2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014,2015,2016 by the GROMACS development team.
5 * Copyright (c) 2017,2018,2019,2020, by the GROMACS development team, led by
6 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
7 * and including many others, as listed in the AUTHORS file in the
8 * top-level source directory and at http://www.gromacs.org.
10 * GROMACS is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU Lesser General Public License
12 * as published by the Free Software Foundation; either version 2.1
13 * of the License, or (at your option) any later version.
15 * GROMACS is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * Lesser General Public License for more details.
20 * You should have received a copy of the GNU Lesser General Public
21 * License along with GROMACS; if not, see
22 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
23 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
25 * If you want to redistribute modifications to GROMACS, please
26 * consider that scientific software is very special. Version
27 * control is crucial - bugs must be traceable. We will be happy to
28 * consider code for inclusion in the official distribution, but
29 * derived work must not be called official GROMACS. Details are found
30 * in the README & COPYING files - if they are missing, get the
31 * official version at http://www.gromacs.org.
33 * To help us fund GROMACS development, we humbly ask that you cite
34 * the research papers on the package. Check out http://www.gromacs.org.
40 * Implements the Grid class.
42 * \author Berk Hess <hess@kth.se>
43 * \ingroup module_nbnxm
55 #include "gromacs/math/utilities.h"
56 #include "gromacs/math/vec.h"
57 #include "gromacs/mdlib/gmx_omp_nthreads.h"
58 #include "gromacs/mdlib/updategroupscog.h"
59 #include "gromacs/mdtypes/forcerec.h" // only for GET_CGINFO_*
60 #include "gromacs/nbnxm/atomdata.h"
61 #include "gromacs/simd/simd.h"
62 #include "gromacs/simd/vector_operations.h"
64 #include "boundingboxes.h"
65 #include "gridsetdata.h"
66 #include "nbnxm_geometry.h"
67 #include "pairlistparams.h"
72 Grid::Geometry::Geometry(const PairlistType pairlistType) :
73 isSimple(pairlistType != PairlistType::HierarchicalNxN),
74 numAtomsICluster(IClusterSizePerListType[pairlistType]),
75 numAtomsJCluster(JClusterSizePerListType[pairlistType]),
76 numAtomsPerCell((isSimple ? 1 : c_gpuNumClusterPerCell) * numAtomsICluster),
77 numAtomsICluster2Log(get_2log(numAtomsICluster))
81 Grid::Grid(const PairlistType pairlistType, const bool& haveFep) :
82 geometry_(pairlistType),
87 /*! \brief Returns the atom density (> 0) of a rectangular grid */
88 static real gridAtomDensity(int numAtoms, const rvec lowerCorner, const rvec upperCorner)
94 /* To avoid zero density we use a minimum of 1 atom */
98 rvec_sub(upperCorner, lowerCorner, size);
100 return static_cast<real>(numAtoms) / (size[XX] * size[YY] * size[ZZ]);
103 void Grid::setDimensions(const int ddZone,
105 gmx::RVec lowerCorner,
106 gmx::RVec upperCorner,
108 const real maxAtomGroupRadius,
110 gmx::PinningPolicy pinningPolicy)
112 /* We allow passing lowerCorner=upperCorner, in which case we need to
113 * create a finite sized bounding box to avoid division by zero.
114 * We use a minimum size such that the volume fits in float with some
115 * margin for computing and using the atom number density.
117 constexpr real c_minimumGridSize = 1e-10;
118 for (int d = 0; d < DIM; d++)
120 GMX_ASSERT(upperCorner[d] >= lowerCorner[d],
121 "Upper corner should be larger than the lower corner");
122 if (upperCorner[d] - lowerCorner[d] < c_minimumGridSize)
124 /* Ensure we apply a correction to the bounding box */
126 std::max(std::abs(lowerCorner[d]) * GMX_REAL_EPS, 0.5_real * c_minimumGridSize);
127 lowerCorner[d] -= correction;
128 upperCorner[d] += correction;
132 /* For the home zone we compute the density when not set (=-1) or when =0 */
133 if (ddZone == 0 && atomDensity <= 0)
135 atomDensity = gridAtomDensity(numAtoms, lowerCorner, upperCorner);
138 dimensions_.atomDensity = atomDensity;
139 dimensions_.maxAtomGroupRadius = maxAtomGroupRadius;
142 rvec_sub(upperCorner, lowerCorner, size);
144 if (numAtoms > geometry_.numAtomsPerCell)
146 GMX_ASSERT(atomDensity > 0, "With one or more atoms, the density should be positive");
148 /* target cell length */
151 if (geometry_.isSimple)
153 /* To minimize the zero interactions, we should make
154 * the largest of the i/j cell cubic.
156 int numAtomsInCell = std::max(geometry_.numAtomsICluster, geometry_.numAtomsJCluster);
158 /* Approximately cubic cells */
159 real tlen = std::cbrt(numAtomsInCell / atomDensity);
165 /* Approximately cubic sub cells */
166 real tlen = std::cbrt(geometry_.numAtomsICluster / atomDensity);
167 tlen_x = tlen * c_gpuNumClusterPerCellX;
168 tlen_y = tlen * c_gpuNumClusterPerCellY;
170 /* We round ncx and ncy down, because we get less cell pairs
171 * in the pairlist when the fixed cell dimensions (x,y) are
172 * larger than the variable one (z) than the other way around.
174 dimensions_.numCells[XX] = std::max(1, static_cast<int>(size[XX] / tlen_x));
175 dimensions_.numCells[YY] = std::max(1, static_cast<int>(size[YY] / tlen_y));
179 dimensions_.numCells[XX] = 1;
180 dimensions_.numCells[YY] = 1;
183 for (int d = 0; d < DIM - 1; d++)
185 dimensions_.cellSize[d] = size[d] / dimensions_.numCells[d];
186 dimensions_.invCellSize[d] = 1 / dimensions_.cellSize[d];
191 /* This is a non-home zone, add an extra row of cells
192 * for particles communicated for bonded interactions.
193 * These can be beyond the cut-off. It doesn't matter where
194 * they end up on the grid, but for performance it's better
195 * if they don't end up in cells that can be within cut-off range.
197 dimensions_.numCells[XX]++;
198 dimensions_.numCells[YY]++;
201 /* We need one additional cell entry for particles moved by DD */
202 cxy_na_.resize(numColumns() + 1);
203 cxy_ind_.resize(numColumns() + 2);
204 changePinningPolicy(&cxy_na_, pinningPolicy);
205 changePinningPolicy(&cxy_ind_, pinningPolicy);
207 /* Worst case scenario of 1 atom in each last cell */
209 if (geometry_.numAtomsJCluster <= geometry_.numAtomsICluster)
211 maxNumCells = numAtoms / geometry_.numAtomsPerCell + numColumns();
215 maxNumCells = numAtoms / geometry_.numAtomsPerCell
216 + numColumns() * geometry_.numAtomsJCluster / geometry_.numAtomsICluster;
219 if (!geometry_.isSimple)
221 numClusters_.resize(maxNumCells);
223 bbcz_.resize(maxNumCells);
225 /* This resize also zeros the contents, this avoid possible
226 * floating exceptions in SIMD with the unused bb elements.
228 if (geometry_.isSimple)
230 bb_.resize(maxNumCells);
235 pbb_.resize(packedBoundingBoxesIndex(maxNumCells * c_gpuNumClusterPerCell));
237 bb_.resize(maxNumCells * c_gpuNumClusterPerCell);
241 if (geometry_.numAtomsJCluster == geometry_.numAtomsICluster)
247 GMX_ASSERT(geometry_.isSimple, "Only CPU lists should have different i/j cluster sizes");
249 bbjStorage_.resize(maxNumCells * geometry_.numAtomsICluster / geometry_.numAtomsJCluster);
253 flags_.resize(maxNumCells);
256 fep_.resize(maxNumCells * geometry_.numAtomsPerCell / geometry_.numAtomsICluster);
259 copy_rvec(lowerCorner, dimensions_.lowerCorner);
260 copy_rvec(upperCorner, dimensions_.upperCorner);
261 copy_rvec(size, dimensions_.gridSize);
264 /* We need to sort paricles in grid columns on z-coordinate.
265 * As particle are very often distributed homogeneously, we use a sorting
266 * algorithm similar to pigeonhole sort. We multiply the z-coordinate
267 * by a factor, cast to an int and try to store in that hole. If the hole
268 * is full, we move this or another particle. A second pass is needed to make
269 * contiguous elements. SORT_GRID_OVERSIZE is the ratio of holes to particles.
270 * 4 is the optimal value for homogeneous particle distribution and allows
271 * for an O(#particles) sort up till distributions were all particles are
272 * concentrated in 1/4 of the space. No NlogN fallback is implemented,
273 * as it can be expensive to detect imhomogeneous particle distributions.
275 /*! \brief Ratio of grid cells to atoms */
276 static constexpr int c_sortGridRatio = 4;
277 /*! \brief Maximum ratio of holes used, in the worst case all particles
278 * end up in the last hole and we need num. atoms extra holes at the end.
280 static constexpr int c_sortGridMaxSizeFactor = c_sortGridRatio + 1;
282 /*! \brief Sorts particle index a on coordinates x along dim.
284 * Backwards tells if we want decreasing iso increasing coordinates.
285 * h0 is the minimum of the coordinate range.
286 * invh is the 1/length of the sorting range.
287 * n_per_h (>=n) is the expected average number of particles per 1/invh
288 * sort is the sorting work array.
289 * sort should have a size of at least n_per_h*c_sortGridRatio + n,
290 * or easier, allocate at least n*c_sortGridMaxSizeFactor elements.
292 static void sort_atoms(int dim,
294 int gmx_unused dd_zone,
295 bool gmx_unused relevantAtomsAreWithinGridBounds,
298 gmx::ArrayRef<const gmx::RVec> x,
302 gmx::ArrayRef<int> sort)
310 GMX_ASSERT(n <= n_per_h, "We require n <= n_per_h");
312 /* Transform the inverse range height into the inverse hole height */
313 invh *= n_per_h * c_sortGridRatio;
315 /* Set nsort to the maximum possible number of holes used.
316 * In worst case all n elements end up in the last bin.
318 int nsort = n_per_h * c_sortGridRatio + n;
320 /* Determine the index range used, so we can limit it for the second pass */
321 int zi_min = INT_MAX;
324 /* Sort the particles using a simple index sort */
325 for (int i = 0; i < n; i++)
327 /* The cast takes care of float-point rounding effects below zero.
328 * This code assumes particles are less than 1/SORT_GRID_OVERSIZE
329 * times the box height out of the box.
331 int zi = static_cast<int>((x[a[i]][dim] - h0) * invh);
334 /* As we can have rounding effect, we use > iso >= here */
335 if (relevantAtomsAreWithinGridBounds && (zi < 0 || (dd_zone == 0 && zi > n_per_h * c_sortGridRatio)))
338 "(int)((x[%d][%c]=%f - %f)*%f) = %d, not in 0 - %d*%d\n",
354 /* In a non-local domain, particles communcated for bonded interactions
355 * can be far beyond the grid size, which is set by the non-bonded
356 * cut-off distance. We sort such particles into the last cell.
358 if (zi > n_per_h * c_sortGridRatio)
360 zi = n_per_h * c_sortGridRatio;
363 /* Ideally this particle should go in sort cell zi,
364 * but that might already be in use,
365 * in that case find the first empty cell higher up
370 zi_min = std::min(zi_min, zi);
371 zi_max = std::max(zi_max, zi);
375 /* We have multiple atoms in the same sorting slot.
376 * Sort on real z for minimal bounding box size.
377 * There is an extra check for identical z to ensure
378 * well-defined output order, independent of input order
379 * to ensure binary reproducibility after restarts.
382 && (x[a[i]][dim] > x[sort[zi]][dim]
383 || (x[a[i]][dim] == x[sort[zi]][dim] && a[i] > sort[zi])))
390 /* Shift all elements by one slot until we find an empty slot */
393 while (sort[zim] >= 0)
401 zi_max = std::max(zi_max, zim);
404 zi_max = std::max(zi_max, zi);
411 for (int zi = 0; zi < nsort; zi++)
422 for (int zi = zi_max; zi >= zi_min; zi--)
433 gmx_incons("Lost particles while sorting");
438 //! Returns double up to one least significant float bit smaller than x
439 static double R2F_D(const float x)
441 return static_cast<float>(x >= 0 ? (1 - GMX_FLOAT_EPS) * x : (1 + GMX_FLOAT_EPS) * x);
443 //! Returns double up to one least significant float bit larger than x
444 static double R2F_U(const float x)
446 return static_cast<float>(x >= 0 ? (1 + GMX_FLOAT_EPS) * x : (1 - GMX_FLOAT_EPS) * x);
450 static float R2F_D(const float x)
455 static float R2F_U(const float x)
461 //! Computes the bounding box for na coordinates in order x,y,z, bb order xyz0
462 static void calc_bounding_box(int na, int stride, const real* x, BoundingBox* bb)
465 real xl, xh, yl, yh, zl, zh;
475 for (int j = 1; j < na; j++)
477 xl = std::min(xl, x[i + XX]);
478 xh = std::max(xh, x[i + XX]);
479 yl = std::min(yl, x[i + YY]);
480 yh = std::max(yh, x[i + YY]);
481 zl = std::min(zl, x[i + ZZ]);
482 zh = std::max(zh, x[i + ZZ]);
485 /* Note: possible double to float conversion here */
486 bb->lower.x = R2F_D(xl);
487 bb->lower.y = R2F_D(yl);
488 bb->lower.z = R2F_D(zl);
489 bb->upper.x = R2F_U(xh);
490 bb->upper.y = R2F_U(yh);
491 bb->upper.z = R2F_U(zh);
494 /*! \brief Computes the bounding box for na packed coordinates, bb order xyz0 */
495 static void calc_bounding_box_x_x4(int na, const real* x, BoundingBox* bb)
497 real xl, xh, yl, yh, zl, zh;
499 xl = x[XX * c_packX4];
500 xh = x[XX * c_packX4];
501 yl = x[YY * c_packX4];
502 yh = x[YY * c_packX4];
503 zl = x[ZZ * c_packX4];
504 zh = x[ZZ * c_packX4];
505 for (int j = 1; j < na; j++)
507 xl = std::min(xl, x[j + XX * c_packX4]);
508 xh = std::max(xh, x[j + XX * c_packX4]);
509 yl = std::min(yl, x[j + YY * c_packX4]);
510 yh = std::max(yh, x[j + YY * c_packX4]);
511 zl = std::min(zl, x[j + ZZ * c_packX4]);
512 zh = std::max(zh, x[j + ZZ * c_packX4]);
514 /* Note: possible double to float conversion here */
515 bb->lower.x = R2F_D(xl);
516 bb->lower.y = R2F_D(yl);
517 bb->lower.z = R2F_D(zl);
518 bb->upper.x = R2F_U(xh);
519 bb->upper.y = R2F_U(yh);
520 bb->upper.z = R2F_U(zh);
523 /*! \brief Computes the bounding box for na coordinates, bb order xyz0 */
524 static void calc_bounding_box_x_x8(int na, const real* x, BoundingBox* bb)
526 real xl, xh, yl, yh, zl, zh;
528 xl = x[XX * c_packX8];
529 xh = x[XX * c_packX8];
530 yl = x[YY * c_packX8];
531 yh = x[YY * c_packX8];
532 zl = x[ZZ * c_packX8];
533 zh = x[ZZ * c_packX8];
534 for (int j = 1; j < na; j++)
536 xl = std::min(xl, x[j + XX * c_packX8]);
537 xh = std::max(xh, x[j + XX * c_packX8]);
538 yl = std::min(yl, x[j + YY * c_packX8]);
539 yh = std::max(yh, x[j + YY * c_packX8]);
540 zl = std::min(zl, x[j + ZZ * c_packX8]);
541 zh = std::max(zh, x[j + ZZ * c_packX8]);
543 /* Note: possible double to float conversion here */
544 bb->lower.x = R2F_D(xl);
545 bb->lower.y = R2F_D(yl);
546 bb->lower.z = R2F_D(zl);
547 bb->upper.x = R2F_U(xh);
548 bb->upper.y = R2F_U(yh);
549 bb->upper.z = R2F_U(zh);
552 /*! \brief Computes the bounding box for na packed coordinates, bb order xyz0 */
553 gmx_unused static void calc_bounding_box_x_x4_halves(int na, const real* x, BoundingBox* bb, BoundingBox* bbj)
555 // TODO: During SIMDv2 transition only some archs use namespace (remove when done)
558 calc_bounding_box_x_x4(std::min(na, 2), x, bbj);
562 calc_bounding_box_x_x4(std::min(na - 2, 2), x + (c_packX4 >> 1), bbj + 1);
566 /* Set the "empty" bounding box to the same as the first one,
567 * so we don't need to treat special cases in the rest of the code.
569 #if NBNXN_SEARCH_BB_SIMD4
570 store4(bbj[1].lower.ptr(), load4(bbj[0].lower.ptr()));
571 store4(bbj[1].upper.ptr(), load4(bbj[0].upper.ptr()));
577 #if NBNXN_SEARCH_BB_SIMD4
578 store4(bb->lower.ptr(), min(load4(bbj[0].lower.ptr()), load4(bbj[1].lower.ptr())));
579 store4(bb->upper.ptr(), max(load4(bbj[0].upper.ptr()), load4(bbj[1].upper.ptr())));
582 bb->lower = BoundingBox::Corner::min(bbj[0].lower, bbj[1].lower);
583 bb->upper = BoundingBox::Corner::max(bbj[0].upper, bbj[1].upper);
590 /*! \brief Computes the bounding box for na coordinates in order xyz, bb order xxxxyyyyzzzz */
591 static void calc_bounding_box_xxxx(int na, int stride, const real* x, float* bb)
594 real xl, xh, yl, yh, zl, zh;
604 for (int j = 1; j < na; j++)
606 xl = std::min(xl, x[i + XX]);
607 xh = std::max(xh, x[i + XX]);
608 yl = std::min(yl, x[i + YY]);
609 yh = std::max(yh, x[i + YY]);
610 zl = std::min(zl, x[i + ZZ]);
611 zh = std::max(zh, x[i + ZZ]);
614 /* Note: possible double to float conversion here */
615 bb[0 * c_packedBoundingBoxesDimSize] = R2F_D(xl);
616 bb[1 * c_packedBoundingBoxesDimSize] = R2F_D(yl);
617 bb[2 * c_packedBoundingBoxesDimSize] = R2F_D(zl);
618 bb[3 * c_packedBoundingBoxesDimSize] = R2F_U(xh);
619 bb[4 * c_packedBoundingBoxesDimSize] = R2F_U(yh);
620 bb[5 * c_packedBoundingBoxesDimSize] = R2F_U(zh);
623 #endif /* NBNXN_BBXXXX */
625 #if NBNXN_SEARCH_SIMD4_FLOAT_X_BB
627 /*! \brief Computes the bounding box for na coordinates in order xyz?, bb order xyz0 */
628 static void calc_bounding_box_simd4(int na, const float* x, BoundingBox* bb)
630 // TODO: During SIMDv2 transition only some archs use namespace (remove when done)
633 static_assert(sizeof(BoundingBox::Corner) == GMX_SIMD4_WIDTH * sizeof(float),
634 "The Corner struct should hold exactly 4 floats");
636 Simd4Float bb_0_S = load4(x);
637 Simd4Float bb_1_S = bb_0_S;
639 for (int i = 1; i < na; i++)
641 Simd4Float x_S = load4(x + i * GMX_SIMD4_WIDTH);
642 bb_0_S = min(bb_0_S, x_S);
643 bb_1_S = max(bb_1_S, x_S);
646 store4(bb->lower.ptr(), bb_0_S);
647 store4(bb->upper.ptr(), bb_1_S);
652 /*! \brief Computes the bounding box for na coordinates in order xyz?, bb order xxxxyyyyzzzz */
653 static void calc_bounding_box_xxxx_simd4(int na, const float* x, BoundingBox* bb_work_aligned, real* bb)
655 calc_bounding_box_simd4(na, x, bb_work_aligned);
657 bb[0 * c_packedBoundingBoxesDimSize] = bb_work_aligned->lower.x;
658 bb[1 * c_packedBoundingBoxesDimSize] = bb_work_aligned->lower.y;
659 bb[2 * c_packedBoundingBoxesDimSize] = bb_work_aligned->lower.z;
660 bb[3 * c_packedBoundingBoxesDimSize] = bb_work_aligned->upper.x;
661 bb[4 * c_packedBoundingBoxesDimSize] = bb_work_aligned->upper.y;
662 bb[5 * c_packedBoundingBoxesDimSize] = bb_work_aligned->upper.z;
665 # endif /* NBNXN_BBXXXX */
667 #endif /* NBNXN_SEARCH_SIMD4_FLOAT_X_BB */
670 /*! \brief Combines pairs of consecutive bounding boxes */
671 static void combine_bounding_box_pairs(const Grid& grid,
672 gmx::ArrayRef<const BoundingBox> bb,
673 gmx::ArrayRef<BoundingBox> bbj)
675 // TODO: During SIMDv2 transition only some archs use namespace (remove when done)
678 for (int i = 0; i < grid.numColumns(); i++)
680 /* Starting bb in a column is expected to be 2-aligned */
681 const int sc2 = grid.firstCellInColumn(i) >> 1;
682 /* For odd numbers skip the last bb here */
683 const int nc2 = (grid.numAtomsInColumn(i) + 3) >> (2 + 1);
685 for (c2 = sc2; c2 < sc2 + nc2; c2++)
687 #if NBNXN_SEARCH_BB_SIMD4
688 Simd4Float min_S, max_S;
690 min_S = min(load4(bb[c2 * 2 + 0].lower.ptr()), load4(bb[c2 * 2 + 1].lower.ptr()));
691 max_S = max(load4(bb[c2 * 2 + 0].upper.ptr()), load4(bb[c2 * 2 + 1].upper.ptr()));
692 store4(bbj[c2].lower.ptr(), min_S);
693 store4(bbj[c2].upper.ptr(), max_S);
695 bbj[c2].lower = BoundingBox::Corner::min(bb[c2 * 2 + 0].lower, bb[c2 * 2 + 1].lower);
696 bbj[c2].upper = BoundingBox::Corner::max(bb[c2 * 2 + 0].upper, bb[c2 * 2 + 1].upper);
699 if (((grid.numAtomsInColumn(i) + 3) >> 2) & 1)
701 /* The bb count in this column is odd: duplicate the last bb */
702 bbj[c2].lower = bb[c2 * 2].lower;
703 bbj[c2].upper = bb[c2 * 2].upper;
709 /*! \brief Prints the average bb size, used for debug output */
710 static void print_bbsizes_simple(FILE* fp, const Grid& grid)
713 for (int c = 0; c < grid.numCells(); c++)
715 const BoundingBox& bb = grid.iBoundingBoxes()[c];
716 ba[XX] += bb.upper.x - bb.lower.x;
717 ba[YY] += bb.upper.y - bb.lower.y;
718 ba[ZZ] += bb.upper.z - bb.lower.z;
720 dsvmul(1.0 / grid.numCells(), ba, ba);
722 const Grid::Dimensions& dims = grid.dimensions();
724 (dims.atomDensity > 0 ? grid.geometry().numAtomsICluster
725 / (dims.atomDensity * dims.cellSize[XX] * dims.cellSize[YY])
729 "ns bb: grid %4.2f %4.2f %4.2f abs %4.2f %4.2f %4.2f rel %4.2f %4.2f %4.2f\n",
736 ba[XX] * dims.invCellSize[XX],
737 ba[YY] * dims.invCellSize[YY],
738 dims.atomDensity > 0 ? ba[ZZ] / avgCellSizeZ : 0.0);
741 /*! \brief Prints the average bb size, used for debug output */
742 static void print_bbsizes_supersub(FILE* fp, const Grid& grid)
749 for (int c = 0; c < grid.numCells(); c++)
752 for (int s = 0; s < grid.numClustersPerCell()[c]; s += c_packedBoundingBoxesDimSize)
754 int cs_w = (c * c_gpuNumClusterPerCell + s) / c_packedBoundingBoxesDimSize;
755 auto boundingBoxes = grid.packedBoundingBoxes().subArray(
756 cs_w * c_packedBoundingBoxesSize, c_packedBoundingBoxesSize);
757 for (int i = 0; i < c_packedBoundingBoxesDimSize; i++)
759 for (int d = 0; d < DIM; d++)
761 ba[d] += boundingBoxes[(DIM + d) * c_packedBoundingBoxesDimSize + i]
762 - boundingBoxes[(0 + d) * c_packedBoundingBoxesDimSize + i];
767 for (int s = 0; s < grid.numClustersPerCell()[c]; s++)
769 const BoundingBox& bb = grid.iBoundingBoxes()[c * c_gpuNumClusterPerCell + s];
770 ba[XX] += bb.upper.x - bb.lower.x;
771 ba[YY] += bb.upper.y - bb.lower.y;
772 ba[ZZ] += bb.upper.z - bb.lower.z;
775 ns += grid.numClustersPerCell()[c];
777 dsvmul(1.0 / ns, ba, ba);
779 const Grid::Dimensions& dims = grid.dimensions();
780 const real avgClusterSizeZ =
781 (dims.atomDensity > 0 ? grid.geometry().numAtomsPerCell
782 / (dims.atomDensity * dims.cellSize[XX]
783 * dims.cellSize[YY] * c_gpuNumClusterPerCellZ)
787 "ns bb: grid %4.2f %4.2f %4.2f abs %4.2f %4.2f %4.2f rel %4.2f %4.2f %4.2f\n",
788 dims.cellSize[XX] / c_gpuNumClusterPerCellX,
789 dims.cellSize[YY] / c_gpuNumClusterPerCellY,
794 ba[XX] * c_gpuNumClusterPerCellX * dims.invCellSize[XX],
795 ba[YY] * c_gpuNumClusterPerCellY * dims.invCellSize[YY],
796 dims.atomDensity > 0 ? ba[ZZ] / avgClusterSizeZ : 0.0);
799 /*!\brief Set non-bonded interaction flags for the current cluster.
801 * Sorts atoms on LJ coefficients: !=0 first, ==0 at the end.
803 static void sort_cluster_on_flag(int numAtomsInCluster,
807 gmx::ArrayRef<int> order,
810 constexpr int c_maxNumAtomsInCluster = 8;
811 int sort1[c_maxNumAtomsInCluster];
812 int sort2[c_maxNumAtomsInCluster];
814 GMX_ASSERT(numAtomsInCluster <= c_maxNumAtomsInCluster,
815 "Need to increase c_maxNumAtomsInCluster to support larger clusters");
820 for (int s = atomStart; s < atomEnd; s += numAtomsInCluster)
822 /* Make lists for this (sub-)cell on atoms with and without LJ */
825 gmx_bool haveQ = FALSE;
827 for (int a = s; a < std::min(s + numAtomsInCluster, atomEnd); a++)
829 haveQ = haveQ || GET_CGINFO_HAS_Q(atinfo[order[a]]);
831 if (GET_CGINFO_HAS_VDW(atinfo[order[a]]))
833 sort1[n1++] = order[a];
838 sort2[n2++] = order[a];
842 /* If we don't have atoms with LJ, there's nothing to sort */
845 *flags |= NBNXN_CI_DO_LJ(subc);
847 if (2 * n1 <= numAtomsInCluster)
849 /* Only sort when strictly necessary. Ordering particles
850 * Ordering particles can lead to less accurate summation
851 * due to rounding, both for LJ and Coulomb interactions.
853 if (2 * (a_lj_max - s) >= numAtomsInCluster)
855 for (int i = 0; i < n1; i++)
857 order[atomStart + i] = sort1[i];
859 for (int j = 0; j < n2; j++)
861 order[atomStart + n1 + j] = sort2[j];
865 *flags |= NBNXN_CI_HALF_LJ(subc);
870 *flags |= NBNXN_CI_DO_COUL(subc);
876 /*! \brief Fill a pair search cell with atoms.
878 * Potentially sorts atoms and sets the interaction flags.
880 void Grid::fillCell(GridSetData* gridSetData,
881 nbnxn_atomdata_t* nbat,
885 gmx::ArrayRef<const gmx::RVec> x,
886 BoundingBox gmx_unused* bb_work_aligned)
888 const int numAtoms = atomEnd - atomStart;
890 const gmx::ArrayRef<int>& cells = gridSetData->cells;
891 const gmx::ArrayRef<int>& atomIndices = gridSetData->atomIndices;
893 if (geometry_.isSimple)
895 /* Note that non-local grids are already sorted.
896 * Then sort_cluster_on_flag will only set the flags and the sorting
897 * will not affect the atom order.
899 sort_cluster_on_flag(geometry_.numAtomsICluster,
904 flags_.data() + atomToCluster(atomStart) - cellOffset_);
909 /* Set the fep flag for perturbed atoms in this (sub-)cell */
911 /* The grid-local cluster/(sub-)cell index */
912 int cell = atomToCluster(atomStart)
913 - cellOffset_ * (geometry_.isSimple ? 1 : c_gpuNumClusterPerCell);
915 for (int at = atomStart; at < atomEnd; at++)
917 if (atomIndices[at] >= 0 && GET_CGINFO_FEP(atinfo[atomIndices[at]]))
919 fep_[cell] |= (1 << (at - atomStart));
924 /* Now we have sorted the atoms, set the cell indices */
925 for (int at = atomStart; at < atomEnd; at++)
927 cells[atomIndices[at]] = at;
930 copy_rvec_to_nbat_real(atomIndices.data() + atomStart,
932 geometry_.numAtomsICluster,
933 as_rvec_array(x.data()),
938 if (nbat->XFormat == nbatX4)
940 /* Store the bounding boxes as xyz.xyz. */
941 size_t offset = atomToCluster(atomStart - cellOffset_ * geometry_.numAtomsICluster);
942 BoundingBox* bb_ptr = bb_.data() + offset;
944 #if GMX_SIMD && GMX_SIMD_REAL_WIDTH == 2
945 if (2 * geometry_.numAtomsJCluster == geometry_.numAtomsICluster)
947 calc_bounding_box_x_x4_halves(numAtoms,
948 nbat->x().data() + atom_to_x_index<c_packX4>(atomStart),
950 bbj_.data() + offset * 2);
955 calc_bounding_box_x_x4(numAtoms, nbat->x().data() + atom_to_x_index<c_packX4>(atomStart), bb_ptr);
958 else if (nbat->XFormat == nbatX8)
960 /* Store the bounding boxes as xyz.xyz. */
961 size_t offset = atomToCluster(atomStart - cellOffset_ * geometry_.numAtomsICluster);
962 BoundingBox* bb_ptr = bb_.data() + offset;
964 calc_bounding_box_x_x8(numAtoms, nbat->x().data() + atom_to_x_index<c_packX8>(atomStart), bb_ptr);
967 else if (!geometry_.isSimple)
969 /* Store the bounding boxes in a format convenient
970 * for SIMD4 calculations: xxxxyyyyzzzz...
972 const int clusterIndex = ((atomStart - cellOffset_ * geometry_.numAtomsPerCell)
973 >> geometry_.numAtomsICluster2Log);
974 float* pbb_ptr = pbb_.data() + packedBoundingBoxesIndex(clusterIndex)
975 + (clusterIndex & (c_packedBoundingBoxesDimSize - 1));
977 # if NBNXN_SEARCH_SIMD4_FLOAT_X_BB
978 if (nbat->XFormat == nbatXYZQ)
980 GMX_ASSERT(bb_work_aligned != nullptr, "Must have valid aligned work structure");
981 calc_bounding_box_xxxx_simd4(
982 numAtoms, nbat->x().data() + atomStart * nbat->xstride, bb_work_aligned, pbb_ptr);
987 calc_bounding_box_xxxx(
988 numAtoms, nbat->xstride, nbat->x().data() + atomStart * nbat->xstride, pbb_ptr);
993 "cell %4d bb %5.2f %5.2f %5.2f %5.2f %5.2f %5.2f\n",
994 atomToCluster(atomStart),
995 pbb_ptr[0 * c_packedBoundingBoxesDimSize],
996 pbb_ptr[3 * c_packedBoundingBoxesDimSize],
997 pbb_ptr[1 * c_packedBoundingBoxesDimSize],
998 pbb_ptr[4 * c_packedBoundingBoxesDimSize],
999 pbb_ptr[2 * c_packedBoundingBoxesDimSize],
1000 pbb_ptr[5 * c_packedBoundingBoxesDimSize]);
1006 /* Store the bounding boxes as xyz.xyz. */
1007 BoundingBox* bb_ptr =
1008 bb_.data() + atomToCluster(atomStart - cellOffset_ * geometry_.numAtomsPerCell);
1010 calc_bounding_box(numAtoms, nbat->xstride, nbat->x().data() + atomStart * nbat->xstride, bb_ptr);
1014 int bbo = atomToCluster(atomStart - cellOffset_ * geometry_.numAtomsPerCell);
1016 "cell %4d bb %5.2f %5.2f %5.2f %5.2f %5.2f %5.2f\n",
1017 atomToCluster(atomStart),
1028 void Grid::sortColumnsCpuGeometry(GridSetData* gridSetData,
1031 gmx::ArrayRef<const gmx::RVec> x,
1032 nbnxn_atomdata_t* nbat,
1033 const gmx::Range<int> columnRange,
1034 gmx::ArrayRef<int> sort_work)
1039 "cell_offset %d sorting columns %d - %d\n",
1041 *columnRange.begin(),
1042 *columnRange.end());
1045 const bool relevantAtomsAreWithinGridBounds = (dimensions_.maxAtomGroupRadius == 0);
1047 const int numAtomsPerCell = geometry_.numAtomsPerCell;
1049 /* Sort the atoms within each x,y column in 3 dimensions */
1050 for (int cxy : columnRange)
1052 const int numAtoms = numAtomsInColumn(cxy);
1053 const int numCellsZ = cxy_ind_[cxy + 1] - cxy_ind_[cxy];
1054 const int atomOffset = firstAtomInColumn(cxy);
1056 /* Sort the atoms within each x,y column on z coordinate */
1060 relevantAtomsAreWithinGridBounds,
1061 gridSetData->atomIndices.data() + atomOffset,
1064 dimensions_.lowerCorner[ZZ],
1065 1.0 / dimensions_.gridSize[ZZ],
1066 numCellsZ * numAtomsPerCell,
1069 /* Fill the ncz cells in this column */
1070 const int firstCell = firstCellInColumn(cxy);
1071 int cellFilled = firstCell;
1072 for (int cellZ = 0; cellZ < numCellsZ; cellZ++)
1074 const int cell = firstCell + cellZ;
1076 const int atomOffsetCell = atomOffset + cellZ * numAtomsPerCell;
1077 const int numAtomsCell = std::min(numAtomsPerCell, numAtoms - (atomOffsetCell - atomOffset));
1079 fillCell(gridSetData, nbat, atomOffsetCell, atomOffsetCell + numAtomsCell, atinfo, x, nullptr);
1081 /* This copy to bbcz is not really necessary.
1082 * But it allows to use the same grid search code
1083 * for the simple and supersub cell setups.
1085 if (numAtomsCell > 0)
1089 bbcz_[cell].lower = bb_[cellFilled].lower.z;
1090 bbcz_[cell].upper = bb_[cellFilled].upper.z;
1093 /* Set the unused atom indices to -1 */
1094 for (int ind = numAtoms; ind < numCellsZ * numAtomsPerCell; ind++)
1096 gridSetData->atomIndices[atomOffset + ind] = -1;
1101 /* Spatially sort the atoms within one grid column */
1102 void Grid::sortColumnsGpuGeometry(GridSetData* gridSetData,
1105 gmx::ArrayRef<const gmx::RVec> x,
1106 nbnxn_atomdata_t* nbat,
1107 const gmx::Range<int> columnRange,
1108 gmx::ArrayRef<int> sort_work)
1110 BoundingBox bb_work_array[2];
1111 BoundingBox* bb_work_aligned = reinterpret_cast<BoundingBox*>(
1112 (reinterpret_cast<std::size_t>(bb_work_array + 1)) & (~(static_cast<std::size_t>(15))));
1117 "cell_offset %d sorting columns %d - %d\n",
1119 *columnRange.begin(),
1120 *columnRange.end());
1123 const bool relevantAtomsAreWithinGridBounds = (dimensions_.maxAtomGroupRadius == 0);
1125 const int numAtomsPerCell = geometry_.numAtomsPerCell;
1127 const int subdiv_x = geometry_.numAtomsICluster;
1128 const int subdiv_y = c_gpuNumClusterPerCellX * subdiv_x;
1129 const int subdiv_z = c_gpuNumClusterPerCellY * subdiv_y;
1131 /* Extract the atom index array that will be filled here */
1132 const gmx::ArrayRef<int>& atomIndices = gridSetData->atomIndices;
1134 /* Sort the atoms within each x,y column in 3 dimensions.
1135 * Loop over all columns on the x/y grid.
1137 for (int cxy : columnRange)
1139 const int gridX = cxy / dimensions_.numCells[YY];
1140 const int gridY = cxy - gridX * dimensions_.numCells[YY];
1142 const int numAtomsInColumn = cxy_na_[cxy];
1143 const int numCellsInColumn = cxy_ind_[cxy + 1] - cxy_ind_[cxy];
1144 const int atomOffset = firstAtomInColumn(cxy);
1146 /* Sort the atoms within each x,y column on z coordinate */
1150 relevantAtomsAreWithinGridBounds,
1151 atomIndices.data() + atomOffset,
1154 dimensions_.lowerCorner[ZZ],
1155 1.0 / dimensions_.gridSize[ZZ],
1156 numCellsInColumn * numAtomsPerCell,
1159 /* This loop goes over the cells and clusters along z at once */
1160 for (int sub_z = 0; sub_z < numCellsInColumn * c_gpuNumClusterPerCellZ; sub_z++)
1162 const int atomOffsetZ = atomOffset + sub_z * subdiv_z;
1163 const int numAtomsZ = std::min(subdiv_z, numAtomsInColumn - (atomOffsetZ - atomOffset));
1165 /* We have already sorted on z */
1167 if (sub_z % c_gpuNumClusterPerCellZ == 0)
1169 cz = sub_z / c_gpuNumClusterPerCellZ;
1170 const int cell = cxy_ind_[cxy] + cz;
1172 /* The number of atoms in this cell/super-cluster */
1173 const int numAtoms =
1174 std::min(numAtomsPerCell, numAtomsInColumn - (atomOffsetZ - atomOffset));
1176 numClusters_[cell] = std::min(
1177 c_gpuNumClusterPerCell,
1178 (numAtoms + geometry_.numAtomsICluster - 1) / geometry_.numAtomsICluster);
1180 /* Store the z-boundaries of the bounding box of the cell */
1181 bbcz_[cell].lower = x[atomIndices[atomOffsetZ]][ZZ];
1182 bbcz_[cell].upper = x[atomIndices[atomOffsetZ + numAtoms - 1]][ZZ];
1185 if (c_gpuNumClusterPerCellY > 1)
1187 /* Sort the atoms along y */
1191 relevantAtomsAreWithinGridBounds,
1192 atomIndices.data() + atomOffsetZ,
1195 dimensions_.lowerCorner[YY] + gridY * dimensions_.cellSize[YY],
1196 dimensions_.invCellSize[YY],
1201 for (int sub_y = 0; sub_y < c_gpuNumClusterPerCellY; sub_y++)
1203 const int atomOffsetY = atomOffsetZ + sub_y * subdiv_y;
1204 const int numAtomsY = std::min(subdiv_y, numAtomsInColumn - (atomOffsetY - atomOffset));
1206 if (c_gpuNumClusterPerCellX > 1)
1208 /* Sort the atoms along x */
1210 ((cz * c_gpuNumClusterPerCellY + sub_y) & 1) != 0,
1212 relevantAtomsAreWithinGridBounds,
1213 atomIndices.data() + atomOffsetY,
1216 dimensions_.lowerCorner[XX] + gridX * dimensions_.cellSize[XX],
1217 dimensions_.invCellSize[XX],
1222 for (int sub_x = 0; sub_x < c_gpuNumClusterPerCellX; sub_x++)
1224 const int atomOffsetX = atomOffsetY + sub_x * subdiv_x;
1225 const int numAtomsX =
1226 std::min(subdiv_x, numAtomsInColumn - (atomOffsetX - atomOffset));
1228 fillCell(gridSetData, nbat, atomOffsetX, atomOffsetX + numAtomsX, atinfo, x, bb_work_aligned);
1233 /* Set the unused atom indices to -1 */
1234 for (int ind = numAtomsInColumn; ind < numCellsInColumn * numAtomsPerCell; ind++)
1236 atomIndices[atomOffset + ind] = -1;
1241 /*! \brief Sets the cell index in the cell array for atom \p atomIndex and increments the atom count for the grid column */
1242 static void setCellAndAtomCount(gmx::ArrayRef<int> cell, int cellIndex, gmx::ArrayRef<int> cxy_na, int atomIndex)
1244 cell[atomIndex] = cellIndex;
1245 cxy_na[cellIndex] += 1;
1248 void Grid::calcColumnIndices(const Grid::Dimensions& gridDims,
1249 const gmx::UpdateGroupsCog* updateGroupsCog,
1250 const gmx::Range<int> atomRange,
1251 gmx::ArrayRef<const gmx::RVec> x,
1256 gmx::ArrayRef<int> cell,
1257 gmx::ArrayRef<int> cxy_na)
1259 const int numColumns = gridDims.numCells[XX] * gridDims.numCells[YY];
1261 /* We add one extra cell for particles which moved during DD */
1262 for (int i = 0; i < numColumns; i++)
1267 int taskAtomStart = *atomRange.begin() + static_cast<int>((thread + 0) * atomRange.size()) / nthread;
1268 int taskAtomEnd = *atomRange.begin() + static_cast<int>((thread + 1) * atomRange.size()) / nthread;
1273 for (int i = taskAtomStart; i < taskAtomEnd; i++)
1275 if (move == nullptr || move[i] >= 0)
1278 const gmx::RVec& coord = (updateGroupsCog ? updateGroupsCog->cogForAtom(i) : x[i]);
1280 /* We need to be careful with rounding,
1281 * particles might be a few bits outside the local zone.
1282 * The int cast takes care of the lower bound,
1283 * we will explicitly take care of the upper bound.
1285 int cx = static_cast<int>((coord[XX] - gridDims.lowerCorner[XX])
1286 * gridDims.invCellSize[XX]);
1287 int cy = static_cast<int>((coord[YY] - gridDims.lowerCorner[YY])
1288 * gridDims.invCellSize[YY]);
1291 if (cx < 0 || cx > gridDims.numCells[XX] || cy < 0 || cy > gridDims.numCells[YY])
1294 "grid cell cx %d cy %d out of range (max %d %d)\n"
1295 "atom %f %f %f, grid->c0 %f %f",
1298 gridDims.numCells[XX],
1299 gridDims.numCells[YY],
1303 gridDims.lowerCorner[XX],
1304 gridDims.lowerCorner[YY]);
1307 /* Take care of potential rouding issues */
1308 cx = std::min(cx, gridDims.numCells[XX] - 1);
1309 cy = std::min(cy, gridDims.numCells[YY] - 1);
1311 /* For the moment cell will contain only the, grid local,
1312 * x and y indices, not z.
1314 setCellAndAtomCount(cell, cx * gridDims.numCells[YY] + cy, cxy_na, i);
1318 /* Put this moved particle after the end of the grid,
1319 * so we can process it later without using conditionals.
1321 setCellAndAtomCount(cell, numColumns, cxy_na, i);
1328 for (int i = taskAtomStart; i < taskAtomEnd; i++)
1330 int cx = static_cast<int>((x[i][XX] - gridDims.lowerCorner[XX]) * gridDims.invCellSize[XX]);
1331 int cy = static_cast<int>((x[i][YY] - gridDims.lowerCorner[YY]) * gridDims.invCellSize[YY]);
1333 /* For non-home zones there could be particles outside
1334 * the non-bonded cut-off range, which have been communicated
1335 * for bonded interactions only. For the result it doesn't
1336 * matter where these end up on the grid. For performance
1337 * we put them in an extra row at the border.
1339 cx = std::max(cx, 0);
1340 cx = std::min(cx, gridDims.numCells[XX] - 1);
1341 cy = std::max(cy, 0);
1342 cy = std::min(cy, gridDims.numCells[YY] - 1);
1344 /* For the moment cell will contain only the, grid local,
1345 * x and y indices, not z.
1347 setCellAndAtomCount(cell, cx * gridDims.numCells[YY] + cy, cxy_na, i);
1352 /*! \brief Resizes grid and atom data which depend on the number of cells */
1353 static void resizeForNumberOfCells(const int numNbnxnAtoms,
1354 const int numAtomsMoved,
1355 GridSetData* gridSetData,
1356 nbnxn_atomdata_t* nbat)
1358 /* Note: gridSetData->cellIndices was already resized before */
1360 /* To avoid conditionals we store the moved particles at the end of a,
1361 * make sure we have enough space.
1363 gridSetData->atomIndices.resize(numNbnxnAtoms + numAtomsMoved);
1365 /* Make space in nbat for storing the atom coordinates */
1366 nbat->resizeCoordinateBuffer(numNbnxnAtoms);
1369 void Grid::setCellIndices(int ddZone,
1371 GridSetData* gridSetData,
1372 gmx::ArrayRef<GridWork> gridWork,
1373 const gmx::Range<int> atomRange,
1375 gmx::ArrayRef<const gmx::RVec> x,
1376 const int numAtomsMoved,
1377 nbnxn_atomdata_t* nbat)
1379 cellOffset_ = cellOffset;
1381 srcAtomBegin_ = *atomRange.begin();
1382 srcAtomEnd_ = *atomRange.end();
1384 const int nthread = gmx_omp_nthreads_get(emntPairsearch);
1386 const int numAtomsPerCell = geometry_.numAtomsPerCell;
1388 /* Make the cell index as a function of x and y */
1392 for (int i = 0; i < numColumns() + 1; i++)
1394 /* We set ncz_max at the beginning of the loop iso at the end
1395 * to skip i=grid->ncx*grid->numCells[YY] which are moved particles
1396 * that do not need to be ordered on the grid.
1402 int cxy_na_i = gridWork[0].numAtomsPerColumn[i];
1403 for (int thread = 1; thread < nthread; thread++)
1405 cxy_na_i += gridWork[thread].numAtomsPerColumn[i];
1407 ncz = (cxy_na_i + numAtomsPerCell - 1) / numAtomsPerCell;
1408 if (nbat->XFormat == nbatX8)
1410 /* Make the number of cell a multiple of 2 */
1411 ncz = (ncz + 1) & ~1;
1413 cxy_ind_[i + 1] = cxy_ind_[i] + ncz;
1414 /* Clear cxy_na_, so we can reuse the array below */
1417 numCellsTotal_ = cxy_ind_[numColumns()] - cxy_ind_[0];
1418 numCellsColumnMax_ = ncz_max;
1420 /* Resize grid and atom data which depend on the number of cells */
1421 resizeForNumberOfCells(atomIndexEnd(), numAtomsMoved, gridSetData, nbat);
1426 "ns na_sc %d na_c %d super-cells: %d x %d y %d z %.1f maxz %d\n",
1428 geometry_.numAtomsICluster,
1430 dimensions_.numCells[XX],
1431 dimensions_.numCells[YY],
1432 numCellsTotal_ / (static_cast<double>(numColumns())),
1437 for (int cy = 0; cy < dimensions_.numCells[YY]; cy++)
1439 for (int cx = 0; cx < dimensions_.numCells[XX]; cx++)
1441 fprintf(debug, " %2d", cxy_ind_[i + 1] - cxy_ind_[i]);
1444 fprintf(debug, "\n");
1449 /* Make sure the work array for sorting is large enough */
1450 const int worstCaseSortBufferSize = ncz_max * numAtomsPerCell * c_sortGridMaxSizeFactor;
1451 if (worstCaseSortBufferSize > gmx::index(gridWork[0].sortBuffer.size()))
1453 for (GridWork& work : gridWork)
1455 /* Elements not in use should be -1 */
1456 work.sortBuffer.resize(worstCaseSortBufferSize, -1);
1460 /* Now we know the dimensions we can fill the grid.
1461 * This is the first, unsorted fill. We sort the columns after this.
1463 gmx::ArrayRef<int> cells = gridSetData->cells;
1464 gmx::ArrayRef<int> atomIndices = gridSetData->atomIndices;
1465 for (int i : atomRange)
1467 /* At this point nbs->cell contains the local grid x,y indices */
1468 const int cxy = cells[i];
1469 atomIndices[firstAtomInColumn(cxy) + cxy_na_[cxy]++] = i;
1474 /* Set the cell indices for the moved particles */
1475 int n0 = numCellsTotal_ * numAtomsPerCell;
1476 int n1 = numCellsTotal_ * numAtomsPerCell + cxy_na_[numColumns()];
1477 for (int i = n0; i < n1; i++)
1479 cells[atomIndices[i]] = i;
1483 /* Sort the super-cell columns along z into the sub-cells. */
1484 #pragma omp parallel for num_threads(nthread) schedule(static)
1485 for (int thread = 0; thread < nthread; thread++)
1489 gmx::Range<int> columnRange(((thread + 0) * numColumns()) / nthread,
1490 ((thread + 1) * numColumns()) / nthread);
1491 if (geometry_.isSimple)
1493 sortColumnsCpuGeometry(
1494 gridSetData, ddZone, atinfo, x, nbat, columnRange, gridWork[thread].sortBuffer);
1498 sortColumnsGpuGeometry(
1499 gridSetData, ddZone, atinfo, x, nbat, columnRange, gridWork[thread].sortBuffer);
1502 GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR
1505 if (geometry_.isSimple && nbat->XFormat == nbatX8)
1507 combine_bounding_box_pairs(*this, bb_, bbj_);
1510 if (!geometry_.isSimple)
1512 numClustersTotal_ = 0;
1513 for (int i = 0; i < numCellsTotal_; i++)
1515 numClustersTotal_ += numClusters_[i];
1521 if (geometry_.isSimple)
1523 print_bbsizes_simple(debug, *this);
1528 "ns non-zero sub-cells: %d average atoms %.2f\n",
1530 atomRange.size() / static_cast<double>(numClustersTotal_));
1532 print_bbsizes_supersub(debug, *this);
1537 } // namespace Nbnxm