?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Use stack buffer for LINCS&bonded gather/scatter
[gromacs.git] / src / gromacs / mdlib / clincs.cpp
blob7be3cef649211d2536bdfd3e541ced058f170be9
1 /*
2 * This file is part of the GROMACS molecular simulation package.
3 *
4 * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
5 * Copyright (c) 2001-2004, The GROMACS development team.
6 * Copyright (c) 2013,2014,2015, by the GROMACS development team, led by
7 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
8 * and including many others, as listed in the AUTHORS file in the
9 * top-level source directory and at http://www.gromacs.org.
10 *
11 * GROMACS is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU Lesser General Public License
13 * as published by the Free Software Foundation; either version 2.1
14 * of the License, or (at your option) any later version.
15 *
16 * GROMACS is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
19 * Lesser General Public License for more details.
20 *
21 * You should have received a copy of the GNU Lesser General Public
22 * License along with GROMACS; if not, see
23 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
24 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
25 *
26 * If you want to redistribute modifications to GROMACS, please
27 * consider that scientific software is very special. Version
28 * control is crucial - bugs must be traceable. We will be happy to
29 * consider code for inclusion in the official distribution, but
30 * derived work must not be called official GROMACS. Details are found
31 * in the README & COPYING files - if they are missing, get the
32 * official version at http://www.gromacs.org.
33 *
34 * To help us fund GROMACS development, we humbly ask that you cite
35 * the research papers on the package. Check out http://www.gromacs.org.
36 */
37 /* This file is completely threadsafe - keep it that way! */
38 #include "gmxpre.h"
39
40 #include "config.h"
41
42 #include <assert.h>
43 #include <math.h>
44 #include <stdlib.h>
45
46 #include <algorithm>
47
48 #include "gromacs/domdec/domdec.h"
49 #include "gromacs/fileio/gmxfio.h"
50 #include "gromacs/legacyheaders/constr.h"
51 #include "gromacs/legacyheaders/copyrite.h"
52 #include "gromacs/legacyheaders/gmx_omp_nthreads.h"
53 #include "gromacs/legacyheaders/mdrun.h"
54 #include "gromacs/legacyheaders/nrnb.h"
55 #include "gromacs/legacyheaders/types/commrec.h"
56 #include "gromacs/math/units.h"
57 #include "gromacs/math/vec.h"
58 #include "gromacs/pbcutil/pbc-simd.h"
59 #include "gromacs/pbcutil/pbc.h"
60 #include "gromacs/simd/simd.h"
61 #include "gromacs/simd/simd_math.h"
62 #include "gromacs/simd/vector_operations.h"
63 #include "gromacs/topology/block.h"
64 #include "gromacs/topology/mtop_util.h"
65 #include "gromacs/utility/bitmask.h"
66 #include "gromacs/utility/fatalerror.h"
67 #include "gromacs/utility/gmxomp.h"
68 #include "gromacs/utility/smalloc.h"
69
70 /* MSVC 2010 produces buggy SIMD PBC code, disable SIMD for MSVC <= 2010 */
71 #if defined GMX_SIMD_HAVE_REAL && !(defined _MSC_VER && _MSC_VER < 1700)
72 #define LINCS_SIMD
73 #endif
74
75
76 #if defined(GMX_SIMD_X86_AVX_256) || defined(GMX_SIMD_X86_AVX2_256)
77
78 // This was originally work-in-progress for augmenting the SIMD module with
79 // masked load/store operations. Instead, that turned into and extended SIMD
80 // interface that supports gather/scatter in all platforms, which will be
81 // part of a future Gromacs version. However, since the code for bonded
82 // interactions and LINCS was already written it would be a pity not to get
83 // the performance gains in Gromacs-5.1. For this reason we have added it as
84 // a bit of a hack in the two files that use it. It will be replaced with the
85 // new generic functionality after version 5.1
86
87 # ifdef GMX_DOUBLE
88 static gmx_inline void gmx_simdcall
89 gmx_hack_simd_transpose4_r(gmx_simd_double_t *row0,
90 gmx_simd_double_t *row1,
91 gmx_simd_double_t *row2,
92 gmx_simd_double_t *row3)
93 {
94 __m256d tmp0, tmp1, tmp2, tmp3;
95
96 tmp0 = _mm256_unpacklo_pd(*row0, *row1);
97 tmp2 = _mm256_unpacklo_pd(*row2, *row3);
98 tmp1 = _mm256_unpackhi_pd(*row0, *row1);
99 tmp3 = _mm256_unpackhi_pd(*row2, *row3);
100 *row0 = _mm256_permute2f128_pd(tmp0, tmp2, 0b00100000);
101 *row1 = _mm256_permute2f128_pd(tmp1, tmp3, 0b00100000);
102 *row2 = _mm256_permute2f128_pd(tmp0, tmp2, 0b00110001);
103 *row3 = _mm256_permute2f128_pd(tmp1, tmp3, 0b00110001);
104 }
105
106 static gmx_inline void gmx_simdcall
107 gmx_hack_simd4_transpose_to_simd_r(const gmx_simd4_double_t *a,
108 gmx_simd_double_t *row0,
109 gmx_simd_double_t *row1,
110 gmx_simd_double_t *row2,
111 gmx_simd_double_t *row3)
112 {
113 *row0 = a[0];
114 *row1 = a[1];
115 *row2 = a[2];
116 *row3 = a[3];
117
118 gmx_hack_simd_transpose4_r(row0, row1, row2, row3);
119 }
120
121 static gmx_inline void gmx_simdcall
122 gmx_hack_simd_transpose_to_simd4_r(gmx_simd_double_t row0,
123 gmx_simd_double_t row1,
124 gmx_simd_double_t row2,
125 gmx_simd_double_t row3,
126 gmx_simd4_double_t *a)
127 {
128 a[0] = row0;
129 a[1] = row1;
130 a[2] = row2;
131 a[3] = row3;
132
133 gmx_hack_simd_transpose4_r(&a[0], &a[1], &a[2], &a[3]);
134 }
135
136
137 # ifdef GMX_SIMD_X86_AVX_GCC_MASKLOAD_BUG
138 # define gmx_hack_simd4_load3_r(mem) _mm256_maskload_pd((mem), _mm_castsi128_ps(_mm256_set_epi32(0, 0, -1, -1, -1, -1, -1, -1)))
139 # define gmx_hack_simd4_store3_r(mem, x) _mm256_maskstore_pd((mem), _mm_castsi128_ps(_mm256_set_epi32(0, 0, -1, -1, -1, -1, -1, -1)), (x))
140 # else
141 # define gmx_hack_simd4_load3_r(mem) _mm256_maskload_pd((mem), _mm256_set_epi32(0, 0, -1, -1, -1, -1, -1, -1))
142 # define gmx_hack_simd4_store3_r(mem, x) _mm256_maskstore_pd((mem), _mm256_set_epi32(0, 0, -1, -1, -1, -1, -1, -1), (x))
143 # endif
144
145 # else /* single instead of double */
146 static gmx_inline void gmx_simdcall
147 gmx_hack_simd_transpose4_r(gmx_simd_float_t *row0,
148 gmx_simd_float_t *row1,
149 gmx_simd_float_t *row2,
150 gmx_simd_float_t *row3)
151 {
152 __m256 tmp0, tmp1, tmp2, tmp3;
153
154 tmp0 = _mm256_unpacklo_ps(*row0, *row1);
155 tmp2 = _mm256_unpacklo_ps(*row2, *row3);
156 tmp1 = _mm256_unpackhi_ps(*row0, *row1);
157 tmp3 = _mm256_unpackhi_ps(*row2, *row3);
158 *row0 = _mm256_shuffle_ps(tmp0, tmp2, 0b0100010001000100);
159 *row1 = _mm256_shuffle_ps(tmp0, tmp2, 0b1110111011101110);
160 *row2 = _mm256_shuffle_ps(tmp1, tmp3, 0b0100010001000100);
161 *row3 = _mm256_shuffle_ps(tmp1, tmp3, 0b1110111011101110);
162 }
163
164 static gmx_inline void gmx_simdcall
165 gmx_hack_simd4_transpose_to_simd_r(const gmx_simd4_float_t *a,
166 gmx_simd_float_t *row0,
167 gmx_simd_float_t *row1,
168 gmx_simd_float_t *row2,
169 gmx_simd_float_t *row3)
170 {
171 *row0 = _mm256_insertf128_ps(_mm256_castps128_ps256(a[0]), a[4], 1);
172 *row1 = _mm256_insertf128_ps(_mm256_castps128_ps256(a[1]), a[5], 1);
173 *row2 = _mm256_insertf128_ps(_mm256_castps128_ps256(a[2]), a[6], 1);
174 *row3 = _mm256_insertf128_ps(_mm256_castps128_ps256(a[3]), a[7], 1);
175
176 gmx_hack_simd_transpose4_r(row0, row1, row2, row3);
177 }
178
179 static gmx_inline void gmx_simdcall
180 gmx_hack_simd_transpose_to_simd4_r(gmx_simd_float_t row0,
181 gmx_simd_float_t row1,
182 gmx_simd_float_t row2,
183 gmx_simd_float_t row3,
184 gmx_simd4_float_t *a)
185 {
186 gmx_hack_simd_transpose4_r(&row0, &row1, &row2, &row3);
187
188 a[0] = _mm256_extractf128_ps(row0, 0);
189 a[1] = _mm256_extractf128_ps(row1, 0);
190 a[2] = _mm256_extractf128_ps(row2, 0);
191 a[3] = _mm256_extractf128_ps(row3, 0);
192 a[4] = _mm256_extractf128_ps(row0, 1);
193 a[5] = _mm256_extractf128_ps(row1, 1);
194 a[6] = _mm256_extractf128_ps(row2, 1);
195 a[7] = _mm256_extractf128_ps(row3, 1);
196 }
197 #ifdef GMX_SIMD_X86_AVX_GCC_MASKLOAD_BUG
198 # define gmx_hack_simd4_load3_r(mem) _mm_maskload_ps((mem), _mm_castsi256_pd(_mm_set_epi32(0, -1, -1, -1)))
199 # define gmx_hack_simd4_store3_r(mem, x) _mm_maskstore_ps((mem), _mm_castsi256_pd(_mm_set_epi32(0, -1, -1, -1)), (x))
200 #else
201 # define gmx_hack_simd4_load3_r(mem) _mm_maskload_ps((mem), _mm_set_epi32(0, -1, -1, -1))
202 # define gmx_hack_simd4_store3_r(mem, x) _mm_maskstore_ps((mem), _mm_set_epi32(0, -1, -1, -1), (x))
203 #endif
204
205 #endif /* double */
206
207 #endif /* AVX */
208
209 #ifdef GMX_SIMD_HAVE_REAL
210 /*! \brief Store differences between indexed rvecs in SIMD registers.
211 *
212 * Returns SIMD register with the difference vectors:
213 * v[pair_index[i*2]] - v[pair_index[i*2 + 1]]
214 *
215 * \param[in] v Array of rvecs
216 * \param[in] pair_index Index pairs for GMX_SIMD_REAL_WIDTH vector pairs
217 * \param[in,out] buf Aligned tmp buffer of size 3*GMX_SIMD_REAL_WIDTH
218 * \param[out] dx SIMD register with x difference
219 * \param[out] dy SIMD register with y difference
220 * \param[out] dz SIMD register with z difference
221 */
222 static gmx_inline void gmx_simdcall
223 gmx_hack_simd_gather_rvec_dist_pair_index(const rvec *v,
224 const int *pair_index,
225 real gmx_unused *buf,
226 gmx_simd_real_t *dx,
227 gmx_simd_real_t *dy,
228 gmx_simd_real_t *dz)
229 {
230 #if defined(GMX_SIMD_X86_AVX_256) || defined(GMX_SIMD_X86_AVX2_256)
231 int i;
232 gmx_simd4_real_t d[GMX_SIMD_REAL_WIDTH];
233 gmx_simd_real_t tmp;
234
235 for (i = 0; i < GMX_SIMD_REAL_WIDTH; i++)
236 {
237 d[i] = gmx_simd4_sub_r(gmx_hack_simd4_load3_r(&(v[pair_index[i*2 + 0]][0])),
238 gmx_hack_simd4_load3_r(&(v[pair_index[i*2 + 1]][0])));
239 }
240
241 gmx_hack_simd4_transpose_to_simd_r(d, dx, dy, dz, &tmp);
242 #else
243 #if GMX_ALIGNMENT
244 GMX_ALIGNED(real, GMX_SIMD_REAL_WIDTH) buf_aligned[3*GMX_SIMD_REAL_WIDTH];
245 #else
246 real* buf_aligned = buf;
247 #endif
248
249 int i, m;
250
251 for (i = 0; i < GMX_SIMD_REAL_WIDTH; i++)
252 {
253 /* Store the distances packed and aligned */
254 for (m = 0; m < DIM; m++)
255 {
256 buf_aligned[m*GMX_SIMD_REAL_WIDTH + i] =
257 v[pair_index[i*2]][m] - v[pair_index[i*2 + 1]][m];
258 }
259 }
260 *dx = gmx_simd_load_r(buf_aligned + 0*GMX_SIMD_REAL_WIDTH);
261 *dy = gmx_simd_load_r(buf_aligned + 1*GMX_SIMD_REAL_WIDTH);
262 *dz = gmx_simd_load_r(buf_aligned + 2*GMX_SIMD_REAL_WIDTH);
263 #endif
264 }
265
266
267 /*! \brief Stores SIMD vector into multiple rvecs.
268 *
269 * \param[in] x SIMD register with x-components of the vectors
270 * \param[in] y SIMD register with y-components of the vectors
271 * \param[in] z SIMD register with z-components of the vectors
272 * \param[in,out] buf Aligned tmp buffer of size 3*GMX_SIMD_REAL_WIDTH
273 * \param[out] v Array of GMX_SIMD_REAL_WIDTH rvecs
274 */
275 static gmx_inline void gmx_simdcall
276 gmx_simd_store_vec_to_rvec(gmx_simd_real_t x,
277 gmx_simd_real_t y,
278 gmx_simd_real_t z,
279 real gmx_unused *buf,
280 rvec *v)
281 {
282 #if defined(GMX_SIMD_X86_AVX_256) || defined(GMX_SIMD_X86_AVX2_256)
283 int i;
284 gmx_simd4_real_t s4[GMX_SIMD_REAL_WIDTH];
285 gmx_simd_real_t zero = gmx_simd_setzero_r();
286
287 gmx_hack_simd_transpose_to_simd4_r(x, y, z, zero, s4);
288
289 for (i = 0; i < GMX_SIMD_REAL_WIDTH; i++)
290 {
291 gmx_hack_simd4_store3_r(v[i], s4[i]);
292 }
293 #else
294 #if GMX_ALIGNMENT
295 GMX_ALIGNED(real, GMX_SIMD_REAL_WIDTH) buf_aligned[3*GMX_SIMD_REAL_WIDTH];
296 #else
297 real* buf_aligned = buf;
298 #endif
299
300 int i, m;
301
302 gmx_simd_store_r(buf_aligned + 0*GMX_SIMD_REAL_WIDTH, x);
303 gmx_simd_store_r(buf_aligned + 1*GMX_SIMD_REAL_WIDTH, y);
304 gmx_simd_store_r(buf_aligned + 2*GMX_SIMD_REAL_WIDTH, z);
305
306 for (i = 0; i < GMX_SIMD_REAL_WIDTH; i++)
307 {
308 for (m = 0; m < DIM; m++)
309 {
310 v[i][m] = buf_aligned[m*GMX_SIMD_REAL_WIDTH + i];
311 }
312 }
313 #endif
314 }
315 #endif /* GMX_SIMD_HAVE_REAL */
316
317
318 typedef struct {
319 int b0; /* first constraint for this task */
320 int b1; /* b1-1 is the last constraint for this task */
321 int ntriangle; /* the number of constraints in triangles */
322 int *triangle; /* the list of triangle constraints */
323 int *tri_bits; /* the bits tell if the matrix element should be used */
324 int tri_alloc; /* allocation size of triangle and tri_bits */
325 int nind; /* number of indices */
326 int *ind; /* constraint index for updating atom data */
327 int nind_r; /* number of indices */
328 int *ind_r; /* constraint index for updating atom data */
329 int ind_nalloc; /* allocation size of ind and ind_r */
330 tensor vir_r_m_dr; /* temporary variable for virial calculation */
331 real dhdlambda; /* temporary variable for lambda derivative */
332 real *simd_buf; /* Aligned work array for SIMD */
333 } lincs_task_t;
334
335 typedef struct gmx_lincsdata {
336 int ncg; /* the global number of constraints */
337 int ncg_flex; /* the global number of flexible constraints */
338 int ncg_triangle; /* the global number of constraints in triangles */
339 int nIter; /* the number of iterations */
340 int nOrder; /* the order of the matrix expansion */
341 int max_connect; /* the maximum number of constrains connected to a single atom */
342
343 int nc_real; /* the number of real constraints */
344 int nc; /* the number of constraints including padding for SIMD */
345 int nc_alloc; /* the number we allocated memory for */
346 int ncc; /* the number of constraint connections */
347 int ncc_alloc; /* the number we allocated memory for */
348 real matlam; /* the FE lambda value used for filling blc and blmf */
349 int *con_index; /* mapping from topology to LINCS constraints */
350 real *bllen0; /* the reference distance in topology A */
351 real *ddist; /* the reference distance in top B - the r.d. in top A */
352 int *bla; /* the atom pairs involved in the constraints */
353 real *blc; /* 1/sqrt(invmass1 + invmass2) */
354 real *blc1; /* as blc, but with all masses 1 */
355 int *blnr; /* index into blbnb and blmf */
356 int *blbnb; /* list of constraint connections */
357 int ntriangle; /* the local number of constraints in triangles */
358 int ncc_triangle; /* the number of constraint connections in triangles */
359 gmx_bool bCommIter; /* communicate before each LINCS interation */
360 real *blmf; /* matrix of mass factors for constraint connections */
361 real *blmf1; /* as blmf, but with all masses 1 */
362 real *bllen; /* the reference bond length */
363 int *nlocat; /* the local atom count per constraint, can be NULL */
364
365 int ntask; /* The number of tasks = #threads for LINCS */
366 lincs_task_t *task; /* LINCS thread division */
367 gmx_bitmask_t *atf; /* atom flags for thread parallelization */
368 int atf_nalloc; /* allocation size of atf */
369 gmx_bool bTaskDep; /* are the LINCS tasks interdependent? */
370 /* arrays for temporary storage in the LINCS algorithm */
371 rvec *tmpv;
372 real *tmpncc;
373 real *tmp1;
374 real *tmp2;
375 real *tmp3;
376 real *tmp4;
377 real *mlambda; /* the Lagrange multipliers * -1 */
378 /* storage for the constraint RMS relative deviation output */
379 real rmsd_data[3];
380 } t_gmx_lincsdata;
381
382 /* Define simd_width for memory allocation used for SIMD code */
383 #ifdef LINCS_SIMD
384 static const int simd_width = GMX_SIMD_REAL_WIDTH;
385 #else
386 static const int simd_width = 1;
387 #endif
388 /* We can't use small memory alignments on many systems, so use min 64 bytes*/
389 static const int align_bytes = std::max<int>(64, simd_width*sizeof(real));
390
391 real *lincs_rmsd_data(struct gmx_lincsdata *lincsd)
392 {
393 return lincsd->rmsd_data;
394 }
395
396 real lincs_rmsd(struct gmx_lincsdata *lincsd, gmx_bool bSD2)
397 {
398 if (lincsd->rmsd_data[0] > 0)
399 {
400 return sqrt(lincsd->rmsd_data[bSD2 ? 2 : 1]/lincsd->rmsd_data[0]);
401 }
402 else
403 {
404 return 0;
405 }
406 }
407
408 /* Do a set of nrec LINCS matrix multiplications.
409 * This function will return with up to date thread-local
410 * constraint data, without an OpenMP barrier.
411 */
412 static void lincs_matrix_expand(const struct gmx_lincsdata *lincsd,
413 const lincs_task_t *li_task,
414 const real *blcc,
415 real *rhs1, real *rhs2, real *sol)
416 {
417 int b0, b1, nrec, rec;
418 const int *blnr = lincsd->blnr;
419 const int *blbnb = lincsd->blbnb;
420
421 b0 = li_task->b0;
422 b1 = li_task->b1;
423 nrec = lincsd->nOrder;
424
425 for (rec = 0; rec < nrec; rec++)
426 {
427 int b;
428
429 if (lincsd->bTaskDep)
430 {
431 #pragma omp barrier
432 }
433 for (b = b0; b < b1; b++)
434 {
435 real mvb;
436 int n;
437
438 mvb = 0;
439 for (n = blnr[b]; n < blnr[b+1]; n++)
440 {
441 mvb = mvb + blcc[n]*rhs1[blbnb[n]];
442 }
443 rhs2[b] = mvb;
444 sol[b] = sol[b] + mvb;
445 }
446
447 real *swap;
448
449 swap = rhs1;
450 rhs1 = rhs2;
451 rhs2 = swap;
452 } /* nrec*(ncons+2*nrtot) flops */
453
454 if (lincsd->ntriangle > 0)
455 {
456 /* Perform an extra nrec recursions for only the constraints
457 * involved in rigid triangles.
458 * In this way their accuracy should come close to those of the other
459 * constraints, since traingles of constraints can produce eigenvalues
460 * around 0.7, while the effective eigenvalue for bond constraints
461 * is around 0.4 (and 0.7*0.7=0.5).
462 */
463
464 if (lincsd->bTaskDep)
465 {
466 /* We need a barrier here, since other threads might still be
467 * reading the contents of rhs1 and/o rhs2.
468 * We could avoid this barrier by introducing two extra rhs
469 * arrays for the triangle constraints only.
470 */
471 #pragma omp barrier
472 }
473
474 /* Constraints involved in a triangle are ensured to be in the same
475 * LINCS task. This means no barriers are required during the extra
476 * iterations for the triangle constraints.
477 */
478 const int *triangle = li_task->triangle;
479 const int *tri_bits = li_task->tri_bits;
480
481 for (rec = 0; rec < nrec; rec++)
482 {
483 int tb;
484
485 for (tb = 0; tb < li_task->ntriangle; tb++)
486 {
487 int b, bits, nr0, nr1, n;
488 real mvb;
489
490 b = triangle[tb];
491 bits = tri_bits[tb];
492 mvb = 0;
493 nr0 = blnr[b];
494 nr1 = blnr[b+1];
495 for (n = nr0; n < nr1; n++)
496 {
497 if (bits & (1 << (n - nr0)))
498 {
499 mvb = mvb + blcc[n]*rhs1[blbnb[n]];
500 }
501 }
502 rhs2[b] = mvb;
503 sol[b] = sol[b] + mvb;
504 }
505
506 real *swap;
507
508 swap = rhs1;
509 rhs1 = rhs2;
510 rhs2 = swap;
511 } /* nrec*(ntriangle + ncc_triangle*2) flops */
512 }
513 }
514
515 static void lincs_update_atoms_noind(int ncons, const int *bla,
516 real prefac,
517 const real *fac, rvec *r,
518 const real *invmass,
519 rvec *x)
520 {
521 int b, i, j;
522 real mvb, im1, im2, tmp0, tmp1, tmp2;
523
524 if (invmass != NULL)
525 {
526 for (b = 0; b < ncons; b++)
527 {
528 i = bla[2*b];
529 j = bla[2*b+1];
530 mvb = prefac*fac[b];
531 im1 = invmass[i];
532 im2 = invmass[j];
533 tmp0 = r[b][0]*mvb;
534 tmp1 = r[b][1]*mvb;
535 tmp2 = r[b][2]*mvb;
536 x[i][0] -= tmp0*im1;
537 x[i][1] -= tmp1*im1;
538 x[i][2] -= tmp2*im1;
539 x[j][0] += tmp0*im2;
540 x[j][1] += tmp1*im2;
541 x[j][2] += tmp2*im2;
542 } /* 16 ncons flops */
543 }
544 else
545 {
546 for (b = 0; b < ncons; b++)
547 {
548 i = bla[2*b];
549 j = bla[2*b+1];
550 mvb = prefac*fac[b];
551 tmp0 = r[b][0]*mvb;
552 tmp1 = r[b][1]*mvb;
553 tmp2 = r[b][2]*mvb;
554 x[i][0] -= tmp0;
555 x[i][1] -= tmp1;
556 x[i][2] -= tmp2;
557 x[j][0] += tmp0;
558 x[j][1] += tmp1;
559 x[j][2] += tmp2;
560 }
561 }
562 }
563
564 static void lincs_update_atoms_ind(int ncons, const int *ind, const int *bla,
565 real prefac,
566 const real *fac, rvec *r,
567 const real *invmass,
568 rvec *x)
569 {
570 int bi, b, i, j;
571 real mvb, im1, im2, tmp0, tmp1, tmp2;
572
573 if (invmass != NULL)
574 {
575 for (bi = 0; bi < ncons; bi++)
576 {
577 b = ind[bi];
578 i = bla[2*b];
579 j = bla[2*b+1];
580 mvb = prefac*fac[b];
581 im1 = invmass[i];
582 im2 = invmass[j];
583 tmp0 = r[b][0]*mvb;
584 tmp1 = r[b][1]*mvb;
585 tmp2 = r[b][2]*mvb;
586 x[i][0] -= tmp0*im1;
587 x[i][1] -= tmp1*im1;
588 x[i][2] -= tmp2*im1;
589 x[j][0] += tmp0*im2;
590 x[j][1] += tmp1*im2;
591 x[j][2] += tmp2*im2;
592 } /* 16 ncons flops */
593 }
594 else
595 {
596 for (bi = 0; bi < ncons; bi++)
597 {
598 b = ind[bi];
599 i = bla[2*b];
600 j = bla[2*b+1];
601 mvb = prefac*fac[b];
602 tmp0 = r[b][0]*mvb;
603 tmp1 = r[b][1]*mvb;
604 tmp2 = r[b][2]*mvb;
605 x[i][0] -= tmp0;
606 x[i][1] -= tmp1;
607 x[i][2] -= tmp2;
608 x[j][0] += tmp0;
609 x[j][1] += tmp1;
610 x[j][2] += tmp2;
611 } /* 16 ncons flops */
612 }
613 }
614
615 static void lincs_update_atoms(struct gmx_lincsdata *li, int th,
616 real prefac,
617 const real *fac, rvec *r,
618 const real *invmass,
619 rvec *x)
620 {
621 if (li->ntask == 1)
622 {
623 /* Single thread, we simply update for all constraints */
624 lincs_update_atoms_noind(li->nc_real,
625 li->bla, prefac, fac, r, invmass, x);
626 }
627 else
628 {
629 /* Update the atom vector components for our thread local
630 * constraints that only access our local atom range.
631 * This can be done without a barrier.
632 */
633 lincs_update_atoms_ind(li->task[th].nind, li->task[th].ind,
634 li->bla, prefac, fac, r, invmass, x);
635
636 if (li->task[li->ntask].nind > 0)
637 {
638 /* Update the constraints that operate on atoms
639 * in multiple thread atom blocks on the master thread.
640 */
641 #pragma omp barrier
642 #pragma omp master
643 {
644 lincs_update_atoms_ind(li->task[li->ntask].nind,
645 li->task[li->ntask].ind,
646 li->bla, prefac, fac, r, invmass, x);
647 }
648 }
649 }
650 }
651
652 #ifdef LINCS_SIMD
653 /* Calculate the constraint distance vectors r to project on from x.
654 * Determine the right-hand side of the matrix equation using quantity f.
655 * This function only differs from calc_dr_x_xp_simd below in that
656 * no constraint length is subtracted and no PBC is used for f.
657 */
658 static void gmx_simdcall
659 calc_dr_x_f_simd(int b0,
660 int b1,
661 const int * bla,
662 const rvec * gmx_restrict x,
663 const rvec * gmx_restrict f,
664 const real * gmx_restrict blc,
665 const pbc_simd_t * pbc_simd,
666 real * gmx_restrict vbuf1,
667 real * gmx_restrict vbuf2,
668 rvec * gmx_restrict r,
669 real * gmx_restrict rhs,
670 real * gmx_restrict sol)
671 {
672 int bs;
673
674 assert(b0 % GMX_SIMD_REAL_WIDTH == 0);
675
676 for (bs = b0; bs < b1; bs += GMX_SIMD_REAL_WIDTH)
677 {
678 gmx_simd_real_t rx_S, ry_S, rz_S, n2_S, il_S;
679 gmx_simd_real_t fx_S, fy_S, fz_S, ip_S, rhs_S;
680
681 gmx_hack_simd_gather_rvec_dist_pair_index(x, bla + bs*2, vbuf1,
682 &rx_S, &ry_S, &rz_S);
683
684 pbc_correct_dx_simd(&rx_S, &ry_S, &rz_S, pbc_simd);
685
686 n2_S = gmx_simd_norm2_r(rx_S, ry_S, rz_S);
687 il_S = gmx_simd_invsqrt_r(n2_S);
688
689 rx_S = gmx_simd_mul_r(rx_S, il_S);
690 ry_S = gmx_simd_mul_r(ry_S, il_S);
691 rz_S = gmx_simd_mul_r(rz_S, il_S);
692
693 gmx_simd_store_vec_to_rvec(rx_S, ry_S, rz_S, vbuf1, r + bs);
694
695 gmx_hack_simd_gather_rvec_dist_pair_index(f, bla + bs*2, vbuf2,
696 &fx_S, &fy_S, &fz_S);
697
698 ip_S = gmx_simd_iprod_r(rx_S, ry_S, rz_S,
699 fx_S, fy_S, fz_S);
700
701 rhs_S = gmx_simd_mul_r(gmx_simd_load_r(blc + bs), ip_S);
702
703 gmx_simd_store_r(rhs + bs, rhs_S);
704 gmx_simd_store_r(sol + bs, rhs_S);
705 }
706 }
707 #endif /* LINCS_SIMD */
708
709 /* LINCS projection, works on derivatives of the coordinates */
710 static void do_lincsp(rvec *x, rvec *f, rvec *fp, t_pbc *pbc,
711 struct gmx_lincsdata *lincsd, int th,
712 real *invmass,
713 int econq, gmx_bool bCalcDHDL,
714 gmx_bool bCalcVir, tensor rmdf)
715 {
716 int b0, b1, b;
717 int *bla, *blnr, *blbnb;
718 rvec *r;
719 real *blc, *blmf, *blcc, *rhs1, *rhs2, *sol;
720
721 b0 = lincsd->task[th].b0;
722 b1 = lincsd->task[th].b1;
723
724 bla = lincsd->bla;
725 r = lincsd->tmpv;
726 blnr = lincsd->blnr;
727 blbnb = lincsd->blbnb;
728 if (econq != econqForce)
729 {
730 /* Use mass-weighted parameters */
731 blc = lincsd->blc;
732 blmf = lincsd->blmf;
733 }
734 else
735 {
736 /* Use non mass-weighted parameters */
737 blc = lincsd->blc1;
738 blmf = lincsd->blmf1;
739 }
740 blcc = lincsd->tmpncc;
741 rhs1 = lincsd->tmp1;
742 rhs2 = lincsd->tmp2;
743 sol = lincsd->tmp3;
744
745 #ifdef LINCS_SIMD
746
747 /* This SIMD code does the same as the plain-C code after the #else.
748 * The only difference is that we always call pbc code, as with SIMD
749 * the overhead of pbc computation (when not needed) is small.
750 */
751 pbc_simd_t pbc_simd;
752
753 /* Convert the pbc struct for SIMD */
754 set_pbc_simd(pbc, &pbc_simd);
755
756 /* Compute normalized x i-j vectors, store in r.
757 * Compute the inner product of r and xp i-j and store in rhs1.
758 */
759 calc_dr_x_f_simd(b0, b1, bla, x, f, blc,
760 &pbc_simd,
761 lincsd->task[th].simd_buf,
762 lincsd->task[th].simd_buf + GMX_SIMD_REAL_WIDTH*DIM,
763 r, rhs1, sol);
764
765 #else /* LINCS_SIMD */
766
767 /* Compute normalized i-j vectors */
768 if (pbc)
769 {
770 for (b = b0; b < b1; b++)
771 {
772 rvec dx;
773
774 pbc_dx_aiuc(pbc, x[bla[2*b]], x[bla[2*b+1]], dx);
775 unitv(dx, r[b]);
776 }
777 }
778 else
779 {
780 for (b = b0; b < b1; b++)
781 {
782 rvec dx;
783
784 rvec_sub(x[bla[2*b]], x[bla[2*b+1]], dx);
785 unitv(dx, r[b]);
786 } /* 16 ncons flops */
787 }
788
789 for (b = b0; b < b1; b++)
790 {
791 int i, j;
792 real mvb;
793
794 i = bla[2*b];
795 j = bla[2*b+1];
796 mvb = blc[b]*(r[b][0]*(f[i][0] - f[j][0]) +
797 r[b][1]*(f[i][1] - f[j][1]) +
798 r[b][2]*(f[i][2] - f[j][2]));
799 rhs1[b] = mvb;
800 sol[b] = mvb;
801 /* 7 flops */
802 }
803
804 #endif /* LINCS_SIMD */
805
806 if (lincsd->bTaskDep)
807 {
808 /* We need a barrier, since the matrix construction below
809 * can access entries in r of other threads.
810 */
811 #pragma omp barrier
812 }
813
814 /* Construct the (sparse) LINCS matrix */
815 for (b = b0; b < b1; b++)
816 {
817 int n;
818
819 for (n = blnr[b]; n < blnr[b+1]; n++)
820 {
821 blcc[n] = blmf[n]*iprod(r[b], r[blbnb[n]]);
822 } /* 6 nr flops */
823 }
824 /* Together: 23*ncons + 6*nrtot flops */
825
826 lincs_matrix_expand(lincsd, &lincsd->task[th], blcc, rhs1, rhs2, sol);
827 /* nrec*(ncons+2*nrtot) flops */
828
829 if (econq == econqDeriv_FlexCon)
830 {
831 /* We only want to constraint the flexible constraints,
832 * so we mask out the normal ones by setting sol to 0.
833 */
834 for (b = b0; b < b1; b++)
835 {
836 if (!(lincsd->bllen0[b] == 0 && lincsd->ddist[b] == 0))
837 {
838 sol[b] = 0;
839 }
840 }
841 }
842
843 /* We multiply sol by blc, so we can use lincs_update_atoms for OpenMP */
844 for (b = b0; b < b1; b++)
845 {
846 sol[b] *= blc[b];
847 }
848
849 /* When constraining forces, we should not use mass weighting,
850 * so we pass invmass=NULL, which results in the use of 1 for all atoms.
851 */
852 lincs_update_atoms(lincsd, th, 1.0, sol, r,
853 (econq != econqForce) ? invmass : NULL, fp);
854
855 if (bCalcDHDL)
856 {
857 real dhdlambda;
858
859 dhdlambda = 0;
860 for (b = b0; b < b1; b++)
861 {
862 dhdlambda -= sol[b]*lincsd->ddist[b];
863 }
864
865 lincsd->task[th].dhdlambda = dhdlambda;
866 }
867
868 if (bCalcVir)
869 {
870 /* Constraint virial,
871 * determines sum r_bond x delta f,
872 * where delta f is the constraint correction
873 * of the quantity that is being constrained.
874 */
875 for (b = b0; b < b1; b++)
876 {
877 real mvb, tmp1;
878 int i, j;
879
880 mvb = lincsd->bllen[b]*sol[b];
881 for (i = 0; i < DIM; i++)
882 {
883 tmp1 = mvb*r[b][i];
884 for (j = 0; j < DIM; j++)
885 {
886 rmdf[i][j] += tmp1*r[b][j];
887 }
888 }
889 } /* 23 ncons flops */
890 }
891 }
892
893 #ifdef LINCS_SIMD
894 /* Calculate the constraint distance vectors r to project on from x.
895 * Determine the right-hand side of the matrix equation using coordinates xp.
896 */
897 static void gmx_simdcall
898 calc_dr_x_xp_simd(int b0,
899 int b1,
900 const int * bla,
901 const rvec * gmx_restrict x,
902 const rvec * gmx_restrict xp,
903 const real * gmx_restrict bllen,
904 const real * gmx_restrict blc,
905 const pbc_simd_t * pbc_simd,
906 real * gmx_restrict vbuf1,
907 real * gmx_restrict vbuf2,
908 rvec * gmx_restrict r,
909 real * gmx_restrict rhs,
910 real * gmx_restrict sol)
911 {
912 int bs;
913
914 assert(b0 % GMX_SIMD_REAL_WIDTH == 0);
915
916 for (bs = b0; bs < b1; bs += GMX_SIMD_REAL_WIDTH)
917 {
918 gmx_simd_real_t rx_S, ry_S, rz_S, n2_S, il_S;
919 gmx_simd_real_t rxp_S, ryp_S, rzp_S, ip_S, rhs_S;
920
921 gmx_hack_simd_gather_rvec_dist_pair_index(x, bla + bs*2, vbuf1,
922 &rx_S, &ry_S, &rz_S);
923
924 pbc_correct_dx_simd(&rx_S, &ry_S, &rz_S, pbc_simd);
925
926 n2_S = gmx_simd_norm2_r(rx_S, ry_S, rz_S);
927 il_S = gmx_simd_invsqrt_r(n2_S);
928
929 rx_S = gmx_simd_mul_r(rx_S, il_S);
930 ry_S = gmx_simd_mul_r(ry_S, il_S);
931 rz_S = gmx_simd_mul_r(rz_S, il_S);
932
933 gmx_simd_store_vec_to_rvec(rx_S, ry_S, rz_S, vbuf1, r + bs);
934
935 gmx_hack_simd_gather_rvec_dist_pair_index(xp, bla + bs*2, vbuf2,
936 &rxp_S, &ryp_S, &rzp_S);
937
938 pbc_correct_dx_simd(&rxp_S, &ryp_S, &rzp_S, pbc_simd);
939
940 ip_S = gmx_simd_iprod_r(rx_S, ry_S, rz_S,
941 rxp_S, ryp_S, rzp_S);
942
943 rhs_S = gmx_simd_mul_r(gmx_simd_load_r(blc + bs),
944 gmx_simd_sub_r(ip_S, gmx_simd_load_r(bllen + bs)));
945
946 gmx_simd_store_r(rhs + bs, rhs_S);
947 gmx_simd_store_r(sol + bs, rhs_S);
948 }
949 }
950 #endif /* LINCS_SIMD */
951
952 /* Determine the distances and right-hand side for the next iteration */
953 static void calc_dist_iter(int b0,
954 int b1,
955 const int * bla,
956 const rvec * gmx_restrict xp,
957 const real * gmx_restrict bllen,
958 const real * gmx_restrict blc,
959 const t_pbc * pbc,
960 real wfac,
961 real * gmx_restrict rhs,
962 real * gmx_restrict sol,
963 gmx_bool * bWarn)
964 {
965 int b;
966
967 for (b = b0; b < b1; b++)
968 {
969 real len, len2, dlen2, mvb;
970 rvec dx;
971
972 len = bllen[b];
973 if (pbc)
974 {
975 pbc_dx_aiuc(pbc, xp[bla[2*b]], xp[bla[2*b+1]], dx);
976 }
977 else
978 {
979 rvec_sub(xp[bla[2*b]], xp[bla[2*b+1]], dx);
980 }
981 len2 = len*len;
982 dlen2 = 2*len2 - norm2(dx);
983 if (dlen2 < wfac*len2)
984 {
985 /* not race free - see detailed comment in caller */
986 *bWarn = TRUE;
987 }
988 if (dlen2 > 0)
989 {
990 mvb = blc[b]*(len - dlen2*gmx_invsqrt(dlen2));
991 }
992 else
993 {
994 mvb = blc[b]*len;
995 }
996 rhs[b] = mvb;
997 sol[b] = mvb;
998 } /* 20*ncons flops */
999 }
1000
1001 #ifdef LINCS_SIMD
1002 /* As the function above, but using SIMD intrinsics */
1003 static void gmx_simdcall
1004 calc_dist_iter_simd(int b0,
1005 int b1,
1006 const int * bla,
1007 const rvec * gmx_restrict x,
1008 const real * gmx_restrict bllen,
1009 const real * gmx_restrict blc,
1010 const pbc_simd_t * pbc_simd,
1011 real wfac,
1012 real * gmx_restrict vbuf,
1013 real * gmx_restrict rhs,
1014 real * gmx_restrict sol,
1015 gmx_bool * bWarn)
1016 {
1017 gmx_simd_real_t min_S = gmx_simd_set1_r(GMX_REAL_MIN);
1018 gmx_simd_real_t two_S = gmx_simd_set1_r(2.0);
1019 gmx_simd_real_t wfac_S = gmx_simd_set1_r(wfac);
1020 gmx_simd_bool_t warn_S;
1021
1022 int bs;
1023
1024 assert(b0 % GMX_SIMD_REAL_WIDTH == 0);
1025
1026 /* Initialize all to FALSE */
1027 warn_S = gmx_simd_cmplt_r(two_S, gmx_simd_setzero_r());
1028
1029 for (bs = b0; bs < b1; bs += GMX_SIMD_REAL_WIDTH)
1030 {
1031 gmx_simd_real_t rx_S, ry_S, rz_S, n2_S;
1032 gmx_simd_real_t len_S, len2_S, dlen2_S, lc_S, blc_S;
1033
1034 gmx_hack_simd_gather_rvec_dist_pair_index(x, bla + bs*2, vbuf,
1035 &rx_S, &ry_S, &rz_S);
1036
1037 pbc_correct_dx_simd(&rx_S, &ry_S, &rz_S, pbc_simd);
1038
1039 n2_S = gmx_simd_norm2_r(rx_S, ry_S, rz_S);
1040
1041 len_S = gmx_simd_load_r(bllen + bs);
1042 len2_S = gmx_simd_mul_r(len_S, len_S);
1043
1044 dlen2_S = gmx_simd_fmsub_r(two_S, len2_S, n2_S);
1045
1046 warn_S = gmx_simd_or_b(warn_S,
1047 gmx_simd_cmplt_r(dlen2_S,
1048 gmx_simd_mul_r(wfac_S, len2_S)));
1049
1050 /* Avoid 1/0 by taking the max with REAL_MIN.
1051 * Note: when dlen2 is close to zero (90 degree constraint rotation),
1052 * the accuracy of the algorithm is no longer relevant.
1053 */
1054 dlen2_S = gmx_simd_max_r(dlen2_S, min_S);
1055
1056 lc_S = gmx_simd_fnmadd_r(dlen2_S, gmx_simd_invsqrt_r(dlen2_S), len_S);
1057
1058 blc_S = gmx_simd_load_r(blc + bs);
1059
1060 lc_S = gmx_simd_mul_r(blc_S, lc_S);
1061
1062 gmx_simd_store_r(rhs + bs, lc_S);
1063 gmx_simd_store_r(sol + bs, lc_S);
1064 }
1065
1066 if (gmx_simd_anytrue_b(warn_S))
1067 {
1068 *bWarn = TRUE;
1069 }
1070 }
1071 #endif /* LINCS_SIMD */
1072
1073 static void do_lincs(rvec *x, rvec *xp, matrix box, t_pbc *pbc,
1074 struct gmx_lincsdata *lincsd, int th,
1075 const real *invmass,
1076 t_commrec *cr,
1077 gmx_bool bCalcDHDL,
1078 real wangle, gmx_bool *bWarn,
1079 real invdt, rvec * gmx_restrict v,
1080 gmx_bool bCalcVir, tensor vir_r_m_dr)
1081 {
1082 int b0, b1, b, i, j, n, iter;
1083 int *bla, *blnr, *blbnb;
1084 rvec *r;
1085 real *blc, *blmf, *bllen, *blcc, *rhs1, *rhs2, *sol, *blc_sol, *mlambda;
1086 int *nlocat;
1087
1088 b0 = lincsd->task[th].b0;
1089 b1 = lincsd->task[th].b1;
1090
1091 bla = lincsd->bla;
1092 r = lincsd->tmpv;
1093 blnr = lincsd->blnr;
1094 blbnb = lincsd->blbnb;
1095 blc = lincsd->blc;
1096 blmf = lincsd->blmf;
1097 bllen = lincsd->bllen;
1098 blcc = lincsd->tmpncc;
1099 rhs1 = lincsd->tmp1;
1100 rhs2 = lincsd->tmp2;
1101 sol = lincsd->tmp3;
1102 blc_sol = lincsd->tmp4;
1103 mlambda = lincsd->mlambda;
1104 nlocat = lincsd->nlocat;
1105
1106 #ifdef LINCS_SIMD
1107
1108 /* This SIMD code does the same as the plain-C code after the #else.
1109 * The only difference is that we always call pbc code, as with SIMD
1110 * the overhead of pbc computation (when not needed) is small.
1111 */
1112 pbc_simd_t pbc_simd;
1113
1114 /* Convert the pbc struct for SIMD */
1115 set_pbc_simd(pbc, &pbc_simd);
1116
1117 /* Compute normalized x i-j vectors, store in r.
1118 * Compute the inner product of r and xp i-j and store in rhs1.
1119 */
1120 calc_dr_x_xp_simd(b0, b1, bla, x, xp, bllen, blc,
1121 &pbc_simd,
1122 lincsd->task[th].simd_buf,
1123 lincsd->task[th].simd_buf + GMX_SIMD_REAL_WIDTH*DIM,
1124 r, rhs1, sol);
1125
1126 #else /* LINCS_SIMD */
1127
1128 if (pbc)
1129 {
1130 /* Compute normalized i-j vectors */
1131 for (b = b0; b < b1; b++)
1132 {
1133 rvec dx;
1134 real mvb;
1135
1136 pbc_dx_aiuc(pbc, x[bla[2*b]], x[bla[2*b+1]], dx);
1137 unitv(dx, r[b]);
1138
1139 pbc_dx_aiuc(pbc, xp[bla[2*b]], xp[bla[2*b+1]], dx);
1140 mvb = blc[b]*(iprod(r[b], dx) - bllen[b]);
1141 rhs1[b] = mvb;
1142 sol[b] = mvb;
1143 }
1144 }
1145 else
1146 {
1147 /* Compute normalized i-j vectors */
1148 for (b = b0; b < b1; b++)
1149 {
1150 real tmp0, tmp1, tmp2, rlen, mvb;
1151
1152 i = bla[2*b];
1153 j = bla[2*b+1];
1154 tmp0 = x[i][0] - x[j][0];
1155 tmp1 = x[i][1] - x[j][1];
1156 tmp2 = x[i][2] - x[j][2];
1157 rlen = gmx_invsqrt(tmp0*tmp0 + tmp1*tmp1 + tmp2*tmp2);
1158 r[b][0] = rlen*tmp0;
1159 r[b][1] = rlen*tmp1;
1160 r[b][2] = rlen*tmp2;
1161 /* 16 ncons flops */
1162
1163 i = bla[2*b];
1164 j = bla[2*b+1];
1165 mvb = blc[b]*(r[b][0]*(xp[i][0] - xp[j][0]) +
1166 r[b][1]*(xp[i][1] - xp[j][1]) +
1167 r[b][2]*(xp[i][2] - xp[j][2]) - bllen[b]);
1168 rhs1[b] = mvb;
1169 sol[b] = mvb;
1170 /* 10 flops */
1171 }
1172 /* Together: 26*ncons + 6*nrtot flops */
1173 }
1174
1175 #endif /* LINCS_SIMD */
1176
1177 if (lincsd->bTaskDep)
1178 {
1179 /* We need a barrier, since the matrix construction below
1180 * can access entries in r of other threads.
1181 */
1182 #pragma omp barrier
1183 }
1184
1185 /* Construct the (sparse) LINCS matrix */
1186 for (b = b0; b < b1; b++)
1187 {
1188 for (n = blnr[b]; n < blnr[b+1]; n++)
1189 {
1190 blcc[n] = blmf[n]*iprod(r[b], r[blbnb[n]]);
1191 }
1192 }
1193 /* Together: 26*ncons + 6*nrtot flops */
1194
1195 lincs_matrix_expand(lincsd, &lincsd->task[th], blcc, rhs1, rhs2, sol);
1196 /* nrec*(ncons+2*nrtot) flops */
1197
1198 #ifndef LINCS_SIMD
1199 for (b = b0; b < b1; b++)
1200 {
1201 mlambda[b] = blc[b]*sol[b];
1202 }
1203 #else
1204 for (b = b0; b < b1; b += GMX_SIMD_REAL_WIDTH)
1205 {
1206 gmx_simd_store_r(mlambda + b,
1207 gmx_simd_mul_r(gmx_simd_load_r(blc + b),
1208 gmx_simd_load_r(sol + b)));
1209 }
1210 #endif
1211
1212 /* Update the coordinates */
1213 lincs_update_atoms(lincsd, th, 1.0, mlambda, r, invmass, xp);
1214
1215 /*
1216 ******** Correction for centripetal effects ********
1217 */
1218
1219 real wfac;
1220
1221 wfac = cos(DEG2RAD*wangle);
1222 wfac = wfac*wfac;
1223
1224 for (iter = 0; iter < lincsd->nIter; iter++)
1225 {
1226 if ((lincsd->bCommIter && DOMAINDECOMP(cr) && cr->dd->constraints))
1227 {
1228 #pragma omp barrier
1229 #pragma omp master
1230 {
1231 /* Communicate the corrected non-local coordinates */
1232 if (DOMAINDECOMP(cr))
1233 {
1234 dd_move_x_constraints(cr->dd, box, xp, NULL, FALSE);
1235 }
1236 }
1237 #pragma omp barrier
1238 }
1239 else if (lincsd->bTaskDep)
1240 {
1241 #pragma omp barrier
1242 }
1243
1244 #ifdef LINCS_SIMD
1245 calc_dist_iter_simd(b0, b1, bla, xp, bllen, blc, &pbc_simd, wfac,
1246 lincsd->task[th].simd_buf, rhs1, sol, bWarn);
1247 #else
1248 calc_dist_iter(b0, b1, bla, xp, bllen, blc, pbc, wfac,
1249 rhs1, sol, bWarn);
1250 /* 20*ncons flops */
1251 #endif
1252
1253 lincs_matrix_expand(lincsd, &lincsd->task[th], blcc, rhs1, rhs2, sol);
1254 /* nrec*(ncons+2*nrtot) flops */
1255
1256 #ifndef LINCS_SIMD
1257 for (b = b0; b < b1; b++)
1258 {
1259 real mvb;
1260
1261 mvb = blc[b]*sol[b];
1262 blc_sol[b] = mvb;
1263 mlambda[b] += mvb;
1264 }
1265 #else
1266 for (b = b0; b < b1; b += GMX_SIMD_REAL_WIDTH)
1267 {
1268 gmx_simd_real_t mvb;
1269
1270 mvb = gmx_simd_mul_r(gmx_simd_load_r(blc + b),
1271 gmx_simd_load_r(sol + b));
1272 gmx_simd_store_r(blc_sol + b, mvb);
1273 gmx_simd_store_r(mlambda + b,
1274 gmx_simd_add_r(gmx_simd_load_r(mlambda + b), mvb));
1275 }
1276 #endif
1277
1278 /* Update the coordinates */
1279 lincs_update_atoms(lincsd, th, 1.0, blc_sol, r, invmass, xp);
1280 }
1281 /* nit*ncons*(37+9*nrec) flops */
1282
1283 if (v != NULL)
1284 {
1285 /* Update the velocities */
1286 lincs_update_atoms(lincsd, th, invdt, mlambda, r, invmass, v);
1287 /* 16 ncons flops */
1288 }
1289
1290 if (nlocat != NULL && (bCalcDHDL || bCalcVir))
1291 {
1292 if (lincsd->bTaskDep)
1293 {
1294 /* In lincs_update_atoms threads might cross-read mlambda */
1295 #pragma omp barrier
1296 }
1297
1298 /* Only account for local atoms */
1299 for (b = b0; b < b1; b++)
1300 {
1301 mlambda[b] *= 0.5*nlocat[b];
1302 }
1303 }
1304
1305 if (bCalcDHDL)
1306 {
1307 real dhdl;
1308
1309 dhdl = 0;
1310 for (b = b0; b < b1; b++)
1311 {
1312 /* Note that this this is dhdl*dt^2, the dt^2 factor is corrected
1313 * later after the contributions are reduced over the threads.
1314 */
1315 dhdl -= lincsd->mlambda[b]*lincsd->ddist[b];
1316 }
1317 lincsd->task[th].dhdlambda = dhdl;
1318 }
1319
1320 if (bCalcVir)
1321 {
1322 /* Constraint virial */
1323 for (b = b0; b < b1; b++)
1324 {
1325 real tmp0, tmp1;
1326
1327 tmp0 = -bllen[b]*mlambda[b];
1328 for (i = 0; i < DIM; i++)
1329 {
1330 tmp1 = tmp0*r[b][i];
1331 for (j = 0; j < DIM; j++)
1332 {
1333 vir_r_m_dr[i][j] -= tmp1*r[b][j];
1334 }
1335 }
1336 } /* 22 ncons flops */
1337 }
1338
1339 /* Total:
1340 * 26*ncons + 6*nrtot + nrec*(ncons+2*nrtot)
1341 * + nit * (20*ncons + nrec*(ncons+2*nrtot) + 17 ncons)
1342 *
1343 * (26+nrec)*ncons + (6+2*nrec)*nrtot
1344 * + nit * ((37+nrec)*ncons + 2*nrec*nrtot)
1345 * if nit=1
1346 * (63+nrec)*ncons + (6+4*nrec)*nrtot
1347 */
1348 }
1349
1350 /* Sets the elements in the LINCS matrix for task li_task */
1351 static void set_lincs_matrix_task(struct gmx_lincsdata *li,
1352 lincs_task_t *li_task,
1353 const real *invmass,
1354 int *ncc_triangle)
1355 {
1356 int i;
1357
1358 /* Construct the coupling coefficient matrix blmf */
1359 li_task->ntriangle = 0;
1360 *ncc_triangle = 0;
1361 for (i = li_task->b0; i < li_task->b1; i++)
1362 {
1363 int a1, a2, n;
1364
1365 a1 = li->bla[2*i];
1366 a2 = li->bla[2*i+1];
1367 for (n = li->blnr[i]; (n < li->blnr[i+1]); n++)
1368 {
1369 int k, sign, center, end;
1370
1371 k = li->blbnb[n];
1372
1373 /* If we are using multiple, independent tasks for LINCS,
1374 * the calls to check_assign_connected should have
1375 * put all connected constraints in our task.
1376 */
1377 assert(li->bTaskDep || (k >= li_task->b0 && k < li_task->b1));
1378
1379 if (a1 == li->bla[2*k] || a2 == li->bla[2*k+1])
1380 {
1381 sign = -1;
1382 }
1383 else
1384 {
1385 sign = 1;
1386 }
1387 if (a1 == li->bla[2*k] || a1 == li->bla[2*k+1])
1388 {
1389 center = a1;
1390 end = a2;
1391 }
1392 else
1393 {
1394 center = a2;
1395 end = a1;
1396 }
1397 li->blmf[n] = sign*invmass[center]*li->blc[i]*li->blc[k];
1398 li->blmf1[n] = sign*0.5;
1399 if (li->ncg_triangle > 0)
1400 {
1401 int nk, kk;
1402
1403 /* Look for constraint triangles */
1404 for (nk = li->blnr[k]; (nk < li->blnr[k+1]); nk++)
1405 {
1406 kk = li->blbnb[nk];
1407 if (kk != i && kk != k &&
1408 (li->bla[2*kk] == end || li->bla[2*kk+1] == end))
1409 {
1410 /* If we are using multiple tasks for LINCS,
1411 * the calls to check_assign_triangle should have
1412 * put all constraints in the triangle in our task.
1413 */
1414 assert(k >= li_task->b0 && k < li_task->b1);
1415 assert(kk >= li_task->b0 && kk < li_task->b1);
1416
1417 if (li_task->ntriangle == 0 ||
1418 li_task->triangle[li_task->ntriangle - 1] < i)
1419 {
1420 /* Add this constraint to the triangle list */
1421 li_task->triangle[li_task->ntriangle] = i;
1422 li_task->tri_bits[li_task->ntriangle] = 0;
1423 li_task->ntriangle++;
1424 if (li->blnr[i+1] - li->blnr[i] > static_cast<int>(sizeof(li_task->tri_bits[0])*8 - 1))
1425 {
1426 gmx_fatal(FARGS, "A constraint is connected to %d constraints, this is more than the %d allowed for constraints participating in triangles",
1427 li->blnr[i+1] - li->blnr[i],
1428 sizeof(li_task->tri_bits[0])*8-1);
1429 }
1430 }
1431 li_task->tri_bits[li_task->ntriangle-1] |= (1 << (n - li->blnr[i]));
1432 (*ncc_triangle)++;
1433 }
1434 }
1435 }
1436 }
1437 }
1438 }
1439
1440 /* Sets the elements in the LINCS matrix */
1441 void set_lincs_matrix(struct gmx_lincsdata *li, real *invmass, real lambda)
1442 {
1443 int i;
1444 const real invsqrt2 = 0.7071067811865475244;
1445
1446 for (i = 0; (i < li->nc); i++)
1447 {
1448 int a1, a2;
1449
1450 a1 = li->bla[2*i];
1451 a2 = li->bla[2*i+1];
1452 li->blc[i] = gmx_invsqrt(invmass[a1] + invmass[a2]);
1453 li->blc1[i] = invsqrt2;
1454 }
1455
1456 /* Construct the coupling coefficient matrix blmf */
1457 int th, ntriangle = 0, ncc_triangle = 0;
1458 #pragma omp parallel for reduction(+: ntriangle, ncc_triangle) num_threads(li->ntask) schedule(static)
1459 for (th = 0; th < li->ntask; th++)
1460 {
1461 set_lincs_matrix_task(li, &li->task[th], invmass, &ncc_triangle);
1462 ntriangle = li->task[th].ntriangle;
1463 }
1464 li->ntriangle = ntriangle;
1465 li->ncc_triangle = ncc_triangle;
1466
1467 if (debug)
1468 {
1469 fprintf(debug, "Of the %d constraints %d participate in triangles\n",
1470 li->nc, li->ntriangle);
1471 fprintf(debug, "There are %d couplings of which %d in triangles\n",
1472 li->ncc, li->ncc_triangle);
1473 }
1474
1475 /* Set matlam,
1476 * so we know with which lambda value the masses have been set.
1477 */
1478 li->matlam = lambda;
1479 }
1480
1481 static int count_triangle_constraints(t_ilist *ilist, t_blocka *at2con)
1482 {
1483 int ncon1, ncon_tot;
1484 int c0, a00, a01, n1, c1, a10, a11, ac1, n2, c2, a20, a21;
1485 int ncon_triangle;
1486 gmx_bool bTriangle;
1487 t_iatom *ia1, *ia2, *iap;
1488
1489 ncon1 = ilist[F_CONSTR].nr/3;
1490 ncon_tot = ncon1 + ilist[F_CONSTRNC].nr/3;
1491
1492 ia1 = ilist[F_CONSTR].iatoms;
1493 ia2 = ilist[F_CONSTRNC].iatoms;
1494
1495 ncon_triangle = 0;
1496 for (c0 = 0; c0 < ncon_tot; c0++)
1497 {
1498 bTriangle = FALSE;
1499 iap = constr_iatomptr(ncon1, ia1, ia2, c0);
1500 a00 = iap[1];
1501 a01 = iap[2];
1502 for (n1 = at2con->index[a01]; n1 < at2con->index[a01+1]; n1++)
1503 {
1504 c1 = at2con->a[n1];
1505 if (c1 != c0)
1506 {
1507 iap = constr_iatomptr(ncon1, ia1, ia2, c1);
1508 a10 = iap[1];
1509 a11 = iap[2];
1510 if (a10 == a01)
1511 {
1512 ac1 = a11;
1513 }
1514 else
1515 {
1516 ac1 = a10;
1517 }
1518 for (n2 = at2con->index[ac1]; n2 < at2con->index[ac1+1]; n2++)
1519 {
1520 c2 = at2con->a[n2];
1521 if (c2 != c0 && c2 != c1)
1522 {
1523 iap = constr_iatomptr(ncon1, ia1, ia2, c2);
1524 a20 = iap[1];
1525 a21 = iap[2];
1526 if (a20 == a00 || a21 == a00)
1527 {
1528 bTriangle = TRUE;
1529 }
1530 }
1531 }
1532 }
1533 }
1534 if (bTriangle)
1535 {
1536 ncon_triangle++;
1537 }
1538 }
1539
1540 return ncon_triangle;
1541 }
1542
1543 static gmx_bool more_than_two_sequential_constraints(const t_ilist *ilist,
1544 const t_blocka *at2con)
1545 {
1546 t_iatom *ia1, *ia2, *iap;
1547 int ncon1, ncon_tot, c;
1548 int a1, a2;
1549 gmx_bool bMoreThanTwoSequentialConstraints;
1550
1551 ncon1 = ilist[F_CONSTR].nr/3;
1552 ncon_tot = ncon1 + ilist[F_CONSTRNC].nr/3;
1553
1554 ia1 = ilist[F_CONSTR].iatoms;
1555 ia2 = ilist[F_CONSTRNC].iatoms;
1556
1557 bMoreThanTwoSequentialConstraints = FALSE;
1558 for (c = 0; c < ncon_tot && !bMoreThanTwoSequentialConstraints; c++)
1559 {
1560 iap = constr_iatomptr(ncon1, ia1, ia2, c);
1561 a1 = iap[1];
1562 a2 = iap[2];
1563 /* Check if this constraint has constraints connected at both atoms */
1564 if (at2con->index[a1+1] - at2con->index[a1] > 1 &&
1565 at2con->index[a2+1] - at2con->index[a2] > 1)
1566 {
1567 bMoreThanTwoSequentialConstraints = TRUE;
1568 }
1569 }
1570
1571 return bMoreThanTwoSequentialConstraints;
1572 }
1573
1574 static int int_comp(const void *a, const void *b)
1575 {
1576 return (*(int *)a) - (*(int *)b);
1577 }
1578
1579 gmx_lincsdata_t init_lincs(FILE *fplog, gmx_mtop_t *mtop,
1580 int nflexcon_global, t_blocka *at2con,
1581 gmx_bool bPLINCS, int nIter, int nProjOrder)
1582 {
1583 struct gmx_lincsdata *li;
1584 int mt, mb;
1585 gmx_moltype_t *molt;
1586 gmx_bool bMoreThanTwoSeq;
1587
1588 if (fplog)
1589 {
1590 fprintf(fplog, "\nInitializing%s LINear Constraint Solver\n",
1591 bPLINCS ? " Parallel" : "");
1592 }
1593
1594 snew(li, 1);
1595
1596 li->ncg =
1597 gmx_mtop_ftype_count(mtop, F_CONSTR) +
1598 gmx_mtop_ftype_count(mtop, F_CONSTRNC);
1599 li->ncg_flex = nflexcon_global;
1600
1601 li->nIter = nIter;
1602 li->nOrder = nProjOrder;
1603
1604 li->max_connect = 0;
1605 for (mt = 0; mt < mtop->nmoltype; mt++)
1606 {
1607 int a;
1608
1609 molt = &mtop->moltype[mt];
1610 for (a = 0; a < molt->atoms.nr; a++)
1611 {
1612 li->max_connect = std::max(li->max_connect,
1613 at2con[mt].index[a + 1] - at2con[mt].index[a]);
1614 }
1615 }
1616
1617 li->ncg_triangle = 0;
1618 bMoreThanTwoSeq = FALSE;
1619 for (mb = 0; mb < mtop->nmolblock; mb++)
1620 {
1621 molt = &mtop->moltype[mtop->molblock[mb].type];
1622
1623 li->ncg_triangle +=
1624 mtop->molblock[mb].nmol*
1625 count_triangle_constraints(molt->ilist,
1626 &at2con[mtop->molblock[mb].type]);
1627
1628 if (!bMoreThanTwoSeq &&
1629 more_than_two_sequential_constraints(molt->ilist, &at2con[mtop->molblock[mb].type]))
1630 {
1631 bMoreThanTwoSeq = TRUE;
1632 }
1633 }
1634
1635 /* Check if we need to communicate not only before LINCS,
1636 * but also before each iteration.
1637 * The check for only two sequential constraints is only
1638 * useful for the common case of H-bond only constraints.
1639 * With more effort we could also make it useful for small
1640 * molecules with nr. sequential constraints <= nOrder-1.
1641 */
1642 li->bCommIter = (bPLINCS && (li->nOrder < 1 || bMoreThanTwoSeq));
1643
1644 if (debug && bPLINCS)
1645 {
1646 fprintf(debug, "PLINCS communication before each iteration: %d\n",
1647 li->bCommIter);
1648 }
1649
1650 /* LINCS can run on any number of threads.
1651 * Currently the number is fixed for the whole simulation,
1652 * but it could be set in set_lincs().
1653 * The current constraint to task assignment code can create independent
1654 * tasks only when not more than two constraints are connected sequentially.
1655 */
1656 li->ntask = gmx_omp_nthreads_get(emntLINCS);
1657 li->bTaskDep = (li->ntask > 1 && bMoreThanTwoSeq);
1658 if (debug)
1659 {
1660 fprintf(debug, "LINCS: using %d threads, tasks are %sdependent\n",
1661 li->ntask, li->bTaskDep ? "" : "in");
1662 }
1663 if (li->ntask == 1)
1664 {
1665 snew(li->task, 1);
1666 }
1667 else
1668 {
1669 /* Allocate an extra elements for "task-overlap" constraints */
1670 snew(li->task, li->ntask + 1);
1671 }
1672 int th;
1673 #pragma omp parallel for num_threads(li->ntask)
1674 for (th = 0; th < li->ntask; th++)
1675 {
1676 /* Per thread SIMD load buffer for loading 2 simd_width rvecs */
1677 snew_aligned(li->task[th].simd_buf, 2*simd_width*DIM,
1678 align_bytes);
1679 }
1680
1681 if (bPLINCS || li->ncg_triangle > 0)
1682 {
1683 please_cite(fplog, "Hess2008a");
1684 }
1685 else
1686 {
1687 please_cite(fplog, "Hess97a");
1688 }
1689
1690 if (fplog)
1691 {
1692 fprintf(fplog, "The number of constraints is %d\n", li->ncg);
1693 if (bPLINCS)
1694 {
1695 fprintf(fplog, "There are inter charge-group constraints,\n"
1696 "will communicate selected coordinates each lincs iteration\n");
1697 }
1698 if (li->ncg_triangle > 0)
1699 {
1700 fprintf(fplog,
1701 "%d constraints are involved in constraint triangles,\n"
1702 "will apply an additional matrix expansion of order %d for couplings\n"
1703 "between constraints inside triangles\n",
1704 li->ncg_triangle, li->nOrder);
1705 }
1706 }
1707
1708 return li;
1709 }
1710
1711 /* Sets up the work division over the threads */
1712 static void lincs_thread_setup(struct gmx_lincsdata *li, int natoms)
1713 {
1714 lincs_task_t *li_m;
1715 int th;
1716 gmx_bitmask_t *atf;
1717 int a;
1718
1719 if (natoms > li->atf_nalloc)
1720 {
1721 li->atf_nalloc = over_alloc_large(natoms);
1722 srenew(li->atf, li->atf_nalloc);
1723 }
1724
1725 atf = li->atf;
1726 /* Clear the atom flags */
1727 for (a = 0; a < natoms; a++)
1728 {
1729 bitmask_clear(&atf[a]);
1730 }
1731
1732 if (li->ntask > BITMASK_SIZE)
1733 {
1734 gmx_fatal(FARGS, "More than %d threads is not supported for LINCS.", BITMASK_SIZE);
1735 }
1736
1737 for (th = 0; th < li->ntask; th++)
1738 {
1739 lincs_task_t *li_task;
1740 int b;
1741
1742 li_task = &li->task[th];
1743
1744 /* For each atom set a flag for constraints from each */
1745 for (b = li_task->b0; b < li_task->b1; b++)
1746 {
1747 bitmask_set_bit(&atf[li->bla[b*2 ]], th);
1748 bitmask_set_bit(&atf[li->bla[b*2 + 1]], th);
1749 }
1750 }
1751
1752 #pragma omp parallel for num_threads(li->ntask) schedule(static)
1753 for (th = 0; th < li->ntask; th++)
1754 {
1755 lincs_task_t *li_task;
1756 gmx_bitmask_t mask;
1757 int b;
1758
1759 li_task = &li->task[th];
1760
1761 if (li_task->b1 - li_task->b0 > li_task->ind_nalloc)
1762 {
1763 li_task->ind_nalloc = over_alloc_large(li_task->b1-li_task->b0);
1764 srenew(li_task->ind, li_task->ind_nalloc);
1765 srenew(li_task->ind_r, li_task->ind_nalloc);
1766 }
1767
1768 bitmask_init_low_bits(&mask, th);
1769
1770 li_task->nind = 0;
1771 li_task->nind_r = 0;
1772 for (b = li_task->b0; b < li_task->b1; b++)
1773 {
1774 /* We let the constraint with the lowest thread index
1775 * operate on atoms with constraints from multiple threads.
1776 */
1777 if (bitmask_is_disjoint(atf[li->bla[b*2]], mask) &&
1778 bitmask_is_disjoint(atf[li->bla[b*2+1]], mask))
1779 {
1780 /* Add the constraint to the local atom update index */
1781 li_task->ind[li_task->nind++] = b;
1782 }
1783 else
1784 {
1785 /* Add the constraint to the rest block */
1786 li_task->ind_r[li_task->nind_r++] = b;
1787 }
1788 }
1789 }
1790
1791 /* We need to copy all constraints which have not be assigned
1792 * to a thread to a separate list which will be handled by one thread.
1793 */
1794 li_m = &li->task[li->ntask];
1795
1796 li_m->nind = 0;
1797 for (th = 0; th < li->ntask; th++)
1798 {
1799 lincs_task_t *li_task;
1800 int b;
1801
1802 li_task = &li->task[th];
1803
1804 if (li_m->nind + li_task->nind_r > li_m->ind_nalloc)
1805 {
1806 li_m->ind_nalloc = over_alloc_large(li_m->nind+li_task->nind_r);
1807 srenew(li_m->ind, li_m->ind_nalloc);
1808 }
1809
1810 for (b = 0; b < li_task->nind_r; b++)
1811 {
1812 li_m->ind[li_m->nind++] = li_task->ind_r[b];
1813 }
1814
1815 if (debug)
1816 {
1817 fprintf(debug, "LINCS thread %d: %d constraints\n",
1818 th, li_task->nind);
1819 }
1820 }
1821
1822 if (debug)
1823 {
1824 fprintf(debug, "LINCS thread r: %d constraints\n",
1825 li_m->nind);
1826 }
1827 }
1828
1829 /* There is no realloc with alignment, so here we make one for reals.
1830 * Note that this function does not preserve the contents of the memory.
1831 */
1832 static void resize_real_aligned(real **ptr, int nelem)
1833 {
1834 sfree_aligned(*ptr);
1835 snew_aligned(*ptr, nelem, align_bytes);
1836 }
1837
1838 static void assign_constraint(struct gmx_lincsdata *li,
1839 int constraint_index,
1840 int a1, int a2,
1841 real lenA, real lenB,
1842 const t_blocka *at2con)
1843 {
1844 int con;
1845
1846 con = li->nc;
1847
1848 /* Make an mapping of local topology constraint index to LINCS index */
1849 li->con_index[constraint_index] = con;
1850
1851 li->bllen0[con] = lenA;
1852 li->ddist[con] = lenB - lenA;
1853 /* Set the length to the topology A length */
1854 li->bllen[con] = lenA;
1855 li->bla[2*con] = a1;
1856 li->bla[2*con+1] = a2;
1857
1858 /* Make space in the constraint connection matrix for constraints
1859 * connected to both end of the current constraint.
1860 */
1861 li->ncc +=
1862 at2con->index[a1 + 1] - at2con->index[a1] - 1 +
1863 at2con->index[a2 + 1] - at2con->index[a2] - 1;
1864
1865 li->blnr[con + 1] = li->ncc;
1866
1867 /* Increase the constraint count */
1868 li->nc++;
1869 }
1870
1871 /* Check if constraint with topology index constraint_index is connected
1872 * to other constraints, and if so add those connected constraints to our task.
1873 */
1874 static void check_assign_connected(struct gmx_lincsdata *li,
1875 const t_iatom *iatom,
1876 const t_idef *idef,
1877 int bDynamics,
1878 int a1, int a2,
1879 const t_blocka *at2con)
1880 {
1881 /* Currently this function only supports constraint groups
1882 * in which all constraints share at least one atom
1883 * (e.g. H-bond constraints).
1884 * Check both ends of the current constraint for
1885 * connected constraints. We need to assign those
1886 * to the same task.
1887 */
1888 int end;
1889
1890 for (end = 0; end < 2; end++)
1891 {
1892 int a, k;
1893
1894 a = (end == 0 ? a1 : a2);
1895
1896 for (k = at2con->index[a]; k < at2con->index[a + 1]; k++)
1897 {
1898 int cc;
1899
1900 cc = at2con->a[k];
1901 /* Check if constraint cc has not yet been assigned */
1902 if (li->con_index[cc] == -1)
1903 {
1904 int type;
1905 real lenA, lenB;
1906
1907 type = iatom[cc*3];
1908 lenA = idef->iparams[type].constr.dA;
1909 lenB = idef->iparams[type].constr.dB;
1910
1911 if (bDynamics || lenA != 0 || lenB != 0)
1912 {
1913 assign_constraint(li, cc, iatom[3*cc + 1], iatom[3*cc + 2], lenA, lenB, at2con);
1914 }
1915 }
1916 }
1917 }
1918 }
1919
1920 /* Check if constraint with topology index constraint_index is involved
1921 * in a constraint triangle, and if so add the other two constraints
1922 * in the triangle to our task.
1923 */
1924 static void check_assign_triangle(struct gmx_lincsdata *li,
1925 const t_iatom *iatom,
1926 const t_idef *idef,
1927 int bDynamics,
1928 int constraint_index,
1929 int a1, int a2,
1930 const t_blocka *at2con)
1931 {
1932 int nca, cc[32], ca[32], k;
1933 int c_triangle[2] = { -1, -1 };
1934
1935 nca = 0;
1936 for (k = at2con->index[a1]; k < at2con->index[a1 + 1]; k++)
1937 {
1938 int c;
1939
1940 c = at2con->a[k];
1941 if (c != constraint_index)
1942 {
1943 int aa1, aa2;
1944
1945 aa1 = iatom[c*3 + 1];
1946 aa2 = iatom[c*3 + 2];
1947 if (aa1 != a1)
1948 {
1949 cc[nca] = c;
1950 ca[nca] = aa1;
1951 nca++;
1952 }
1953 if (aa2 != a1)
1954 {
1955 cc[nca] = c;
1956 ca[nca] = aa2;
1957 nca++;
1958 }
1959 }
1960 }
1961
1962 for (k = at2con->index[a2]; k < at2con->index[a2 + 1]; k++)
1963 {
1964 int c;
1965
1966 c = at2con->a[k];
1967 if (c != constraint_index)
1968 {
1969 int aa1, aa2, i;
1970
1971 aa1 = iatom[c*3 + 1];
1972 aa2 = iatom[c*3 + 2];
1973 if (aa1 != a2)
1974 {
1975 for (i = 0; i < nca; i++)
1976 {
1977 if (aa1 == ca[i])
1978 {
1979 c_triangle[0] = cc[i];
1980 c_triangle[1] = c;
1981 }
1982 }
1983 }
1984 if (aa2 != a2)
1985 {
1986 for (i = 0; i < nca; i++)
1987 {
1988 if (aa2 == ca[i])
1989 {
1990 c_triangle[0] = cc[i];
1991 c_triangle[1] = c;
1992 }
1993 }
1994 }
1995 }
1996 }
1997
1998 if (c_triangle[0] >= 0)
1999 {
2000 int end;
2001
2002 for (end = 0; end < 2; end++)
2003 {
2004 /* Check if constraint c_triangle[end] has not yet been assigned */
2005 if (li->con_index[c_triangle[end]] == -1)
2006 {
2007 int i, type;
2008 real lenA, lenB;
2009
2010 i = c_triangle[end]*3;
2011 type = iatom[i];
2012 lenA = idef->iparams[type].constr.dA;
2013 lenB = idef->iparams[type].constr.dB;
2014
2015 if (bDynamics || lenA != 0 || lenB != 0)
2016 {
2017 assign_constraint(li, c_triangle[end], iatom[i + 1], iatom[i + 2], lenA, lenB, at2con);
2018 }
2019 }
2020 }
2021 }
2022 }
2023
2024 static void set_matrix_indices(struct gmx_lincsdata *li,
2025 const lincs_task_t *li_task,
2026 const t_blocka *at2con,
2027 gmx_bool bSortMatrix)
2028 {
2029 int b;
2030
2031 for (b = li_task->b0; b < li_task->b1; b++)
2032 {
2033 int a1, a2, i, k;
2034
2035 a1 = li->bla[b*2];
2036 a2 = li->bla[b*2 + 1];
2037
2038 i = li->blnr[b];
2039 for (k = at2con->index[a1]; k < at2con->index[a1 + 1]; k++)
2040 {
2041 int concon;
2042
2043 concon = li->con_index[at2con->a[k]];
2044 if (concon != b)
2045 {
2046 li->blbnb[i++] = concon;
2047 }
2048 }
2049 for (k = at2con->index[a2]; k < at2con->index[a2 + 1]; k++)
2050 {
2051 int concon;
2052
2053 concon = li->con_index[at2con->a[k]];
2054 if (concon != b)
2055 {
2056 li->blbnb[i++] = concon;
2057 }
2058 }
2059
2060 if (bSortMatrix)
2061 {
2062 /* Order the blbnb matrix to optimize memory access */
2063 qsort(&(li->blbnb[li->blnr[b]]), li->blnr[b + 1] - li->blnr[b],
2064 sizeof(li->blbnb[0]), int_comp);
2065 }
2066 }
2067 }
2068
2069 void set_lincs(t_idef *idef, t_mdatoms *md,
2070 gmx_bool bDynamics, t_commrec *cr,
2071 struct gmx_lincsdata *li)
2072 {
2073 int natoms, nflexcon;
2074 t_blocka at2con;
2075 t_iatom *iatom;
2076 int i, ncc_alloc_old, ncon_tot;
2077
2078 li->nc_real = 0;
2079 li->nc = 0;
2080 li->ncc = 0;
2081 /* Zero the thread index ranges.
2082 * Otherwise without local constraints we could return with old ranges.
2083 */
2084 for (i = 0; i < li->ntask; i++)
2085 {
2086 li->task[i].b0 = 0;
2087 li->task[i].b1 = 0;
2088 li->task[i].nind = 0;
2089 }
2090 if (li->ntask > 1)
2091 {
2092 li->task[li->ntask].nind = 0;
2093 }
2094
2095 /* This is the local topology, so there are only F_CONSTR constraints */
2096 if (idef->il[F_CONSTR].nr == 0)
2097 {
2098 /* There are no constraints,
2099 * we do not need to fill any data structures.
2100 */
2101 return;
2102 }
2103
2104 if (debug)
2105 {
2106 fprintf(debug, "Building the LINCS connectivity\n");
2107 }
2108
2109 if (DOMAINDECOMP(cr))
2110 {
2111 if (cr->dd->constraints)
2112 {
2113 int start;
2114
2115 dd_get_constraint_range(cr->dd, &start, &natoms);
2116 }
2117 else
2118 {
2119 natoms = cr->dd->nat_home;
2120 }
2121 }
2122 else
2123 {
2124 natoms = md->homenr;
2125 }
2126 at2con = make_at2con(0, natoms, idef->il, idef->iparams, bDynamics,
2127 &nflexcon);
2128
2129 ncon_tot = idef->il[F_CONSTR].nr/3;
2130
2131 /* Ensure we have enough padding for aligned loads for each thread */
2132 if (ncon_tot + li->ntask*simd_width > li->nc_alloc || li->nc_alloc == 0)
2133 {
2134 li->nc_alloc = over_alloc_dd(ncon_tot + li->ntask*simd_width);
2135 srenew(li->con_index, li->nc_alloc);
2136 resize_real_aligned(&li->bllen0, li->nc_alloc);
2137 resize_real_aligned(&li->ddist, li->nc_alloc);
2138 srenew(li->bla, 2*li->nc_alloc);
2139 resize_real_aligned(&li->blc, li->nc_alloc);
2140 resize_real_aligned(&li->blc1, li->nc_alloc);
2141 srenew(li->blnr, li->nc_alloc + 1);
2142 resize_real_aligned(&li->bllen, li->nc_alloc);
2143 srenew(li->tmpv, li->nc_alloc);
2144 if (DOMAINDECOMP(cr))
2145 {
2146 srenew(li->nlocat, li->nc_alloc);
2147 }
2148 resize_real_aligned(&li->tmp1, li->nc_alloc);
2149 resize_real_aligned(&li->tmp2, li->nc_alloc);
2150 resize_real_aligned(&li->tmp3, li->nc_alloc);
2151 resize_real_aligned(&li->tmp4, li->nc_alloc);
2152 resize_real_aligned(&li->mlambda, li->nc_alloc);
2153 }
2154
2155 iatom = idef->il[F_CONSTR].iatoms;
2156
2157 ncc_alloc_old = li->ncc_alloc;
2158 li->blnr[0] = li->ncc;
2159
2160 /* Assign the constraints for li->ntask LINCS tasks.
2161 * We target a uniform distribution of constraints over the tasks.
2162 * Note that when flexible constraints are present, but are removed here
2163 * (e.g. because we are doing EM) we get imbalance, but since that doesn't
2164 * happen during normal MD, that's ok.
2165 */
2166 int ncon_assign, ncon_target, con, th;
2167
2168 /* Determine the number of constraints we need to assign here */
2169 ncon_assign = ncon_tot;
2170 if (!bDynamics)
2171 {
2172 /* With energy minimization, flexible constraints are ignored
2173 * (and thus minimized, as they should be).
2174 */
2175 ncon_assign -= nflexcon;
2176 }
2177
2178 /* Set the target constraint count per task to exactly uniform,
2179 * this might be overridden below.
2180 */
2181 ncon_target = (ncon_assign + li->ntask - 1)/li->ntask;
2182
2183 /* Mark all constraints as unassigned by setting their index to -1 */
2184 for (con = 0; con < ncon_tot; con++)
2185 {
2186 li->con_index[con] = -1;
2187 }
2188
2189 con = 0;
2190 for (th = 0; th < li->ntask; th++)
2191 {
2192 lincs_task_t *li_task;
2193
2194 li_task = &li->task[th];
2195
2196 #ifdef LINCS_SIMD
2197 /* With indepedent tasks we likely have H-bond constraints or constraint
2198 * pairs. The connected constraints will be pulled into the task, so the
2199 * constraints per task will often exceed ncon_target.
2200 * Triangle constraints can also increase the count, but there are
2201 * relatively few of those, so we usually expect to get ncon_target.
2202 */
2203 if (li->bTaskDep)
2204 {
2205 /* We round ncon_target to a multiple of GMX_SIMD_WIDTH,
2206 * since otherwise a lot of operations can be wasted.
2207 * There are several ways to round here, we choose the one
2208 * that alternates block sizes, which helps with Intel HT.
2209 */
2210 ncon_target = ((ncon_assign*(th + 1))/li->ntask - li->nc_real + GMX_SIMD_REAL_WIDTH - 1) & ~(GMX_SIMD_REAL_WIDTH - 1);
2211 }
2212 #endif
2213
2214 /* Continue filling the arrays where we left off with the previous task,
2215 * including padding for SIMD.
2216 */
2217 li_task->b0 = li->nc;
2218
2219 while (con < ncon_tot && li->nc - li_task->b0 < ncon_target)
2220 {
2221 if (li->con_index[con] == -1)
2222 {
2223 int type, a1, a2;
2224 real lenA, lenB;
2225
2226 type = iatom[3*con];
2227 a1 = iatom[3*con + 1];
2228 a2 = iatom[3*con + 2];
2229 lenA = idef->iparams[type].constr.dA;
2230 lenB = idef->iparams[type].constr.dB;
2231 /* Skip the flexible constraints when not doing dynamics */
2232 if (bDynamics || lenA != 0 || lenB != 0)
2233 {
2234 assign_constraint(li, con, a1, a2, lenA, lenB, &at2con);
2235
2236 if (li->ntask > 1 && !li->bTaskDep)
2237 {
2238 /* We can generate independent tasks. Check if we
2239 * need to assign connected constraints to our task.
2240 */
2241 check_assign_connected(li, iatom, idef, bDynamics,
2242 a1, a2, &at2con);
2243 }
2244 if (li->ntask > 1 && li->ncg_triangle > 0)
2245 {
2246 /* Ensure constraints in one triangle are assigned
2247 * to the same task.
2248 */
2249 check_assign_triangle(li, iatom, idef, bDynamics,
2250 con, a1, a2, &at2con);
2251 }
2252 }
2253 }
2254
2255 con++;
2256 }
2257
2258 li_task->b1 = li->nc;
2259
2260 if (simd_width > 1)
2261 {
2262 /* Copy the last atom pair indices and lengths for constraints
2263 * up to a multiple of simd_width, such that we can do all
2264 * SIMD operations without having to worry about end effects.
2265 */
2266 int i, last;
2267
2268 li->nc = ((li_task->b1 + simd_width - 1)/simd_width)*simd_width;
2269 last = li_task->b1 - 1;
2270 for (i = li_task->b1; i < li->nc; i++)
2271 {
2272 li->bla[i*2 ] = li->bla[last*2 ];
2273 li->bla[i*2 + 1] = li->bla[last*2 + 1];
2274 li->bllen0[i] = li->bllen0[last];
2275 li->ddist[i] = li->ddist[last];
2276 li->bllen[i] = li->bllen[last];
2277 li->blnr[i + 1] = li->blnr[last + 1];
2278 }
2279 }
2280
2281 /* Keep track of how many constraints we assigned */
2282 li->nc_real += li_task->b1 - li_task->b0;
2283
2284 if (debug)
2285 {
2286 fprintf(debug, "LINCS task %d constraints %d - %d\n",
2287 th, li_task->b0, li_task->b1);
2288 }
2289 }
2290
2291 assert(li->nc_real == ncon_assign);
2292
2293 gmx_bool bSortMatrix;
2294
2295 /* Without DD we order the blbnb matrix to optimize memory access.
2296 * With DD the overhead of sorting is more than the gain during access.
2297 */
2298 bSortMatrix = !DOMAINDECOMP(cr);
2299
2300 if (li->ncc > li->ncc_alloc)
2301 {
2302 li->ncc_alloc = over_alloc_small(li->ncc);
2303 srenew(li->blbnb, li->ncc_alloc);
2304 }
2305
2306 #pragma omp parallel for num_threads(li->ntask) schedule(static)
2307 for (th = 0; th < li->ntask; th++)
2308 {
2309 lincs_task_t *li_task;
2310
2311 li_task = &li->task[th];
2312
2313 if (li->ncg_triangle > 0 &&
2314 li_task->b1 - li_task->b0 > li_task->tri_alloc)
2315 {
2316 /* This is allocating too much, but it is difficult to improve */
2317 li_task->tri_alloc = over_alloc_dd(li_task->b1 - li_task->b0);
2318 srenew(li_task->triangle, li_task->tri_alloc);
2319 srenew(li_task->tri_bits, li_task->tri_alloc);
2320 }
2321
2322 set_matrix_indices(li, li_task, &at2con, bSortMatrix);
2323 }
2324
2325 done_blocka(&at2con);
2326
2327 if (cr->dd == NULL)
2328 {
2329 /* Since the matrix is static, we should free some memory */
2330 li->ncc_alloc = li->ncc;
2331 srenew(li->blbnb, li->ncc_alloc);
2332 }
2333
2334 if (li->ncc_alloc > ncc_alloc_old)
2335 {
2336 srenew(li->blmf, li->ncc_alloc);
2337 srenew(li->blmf1, li->ncc_alloc);
2338 srenew(li->tmpncc, li->ncc_alloc);
2339 }
2340
2341 if (DOMAINDECOMP(cr) && dd_constraints_nlocalatoms(cr->dd) != NULL)
2342 {
2343 int *nlocat_dd;
2344
2345 nlocat_dd = dd_constraints_nlocalatoms(cr->dd);
2346
2347 /* Convert nlocat from local topology to LINCS constraint indexing */
2348 for (con = 0; con < ncon_tot; con++)
2349 {
2350 li->nlocat[li->con_index[con]] = nlocat_dd[con];
2351 }
2352 }
2353 else
2354 {
2355 li->nlocat = NULL;
2356 }
2357
2358 if (debug)
2359 {
2360 fprintf(debug, "Number of constraints is %d, padded %d, couplings %d\n",
2361 li->nc_real, li->nc, li->ncc);
2362 }
2363
2364 if (li->ntask > 1)
2365 {
2366 lincs_thread_setup(li, md->nr);
2367 }
2368
2369 set_lincs_matrix(li, md->invmass, md->lambda);
2370 }
2371
2372 static void lincs_warning(FILE *fplog,
2373 gmx_domdec_t *dd, rvec *x, rvec *xprime, t_pbc *pbc,
2374 int ncons, int *bla, real *bllen, real wangle,
2375 int maxwarn, int *warncount)
2376 {
2377 int b, i, j;
2378 rvec v0, v1;
2379 real wfac, d0, d1, cosine;
2380 char buf[STRLEN];
2381
2382 wfac = cos(DEG2RAD*wangle);
2383
2384 sprintf(buf, "bonds that rotated more than %g degrees:\n"
2385 " atom 1 atom 2 angle previous, current, constraint length\n",
2386 wangle);
2387 fprintf(stderr, "%s", buf);
2388 if (fplog)
2389 {
2390 fprintf(fplog, "%s", buf);
2391 }
2392
2393 for (b = 0; b < ncons; b++)
2394 {
2395 i = bla[2*b];
2396 j = bla[2*b+1];
2397 if (pbc)
2398 {
2399 pbc_dx_aiuc(pbc, x[i], x[j], v0);
2400 pbc_dx_aiuc(pbc, xprime[i], xprime[j], v1);
2401 }
2402 else
2403 {
2404 rvec_sub(x[i], x[j], v0);
2405 rvec_sub(xprime[i], xprime[j], v1);
2406 }
2407 d0 = norm(v0);
2408 d1 = norm(v1);
2409 cosine = iprod(v0, v1)/(d0*d1);
2410 if (cosine < wfac)
2411 {
2412 sprintf(buf, " %6d %6d %5.1f %8.4f %8.4f %8.4f\n",
2413 ddglatnr(dd, i), ddglatnr(dd, j),
2414 RAD2DEG*acos(cosine), d0, d1, bllen[b]);
2415 fprintf(stderr, "%s", buf);
2416 if (fplog)
2417 {
2418 fprintf(fplog, "%s", buf);
2419 }
2420 if (!gmx_isfinite(d1))
2421 {
2422 gmx_fatal(FARGS, "Bond length not finite.");
2423 }
2424
2425 (*warncount)++;
2426 }
2427 }
2428 if (*warncount > maxwarn)
2429 {
2430 too_many_constraint_warnings(econtLINCS, *warncount);
2431 }
2432 }
2433
2434 static void cconerr(const struct gmx_lincsdata *lincsd,
2435 rvec *x, t_pbc *pbc,
2436 real *ncons_loc, real *ssd, real *max, int *imax)
2437 {
2438 const int *bla, *nlocat;
2439 const real *bllen;
2440 real ma, ssd2;
2441 int count, im, task;
2442
2443 bla = lincsd->bla;
2444 bllen = lincsd->bllen;
2445 nlocat = lincsd->nlocat;
2446
2447 ma = 0;
2448 ssd2 = 0;
2449 im = 0;
2450 count = 0;
2451 for (task = 0; task < lincsd->ntask; task++)
2452 {
2453 int b;
2454
2455 for (b = lincsd->task[task].b0; b < lincsd->task[task].b1; b++)
2456 {
2457 real len, d, r2;
2458 rvec dx;
2459
2460 if (pbc)
2461 {
2462 pbc_dx_aiuc(pbc, x[bla[2*b]], x[bla[2*b+1]], dx);
2463 }
2464 else
2465 {
2466 rvec_sub(x[bla[2*b]], x[bla[2*b+1]], dx);
2467 }
2468 r2 = norm2(dx);
2469 len = r2*gmx_invsqrt(r2);
2470 d = fabs(len/bllen[b]-1);
2471 if (d > ma && (nlocat == NULL || nlocat[b]))
2472 {
2473 ma = d;
2474 im = b;
2475 }
2476 if (nlocat == NULL)
2477 {
2478 ssd2 += d*d;
2479 count++;
2480 }
2481 else
2482 {
2483 ssd2 += nlocat[b]*d*d;
2484 count += nlocat[b];
2485 }
2486 }
2487 }
2488
2489 *ncons_loc = (nlocat ? 0.5 : 1)*count;
2490 *ssd = (nlocat ? 0.5 : 1)*ssd2;
2491 *max = ma;
2492 *imax = im;
2493 }
2494
2495 gmx_bool constrain_lincs(FILE *fplog, gmx_bool bLog, gmx_bool bEner,
2496 t_inputrec *ir,
2497 gmx_int64_t step,
2498 struct gmx_lincsdata *lincsd, t_mdatoms *md,
2499 t_commrec *cr,
2500 rvec *x, rvec *xprime, rvec *min_proj,
2501 matrix box, t_pbc *pbc,
2502 real lambda, real *dvdlambda,
2503 real invdt, rvec *v,
2504 gmx_bool bCalcVir, tensor vir_r_m_dr,
2505 int econq,
2506 t_nrnb *nrnb,
2507 int maxwarn, int *warncount)
2508 {
2509 gmx_bool bCalcDHDL;
2510 char buf[STRLEN], buf2[22], buf3[STRLEN];
2511 int i, p_imax;
2512 real ncons_loc, p_ssd, p_max = 0;
2513 rvec dx;
2514 gmx_bool bOK, bWarn;
2515
2516 bOK = TRUE;
2517
2518 /* This boolean should be set by a flag passed to this routine.
2519 * We can also easily check if any constraint length is changed,
2520 * if not dH/dlambda=0 and we can also set the boolean to FALSE.
2521 */
2522 bCalcDHDL = (ir->efep != efepNO && dvdlambda != NULL);
2523
2524 if (lincsd->nc == 0 && cr->dd == NULL)
2525 {
2526 if (bLog || bEner)
2527 {
2528 lincsd->rmsd_data[0] = 0;
2529 if (ir->eI == eiSD2 && v == NULL)
2530 {
2531 i = 2;
2532 }
2533 else
2534 {
2535 i = 1;
2536 }
2537 lincsd->rmsd_data[i] = 0;
2538 }
2539
2540 return bOK;
2541 }
2542
2543 if (econq == econqCoord)
2544 {
2545 /* We can't use bCalcDHDL here, since NULL can be passed for dvdlambda
2546 * also with efep!=fepNO.
2547 */
2548 if (ir->efep != efepNO)
2549 {
2550 if (md->nMassPerturbed && lincsd->matlam != md->lambda)
2551 {
2552 set_lincs_matrix(lincsd, md->invmass, md->lambda);
2553 }
2554
2555 for (i = 0; i < lincsd->nc; i++)
2556 {
2557 lincsd->bllen[i] = lincsd->bllen0[i] + lambda*lincsd->ddist[i];
2558 }
2559 }
2560
2561 if (lincsd->ncg_flex)
2562 {
2563 /* Set the flexible constraint lengths to the old lengths */
2564 if (pbc != NULL)
2565 {
2566 for (i = 0; i < lincsd->nc; i++)
2567 {
2568 if (lincsd->bllen[i] == 0)
2569 {
2570 pbc_dx_aiuc(pbc, x[lincsd->bla[2*i]], x[lincsd->bla[2*i+1]], dx);
2571 lincsd->bllen[i] = norm(dx);
2572 }
2573 }
2574 }
2575 else
2576 {
2577 for (i = 0; i < lincsd->nc; i++)
2578 {
2579 if (lincsd->bllen[i] == 0)
2580 {
2581 lincsd->bllen[i] =
2582 sqrt(distance2(x[lincsd->bla[2*i]],
2583 x[lincsd->bla[2*i+1]]));
2584 }
2585 }
2586 }
2587 }
2588
2589 if (bLog && fplog)
2590 {
2591 cconerr(lincsd, xprime, pbc,
2592 &ncons_loc, &p_ssd, &p_max, &p_imax);
2593 }
2594
2595 /* This bWarn var can be updated by multiple threads
2596 * at the same time. But as we only need to detect
2597 * if a warning occured or not, this is not an issue.
2598 */
2599 bWarn = FALSE;
2600
2601 /* The OpenMP parallel region of constrain_lincs for coords */
2602 #pragma omp parallel num_threads(lincsd->ntask)
2603 {
2604 int th = gmx_omp_get_thread_num();
2605
2606 clear_mat(lincsd->task[th].vir_r_m_dr);
2607
2608 do_lincs(x, xprime, box, pbc, lincsd, th,
2609 md->invmass, cr,
2610 bCalcDHDL,
2611 ir->LincsWarnAngle, &bWarn,
2612 invdt, v, bCalcVir,
2613 th == 0 ? vir_r_m_dr : lincsd->task[th].vir_r_m_dr);
2614 }
2615
2616 if (bLog && fplog && lincsd->nc > 0)
2617 {
2618 fprintf(fplog, " Rel. Constraint Deviation: RMS MAX between atoms\n");
2619 fprintf(fplog, " Before LINCS %.6f %.6f %6d %6d\n",
2620 sqrt(p_ssd/ncons_loc), p_max,
2621 ddglatnr(cr->dd, lincsd->bla[2*p_imax]),
2622 ddglatnr(cr->dd, lincsd->bla[2*p_imax+1]));
2623 }
2624 if (bLog || bEner)
2625 {
2626 cconerr(lincsd, xprime, pbc,
2627 &ncons_loc, &p_ssd, &p_max, &p_imax);
2628 /* Check if we are doing the second part of SD */
2629 if (ir->eI == eiSD2 && v == NULL)
2630 {
2631 i = 2;
2632 }
2633 else
2634 {
2635 i = 1;
2636 }
2637 lincsd->rmsd_data[0] = ncons_loc;
2638 lincsd->rmsd_data[i] = p_ssd;
2639 }
2640 else
2641 {
2642 lincsd->rmsd_data[0] = 0;
2643 lincsd->rmsd_data[1] = 0;
2644 lincsd->rmsd_data[2] = 0;
2645 }
2646 if (bLog && fplog && lincsd->nc > 0)
2647 {
2648 fprintf(fplog,
2649 " After LINCS %.6f %.6f %6d %6d\n\n",
2650 sqrt(p_ssd/ncons_loc), p_max,
2651 ddglatnr(cr->dd, lincsd->bla[2*p_imax]),
2652 ddglatnr(cr->dd, lincsd->bla[2*p_imax+1]));
2653 }
2654
2655 if (bWarn)
2656 {
2657 if (maxwarn >= 0)
2658 {
2659 cconerr(lincsd, xprime, pbc,
2660 &ncons_loc, &p_ssd, &p_max, &p_imax);
2661 if (MULTISIM(cr))
2662 {
2663 sprintf(buf3, " in simulation %d", cr->ms->sim);
2664 }
2665 else
2666 {
2667 buf3[0] = 0;
2668 }
2669 sprintf(buf, "\nStep %s, time %g (ps) LINCS WARNING%s\n"
2670 "relative constraint deviation after LINCS:\n"
2671 "rms %.6f, max %.6f (between atoms %d and %d)\n",
2672 gmx_step_str(step, buf2), ir->init_t+step*ir->delta_t,
2673 buf3,
2674 sqrt(p_ssd/ncons_loc), p_max,
2675 ddglatnr(cr->dd, lincsd->bla[2*p_imax]),
2676 ddglatnr(cr->dd, lincsd->bla[2*p_imax+1]));
2677 if (fplog)
2678 {
2679 fprintf(fplog, "%s", buf);
2680 }
2681 fprintf(stderr, "%s", buf);
2682 lincs_warning(fplog, cr->dd, x, xprime, pbc,
2683 lincsd->nc, lincsd->bla, lincsd->bllen,
2684 ir->LincsWarnAngle, maxwarn, warncount);
2685 }
2686 bOK = (p_max < 0.5);
2687 }
2688
2689 if (lincsd->ncg_flex)
2690 {
2691 for (i = 0; (i < lincsd->nc); i++)
2692 {
2693 if (lincsd->bllen0[i] == 0 && lincsd->ddist[i] == 0)
2694 {
2695 lincsd->bllen[i] = 0;
2696 }
2697 }
2698 }
2699 }
2700 else
2701 {
2702 /* The OpenMP parallel region of constrain_lincs for derivatives */
2703 #pragma omp parallel num_threads(lincsd->ntask)
2704 {
2705 int th = gmx_omp_get_thread_num();
2706
2707 do_lincsp(x, xprime, min_proj, pbc, lincsd, th,
2708 md->invmass, econq, bCalcDHDL,
2709 bCalcVir, th == 0 ? vir_r_m_dr : lincsd->task[th].vir_r_m_dr);
2710 }
2711 }
2712
2713 if (bCalcDHDL)
2714 {
2715 /* Reduce the dH/dlambda contributions over the threads */
2716 real dhdlambda;
2717 int th;
2718
2719 dhdlambda = 0;
2720 for (th = 0; th < lincsd->ntask; th++)
2721 {
2722 dhdlambda += lincsd->task[th].dhdlambda;
2723 }
2724 if (econqCoord)
2725 {
2726 /* dhdlambda contains dH/dlambda*dt^2, correct for this */
2727 /* TODO This should probably use invdt, so that sd integrator scaling works properly */
2728 dhdlambda /= ir->delta_t*ir->delta_t;
2729 }
2730 *dvdlambda += dhdlambda;
2731 }
2732
2733 if (bCalcVir && lincsd->ntask > 1)
2734 {
2735 for (i = 1; i < lincsd->ntask; i++)
2736 {
2737 m_add(vir_r_m_dr, lincsd->task[i].vir_r_m_dr, vir_r_m_dr);
2738 }
2739 }
2740
2741 /* count assuming nit=1 */
2742 inc_nrnb(nrnb, eNR_LINCS, lincsd->nc_real);
2743 inc_nrnb(nrnb, eNR_LINCSMAT, (2+lincsd->nOrder)*lincsd->ncc);
2744 if (lincsd->ntriangle > 0)
2745 {
2746 inc_nrnb(nrnb, eNR_LINCSMAT, lincsd->nOrder*lincsd->ncc_triangle);
2747 }
2748 if (v)
2749 {
2750 inc_nrnb(nrnb, eNR_CONSTR_V, lincsd->nc_real*2);
2751 }
2752 if (bCalcVir)
2753 {
2754 inc_nrnb(nrnb, eNR_CONSTR_VIR, lincsd->nc_real);
2755 }
2756
2757 return bOK;
2758 }
The ultimate molecular dynamics simulation package