-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathCudaBondedForce.cu
1438 lines (1245 loc) · 46.5 KB
/
CudaBondedForce.cu
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
#include <iostream>
#include <iomanip>
#include <cmath>
#include <cassert>
#include <cuda.h>
#include "cuda_utils.h"
#include "gpu_utils.h"
#include "CudaBondedForce.h"
// Energy and virial in device memory
//static __device__ BondedEnergyVirial_t d_energy_virial;
//
// Reduces energy values
//
__forceinline__ __device__
void reduce_energy(const double epot, volatile double *sh_epot, double *global_epot) {
#if __CUDA_ARCH__ < 300
sh_epot[threadIdx.x] = epot;
__syncthreads();
for (int i=1;i < blockDim.x;i *= 2) {
int t = threadIdx.x + i;
double epot_val = (t < blockDim.x) ? sh_epot[t] : 0.0;
__syncthreads();
sh_epot[threadIdx.x] += epot_val;
__syncthreads();
}
if (threadIdx.x == 0) atomicAdd(global_epot, sh_epot[0]);
#else
const int tid = threadIdx.x & (warpsize-1);
const int base = (threadIdx.x/warpsize);
// Reduce within warps
double energy = epot;
for (int d=warpsize/2;d >= 1;d /= 2) {
energy += __hiloint2double(__shfl(__double2hiint(energy), tid+d),
__shfl(__double2loint(energy), tid+d));
}
// Reduce between warps
// NOTE: this __syncthreads() is needed because we're using a single shared memory buffer
__syncthreads();
if (tid == 0) sh_epot[base] = energy;
__syncthreads();
if (base == 0) {
energy = (tid < blockDim.x/warpsize) ? sh_epot[tid] : 0.0;
for (int d=warpsize/2;d >= 1;d /= 2) {
energy += __hiloint2double(__shfl(__double2hiint(energy), tid+d),
__shfl(__double2loint(energy), tid+d));
}
}
if (threadIdx.x == 0) atomicAdd(global_epot, energy);
#endif
}
//
// Templated sqrt() -function
//
template <typename T>
__forceinline__ __device__
double sqrt_template(const T x) {
if (sizeof(T) == 4) {
return sqrtf(x);
} else {
return sqrt(x);
}
}
template <typename AT, typename CT, bool calc_energy, bool calc_virial>
__device__ void calc_bond_force_device(const int pos,
const bondlist_t* __restrict__ bondlist,
const float2* __restrict__ bondcoef,
const float4* __restrict__ xyzq,
const int stride,
const CT boxx, const CT boxy, const CT boxz,
AT* __restrict__ force, double &epot, Virial_t* __restrict__ virial) {
int ii = bondlist[pos].i;
int jj = bondlist[pos].j;
int ic = bondlist[pos].itype;
int ish = bondlist[pos].ishift;
// Calculate shift for i-atom
CT shx, shy, shz;
calc_box_shift<CT>(ish, boxx, boxy, boxz, shx, shy, shz);
float4 xyzqi = xyzq[ii];
float4 xyzqj = xyzq[jj];
CT dx = xyzqi.x + shx - xyzqj.x;
CT dy = xyzqi.y + shy - xyzqj.y;
CT dz = xyzqi.z + shz - xyzqj.z;
CT r = sqrt_template<CT>(dx*dx + dy*dy + dz*dz);
float2 bondcoef_val = bondcoef[ic];
CT db = r - (CT)bondcoef_val.x;
CT fij = db*(CT)bondcoef_val.y;
if (calc_energy) {
epot += (double)(fij*db);
}
fij *= ((CT)2)/r;
AT fxij, fyij, fzij;
calc_component_force<AT, CT>(fij, dx, dy, dz, fxij, fyij, fzij);
// Store forces
write_force<AT>(fxij, fyij, fzij, ii, stride, force);
write_force<AT>(-fxij, -fyij, -fzij, jj, stride, force);
// Store virial
if (calc_virial) {
#ifdef USE_DP_SFORCE
if (ish != 13) {
atomicAdd(&virial->sforce_dp[ish][0], (double)(fij*dx));
atomicAdd(&virial->sforce_dp[ish][1], (double)(fij*dy));
atomicAdd(&virial->sforce_dp[ish][2], (double)(fij*dz));
}
#else
if (ish != 13) {
fxij /= CONVERT_TO_VIR;
fyij /= CONVERT_TO_VIR;
fzij /= CONVERT_TO_VIR;
atomicAdd((unsigned long long int *)&virial->sforce_fp[ish][0], llitoulli(fxij));
atomicAdd((unsigned long long int *)&virial->sforce_fp[ish][1], llitoulli(fyij));
atomicAdd((unsigned long long int *)&virial->sforce_fp[ish][2], llitoulli(fzij));
}
#endif
}
}
//
// bondcoef.x = cbb
// bondcoef.y = cbc
//
template <typename AT, typename CT, bool calc_energy, bool calc_virial>
__global__ void calc_bond_force_kernel(const int nbondlist,
const bondlist_t* __restrict__ bondlist,
const float2* __restrict__ bondcoef,
const float4* __restrict__ xyzq,
const int stride,
const CT boxx, const CT boxy, const CT boxz,
AT* __restrict__ force,
double* __restrict__ energy_bond,
Virial_t* __restrict__ virial) {
// Amount of shared memory required:
// CUDA_ARCH < 300: blockDim.x*sizeof(double)
// CUDA_ARCH >= 300: (blockDim.x/warpsize)*sizeof(double)
extern __shared__ double sh_epot[];
int pos = threadIdx.x + blockIdx.x*blockDim.x;
double epot;
if (calc_energy) {
epot = 0.0;
}
while (pos < nbondlist) {
calc_bond_force_device<AT, CT, calc_energy, calc_virial>
(pos, bondlist, bondcoef, xyzq, stride, boxx, boxy, boxz, force, epot, virial);
pos += blockDim.x*gridDim.x;
}
// Reduce energy
if (calc_energy) {
reduce_energy(epot, sh_epot, energy_bond);
}
}
//
// bondcoef.x = cbb
// bondcoef.y = cbc
//
template <typename AT, typename CT, bool calc_energy, bool calc_virial>
__global__ void calc_ureyb_force_kernel(
const int nureyblist,
const bondlist_t* __restrict__ ureyblist,
const float2* __restrict__ ureybcoef,
const float4* __restrict__ xyzq,
const int stride,
const CT boxx, const CT boxy, const CT boxz,
AT* __restrict__ force,
double* __restrict__ energy_ureyb,
Virial_t* __restrict__ virial) {
// Amount of shared memory required:
// sh_epot: blockDim.x*sizeof(double)
extern __shared__ double sh_epot[];
int pos = threadIdx.x + blockIdx.x*blockDim.x;
double epot;
if (calc_energy) {
epot = 0.0;
}
while (pos < nureyblist) {
calc_bond_force_device<AT, CT, calc_energy, calc_virial>
(pos, ureyblist, ureybcoef, xyzq, stride, boxx, boxy, boxz, force, epot, virial);
pos += blockDim.x*gridDim.x;
}
// Reduce energy
if (calc_energy) {
reduce_energy(epot, sh_epot, energy_ureyb);
}
}
template <typename AT, typename CT, bool calc_energy, bool calc_virial>
__device__ void calc_angle_force_device(const int pos,
const anglelist_t* __restrict__ anglelist,
const float2* __restrict__ anglecoef,
const float4* __restrict__ xyzq,
const int stride,
const CT boxx, const CT boxy, const CT boxz,
AT* __restrict__ force, double &epot,
Virial_t* __restrict__ virial) {
int ii = anglelist[pos].i;
int jj = anglelist[pos].j;
int kk = anglelist[pos].k;
int ic = anglelist[pos].itype;
int ish = anglelist[pos].ishift1;
int ksh = anglelist[pos].ishift2;
// Calculate shift for i-atom
CT ishx, ishy, ishz;
calc_box_shift<CT>(ish, boxx, boxy, boxz, ishx, ishy, ishz);
// Calculate shift for k-atom
CT kshx, kshy, kshz;
calc_box_shift<CT>(ksh, boxx, boxy, boxz, kshx, kshy, kshz);
CT dxij = xyzq[ii].x + ishx - xyzq[jj].x;
CT dyij = xyzq[ii].y + ishy - xyzq[jj].y;
CT dzij = xyzq[ii].z + ishz - xyzq[jj].z;
CT dxkj = xyzq[kk].x + kshx - xyzq[jj].x;
CT dykj = xyzq[kk].y + kshy - xyzq[jj].y;
CT dzkj = xyzq[kk].z + kshz - xyzq[jj].z;
CT rij = sqrtf(dxij*dxij + dyij*dyij + dzij*dzij);
CT rkj = sqrtf(dxkj*dxkj + dykj*dykj + dzkj*dzkj);
CT rij_inv = ((CT)1)/rij;
CT rkj_inv = ((CT)1)/rkj;
CT dxijr = dxij*rij_inv;
CT dyijr = dyij*rij_inv;
CT dzijr = dzij*rij_inv;
CT dxkjr = dxkj*rkj_inv;
CT dykjr = dykj*rkj_inv;
CT dzkjr = dzkj*rkj_inv;
CT cst = dxijr*dxkjr + dyijr*dykjr + dzijr*dzkjr;
// anglecoef.x = ctb
// anglecoef.y = ctc
float2 anglecoef_val = anglecoef[ic];
// Restrict values of cst to the interval [-0.999 ... 0.999]
// NOTE: we are ignoring the fancy stuff that is done on the CPU version
cst = min((CT)0.999, max(-(CT)0.999, cst));
CT at = acosf(cst);
CT da = at - (CT)anglecoef_val.x;
CT df = ((CT)anglecoef_val.y)*da;
if (calc_energy) {
epot += epot + (double)(df*da);
}
CT st2r = ((CT)1.0)/(((CT)1.0) - cst*cst);
CT str = sqrtf(st2r);
df = -((CT)2.0)*df*str;
CT dtxi = rij_inv*(dxkjr - cst*dxijr);
CT dtxj = rkj_inv*(dxijr - cst*dxkjr);
CT dtyi = rij_inv*(dykjr - cst*dyijr);
CT dtyj = rkj_inv*(dyijr - cst*dykjr);
CT dtzi = rij_inv*(dzkjr - cst*dzijr);
CT dtzj = rkj_inv*(dzijr - cst*dzkjr);
AT AT_dtxi, AT_dtyi, AT_dtzi;
AT AT_dtxj, AT_dtyj, AT_dtzj;
calc_component_force<AT, CT>(df, dtxi, dtyi, dtzi, AT_dtxi, AT_dtyi, AT_dtzi);
calc_component_force<AT, CT>(df, dtxj, dtyj, dtzj, AT_dtxj, AT_dtyj, AT_dtzj);
write_force<AT>(AT_dtxi, AT_dtyi, AT_dtzi, ii, stride, force);
write_force<AT>(AT_dtxj, AT_dtyj, AT_dtzj, kk, stride, force);
write_force<AT>(-AT_dtxi-AT_dtxj, -AT_dtyi-AT_dtyj, -AT_dtzi-AT_dtzj, jj, stride, force);
// Store virial
if (calc_virial) {
#ifdef USE_DP_SFORCE
if (ish != 13) {
atomicAdd(&virial->sforce_dp[ish][0], (double)(df*dtxi));
atomicAdd(&virial->sforce_dp[ish][1], (double)(df*dtyi));
atomicAdd(&virial->sforce_dp[ish][2], (double)(df*dtzi));
}
if (ksh != 13) {
atomicAdd(&virial->sforce_dp[ksh][0], (double)(df*dtxj));
atomicAdd(&virial->sforce_dp[ksh][1], (double)(df*dtyj));
atomicAdd(&virial->sforce_dp[ksh][2], (double)(df*dtzj));
}
#else
if (ish != 13) {
AT_dtxi /= CONVERT_TO_VIR;
AT_dtyi /= CONVERT_TO_VIR;
AT_dtzi /= CONVERT_TO_VIR;
atomicAdd((unsigned long long int *)&virial->sforce_fp[ish][0], llitoulli(AT_dtxi));
atomicAdd((unsigned long long int *)&virial->sforce_fp[ish][1], llitoulli(AT_dtyi));
atomicAdd((unsigned long long int *)&virial->sforce_fp[ish][2], llitoulli(AT_dtzi));
}
if (ksh != 13) {
AT_dtxj /= CONVERT_TO_VIR;
AT_dtyj /= CONVERT_TO_VIR;
AT_dtzj /= CONVERT_TO_VIR;
atomicAdd((unsigned long long int *)&virial->sforce_fp[ksh][0], llitoulli(AT_dtxj));
atomicAdd((unsigned long long int *)&virial->sforce_fp[ksh][1], llitoulli(AT_dtyj));
atomicAdd((unsigned long long int *)&virial->sforce_fp[ksh][2], llitoulli(AT_dtzj));
}
#endif
}
}
//
// anglecoef.x = ctb
// anglecoef.y = ctc
//
template <typename AT, typename CT, bool calc_energy, bool calc_virial>
__global__ void calc_angle_force_kernel(const int nanglelist,
const anglelist_t* __restrict__ anglelist,
const float2* __restrict__ anglecoef,
const float4* __restrict__ xyzq,
const int stride,
const CT boxx, const CT boxy, const CT boxz,
AT* __restrict__ force,
double* __restrict__ energy_angle,
Virial_t* __restrict__ virial) {
// Amount of shared memory required:
// sh_epot: blockDim.x*sizeof(double)
extern __shared__ double sh_epot[];
int pos = threadIdx.x + blockIdx.x*blockDim.x;
double epot;
if (calc_energy) epot = 0.0;
while (pos < nanglelist) {
calc_angle_force_device<AT, CT, calc_energy, calc_virial>
(pos, anglelist, anglecoef, xyzq, stride,
boxx, boxy, boxz, (AT *)force, epot, virial);
pos += blockDim.x*gridDim.x;
}
// Reduce energy
if (calc_energy) {
reduce_energy(epot, sh_epot, energy_angle);
}
}
//
// Dihedral potential
//
// dihecoef.x = cpd (integer)
// dihecoef.y = cpc
// dihecoef.z = cpsin
// dihecoef.w = cpcos
//
// Out: df, e
//
template <typename T, bool calc_energy>
__forceinline__ __device__
void dihe_pot(const float4* dihecoef, const int ic_in,
const T st, const T ct, T& df, double& e) {
df = (T)0;
if (calc_energy) e = 0.0;
int ic = ic_in;
bool lrep = true;
while (lrep) {
float4 dihecoef_val = dihecoef[ic];
int iper = (int)dihecoef_val.x;
lrep = (iper > 0) ? false : true;
iper = abs(iper);
T e1 = (T)1;
T df1 = (T)0;
T ddf1 = (T)0;
// Calculation of cos(n*phi-phi0) and sin(n*phi-phi0).
for (int nper=1;nper <= iper;nper++) {
ddf1 = e1*ct - df1*st;
df1 = e1*st + df1*ct;
e1 = ddf1;
}
if (calc_energy) e1 = e1*dihecoef_val.w + df1*dihecoef_val.z;
df1 = df1*dihecoef_val.w - ddf1*dihecoef_val.z;
df1 = -iper*df1;
if (calc_energy) {
e1 += (T)1;
if (iper == 0) e1 = (T)1;
}
float arg = dihecoef_val.y;
if (calc_energy) e += arg*e1;
df += arg*df1;
ic++;
}
}
//
// Improper dihedral potential
//
// imdihecoef.x = cid (integer)
// imdihecoef.y = cic
// imdihecoef.z = cisin
// imdihecoef.w = cicos
//
// Out: df, e
//
template<typename T, bool calc_energy>
__forceinline__ __device__
void imdihe_pot(const float4 *dihecoef, const int ic_in,
const T st, const T ct, T& df, double& e) {
df = (T)0;
if (calc_energy) e = 0.0;
float4 dihecoef_val = dihecoef[ic_in];
if ((int)dihecoef_val.x != 0) {
int ic = ic_in;
bool lrep = true;
while (lrep) {
int iper = (int)dihecoef_val.x;
lrep = (iper > 0) ? false : true;
iper = abs(iper);
T e1 = (T)1;
T df1 = (T)0;
T ddf1 = (T)0;
// Calculation of cos(n*phi-phi0) and sin(n*phi-phi0).
for (int nper=1;nper <= iper;nper++) {
ddf1 = e1*ct - df1*st;
df1 = e1*st + df1*ct;
e1 = ddf1;
}
if (calc_energy) e1 = e1*dihecoef_val.w + df1*dihecoef_val.z;
df1 = df1*dihecoef_val.w - ddf1*dihecoef_val.z;
df1 = -iper*df1;
if (calc_energy) {
e1 += (T)1;
if (iper == 0) e1 = (T)1;
}
float arg = dihecoef_val.y;
if (calc_energy) e += arg*e1;
df += arg*df1;
ic++;
if (lrep) dihecoef_val = dihecoef[ic];
}
// use harmonic potential
} else {
// calcul of cos(phi-phi0),sin(phi-phi0) and (phi-phi0).
T ca = ct*dihecoef_val.w + st*dihecoef_val.z;
T sa = st*dihecoef_val.w - ct*dihecoef_val.z;
T ap;
if (ca > (T)0.1) {
ap = asinf(sa);
} else {
//ap = sign(acos(max(ca,-(T)1)),sa);
ap = acosf(max(ca,-(T)1));
ap = (sa > (T)0) ? ap : -ap;
// warning is now triggered at deltaphi=84.26...deg (used to be 90).
//nbent = nbent + 1;
}
df = dihecoef_val.y*ap;
if (calc_energy) e = df*ap;
df *= (T)2;
}
}
template <typename AT, typename CT, bool q_dihe, bool calc_energy, bool calc_virial>
__device__ void calc_dihe_force_device(const int pos,
const dihelist_t* __restrict__ dihelist,
const float4* __restrict__ dihecoef,
const float4* __restrict__ xyzq,
const int stride,
const CT boxx, const CT boxy, const CT boxz,
AT* __restrict__ force, double &epot,
Virial_t* __restrict__ virial) {
int ii = dihelist[pos].i;
int jj = dihelist[pos].j;
int kk = dihelist[pos].k;
int ll = dihelist[pos].l;
int ic = dihelist[pos].itype;
int ish = dihelist[pos].ishift1;
int jsh = dihelist[pos].ishift2;
int lsh = dihelist[pos].ishift3;
// Calculate shift for i-atom
CT ishx, ishy, ishz;
calc_box_shift<CT>(ish, boxx, boxy, boxz, ishx, ishy, ishz);
// Calculate shift for j-atom
CT jshx, jshy, jshz;
calc_box_shift<CT>(jsh, boxx, boxy, boxz, jshx, jshy, jshz);
// Calculate shift for l-atom
CT lshx, lshy, lshz;
calc_box_shift<CT>(lsh, boxx, boxy, boxz, lshx, lshy, lshz);
CT fx = (xyzq[ii].x + ishx) - (xyzq[jj].x + jshx);
CT fy = (xyzq[ii].y + ishy) - (xyzq[jj].y + jshy);
CT fz = (xyzq[ii].z + ishz) - (xyzq[jj].z + jshz);
CT gx = xyzq[jj].x + jshx - xyzq[kk].x;
CT gy = xyzq[jj].y + jshy - xyzq[kk].y;
CT gz = xyzq[jj].z + jshz - xyzq[kk].z;
CT hx = xyzq[ll].x + lshx - xyzq[kk].x;
CT hy = xyzq[ll].y + lshy - xyzq[kk].y;
CT hz = xyzq[ll].z + lshz - xyzq[kk].z;
// A=F^G, B=H^G.
CT ax = fy*gz - fz*gy;
CT ay = fz*gx - fx*gz;
CT az = fx*gy - fy*gx;
CT bx = hy*gz - hz*gy;
CT by = hz*gx - hx*gz;
CT bz = hx*gy - hy*gx;
CT ra2 = ax*ax + ay*ay + az*az;
CT rb2 = bx*bx + by*by + bz*bz;
CT rg = sqrtf(gx*gx + gy*gy + gz*gz);
// if((ra2 <= rxmin2) .or. (rb2 <= rxmin2) .or. (rg <= rxmin)) then
// nlinear = nlinear + 1
// endif
CT rgr = 1.0f / rg;
CT ra2r = 1.0f / ra2;
CT rb2r = 1.0f / rb2;
CT rabr = sqrtf(ra2r*rb2r);
// ct=cos(phi)
CT ct = (ax*bx + ay*by + az*bz)*rabr;
//
// ST=sin(phi), Note that sin(phi).G/|G|=B^A/(|A|.|B|)
// which can be simplify to sin(phi)=|G|H.A/(|A|.|B|)
CT st = rg*rabr*(ax*hx + ay*hy + az*hz);
//
// Energy and derivative contributions.
CT df;
double e;
if (q_dihe) {
dihe_pot<CT, calc_energy>(dihecoef, ic, st, ct, df, e);
} else {
imdihe_pot<CT, calc_energy>(dihecoef, ic, st, ct, df, e);
}
if (calc_energy) epot += e;
//
// Compute derivatives wrt catesian coordinates.
//
// GAA=dE/dphi.|G|/A^2, GBB=dE/dphi.|G|/B^2, FG=F.G, HG=H.G
// FGA=dE/dphi*F.G/(|G|A^2), HGB=dE/dphi*H.G/(|G|B^2)
CT fg = fx*gx + fy*gy + fz*gz;
CT hg = hx*gx + hy*gy + hz*gz;
ra2r *= df;
rb2r *= df;
CT fga = fg*ra2r*rgr;
CT hgb = hg*rb2r*rgr;
CT gaa = ra2r*rg;
CT gbb = rb2r*rg;
// DFi=dE/dFi, DGi=dE/dGi, DHi=dE/dHi.
// Store forces
AT dfx, dfy, dfz;
calc_component_force<AT, CT>(-gaa, ax, ay, az, dfx, dfy, dfz);
write_force<AT>(dfx, dfy, dfz, ii, stride, force);
AT dgx, dgy, dgz;
calc_component_force<AT, CT>(fga, ax, ay, az, -hgb, bx, by, bz,
dgx, dgy, dgz);
write_force<AT>(dgx-dfx, dgy-dfy, dgz-dfz, jj, stride, force);
AT dhx, dhy, dhz;
calc_component_force<AT, CT>(gbb, bx, by, bz, dhx, dhy, dhz);
write_force<AT>(-dhx-dgx, -dhy-dgy, -dhz-dgz, kk, stride, force);
write_force<AT>(dhx, dhy, dhz, ll, stride, force);
// Store virial
if (calc_virial) {
#ifdef USE_DP_SFORCE
if (ish != 13) {
atomicAdd(&virial->sforce_dp[ish][0], (double)(-gaa*ax));
atomicAdd(&virial->sforce_dp[ish][1], (double)(-gaa*ay));
atomicAdd(&virial->sforce_dp[ish][2], (double)(-gaa*az));
}
if (jsh != 13) {
atomicAdd(&virial->sforce_dp[jsh][0], (double)(fga*ax - hgb*bx + gaa*ax));
atomicAdd(&virial->sforce_dp[jsh][1], (double)(fga*ay - hgb*by + gaa*ay));
atomicAdd(&virial->sforce_dp[jsh][2], (double)(fga*az - hgb*bz + gaa*az));
}
if (lsh != 13) {
atomicAdd(&virial->sforce_dp[lsh][0], (double )(gbb*bx));
atomicAdd(&virial->sforce_dp[lsh][1], (double )(gbb*by));
atomicAdd(&virial->sforce_dp[lsh][2], (double )(gbb*bz));
}
#else
dfx /= CONVERT_TO_VIR;
dfy /= CONVERT_TO_VIR;
dfz /= CONVERT_TO_VIR;
dgx /= CONVERT_TO_VIR;
dgy /= CONVERT_TO_VIR;
dgz /= CONVERT_TO_VIR;
if (ish != 13) {
atomicAdd((unsigned long long int *)&virial->sforce_fp[ish][0], llitoulli(dfx));
atomicAdd((unsigned long long int *)&virial->sforce_fp[ish][1], llitoulli(dfy));
atomicAdd((unsigned long long int *)&virial->sforce_fp[ish][2], llitoulli(dfz));
}
if (jsh != 13) {
atomicAdd((unsigned long long int *)&virial->sforce_fp[jsh][0], llitoulli(dgx-dfx));
atomicAdd((unsigned long long int *)&virial->sforce_fp[jsh][1], llitoulli(dgy-dfy));
atomicAdd((unsigned long long int *)&virial->sforce_fp[jsh][2], llitoulli(dgz-dfz));
}
if (lsh != 13) {
dhx /= CONVERT_TO_VIR;
dhy /= CONVERT_TO_VIR;
dhz /= CONVERT_TO_VIR;
atomicAdd((unsigned long long int *)&virial->sforce_fp[lsh][0], llitoulli(dhx));
atomicAdd((unsigned long long int *)&virial->sforce_fp[lsh][1], llitoulli(dhy));
atomicAdd((unsigned long long int *)&virial->sforce_fp[lsh][2], llitoulli(dhz));
}
#endif
}
}
//
// dihecoef.x = cpd (integer)
// dihecoef.y = cpc
// dihecoef.z = cpsin
// dihecoef.w = cpcos
//
//
template <typename AT, typename CT, bool calc_energy, bool calc_virial>
__global__ void calc_dihe_force_kernel(const int ndihelist,
const dihelist_t* __restrict__ dihelist,
const float4* __restrict__ dihecoef,
const float4* __restrict__ xyzq,
const int stride,
const CT boxx, const CT boxy, const CT boxz,
AT* __restrict__ force,
double* __restrict__ energy_dihe,
Virial_t* __restrict__ virial) {
// Amount of shared memory required:
// sh_epot: blockDim.x*sizeof(double)
extern __shared__ double sh_epot[];
int pos = threadIdx.x + blockIdx.x*blockDim.x;
double epot;
if (calc_energy) epot = 0.0;
while (pos < ndihelist) {
calc_dihe_force_device<AT, CT, true, calc_energy, calc_virial>
(pos, dihelist, dihecoef, xyzq, stride,
boxx, boxy, boxz, (AT *)force, epot, virial);
pos += blockDim.x*gridDim.x;
}
// Reduce energy
if (calc_energy) {
reduce_energy(epot, sh_epot, energy_dihe);
}
}
//
// dihecoef.x = cpd (integer)
// dihecoef.y = cpc
// dihecoef.z = cpsin
// dihecoef.w = cpcos
//
//
template <typename AT, typename CT, bool calc_energy, bool calc_virial>
__global__ void calc_imdihe_force_kernel(const int nimdihelist,
const dihelist_t* __restrict__ imdihelist,
const float4* __restrict__ imdihecoef,
const float4* __restrict__ xyzq,
const int stride,
const CT boxx, const CT boxy, const CT boxz,
AT* __restrict__ force,
double* __restrict__ energy_imdihe,
Virial_t* __restrict__ virial) {
// Amount of shared memory required:
// sh_epot: blockDim.x*sizeof(double)
extern __shared__ double sh_epot[];
int pos = threadIdx.x + blockIdx.x*blockDim.x;
double epot;
if (calc_energy) epot = 0.0;
while (pos < nimdihelist) {
calc_dihe_force_device<AT, CT, false, calc_energy, calc_virial>
(pos, imdihelist, imdihecoef, xyzq, stride,
boxx, boxy, boxz, (AT *)force, epot, virial);
pos += blockDim.x*gridDim.x;
}
// Reduce energy
if (calc_energy) {
reduce_energy(epot, sh_epot, energy_imdihe);
}
}
/*
template <typename AT, typename CT, bool calc_energy, bool calc_virial>
__global__ void calc_all_forces_kernel() {
// Amount of shared memory required:
// sh_epot: blockDim.x*sizeof(double)
extern __shared__ double sh_epot[];
int pos = threadIdx.x + blockIdx.x*blockDim.x;
double epot;
if (calc_energy) {
epot = 0.0;
}
if (pos < d_setup.nbondlist) {
calc_bond_force_device<AT, CT, calc_energy, calc_virial>
(pos, d_setup.bondlist, d_setup.bondcoef, d_setup.xyzq,
d_setup.stride, d_setup.boxx, d_setup.boxy, d_setup.boxz,
(AT *)d_setup.force, epot);
} else if (pos < d_setup.nureyblist + d_setup.nbondlist) {
calc_bond_force_device<AT, CT, calc_energy, calc_virial>
(pos - d_setup.nbondlist, d_setup.ureyblist, d_setup.ureybcoef, d_setup.xyzq,
d_setup.stride, d_setup.boxx, d_setup.boxy, d_setup.boxz,
(AT *)d_setup.force, epot);
} else if (pos < d_setup.nanglelist + d_setup.nureyblist + d_setup.nbondlist) {
calc_angle_force_device<AT, CT, calc_energy, calc_virial>
(pos - d_setup.nureyblist - d_setup.nbondlist,
d_setup.anglelist, d_setup.anglecoef, d_setup.xyzq, d_setup.stride,
d_setup.boxx, d_setup.boxy, d_setup.boxz, (AT *)d_setup.force, epot);
} else if (pos < d_setup.ndihelist + d_setup.nanglelist +
d_setup.nureyblist + d_setup.nbondlist) {
calc_dihe_force_device<AT, CT, true, calc_energy, calc_virial>
(pos - d_setup.nanglelist - d_setup.nureyblist - d_setup.nbondlist,
d_setup.dihelist, d_setup.dihecoef, d_setup.xyzq, d_setup.stride,
d_setup.boxx, d_setup.boxy, d_setup.boxz, (AT *)d_setup.force, epot);
} else if (pos < d_setup.nimdihelist + d_setup.ndihelist + d_setup.nanglelist +
d_setup.nureyblist + d_setup.nbondlist) {
calc_dihe_force_device<AT, CT, false, calc_energy, calc_virial>
(pos - d_setup.ndihelist - d_setup.nanglelist - d_setup.nureyblist - d_setup.nbondlist,
d_setup.imdihelist, d_setup.imdihecoef, d_setup.xyzq, d_setup.stride,
d_setup.boxx, d_setup.boxy, d_setup.boxz, (AT *)d_setup.force, epot);
}
// Reduce energy
if (calc_energy) {
reduce_energy(epot, sh_epot, &d_energy_virial.energy_bond);
}
}
*/
//
// Calculates all forces in a single kernel call
// NOTE: Energy calculation is disabled here because we cannot get individual energy terms out
//
template <typename AT, typename CT, bool calc_virial>
__global__ void calc_all_forces_kernel(
const int nbondlist,
const bondlist_t* __restrict__ bondlist,
const float2* __restrict__ bondcoef,
const int nureyblist,
const bondlist_t* __restrict__ ureyblist,
const float2* __restrict__ ureybcoef,
const int nanglelist,
const anglelist_t* __restrict__ anglelist,
const float2* __restrict__ anglecoef,
const int ndihelist,
const dihelist_t* __restrict__ dihelist,
const float4* __restrict__ dihecoef,
const int nimdihelist,
const dihelist_t* __restrict__ imdihelist,
const float4* __restrict__ imdihecoef,
const float4* __restrict__ xyzq,
const int stride,
const CT boxx, const CT boxy, const CT boxz,
AT* __restrict__ force,
Virial_t* __restrict__ virial) {
int pos = threadIdx.x + blockIdx.x*blockDim.x;
// Dummy variable
double epot;
if (pos < nbondlist) {
calc_bond_force_device<AT, CT, false, calc_virial>
(pos, bondlist, bondcoef, xyzq,
stride, boxx, boxy, boxz,
force, epot, virial);
} else if (pos < nureyblist + nbondlist) {
calc_bond_force_device<AT, CT, false, calc_virial>
(pos - nbondlist, ureyblist, ureybcoef, xyzq,
stride, boxx, boxy, boxz,
force, epot, virial);
} else if (pos < nanglelist + nureyblist + nbondlist) {
calc_angle_force_device<AT, CT, false, calc_virial>
(pos - nureyblist - nbondlist,
anglelist, anglecoef, xyzq, stride,
boxx, boxy, boxz, force, epot, virial);
} else if (pos < ndihelist + nanglelist +
nureyblist + nbondlist) {
calc_dihe_force_device<AT, CT, true, false, calc_virial>
(pos - nanglelist - nureyblist - nbondlist,
dihelist, dihecoef, xyzq, stride,
boxx, boxy, boxz, force, epot, virial);
} else if (pos < nimdihelist + ndihelist + nanglelist +
nureyblist + nbondlist) {
calc_dihe_force_device<AT, CT, false, false, calc_virial>
(pos - ndihelist - nanglelist - nureyblist - nbondlist,
imdihelist, imdihecoef, xyzq, stride,
boxx, boxy, boxz, force, epot, virial);
}
}
//---------------------------------------------------------------------------------------------------------
//
// Setups lists
//
__device__ void setup_bondlist_kernel(const int i,
const int* __restrict__ bond_tbl, const bond_t* __restrict__ bond,
bondlist_t* __restrict__ bondlist, const float4* __restrict__ xyzq,
const float3 half_box, const int*__restrict__ glo2loc_ind) {
int j = bond_tbl[i];
bond_t bondv = bond[j];
bondlist_t bondlistv;
bondlistv.i = glo2loc_ind[bondv.i];
bondlistv.j = glo2loc_ind[bondv.j];
bondlistv.itype = bondv.itype;
float4 xyzq_i = xyzq[bondlistv.i];
float4 xyzq_j = xyzq[bondlistv.j];
bondlistv.ishift = calc_ishift(xyzq_i, xyzq_j, half_box);
bondlist[i] = bondlistv;
}
__device__ void setup_anglelist_kernel(const int i,
const int* __restrict__ angle_tbl, const angle_t* __restrict__ angle,
anglelist_t* __restrict__ anglelist, const float4* __restrict__ xyzq,
const float3 half_box, const int*__restrict__ glo2loc_ind) {
int j = angle_tbl[i];
angle_t anglev = angle[j];
anglelist_t anglelistv;
anglelistv.i = glo2loc_ind[anglev.i];
anglelistv.j = glo2loc_ind[anglev.j];
anglelistv.k = glo2loc_ind[anglev.k];
anglelistv.itype = anglev.itype;
float4 xyzq_i = xyzq[anglelistv.i];
float4 xyzq_j = xyzq[anglelistv.j];
float4 xyzq_k = xyzq[anglelistv.k];
anglelistv.ishift1 = calc_ishift(xyzq_i, xyzq_j, half_box);
anglelistv.ishift2 = calc_ishift(xyzq_k, xyzq_j, half_box);
anglelist[i] = anglelistv;
}
__device__ void setup_dihelist_kernel(const int i,
const int* __restrict__ dihe_tbl, const dihe_t* __restrict__ dihe,
dihelist_t* __restrict__ dihelist, const float4* __restrict__ xyzq,
const float3 half_box, const int*__restrict__ glo2loc_ind) {
int j = dihe_tbl[i];
dihe_t dihev = dihe[j];
dihelist_t dihelistv;
dihelistv.i = glo2loc_ind[dihev.i];
dihelistv.j = glo2loc_ind[dihev.j];
dihelistv.k = glo2loc_ind[dihev.k];
dihelistv.l = glo2loc_ind[dihev.l];
dihelistv.itype = dihev.itype;
float4 xyzq_i = xyzq[dihelistv.i];
float4 xyzq_j = xyzq[dihelistv.j];
float4 xyzq_k = xyzq[dihelistv.k];
float4 xyzq_l = xyzq[dihelistv.l];
dihelistv.ishift1 = calc_ishift(xyzq_i, xyzq_k, half_box);
dihelistv.ishift2 = calc_ishift(xyzq_j, xyzq_k, half_box);
dihelistv.ishift3 = calc_ishift(xyzq_l, xyzq_k, half_box);
dihelist[i] = dihelistv;
}
__device__ void setup_cmaplist_kernel(const int i,
const int* __restrict__ cmap_tbl, const cmap_t* __restrict__ cmap,
cmaplist_t* __restrict__ cmaplist, const float4* __restrict__ xyzq,
const float3 half_box, const int*__restrict__ glo2loc_ind) {
int j = cmap_tbl[i];
cmap_t cmapv = cmap[j];
cmaplist_t cmaplistv;
cmaplistv.i1 = glo2loc_ind[cmapv.i1];
cmaplistv.j1 = glo2loc_ind[cmapv.j1];
cmaplistv.k1 = glo2loc_ind[cmapv.k1];
cmaplistv.l1 = glo2loc_ind[cmapv.l1];
cmaplistv.i2 = glo2loc_ind[cmapv.i2];
cmaplistv.j2 = glo2loc_ind[cmapv.j2];
cmaplistv.k2 = glo2loc_ind[cmapv.k2];
cmaplistv.l2 = glo2loc_ind[cmapv.l2];
cmaplistv.itype = cmapv.itype;
float4 xyzq_i1 = xyzq[cmaplistv.i1];
float4 xyzq_j1 = xyzq[cmaplistv.j1];
float4 xyzq_k1 = xyzq[cmaplistv.k1];
float4 xyzq_l1 = xyzq[cmaplistv.l1];
float4 xyzq_i2 = xyzq[cmaplistv.i2];
float4 xyzq_j2 = xyzq[cmaplistv.j2];
float4 xyzq_k2 = xyzq[cmaplistv.k2];
float4 xyzq_l2 = xyzq[cmaplistv.l2];
cmaplistv.ishift1 = calc_ishift(xyzq_i1, xyzq_k1, half_box);
cmaplistv.ishift2 = calc_ishift(xyzq_j1, xyzq_k1, half_box);
cmaplistv.ishift3 = calc_ishift(xyzq_l1, xyzq_k1, half_box);
cmaplist[i] = cmaplistv;
}
__global__ void setup_list_kernel(const int nbond_tbl, const int* __restrict__ bond_tbl,
const bond_t* __restrict__ bond, bondlist_t* __restrict__ bondlist,
const int nureyb_tbl, const int* __restrict__ ureyb_tbl,
const bond_t* __restrict__ ureyb, bondlist_t* __restrict__ ureyblist,
const int nangle_tbl, const int* __restrict__ angle_tbl,
const angle_t* __restrict__ angle, anglelist_t* __restrict__ anglelist,
const int ndihe_tbl, const int* __restrict__ dihe_tbl,
const dihe_t* __restrict__ dihe, dihelist_t* __restrict__ dihelist,
const int nimdihe_tbl, const int* __restrict__ imdihe_tbl,
const dihe_t* __restrict__ imdihe, dihelist_t* __restrict__ imdihelist,
const int ncmap_tbl, const int* __restrict__ cmap_tbl,
const cmap_t* __restrict__ cmap, cmaplist_t* __restrict__ cmaplist,
const float4* __restrict__ xyzq,
const float3 half_box, const int*__restrict__ glo2loc_ind) {
int pos = threadIdx.x + blockIdx.x*blockDim.x;
if (pos < nbond_tbl) {
setup_bondlist_kernel(pos, bond_tbl, bond, bondlist, xyzq, half_box, glo2loc_ind);
} else if (pos < nbond_tbl + nureyb_tbl) {
setup_bondlist_kernel(pos-nbond_tbl, ureyb_tbl, ureyb, ureyblist, xyzq, half_box, glo2loc_ind);
} else if (pos < nbond_tbl + nureyb_tbl + nangle_tbl) {
setup_anglelist_kernel(pos-nbond_tbl-nureyb_tbl, angle_tbl, angle, anglelist, xyzq, half_box, glo2loc_ind);
} else if (pos < nbond_tbl + nureyb_tbl + nangle_tbl + ndihe_tbl) {
setup_dihelist_kernel(pos-nbond_tbl-nureyb_tbl-nangle_tbl,
dihe_tbl, dihe, dihelist, xyzq, half_box, glo2loc_ind);
} else if (pos < nbond_tbl + nureyb_tbl + nangle_tbl + ndihe_tbl + nimdihe_tbl) {
setup_dihelist_kernel(pos-nbond_tbl-nureyb_tbl-nangle_tbl-ndihe_tbl,
imdihe_tbl, imdihe, imdihelist, xyzq, half_box, glo2loc_ind);
} else if (pos < nbond_tbl + nureyb_tbl + nangle_tbl + ndihe_tbl + nimdihe_tbl + ncmap_tbl) {
setup_cmaplist_kernel(pos-nbond_tbl-nureyb_tbl-nangle_tbl-ndihe_tbl-nimdihe_tbl,
cmap_tbl, cmap, cmaplist, xyzq, half_box, glo2loc_ind);
}
}
//-----------------------------------------------------------------------------------------------------------
//#############################################################################################
//
// Class creator
//
template <typename AT, typename CT>
CudaBondedForce<AT, CT>::CudaBondedForce(CudaEnergyVirial &energyVirial,
const char *nameBond, const char *nameUreyb, const char *nameAngle,
const char *nameDihe, const char *nameImdihe, const char *nameCmap) :
energyVirial(energyVirial) {
// Insert energy terms
if (nameBond != NULL) {
energyVirial.insert(nameBond);
strBond = nameBond;
}