PLASMA  2.4.5
PLASMA - Parallel Linear Algebra for Scalable Multi-core Architectures
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups
core_stsmlq_sytra1.c
Go to the documentation of this file.
1 
17 #include <lapacke.h>
18 #include "common.h"
19 #undef COMPLEX
20 #define REAL
21 
22 /***************************************************************************/
113 #if defined(PLASMA_HAVE_WEAK)
114 #pragma weak CORE_stsmlq_sytra1 = PCORE_stsmlq_sytra1
115 #define CORE_stsmlq_sytra1 PCORE_stsmlq_sytra1
116 #define CORE_stsmlq PCORE_stsmlq
117 int CORE_stsmlq(int side, int trans,
118  int M1, int N1, int M2, int N2, int K, int IB,
119  float *A1, int LDA1,
120  float *A2, int LDA2,
121  float *V, int LDV,
122  float *T, int LDT,
123  float *WORK, int LDWORK);
124 #endif
125 int CORE_stsmlq_sytra1( int side, int trans,
126  int m1, int n1, int m2, int n2,
127  int k, int ib,
128  float *A1, int lda1,
129  float *A2, int lda2,
130  float *V, int ldv,
131  float *T, int ldt,
132  float *WORK, int ldwork)
133 {
134  int i, j;
135 
136  if ( (m1 != n1) ) {
137  coreblas_error(3, "Illegal value of M1, N1");
138  return -3;
139  }
140 
141  /* in-place transposition of A1 */
142  for (j = 0; j < n1; j++){
143  A1[j + j*lda1] = (A1[j + j*lda1]);
144 
145  for (i = j+1; i < m1; i++){
146  *WORK = *(A1 + i + j*lda1);
147  *(A1 + i + j*lda1) = (*(A1 + j + i*lda1));
148  *(A1 + j + i*lda1) = (*WORK);
149  }
150  }
151 
152  CORE_stsmlq(side, trans, m1, n1, m2, n2, k, ib,
153  A1, lda1, A2, lda2,
154  V, ldv, T, ldt,
155  WORK, ldwork);
156 
157  /* in-place transposition of A1 */
158  for (j = 0; j < n1; j++){
159  A1[j + j*lda1] = (A1[j + j*lda1]);
160 
161  for (i = j+1; i < m1; i++){
162  *WORK = *(A1 + i + j*lda1);
163  *(A1 + i + j*lda1) = (*(A1 + j + i*lda1));
164  *(A1 + j + i*lda1) = (*WORK);
165  }
166  }
167 
168  return PLASMA_SUCCESS;
169 }
170 
171 /***************************************************************************/
174 void QUARK_CORE_stsmlq_sytra1(Quark *quark, Quark_Task_Flags *task_flags,
175  int side, int trans,
176  int m1, int n1, int m2, int n2, int k, int ib, int nb,
177  float *A1, int lda1,
178  float *A2, int lda2,
179  float *V, int ldv,
180  float *T, int ldt)
181 {
182  int ldwork = side == PlasmaLeft ? ib : nb;
183 
184  QUARK_Insert_Task(quark, CORE_stsmlq_sytra1_quark, task_flags,
185  sizeof(PLASMA_enum), &side, VALUE,
186  sizeof(PLASMA_enum), &trans, VALUE,
187  sizeof(int), &m1, VALUE,
188  sizeof(int), &n1, VALUE,
189  sizeof(int), &m2, VALUE,
190  sizeof(int), &n2, VALUE,
191  sizeof(int), &k, VALUE,
192  sizeof(int), &ib, VALUE,
193  sizeof(float)*nb*nb, A1, INOUT|QUARK_REGION_U|QUARK_REGION_D,
194  sizeof(int), &lda1, VALUE,
195  sizeof(float)*nb*nb, A2, INOUT,
196  sizeof(int), &lda2, VALUE,
197  sizeof(float)*nb*nb, V, INPUT,
198  sizeof(int), &ldv, VALUE,
199  sizeof(float)*ib*nb, T, INPUT,
200  sizeof(int), &ldt, VALUE,
201  sizeof(float)*ib*nb, NULL, SCRATCH,
202  sizeof(int), &ldwork, VALUE,
203  0);
204 }
205 
206 /***************************************************************************/
214 #if defined(PLASMA_HAVE_WEAK)
215 #pragma weak CORE_stsmlq_sytra1_quark = PCORE_stsmlq_sytra1_quark
216 #define CORE_stsmlq_sytra1_quark PCORE_stsmlq_sytra1_quark
217 #endif
218 void CORE_stsmlq_sytra1_quark(Quark *quark)
219 {
220  int side;
221  int trans;
222  int m1;
223  int n1;
224  int m2;
225  int n2;
226  int k;
227  int ib;
228  float *A1;
229  int lda1;
230  float *A2;
231  int lda2;
232  float *V;
233  int ldv;
234  float *T;
235  int ldt;
236  float *WORK;
237  int ldwork;
238 
239  quark_unpack_args_18(quark, side, trans, m1, n1, m2, n2, k, ib,
240  A1, lda1, A2, lda2, V, ldv, T, ldt, WORK, ldwork);
241  CORE_stsmlq_sytra1(side, trans, m1, n1, m2, n2, k, ib,
242  A1, lda1, A2, lda2, V, ldv, T, ldt, WORK, ldwork);
243 }
244