2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
5 * Copyright (c) 2001-2012, The GROMACS development team,
6 * check out http://www.gromacs.org for more information.
7 * Copyright (c) 2012, by the GROMACS development team, led by
8 * David van der Spoel, Berk Hess, Erik Lindahl, and including many
9 * others, as listed in the AUTHORS file in the top-level source
10 * directory and at http://www.gromacs.org.
12 * GROMACS is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU Lesser General Public License
14 * as published by the Free Software Foundation; either version 2.1
15 * of the License, or (at your option) any later version.
17 * GROMACS is distributed in the hope that it will be useful,
18 * but WITHOUT ANY WARRANTY; without even the implied warranty of
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20 * Lesser General Public License for more details.
22 * You should have received a copy of the GNU Lesser General Public
23 * License along with GROMACS; if not, see
24 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
25 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
27 * If you want to redistribute modifications to GROMACS, please
28 * consider that scientific software is very special. Version
29 * control is crucial - bugs must be traceable. We will be happy to
30 * consider code for inclusion in the official distribution, but
31 * derived work must not be called official GROMACS. Details are found
32 * in the README & COPYING files - if they are missing, get the
33 * official version at http://www.gromacs.org.
35 * To help us fund GROMACS development, we humbly ask that you cite
36 * the research papers on the package. Check out http://www.gromacs.org.
39 #if GMX_NBNXN_SIMD_BITWIDTH == 128
40 #define GMX_MM128_HERE
42 #if GMX_NBNXN_SIMD_BITWIDTH == 256
43 #define GMX_MM256_HERE
45 #error "unsupported GMX_NBNXN_SIMD_BITWIDTH"
48 #include "gmx_simd_macros.h"
50 #if GMX_SIMD_WIDTH_HERE >= 2*NBNXN_CPU_CLUSTER_I_SIZE
51 #define STRIDE_S (GMX_SIMD_WIDTH_HERE/2)
53 #define STRIDE_S NBNXN_CPU_CLUSTER_I_SIZE
56 static gmx_inline gmx_mm_pr gmx_load_hpr_hilo_pr(const real *a)
60 a_SSE = _mm_load_ps(a);
62 return gmx_2hpr_to_pr(a_SSE, a_SSE);
65 static gmx_inline gmx_mm_pr gmx_set_2real_shift_pr(const real *a, real shift)
69 a0 = _mm_set1_ps(a[0] + shift);
70 a1 = _mm_set1_ps(a[1] + shift);
72 return gmx_2hpr_to_pr(a1, a0);
75 /* Copies PBC shifted i-cell packed atom coordinates to working array */
76 static gmx_inline void
77 icell_set_x_simd_2xnn(int ci,
78 real shx, real shy, real shz,
80 int stride, const real *x,
81 nbnxn_list_work_t *work)
84 nbnxn_x_ci_simd_2xnn_t *x_ci;
86 x_ci = work->x_ci_simd_2xnn;
88 ia = X_IND_CI_SIMD_2XNN(ci);
90 x_ci->ix_SSE0 = gmx_set_2real_shift_pr(x + ia + 0*STRIDE_S + 0, shx);
91 x_ci->iy_SSE0 = gmx_set_2real_shift_pr(x + ia + 1*STRIDE_S + 0, shy);
92 x_ci->iz_SSE0 = gmx_set_2real_shift_pr(x + ia + 2*STRIDE_S + 0, shz);
93 x_ci->ix_SSE2 = gmx_set_2real_shift_pr(x + ia + 0*STRIDE_S + 2, shx);
94 x_ci->iy_SSE2 = gmx_set_2real_shift_pr(x + ia + 1*STRIDE_S + 2, shy);
95 x_ci->iz_SSE2 = gmx_set_2real_shift_pr(x + ia + 2*STRIDE_S + 2, shz);
98 /* SIMD code for making a pair list of cell ci vs cell cjf-cjl
99 * for coordinates in packed format.
100 * Checks bouding box distances and possibly atom pair distances.
101 * This is an accelerated version of make_cluster_list_simple.
103 static gmx_inline void
104 make_cluster_list_simd_2xnn(const nbnxn_grid_t *gridj,
105 nbnxn_pairlist_t *nbl,
106 int ci, int cjf, int cjl,
107 gmx_bool remove_sub_diag,
109 real rl2, float rbb2,
112 const nbnxn_x_ci_simd_2xnn_t *work;
115 gmx_mm_pr jx_SSE, jy_SSE, jz_SSE;
117 gmx_mm_pr dx_SSE0, dy_SSE0, dz_SSE0;
118 gmx_mm_pr dx_SSE2, dy_SSE2, dz_SSE2;
125 gmx_mm_pr wco_any_SSE;
131 int xind_f, xind_l, cj;
133 cjf = CI_TO_CJ_SIMD_2XNN(cjf);
134 cjl = CI_TO_CJ_SIMD_2XNN(cjl+1) - 1;
136 work = nbl->work->x_ci_simd_2xnn;
138 bb_ci = nbl->work->bb_ci;
140 rc2_SSE = gmx_set1_pr(rl2);
143 while (!InRange && cjf <= cjl)
145 d2 = subc_bb_dist2_sse(4, 0, bb_ci, cjf, gridj->bbj);
148 /* Check if the distance is within the distance where
149 * we use only the bounding box distance rbb,
150 * or within the cut-off and there is at least one atom pair
151 * within the cut-off.
159 xind_f = X_IND_CJ_SIMD_2XNN(CI_TO_CJ_SIMD_2XNN(gridj->cell0) + cjf);
161 jx_SSE = gmx_load_hpr_hilo_pr(x_j+xind_f+0*STRIDE_S);
162 jy_SSE = gmx_load_hpr_hilo_pr(x_j+xind_f+1*STRIDE_S);
163 jz_SSE = gmx_load_hpr_hilo_pr(x_j+xind_f+2*STRIDE_S);
165 /* Calculate distance */
166 dx_SSE0 = gmx_sub_pr(work->ix_SSE0, jx_SSE);
167 dy_SSE0 = gmx_sub_pr(work->iy_SSE0, jy_SSE);
168 dz_SSE0 = gmx_sub_pr(work->iz_SSE0, jz_SSE);
169 dx_SSE2 = gmx_sub_pr(work->ix_SSE2, jx_SSE);
170 dy_SSE2 = gmx_sub_pr(work->iy_SSE2, jy_SSE);
171 dz_SSE2 = gmx_sub_pr(work->iz_SSE2, jz_SSE);
173 /* rsq = dx*dx+dy*dy+dz*dz */
174 rsq_SSE0 = gmx_calc_rsq_pr(dx_SSE0, dy_SSE0, dz_SSE0);
175 rsq_SSE2 = gmx_calc_rsq_pr(dx_SSE2, dy_SSE2, dz_SSE2);
177 wco_SSE0 = gmx_cmplt_pr(rsq_SSE0, rc2_SSE);
178 wco_SSE2 = gmx_cmplt_pr(rsq_SSE2, rc2_SSE);
180 wco_any_SSE = gmx_or_pr(wco_SSE0, wco_SSE2);
182 InRange = gmx_movemask_pr(wco_any_SSE);
184 *ndistc += 2*GMX_SIMD_WIDTH_HERE;
197 while (!InRange && cjl > cjf)
199 d2 = subc_bb_dist2_sse(4, 0, bb_ci, cjl, gridj->bbj);
202 /* Check if the distance is within the distance where
203 * we use only the bounding box distance rbb,
204 * or within the cut-off and there is at least one atom pair
205 * within the cut-off.
213 xind_l = X_IND_CJ_SIMD_2XNN(CI_TO_CJ_SIMD_2XNN(gridj->cell0) + cjl);
215 jx_SSE = gmx_load_hpr_hilo_pr(x_j+xind_l+0*STRIDE_S);
216 jy_SSE = gmx_load_hpr_hilo_pr(x_j+xind_l+1*STRIDE_S);
217 jz_SSE = gmx_load_hpr_hilo_pr(x_j+xind_l+2*STRIDE_S);
219 /* Calculate distance */
220 dx_SSE0 = gmx_sub_pr(work->ix_SSE0, jx_SSE);
221 dy_SSE0 = gmx_sub_pr(work->iy_SSE0, jy_SSE);
222 dz_SSE0 = gmx_sub_pr(work->iz_SSE0, jz_SSE);
223 dx_SSE2 = gmx_sub_pr(work->ix_SSE2, jx_SSE);
224 dy_SSE2 = gmx_sub_pr(work->iy_SSE2, jy_SSE);
225 dz_SSE2 = gmx_sub_pr(work->iz_SSE2, jz_SSE);
227 /* rsq = dx*dx+dy*dy+dz*dz */
228 rsq_SSE0 = gmx_calc_rsq_pr(dx_SSE0, dy_SSE0, dz_SSE0);
229 rsq_SSE2 = gmx_calc_rsq_pr(dx_SSE2, dy_SSE2, dz_SSE2);
231 wco_SSE0 = gmx_cmplt_pr(rsq_SSE0, rc2_SSE);
232 wco_SSE2 = gmx_cmplt_pr(rsq_SSE2, rc2_SSE);
234 wco_any_SSE = gmx_or_pr(wco_SSE0, wco_SSE2);
236 InRange = gmx_movemask_pr(wco_any_SSE);
238 *ndistc += 2*GMX_SIMD_WIDTH_HERE;
248 for (cj = cjf; cj <= cjl; cj++)
250 /* Store cj and the interaction mask */
251 nbl->cj[nbl->ncj].cj = CI_TO_CJ_SIMD_2XNN(gridj->cell0) + cj;
252 nbl->cj[nbl->ncj].excl = get_imask_x86_simd_2xnn(remove_sub_diag, ci, cj);
255 /* Increase the closing index in i super-cell list */
256 nbl->ci[nbl->nci].cj_ind_end = nbl->ncj;
261 #undef GMX_MM128_HERE
262 #undef GMX_MM256_HERE