root / src / pfact / HPL_dlocswpN.c @ 9
Historique | Voir | Annoter | Télécharger (17,35 ko)
1 |
/*
|
---|---|
2 |
* -- High Performance Computing Linpack Benchmark (HPL)
|
3 |
* HPL - 2.0 - September 10, 2008
|
4 |
* Antoine P. Petitet
|
5 |
* University of Tennessee, Knoxville
|
6 |
* Innovative Computing Laboratory
|
7 |
* (C) Copyright 2000-2008 All Rights Reserved
|
8 |
*
|
9 |
* -- Copyright notice and Licensing terms:
|
10 |
*
|
11 |
* Redistribution and use in source and binary forms, with or without
|
12 |
* modification, are permitted provided that the following conditions
|
13 |
* are met:
|
14 |
*
|
15 |
* 1. Redistributions of source code must retain the above copyright
|
16 |
* notice, this list of conditions and the following disclaimer.
|
17 |
*
|
18 |
* 2. Redistributions in binary form must reproduce the above copyright
|
19 |
* notice, this list of conditions, and the following disclaimer in the
|
20 |
* documentation and/or other materials provided with the distribution.
|
21 |
*
|
22 |
* 3. All advertising materials mentioning features or use of this
|
23 |
* software must display the following acknowledgement:
|
24 |
* This product includes software developed at the University of
|
25 |
* Tennessee, Knoxville, Innovative Computing Laboratory.
|
26 |
*
|
27 |
* 4. The name of the University, the name of the Laboratory, or the
|
28 |
* names of its contributors may not be used to endorse or promote
|
29 |
* products derived from this software without specific written
|
30 |
* permission.
|
31 |
*
|
32 |
* -- Disclaimer:
|
33 |
*
|
34 |
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
|
35 |
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
|
36 |
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
|
37 |
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE UNIVERSITY
|
38 |
* OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
|
39 |
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
|
40 |
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
|
41 |
* DATA OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
|
42 |
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
43 |
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
44 |
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
45 |
* ---------------------------------------------------------------------
|
46 |
*/
|
47 |
/*
|
48 |
* Include files
|
49 |
*/
|
50 |
#include "hpl.h" |
51 |
/*
|
52 |
* Define default value for unrolling factor
|
53 |
*/
|
54 |
#ifndef HPL_LOCSWP_DEPTH
|
55 |
#define HPL_LOCSWP_DEPTH 32 |
56 |
#define HPL_LOCSWP_LOG2_DEPTH 5 |
57 |
#endif
|
58 |
|
59 |
#ifdef STDC_HEADERS
|
60 |
void HPL_dlocswpN
|
61 |
( |
62 |
HPL_T_panel * PANEL, |
63 |
const int II, |
64 |
const int JJ, |
65 |
double * WORK
|
66 |
) |
67 |
#else
|
68 |
void HPL_dlocswpN
|
69 |
( PANEL, II, JJ, WORK ) |
70 |
HPL_T_panel * PANEL; |
71 |
const int II; |
72 |
const int JJ; |
73 |
double * WORK;
|
74 |
#endif
|
75 |
{ |
76 |
/*
|
77 |
* Purpose
|
78 |
* =======
|
79 |
*
|
80 |
* HPL_dlocswpN performs the local swapping operations within a panel.
|
81 |
* The lower triangular N0-by-N0 upper block of the panel is stored in
|
82 |
* no-transpose form (i.e. just like the input matrix itself).
|
83 |
*
|
84 |
* Arguments
|
85 |
* =========
|
86 |
*
|
87 |
* PANEL (local input/output) HPL_T_panel *
|
88 |
* On entry, PANEL points to the data structure containing the
|
89 |
* panel information.
|
90 |
*
|
91 |
* II (local input) const int
|
92 |
* On entry, II specifies the row offset where the column to be
|
93 |
* operated on starts with respect to the panel.
|
94 |
*
|
95 |
* JJ (local input) const int
|
96 |
* On entry, JJ specifies the column offset where the column to
|
97 |
* be operated on starts with respect to the panel.
|
98 |
*
|
99 |
* WORK (local workspace) double *
|
100 |
* On entry, WORK is a workarray of size at least 2 * (4+2*N0).
|
101 |
* WORK[0] contains the local maximum absolute value scalar,
|
102 |
* WORK[1] contains the corresponding local row index, WORK[2]
|
103 |
* contains the corresponding global row index, and WORK[3] is
|
104 |
* the coordinate of process owning this max. The N0 length max
|
105 |
* row is stored in WORK[4:4+N0-1]; Note that this is also the
|
106 |
* JJth row (or column) of L1. The remaining part of this array
|
107 |
* is used as workspace.
|
108 |
*
|
109 |
* ---------------------------------------------------------------------
|
110 |
*/
|
111 |
/*
|
112 |
* .. Local Variables ..
|
113 |
*/
|
114 |
double gmax;
|
115 |
double * A1, * A2, * L, * Wr0, * Wmx;
|
116 |
int ilindx, lda, myrow, n0, nr, nu;
|
117 |
register int i; |
118 |
/* ..
|
119 |
* .. Executable Statements ..
|
120 |
*/
|
121 |
myrow = PANEL->grid->myrow; n0 = PANEL->jb; lda = PANEL->lda; |
122 |
|
123 |
Wr0 = ( Wmx = WORK + 4 ) + n0; Wmx[JJ] = gmax = WORK[0]; |
124 |
nu = (int)( ( (unsigned int)(n0) >> HPL_LOCSWP_LOG2_DEPTH ) |
125 |
<< HPL_LOCSWP_LOG2_DEPTH ); |
126 |
nr = n0 - nu; |
127 |
/*
|
128 |
* Replicated swap and copy of the current (new) row of A into L1
|
129 |
*/
|
130 |
L = Mptr( PANEL->L1, JJ, 0, n0 );
|
131 |
/*
|
132 |
* If the pivot is non-zero ...
|
133 |
*/
|
134 |
if( gmax != HPL_rzero )
|
135 |
{ |
136 |
/*
|
137 |
* and if I own the current row of A ...
|
138 |
*/
|
139 |
if( myrow == PANEL->prow )
|
140 |
{ |
141 |
/*
|
142 |
* and if I also own the row to be swapped with the current row of A ...
|
143 |
*/
|
144 |
if( myrow == (int)(WORK[3]) ) |
145 |
{ |
146 |
/*
|
147 |
* and if the current row of A is not to swapped with itself ...
|
148 |
*/
|
149 |
if( ( ilindx = (int)(WORK[1]) ) != 0 ) |
150 |
{ |
151 |
/*
|
152 |
* then copy the max row into L1 and locally swap the 2 rows of A.
|
153 |
*/
|
154 |
A1 = Mptr( PANEL->A, II, 0, lda );
|
155 |
A2 = Mptr( A1, ilindx, 0, lda );
|
156 |
|
157 |
for( i = 0; i < nu; i += HPL_LOCSWP_DEPTH, |
158 |
Wmx += HPL_LOCSWP_DEPTH, Wr0 += HPL_LOCSWP_DEPTH ) |
159 |
{ |
160 |
*L=*A1=Wmx[ 0]; *A2=Wr0[ 0]; L+=n0; A1+=lda; A2+=lda; |
161 |
#if ( HPL_LOCSWP_DEPTH > 1 ) |
162 |
*L=*A1=Wmx[ 1]; *A2=Wr0[ 1]; L+=n0; A1+=lda; A2+=lda; |
163 |
#endif
|
164 |
#if ( HPL_LOCSWP_DEPTH > 2 ) |
165 |
*L=*A1=Wmx[ 2]; *A2=Wr0[ 2]; L+=n0; A1+=lda; A2+=lda; |
166 |
*L=*A1=Wmx[ 3]; *A2=Wr0[ 3]; L+=n0; A1+=lda; A2+=lda; |
167 |
#endif
|
168 |
#if ( HPL_LOCSWP_DEPTH > 4 ) |
169 |
*L=*A1=Wmx[ 4]; *A2=Wr0[ 4]; L+=n0; A1+=lda; A2+=lda; |
170 |
*L=*A1=Wmx[ 5]; *A2=Wr0[ 5]; L+=n0; A1+=lda; A2+=lda; |
171 |
*L=*A1=Wmx[ 6]; *A2=Wr0[ 6]; L+=n0; A1+=lda; A2+=lda; |
172 |
*L=*A1=Wmx[ 7]; *A2=Wr0[ 7]; L+=n0; A1+=lda; A2+=lda; |
173 |
#endif
|
174 |
#if ( HPL_LOCSWP_DEPTH > 8 ) |
175 |
*L=*A1=Wmx[ 8]; *A2=Wr0[ 8]; L+=n0; A1+=lda; A2+=lda; |
176 |
*L=*A1=Wmx[ 9]; *A2=Wr0[ 9]; L+=n0; A1+=lda; A2+=lda; |
177 |
*L=*A1=Wmx[10]; *A2=Wr0[10]; L+=n0; A1+=lda; A2+=lda; |
178 |
*L=*A1=Wmx[11]; *A2=Wr0[11]; L+=n0; A1+=lda; A2+=lda; |
179 |
*L=*A1=Wmx[12]; *A2=Wr0[12]; L+=n0; A1+=lda; A2+=lda; |
180 |
*L=*A1=Wmx[13]; *A2=Wr0[13]; L+=n0; A1+=lda; A2+=lda; |
181 |
*L=*A1=Wmx[14]; *A2=Wr0[14]; L+=n0; A1+=lda; A2+=lda; |
182 |
*L=*A1=Wmx[15]; *A2=Wr0[15]; L+=n0; A1+=lda; A2+=lda; |
183 |
#endif
|
184 |
#if ( HPL_LOCSWP_DEPTH > 16 ) |
185 |
*L=*A1=Wmx[16]; *A2=Wr0[16]; L+=n0; A1+=lda; A2+=lda; |
186 |
*L=*A1=Wmx[17]; *A2=Wr0[17]; L+=n0; A1+=lda; A2+=lda; |
187 |
*L=*A1=Wmx[18]; *A2=Wr0[18]; L+=n0; A1+=lda; A2+=lda; |
188 |
*L=*A1=Wmx[19]; *A2=Wr0[19]; L+=n0; A1+=lda; A2+=lda; |
189 |
*L=*A1=Wmx[20]; *A2=Wr0[20]; L+=n0; A1+=lda; A2+=lda; |
190 |
*L=*A1=Wmx[21]; *A2=Wr0[21]; L+=n0; A1+=lda; A2+=lda; |
191 |
*L=*A1=Wmx[22]; *A2=Wr0[22]; L+=n0; A1+=lda; A2+=lda; |
192 |
*L=*A1=Wmx[23]; *A2=Wr0[23]; L+=n0; A1+=lda; A2+=lda; |
193 |
*L=*A1=Wmx[24]; *A2=Wr0[24]; L+=n0; A1+=lda; A2+=lda; |
194 |
*L=*A1=Wmx[25]; *A2=Wr0[25]; L+=n0; A1+=lda; A2+=lda; |
195 |
*L=*A1=Wmx[26]; *A2=Wr0[26]; L+=n0; A1+=lda; A2+=lda; |
196 |
*L=*A1=Wmx[27]; *A2=Wr0[27]; L+=n0; A1+=lda; A2+=lda; |
197 |
*L=*A1=Wmx[28]; *A2=Wr0[28]; L+=n0; A1+=lda; A2+=lda; |
198 |
*L=*A1=Wmx[29]; *A2=Wr0[29]; L+=n0; A1+=lda; A2+=lda; |
199 |
*L=*A1=Wmx[30]; *A2=Wr0[30]; L+=n0; A1+=lda; A2+=lda; |
200 |
*L=*A1=Wmx[31]; *A2=Wr0[31]; L+=n0; A1+=lda; A2+=lda; |
201 |
#endif
|
202 |
} |
203 |
for( i = 0; i < nr; i++, L += n0, A1 += lda, A2 += lda ) |
204 |
{ *L = *A1 = Wmx[i]; *A2 = Wr0[i]; } |
205 |
} |
206 |
else
|
207 |
{ |
208 |
/*
|
209 |
* otherwise the current row of A is swapped with itself, so just copy
|
210 |
* the current of A into L1.
|
211 |
*/
|
212 |
*Mptr( PANEL->A, II, JJ, lda ) = gmax; |
213 |
|
214 |
for( i = 0; i < nu; i += HPL_LOCSWP_DEPTH, |
215 |
Wmx += HPL_LOCSWP_DEPTH, Wr0 += HPL_LOCSWP_DEPTH ) |
216 |
{ |
217 |
*L = Wmx[ 0]; L+=n0;
|
218 |
#if ( HPL_LOCSWP_DEPTH > 1 ) |
219 |
*L = Wmx[ 1]; L+=n0;
|
220 |
#endif
|
221 |
#if ( HPL_LOCSWP_DEPTH > 2 ) |
222 |
*L = Wmx[ 2]; L+=n0; *L = Wmx[ 3]; L+=n0; |
223 |
#endif
|
224 |
#if ( HPL_LOCSWP_DEPTH > 4 ) |
225 |
*L = Wmx[ 4]; L+=n0; *L = Wmx[ 5]; L+=n0; |
226 |
*L = Wmx[ 6]; L+=n0; *L = Wmx[ 7]; L+=n0; |
227 |
#endif
|
228 |
#if ( HPL_LOCSWP_DEPTH > 8 ) |
229 |
*L = Wmx[ 8]; L+=n0; *L = Wmx[ 9]; L+=n0; |
230 |
*L = Wmx[10]; L+=n0; *L = Wmx[11]; L+=n0; |
231 |
*L = Wmx[12]; L+=n0; *L = Wmx[13]; L+=n0; |
232 |
*L = Wmx[14]; L+=n0; *L = Wmx[15]; L+=n0; |
233 |
#endif
|
234 |
#if ( HPL_LOCSWP_DEPTH > 16 ) |
235 |
*L = Wmx[16]; L+=n0; *L = Wmx[17]; L+=n0; |
236 |
*L = Wmx[18]; L+=n0; *L = Wmx[19]; L+=n0; |
237 |
*L = Wmx[20]; L+=n0; *L = Wmx[21]; L+=n0; |
238 |
*L = Wmx[22]; L+=n0; *L = Wmx[23]; L+=n0; |
239 |
*L = Wmx[24]; L+=n0; *L = Wmx[25]; L+=n0; |
240 |
*L = Wmx[26]; L+=n0; *L = Wmx[27]; L+=n0; |
241 |
*L = Wmx[28]; L+=n0; *L = Wmx[29]; L+=n0; |
242 |
*L = Wmx[30]; L+=n0; *L = Wmx[31]; L+=n0; |
243 |
#endif
|
244 |
} |
245 |
for( i = 0; i < nr; i++, L += n0 ) { *L = Wmx[i]; } |
246 |
} |
247 |
} |
248 |
else
|
249 |
{ |
250 |
/*
|
251 |
* otherwise, the row to be swapped with the current row of A is in Wmx,
|
252 |
* so copy Wmx into L1 and A.
|
253 |
*/
|
254 |
A1 = Mptr( PANEL->A, II, 0, lda );
|
255 |
|
256 |
for( i = 0; i < nu; i += HPL_LOCSWP_DEPTH, |
257 |
Wmx += HPL_LOCSWP_DEPTH ) |
258 |
{ |
259 |
*L = *A1 = Wmx[ 0]; L += n0; A1 += lda;
|
260 |
#if ( HPL_LOCSWP_DEPTH > 1 ) |
261 |
*L = *A1 = Wmx[ 1]; L += n0; A1 += lda;
|
262 |
#endif
|
263 |
#if ( HPL_LOCSWP_DEPTH > 2 ) |
264 |
*L = *A1 = Wmx[ 2]; L += n0; A1 += lda;
|
265 |
*L = *A1 = Wmx[ 3]; L += n0; A1 += lda;
|
266 |
#endif
|
267 |
#if ( HPL_LOCSWP_DEPTH > 4 ) |
268 |
*L = *A1 = Wmx[ 4]; L += n0; A1 += lda;
|
269 |
*L = *A1 = Wmx[ 5]; L += n0; A1 += lda;
|
270 |
*L = *A1 = Wmx[ 6]; L += n0; A1 += lda;
|
271 |
*L = *A1 = Wmx[ 7]; L += n0; A1 += lda;
|
272 |
#endif
|
273 |
#if ( HPL_LOCSWP_DEPTH > 8 ) |
274 |
*L = *A1 = Wmx[ 8]; L += n0; A1 += lda;
|
275 |
*L = *A1 = Wmx[ 9]; L += n0; A1 += lda;
|
276 |
*L = *A1 = Wmx[10]; L += n0; A1 += lda;
|
277 |
*L = *A1 = Wmx[11]; L += n0; A1 += lda;
|
278 |
*L = *A1 = Wmx[12]; L += n0; A1 += lda;
|
279 |
*L = *A1 = Wmx[13]; L += n0; A1 += lda;
|
280 |
*L = *A1 = Wmx[14]; L += n0; A1 += lda;
|
281 |
*L = *A1 = Wmx[15]; L += n0; A1 += lda;
|
282 |
#endif
|
283 |
#if ( HPL_LOCSWP_DEPTH > 16 ) |
284 |
*L = *A1 = Wmx[16]; L += n0; A1 += lda;
|
285 |
*L = *A1 = Wmx[17]; L += n0; A1 += lda;
|
286 |
*L = *A1 = Wmx[18]; L += n0; A1 += lda;
|
287 |
*L = *A1 = Wmx[19]; L += n0; A1 += lda;
|
288 |
*L = *A1 = Wmx[20]; L += n0; A1 += lda;
|
289 |
*L = *A1 = Wmx[21]; L += n0; A1 += lda;
|
290 |
*L = *A1 = Wmx[22]; L += n0; A1 += lda;
|
291 |
*L = *A1 = Wmx[23]; L += n0; A1 += lda;
|
292 |
*L = *A1 = Wmx[24]; L += n0; A1 += lda;
|
293 |
*L = *A1 = Wmx[25]; L += n0; A1 += lda;
|
294 |
*L = *A1 = Wmx[26]; L += n0; A1 += lda;
|
295 |
*L = *A1 = Wmx[27]; L += n0; A1 += lda;
|
296 |
*L = *A1 = Wmx[28]; L += n0; A1 += lda;
|
297 |
*L = *A1 = Wmx[29]; L += n0; A1 += lda;
|
298 |
*L = *A1 = Wmx[30]; L += n0; A1 += lda;
|
299 |
*L = *A1 = Wmx[31]; L += n0; A1 += lda;
|
300 |
#endif
|
301 |
} |
302 |
|
303 |
for( i = 0; i < nr; i++, L += n0, A1 += lda ) |
304 |
{ *L = *A1 = Wmx[i]; } |
305 |
} |
306 |
} |
307 |
else
|
308 |
{ |
309 |
/*
|
310 |
* otherwise I do not own the current row of A, so copy the max row Wmx
|
311 |
* into L1.
|
312 |
*/
|
313 |
for( i = 0; i < nu; i += HPL_LOCSWP_DEPTH, |
314 |
Wmx += HPL_LOCSWP_DEPTH ) |
315 |
{ |
316 |
*L = Wmx[ 0]; L+=n0;
|
317 |
#if ( HPL_LOCSWP_DEPTH > 1 ) |
318 |
*L = Wmx[ 1]; L+=n0;
|
319 |
#endif
|
320 |
#if ( HPL_LOCSWP_DEPTH > 2 ) |
321 |
*L = Wmx[ 2]; L+=n0; *L = Wmx[ 3]; L+=n0; |
322 |
#endif
|
323 |
#if ( HPL_LOCSWP_DEPTH > 4 ) |
324 |
*L = Wmx[ 4]; L+=n0; *L = Wmx[ 5]; L+=n0; |
325 |
*L = Wmx[ 6]; L+=n0; *L = Wmx[ 7]; L+=n0; |
326 |
#endif
|
327 |
#if ( HPL_LOCSWP_DEPTH > 8 ) |
328 |
*L = Wmx[ 8]; L+=n0; *L = Wmx[ 9]; L+=n0; |
329 |
*L = Wmx[10]; L+=n0; *L = Wmx[11]; L+=n0; |
330 |
*L = Wmx[12]; L+=n0; *L = Wmx[13]; L+=n0; |
331 |
*L = Wmx[14]; L+=n0; *L = Wmx[15]; L+=n0; |
332 |
#endif
|
333 |
#if ( HPL_LOCSWP_DEPTH > 16 ) |
334 |
*L = Wmx[16]; L+=n0; *L = Wmx[17]; L+=n0; |
335 |
*L = Wmx[18]; L+=n0; *L = Wmx[19]; L+=n0; |
336 |
*L = Wmx[20]; L+=n0; *L = Wmx[21]; L+=n0; |
337 |
*L = Wmx[22]; L+=n0; *L = Wmx[23]; L+=n0; |
338 |
*L = Wmx[24]; L+=n0; *L = Wmx[25]; L+=n0; |
339 |
*L = Wmx[26]; L+=n0; *L = Wmx[27]; L+=n0; |
340 |
*L = Wmx[28]; L+=n0; *L = Wmx[29]; L+=n0; |
341 |
*L = Wmx[30]; L+=n0; *L = Wmx[31]; L+=n0; |
342 |
#endif
|
343 |
} |
344 |
for( i = 0; i < nr; i++, L += n0 ) { *L = Wmx[i]; } |
345 |
/*
|
346 |
* and if I own the max row, overwrite it with the current row Wr0.
|
347 |
*/
|
348 |
if( myrow == (int)(WORK[3]) ) |
349 |
{ |
350 |
A2 = Mptr( PANEL->A, II + (size_t)(WORK[1]), 0, lda ); |
351 |
|
352 |
for( i = 0; i < nu; i += HPL_LOCSWP_DEPTH, |
353 |
Wr0 += HPL_LOCSWP_DEPTH ) |
354 |
{ |
355 |
*A2 = Wr0[ 0]; A2+=lda;
|
356 |
#if ( HPL_LOCSWP_DEPTH > 1 ) |
357 |
*A2 = Wr0[ 1]; A2+=lda;
|
358 |
#endif
|
359 |
#if ( HPL_LOCSWP_DEPTH > 2 ) |
360 |
*A2 = Wr0[ 2]; A2+=lda; *A2 = Wr0[ 3]; A2+=lda; |
361 |
#endif
|
362 |
#if ( HPL_LOCSWP_DEPTH > 4 ) |
363 |
*A2 = Wr0[ 4]; A2+=lda; *A2 = Wr0[ 5]; A2+=lda; |
364 |
*A2 = Wr0[ 6]; A2+=lda; *A2 = Wr0[ 7]; A2+=lda; |
365 |
#endif
|
366 |
#if ( HPL_LOCSWP_DEPTH > 8 ) |
367 |
*A2 = Wr0[ 8]; A2+=lda; *A2 = Wr0[ 9]; A2+=lda; |
368 |
*A2 = Wr0[10]; A2+=lda; *A2 = Wr0[11]; A2+=lda; |
369 |
*A2 = Wr0[12]; A2+=lda; *A2 = Wr0[13]; A2+=lda; |
370 |
*A2 = Wr0[14]; A2+=lda; *A2 = Wr0[15]; A2+=lda; |
371 |
#endif
|
372 |
#if ( HPL_LOCSWP_DEPTH > 16 ) |
373 |
*A2 = Wr0[16]; A2+=lda; *A2 = Wr0[17]; A2+=lda; |
374 |
*A2 = Wr0[18]; A2+=lda; *A2 = Wr0[19]; A2+=lda; |
375 |
*A2 = Wr0[20]; A2+=lda; *A2 = Wr0[21]; A2+=lda; |
376 |
*A2 = Wr0[22]; A2+=lda; *A2 = Wr0[23]; A2+=lda; |
377 |
*A2 = Wr0[24]; A2+=lda; *A2 = Wr0[25]; A2+=lda; |
378 |
*A2 = Wr0[26]; A2+=lda; *A2 = Wr0[27]; A2+=lda; |
379 |
*A2 = Wr0[28]; A2+=lda; *A2 = Wr0[29]; A2+=lda; |
380 |
*A2 = Wr0[30]; A2+=lda; *A2 = Wr0[31]; A2+=lda; |
381 |
#endif
|
382 |
} |
383 |
|
384 |
for( i = 0; i < nr; i++, A2 += lda ) { *A2 = Wr0[i]; } |
385 |
} |
386 |
} |
387 |
} |
388 |
else
|
389 |
{ |
390 |
/*
|
391 |
* Otherwise the max element in the current column is zero, simply copy
|
392 |
* the current row Wr0 into L1. The matrix is singular.
|
393 |
*/
|
394 |
for( i = 0; i < nu; i += HPL_LOCSWP_DEPTH, |
395 |
Wr0 += HPL_LOCSWP_DEPTH ) |
396 |
{ |
397 |
*L = Wr0[ 0]; L+=n0;
|
398 |
#if ( HPL_LOCSWP_DEPTH > 1 ) |
399 |
*L = Wr0[ 1]; L+=n0;
|
400 |
#endif
|
401 |
#if ( HPL_LOCSWP_DEPTH > 2 ) |
402 |
*L = Wr0[ 2]; L+=n0; *L = Wr0[ 3]; L+=n0; |
403 |
#endif
|
404 |
#if ( HPL_LOCSWP_DEPTH > 4 ) |
405 |
*L = Wr0[ 4]; L+=n0; *L = Wr0[ 5]; L+=n0; |
406 |
*L = Wr0[ 6]; L+=n0; *L = Wr0[ 7]; L+=n0; |
407 |
#endif
|
408 |
#if ( HPL_LOCSWP_DEPTH > 8 ) |
409 |
*L = Wr0[ 8]; L+=n0; *L = Wr0[ 9]; L+=n0; |
410 |
*L = Wr0[10]; L+=n0; *L = Wr0[11]; L+=n0; |
411 |
*L = Wr0[12]; L+=n0; *L = Wr0[13]; L+=n0; |
412 |
*L = Wr0[14]; L+=n0; *L = Wr0[15]; L+=n0; |
413 |
#endif
|
414 |
#if ( HPL_LOCSWP_DEPTH > 16 ) |
415 |
*L = Wr0[16]; L+=n0; *L = Wr0[17]; L+=n0; |
416 |
*L = Wr0[18]; L+=n0; *L = Wr0[19]; L+=n0; |
417 |
*L = Wr0[20]; L+=n0; *L = Wr0[21]; L+=n0; |
418 |
*L = Wr0[22]; L+=n0; *L = Wr0[23]; L+=n0; |
419 |
*L = Wr0[24]; L+=n0; *L = Wr0[25]; L+=n0; |
420 |
*L = Wr0[26]; L+=n0; *L = Wr0[27]; L+=n0; |
421 |
*L = Wr0[28]; L+=n0; *L = Wr0[29]; L+=n0; |
422 |
*L = Wr0[30]; L+=n0; *L = Wr0[31]; L+=n0; |
423 |
#endif
|
424 |
} |
425 |
|
426 |
for( i = 0; i < nr; i++, L += n0 ) { *L = Wr0[i]; } |
427 |
/*
|
428 |
* set INFO.
|
429 |
*/
|
430 |
if( *(PANEL->DINFO) == 0.0 ) |
431 |
*(PANEL->DINFO) = (double)(PANEL->ia + JJ + 1); |
432 |
} |
433 |
/*
|
434 |
* End of HPL_dlocswpN
|
435 |
*/
|
436 |
} |