const char *side, const char *uplo, const int *m, const int *n,
219 const cuComplex *alpha, const cuComplex *a, const int *lda,
220 const cuComplex *b, const int *ldb, const cuComplex *beta,
221 cuComplex *c, const int *ldc);
223 void zhemm_(const char *side, const char *uplo,
224 const int *m, const int *n, const cuDoubleComplex *alpha, const cuDoubleComplex *a, const int *lda,
225 const cuDoubleComplex *b, const int *ldb, const cuDoubleComplex *beta, cuDoubleComplex *c, const int *ldc);
227 /* HEMM with no underscore*/
228 void chemm(const char *side, const char *uplo,
229 const int *m, const int *n, const cuComplex *alpha, const cuComplex *a, const int *lda,
230 const cuComplex *b, const int *ldb, const cuComplex *beta, cuComplex *c, const int *ldc);
232 void zhemm(const char *side, const char *uplo,
233 const int *m, const int *n, const cuDoubleComplex *alpha, const cuDoubleComplex *a, const int *lda,
234 const cuDoubleComplex *b, const int *ldb, const cuDoubleComplex *beta, cuDoubleComplex *c, const int *ldc);
237 void ssyr2k_(const char *uplo, const char *trans, const int *n, const int *k,
239 const float *a, const int *lda,
240 const float *b, const int *ldb,
241 const float *beta, float *c, const int *ldc);
243 void dsyr2k_(const char *uplo, const char *trans, const int *n, const int *k,
245 const double *a, const int *lda,
246 const double *b, const int *ldb,
248 double *c, const int *ldc);
250 void csyr2k_(const char *uplo, const char *trans, const int *n, const int *k,
251 const cuComplex *alpha,
252 const cuComplex *a, const int *lda,
253 const cuComplex *b, const int *ldb,
254 const cuComplex *beta, cuComplex *c, const int *ldc);
256 void zsyr2k_(const char *uplo, const char *trans, const int *n, const int *k,
257 const cuDoubleComplex *alpha,
258 const cuDoubleComplex *a, const int *lda,
259 const cuDoubleComplex *b, const int *ldb,
260 const cuDoubleComplex *beta, cuDoubleComplex *c, const int *ldc);
262 /* SYR2K no_underscore*/
263 void ssyr2k(const char *uplo, const char *trans, const int *n, const int *k,
264 const float *alpha, const float *a, const int *lda,
265 const float *b, const int *ldb,
266 const float *beta, float *c, const int *ldc);
268 void dsyr2k(const char *uplo, const char *trans, const int *n, const int *k,
270 const double *a, const int *lda,
271 const double *b, const int *ldb,
273 double *c, const int *ldc);
275 void csyr2k(const char *uplo, const char *trans, const int *n, const int *k,
276 const cuComplex *alpha,
277 const cuComplex *a, const int *lda,
278 const cuComplex *b, const int *ldb,
279 const cuComplex *beta, cuComplex *c, const int *ldc);
281 void zsyr2k(const char *uplo, const char *trans, const int *n, const int *k,
282 const cuDoubleComplex *alpha,
283 const cuDoubleComplex *a, const int *lda,
284 const cuDoubleComplex *b, const int *ldb,
285 const cuDoubleComplex *beta, cuDoubleComplex *c, const int *ldc);
288 void cher2k_(const char *uplo, const char *trans, const int *n, const int *k,
289 const cuComplex *alpha,
290 const cuComplex *a, const int *lda,
291 const cuComplex *b, const int *ldb,
292 const float *beta, cuComplex *c, const int *ldc);
294 void zher2k_(const char *uplo, const char *trans, const int *n, const int *k,
295 const cuDoubleComplex *alpha,
296 const cuDoubleComplex *a, const int *lda,
297 const cuDoubleComplex *b, const int *ldb,
298 const double *beta, cuDoubleComplex *c, const int *ldc);\
300 /* HER2K with no underscore */
301 void cher2k(const char *uplo, const char *trans, const int *n, const int *k,
302 const cuComplex *alpha,
303 const cuComplex *a, const int *lda,
304 const cuComplex *b, const int *ldb,
305 const float *beta, cuComplex *c, const int *ldc);
307 void zher2k(const char *uplo, const char *trans, const int *n, const int *k,
308 const cuDoubleComplex *alpha,
309 const cuDoubleComplex *a, const int *lda,
310 const cuDoubleComplex *b, const int *ldb,
311 const double *beta, cuDoubleComplex *c, const int *ldc);
314 void strmm_(const char *side, const char *uplo, const char *transa, const char *diag,
315 const int *m, const int *n, const float *alpha, const float *a, const int *lda,
316 float *b, const int *ldb);
318 void dtrmm_(const char *side, const char *uplo, const char *transa, const char *diag,
319 const int *m, const int *n, const double *alpha, const double *a, const int *lda,
320 double *b, const int *ldb);
322 void ctrmm_(const char *side, const char *uplo, const char *transa, const char *diag,
323 const int *m, const int *n, const cuComplex *alpha,
324 const cuComplex *a, const int *lda, cuComplex *b, const int *ldb);
326 void ztrmm_(const char *side, const char *uplo, const char *transa, const char *diag,
327 const int *m, const int *n, const cuDoubleComplex *alpha,
328 const cuDoubleComplex *a, const int *lda, cuDoubleComplex *b, const int *ldb);
330 void strmm(const char *side, const char *uplo, const char *transa, const char *diag,
331 const int *m, const int *n, const float *alpha, const float *a, const int *lda,
332 float *b, const int *ldb);
334 void dtrmm(const char *side, const char *uplo, const char *transa, const char *diag,
335 const int *m, const int *n, const double *alpha, const double *a, const int *lda,
336 double *b, const int *ldb);
338 void ctrmm(const char *side, const char *uplo, const char *transa, const char *diag,
339 const int *m, const int *n, const cuComplex *alpha,
340 const cuComplex *a, const int *lda, cuComplex *b, const int *ldb);
342 void ztrmm(const char *side, const char *uplo, const char *transa, const char *diag,
343 const int *m, const int *n, const cuDoubleComplex *alpha,
344 const cuDoubleComplex *a, const int *lda, cuDoubleComplex *b, const int *ldb);
347 #if defined(__cplusplus)
349 #endif /* __cplusplus */
351 #endif /* !defined(NVBLAS_H_) */