forked from christophmschaefer/miluphcuda
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathpredictor_corrector.cu
892 lines (823 loc) · 32.4 KB
/
predictor_corrector.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
/**
* @author Christoph Schaefer [email protected]
*
* @section LICENSE
* Copyright (c) 2019 Christoph Schaefer
*
* This file is part of miluphcuda.
*
* miluphcuda is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* miluphcuda is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with miluphcuda. If not, see <http://www.gnu.org/licenses/>.
*
*/
#include "predictor_corrector_euler.h"
#include "timeintegration.h"
#include "parameter.h"
#include "memory_handling.h"
#include "miluph.h"
#include "pressure.h"
#include "rhs.h"
#include <float.h>
/* predictor corrector scheme with an initial step of dt/2 and the corrector step with dt */
extern __device__ double endTimeD, currentTimeD;
extern __device__ double substep_currentTimeD;
extern __device__ double dt;
extern __device__ int blockCount;
extern __device__ double emin_d;
extern __device__ double Smin_d;
extern __device__ double rhomin_d;
extern __device__ double damagemin_d;
extern __device__ double alphamin_d;
extern __device__ double betamin_d;
extern __device__ double alpha_epspormin_d;
extern __device__ double epsilon_vmin_d;
extern __device__ int pressureChangeSmallEnough;
extern __device__ double maxpressureDiff;
extern double L_ini;
__global__ void CorrectorStep()
{
register int i;
#if SOLID
register int j;
register int k;
#endif
#if GRAVITATING_POINT_MASSES
// pointmass loop
for (i = threadIdx.x + blockIdx.x * blockDim.x; i < numPointmasses; i+= blockDim.x * gridDim.x) {
pointmass.x[i] = pointmass.x[i] + dt * predictor_pointmass.vx[i];
#if DIM > 1
pointmass.y[i] = pointmass.y[i] + dt * predictor_pointmass.vy[i];
pointmass.vy[i] = pointmass.vy[i] + dt * predictor_pointmass.ay[i];
pointmass.ay[i] = predictor_pointmass.ay[i];
#endif
pointmass.vx[i] = pointmass.vx[i] + dt * predictor_pointmass.ax[i];
pointmass.ax[i] = predictor_pointmass.ax[i];
#if DIM == 3
pointmass.z[i] = pointmass.z[i] + dt * predictor_pointmass.vz[i];
pointmass.vz[i] = pointmass.vz[i] + dt * predictor_pointmass.az[i];
pointmass.az[i] = predictor_pointmass.az[i];
#endif
}
#endif // GRAVITATING_POINT_MASSES
// particle loop
for (i = threadIdx.x + blockIdx.x * blockDim.x; i < numParticles; i+= blockDim.x * gridDim.x) {
p.x[i] = p.x[i] + dt * predictor.dxdt[i];
#if DIM > 1
p.y[i] = p.y[i] + dt * predictor.dydt[i];
p.vy[i] = p.vy[i] + dt * predictor.ay[i];
p.ay[i] = predictor.ay[i];
#endif
p.vx[i] = p.vx[i] + dt * predictor.ax[i];
p.ax[i] = predictor.ax[i];
#if DIM == 3
p.z[i] = p.z[i] + dt * predictor.dzdt[i];
p.vz[i] = p.vz[i] + dt * predictor.az[i];
p.az[i] = predictor.az[i];
#endif
#if INTEGRATE_DENSITY
p.rho[i] = p.rho[i] + dt * predictor.drhodt[i];
p.drhodt[i] = predictor.drhodt[i];
#else
p.rho[i] = p.rho[i];
#endif
#if INTEGRATE_ENERGY
p.e[i] = p.e[i] + dt * predictor.dedt[i];
p.dedt[i] = predictor.dedt[i];
#endif
#if FRAGMENTATION
p.d[i] = p.d[i] + dt * predictor.dddt[i];
p.dddt[i] = predictor.dddt[i];
#endif
#if INTEGRATE_SML
p.h[i] = p.h[i] + dt * predictor.dhdt[i];
p.dhdt[i] = predictor.dhdt[i];
#else
p.h[i] = predictor.h[i];
#endif
#if JC_PLASTICITY
p.ep[i] = p.ep[i] + dt * predictor.edotp[i];
p.edotp[i] = predictor.edotp[i];
p.T[i] = p.T[i] + dt * predictor.dTdt[i];
p.dTdt[i] = predictor.dTdt[i];
#endif
#if SIRONO_POROSITY
p.rho_0prime[i] = p.rho_0prime[i];
p.rho_c_plus[i] = p.rho_c_plus[i];
p.rho_c_minus[i] = p.rho_c_minus[i];
p.compressive_strength[i] = p.compressive_strength[i];
p.tensile_strength[i] = p.tensile_strength[i];
p.shear_strength[i] = p.shear_strength[i];
p.K[i] = p.K[i];
p.flag_rho_0prime[i] = p.flag_rho_0prime[i];
p.flag_plastic[i] = p.flag_plastic[i];
#endif
#if EPSALPHA_POROSITY
p.alpha_epspor[i] = p.alpha_epspor[i] + dt * predictor.dalpha_epspordt[i];
p.epsilon_v[i] = p.epsilon_v[i] + dt * predictor.depsilon_vdt[i];
p.dalpha_epspordt[i] = predictor.dalpha_epspordt[i];
p.depsilon_vdt[i] = predictor.depsilon_vdt[i];
#endif
#if INVISCID_SPH
p.beta[i] = p.beta[i] + dt * predictor.dbetadt[i];
p.dbetadt[i] = predictor.dbetadt[i];
#endif
#if SOLID
for (j = 0; j < DIM; j++) {
for (k = 0; k < DIM; k++) {
p.S[stressIndex(i,j,k)] = p.S[stressIndex(i,j,k)] + dt * predictor.dSdt[stressIndex(i,j,k)];
p.dSdt[stressIndex(i,j,k)] = predictor.dSdt[stressIndex(i,j,k)];
}
}
#endif
}
}
__global__ void PredictorStep()
{
register int i;
#if SOLID
register int j;
register int k;
#endif
#if GRAVITATING_POINT_MASSES
// pointmass loop
for (i = threadIdx.x + blockIdx.x * blockDim.x; i < numPointmasses; i+= blockDim.x * gridDim.x) {
predictor_pointmass.x[i] = pointmass.x[i] + dt/2 * pointmass.vx[i];
predictor_pointmass.vx[i] = pointmass.vx[i] + dt/2 * pointmass.ax[i];
#if DIM > 1
predictor_pointmass.y[i] = pointmass.y[i] + dt/2 * pointmass.vy[i];
predictor_pointmass.vy[i] = pointmass.vy[i] + dt/2 * pointmass.ay[i];
#endif
#if DIM > 2
predictor_pointmass.z[i] = pointmass.z[i] + dt/2 * pointmass.vz[i];
predictor_pointmass.vz[i] = pointmass.vz[i] + dt/2 * pointmass.az[i];
#endif
}
#endif // GRAVITATING_POINT_MASSES
// particle loop
for (i = threadIdx.x + blockIdx.x * blockDim.x; i < numParticles; i+= blockDim.x * gridDim.x) {
predictor.x[i] = p.x[i] + dt/2 * p.dxdt[i];
predictor.vx[i] = p.vx[i] + dt/2 * p.ax[i];
#if DIM > 1
predictor.y[i] = p.y[i] + dt/2 * p.dydt[i];
predictor.vy[i] = p.vy[i] + dt/2 * p.ay[i];
#endif
#if DIM > 2
predictor.z[i] = p.z[i] + dt/2 * p.dzdt[i];
predictor.vz[i] = p.vz[i] + dt/2 * p.az[i];
#endif
#if INTEGRATE_DENSITY
predictor.rho[i] = p.rho[i] + dt/2 * p.drhodt[i];
#else
predictor.rho[i] = p.rho[i];
#endif
#if INTEGRATE_ENERGY
predictor.e[i] = p.e[i] + dt/2 * p.dedt[i];
#endif
#if INTEGRATE_SML
predictor.h[i] = p.h[i] + dt/2 * p.dhdt[i];
#else
predictor.h[i] = p.h[i];
#endif
#if PALPHA_POROSITY
// p points to p_device and p.p is the pressure at the start of the timestep
// while predictor changes during the adaptive time step
predictor.pold[i] = p.p[i];
predictor.alpha_jutzi[i] = p.alpha_jutzi[i] + dt/2 * p.dalphadt[i];
predictor.alpha_jutzi_old[i] = p.alpha_jutzi_old[i];
#endif
#if FRAGMENTATION
predictor.d[i] = p.d[i] + dt/2 * p.dddt[i];
predictor.numActiveFlaws[i] = p.numActiveFlaws[i];
#if PALPHA_POROSITY
predictor.damage_porjutzi[i] = p.damage_porjutzi[i] + dt/2 * p.ddamage_porjutzidt[i];
#endif
#endif
#if JC_PLASTICITY
predictor.ep[i] = p.ep[i] + dt/2 * p.edotp[i];
predictor.T[i] = p.T[i] + dt/2 * p.dTdt[i];
#endif
#if SIRONO_POROSITY
predictor.rho_0prime[i] = p.rho_0prime[i];
predictor.rho_c_plus[i] = p.rho_c_plus[i];
predictor.rho_c_minus[i] = p.rho_c_minus[i];
predictor.compressive_strength[i] = p.compressive_strength[i];
predictor.tensile_strength[i] = p.tensile_strength[i];
predictor.shear_strength[i] = p.shear_strength[i];
predictor.K[i] = p.K[i];
predictor.flag_rho_0prime[i] = p.flag_rho_0prime[i];
predictor.flag_plastic[i] = p.flag_plastic[i];
#endif
#if EPSALPHA_POROSITY
predictor.alpha_epspor[i] = p.alpha_epspor[i] + dt/2 * p.dalpha_epspordt[i];
predictor.epsilon_v[i] = p.epsilon_v[i] + dt/2 * p.depsilon_vdt[i];
#endif
#if INVISCID_SPH
predictor.beta[i] = p.beta[i] + dt/2 * p.dbetadt[i];
#endif
#if SOLID
for (j = 0; j < DIM; j++) {
for (k = 0; k < DIM; k++) {
predictor.S[stressIndex(i,j,k)] = p.S[stressIndex(i,j,k)] + dt/2 * p.dSdt[stressIndex(i,j,k)];
}
}
#endif
}
}
#if PALPHA_POROSITY
__global__ void CorrectorStepPorous()
{
register int i;
#if SOLID
register int j;
register int k;
#endif
#if GRAVITATING_POINT_MASSES
// pointmass loop
for (i = threadIdx.x + blockIdx.x * blockDim.x; i < numPointmasses; i+= blockDim.x * gridDim.x) {
predictor_pointmass.x[i] = pointmass.x[i] + dt * predictor_pointmass.vx[i];
#if DIM > 1
predictor_pointmass.y[i] = pointmass.y[i] + dt * predictor_pointmass.vy[i];
predictor_pointmass.vy[i] = pointmass.vy[i] + dt * predictor_pointmass.ay[i];
predictor_pointmass.ay[i] = predictor_pointmass.ay[i];
#endif
predictor_pointmass.vx[i] = pointmass.vx[i] + dt * predictor_pointmass.ax[i];
predictor_pointmass.ax[i] = predictor_pointmass.ax[i];
#if DIM == 3
predictor_pointmass.z[i] = pointmass.z[i] + dt * predictor_pointmass.vz[i];
predictor_pointmass.vz[i] = pointmass.vz[i] + dt * predictor_pointmass.az[i];
predictor_pointmass.az[i] = predictor_pointmass.az[i];
#endif
}
#endif // GRAVITATING_POINT_MASSES
for (i = threadIdx.x + blockIdx.x * blockDim.x; i < numParticles; i+= blockDim.x * gridDim.x) {
predictor.x[i] = p.x[i] + dt * predictor.dxdt[i];
predictor.y[i] = p.y[i] + dt * predictor.dydt[i];
predictor.vx[i] = p.vx[i] + dt * predictor.ax[i];
predictor.vy[i] = p.vy[i] + dt * predictor.ay[i];
#if DIM == 3
predictor.z[i] = p.z[i] + dt * predictor.dzdt[i];
predictor.vz[i] = p.vz[i] + dt * predictor.az[i];
#endif
#if INTEGRATE_DENSITY
predictor.rho[i] = p.rho[i] + dt * predictor.drhodt[i];
#else
predictor.rho[i] = p.rho[i];
#endif
#if INTEGRATE_SML
predictor.h[i] = p.h[i] + dt * predictor.dhdt[i];
#else
predictor.h[i] = p.h[i];
#endif
#if INTEGRATE_ENERGY
predictor.e[i] = p.e[i] + dt * predictor.dedt[i];
#endif
#if FRAGMENTATION
predictor.d[i] = p.d[i] + dt * predictor.dddt[i];
#if PALPHA_POROSITY
if (predictor.p[i] > p.p[i]) {
predictor.damage_porjutzi[i] = p.damage_porjutzi[i] + dt * predictor.ddamage_porjutzidt[i];
} else {
predictor.d[i] = predictor.d[i];
predictor.damage_porjutzi[i] = p.damage_porjutzi[i];
}
#endif
#endif
#if INVISCID_SPH
predictor.beta[i] = p.beta[i] + dt * predictor.dbetadt[i];
#endif
#if SOLID
for (j = 0; j < DIM; j++) {
for (k = 0; k < DIM; k++) {
predictor.S[stressIndex(i,j,k)] = p.S[stressIndex(i,j,k)] + dt * predictor.dSdt[stressIndex(i,j,k)];
}
}
#endif
#if PALPHA_POROSITY
/* check if we have compaction and change alpha accordingly */
if (predictor.p[i] > p.p[i]) {
predictor.alpha_jutzi[i] = p.alpha_jutzi[i] + dt * predictor.dalphadt[i];
} else {
predictor.alpha_jutzi[i] = p.alpha_jutzi[i];
}
predictor.alpha_jutzi_old[i] = p.alpha_jutzi_old[i];
#endif
}
}
/* check the pressure change to avoid large deviation from the crush-curve */
__global__ void pressureChangeCheck(double *maxpressureDiffPerBlock)
{
__shared__ double sharedMaxpressureDiff[NUM_THREADS_PC_INTEGRATOR];
double localMaxpressureDiff = 0.0;
double tmp = 0.0;
int i, j, k, m;
maxpressureDiff = 0.0;
for (i = threadIdx.x + blockIdx.x * blockDim.x; i < numParticles; i+= blockDim.x * gridDim.x) {
if (matEOS[p_rhs.materialId[i]] == EOS_TYPE_JUTZI || matEOS[p_rhs.materialId[i]] == EOS_TYPE_JUTZI_MURNAGHAN) {
// cms - 20190626
// first rhs is called at beginning of timestep with predictor
// and at the end with p_device
tmp = (p.p[i] - predictor.pold[i]);
localMaxpressureDiff = max(tmp, localMaxpressureDiff);
}
}
i = threadIdx.x;
sharedMaxpressureDiff[i] = localMaxpressureDiff;
for (j = NUM_THREADS_PC_INTEGRATOR / 2; j > 0; j /= 2) {
__syncthreads();
if (i < j) {
k = i + j;
sharedMaxpressureDiff[i] = localMaxpressureDiff = max(localMaxpressureDiff, sharedMaxpressureDiff[k]);
}
}
// write block result to global memory
if (i == 0) {
k = blockIdx.x;
maxpressureDiffPerBlock[k] = localMaxpressureDiff;
m = gridDim.x - 1;
if (m == atomicInc((unsigned int *)&blockCount, m)) {
// last block, so combine all block results
for (j = 0; j <= m; j++) {
localMaxpressureDiff = max(localMaxpressureDiff, maxpressureDiffPerBlock[j]);
}
maxpressureDiff = localMaxpressureDiff;
// reset block count
blockCount = 0;
}
if (maxpressureDiff > max_abs_pressure_change) {
printf("maxpressure change %e\n", maxpressureDiff);
pressureChangeSmallEnough = FALSE;
dt = 0.25 * dt;
dt = min(dt, endTimeD - currentTimeD);
} else {
pressureChangeSmallEnough = TRUE;
// currentTimeD += dt;
}
}
}
#endif
__global__ void setTimestep(double *forcesPerBlock, double *courantPerBlock, double *dtSPerBlock, double *dtePerBlock, double *dtrhoPerBlock, double *dtdamagePerBlock, double *dtalphaPerBlock, double *dtartviscPerBlock, double *dtbetaPerBlock, double *dtalpha_epsporPerBlock, double *dtepsilon_vPerBlock)
{
#define SAFETY_FIRST 0.7
__shared__ double sharedForces[NUM_THREADS_LIMITTIMESTEP];
__shared__ double sharedCourant[NUM_THREADS_LIMITTIMESTEP];
__shared__ double sharedArtVisc[NUM_THREADS_LIMITTIMESTEP];
__shared__ double sharedS[NUM_THREADS_LIMITTIMESTEP];
__shared__ double sharede[NUM_THREADS_LIMITTIMESTEP];
__shared__ double sharedrho[NUM_THREADS_LIMITTIMESTEP];
__shared__ double shareddamage[NUM_THREADS_LIMITTIMESTEP];
__shared__ double sharedalpha[NUM_THREADS_LIMITTIMESTEP];
__shared__ double sharedbeta[NUM_THREADS_LIMITTIMESTEP];
__shared__ double sharedalpha_epspor[NUM_THREADS_LIMITTIMESTEP];
__shared__ double sharedepsilon_v[NUM_THREADS_LIMITTIMESTEP];
int i, j, k, m;
int d, dd;
int index;
int hasEnergy;
double forces = DBL_MAX, courant = DBL_MAX;
double dtx = DBL_MAX;
double dtS = DBL_MAX;
double dtrho = DBL_MAX;
double dte = DBL_MAX;
double dtdamage = DBL_MAX;
double dtalpha = DBL_MAX;
double dtbeta = DBL_MAX;
double dtalpha_epspor = DBL_MAX;
double dtepsilon_v = DBL_MAX;
double temp;
double sml;
int matId;
#if SOLID
double myS, dS;
#endif
double ax, ay;
#if DIM == 3
double az;
#endif
double dtartvisc = DBL_MAX;
for (i = threadIdx.x + blockIdx.x * blockDim.x; i < numParticles; i+= blockDim.x * gridDim.x) {
matId = p_rhs.materialId[i];
#if INTEGRATE_ENERGY
hasEnergy = 0;
switch (matEOS[matId]) {
case (EOS_TYPE_TILLOTSON):
hasEnergy = 1;
break;
case (EOS_TYPE_JUTZI):
hasEnergy = 1;
break;
case (EOS_TYPE_SIRONO):
hasEnergy = 1;
break;
case (EOS_TYPE_EPSILON):
hasEnergy = 1;
break;
case (EOS_TYPE_ANEOS):
hasEnergy = 1;
break;
default:
hasEnergy = 0;
break;
}
#endif
ax = p.ax[i];
#if DIM > 1
ay = p.ay[i];
#endif
#if DIM == 3
az = p.az[i];
#endif
temp = ax*ax;
#if DIM > 1
temp += + ay*ay;
#endif
#if DIM == 3
temp += az*az;
#endif
;
sml = p.h[i];
temp = sqrt(sml / sqrt(temp));
forces = min(forces, temp);
temp = sml / p.cs[i];
courant = min(courant, temp);
#define CFL 0.7
#if ARTIFICIAL_VISCOSITY
temp = CFL * sml / (p.cs[i] + 1.2 * (matAlpha[matId]) * p.cs[i] + matBeta[matId] *
p.muijmax[i]);
dtartvisc = min(dtartvisc, temp);
#endif
#if INVISCID_SPH
if (p.dbetadt[i] != 0) {
temp = SAFETY_FIRST * (fabs(p.beta[i])+betamin_d)/fabs(p.dbetadt[i]);
dtbeta = min(temp, dtbeta);
}
#endif
#if SOLID
myS = 0;
dS = 0;
for (d = 0; d < DIM; d++) {
for (dd = 0; dd < DIM; dd++) {
index = i*DIM*DIM+d*DIM+dd;
myS += p.S[index]*p.S[index];
dS += p.dSdt[index]*p.dSdt[index];
}
}
if (dS != 0) {
temp = SAFETY_FIRST * sqrt((myS+Smin_d)/dS);
dtS = min(temp, dtS);
}
#endif
#if INTEGRATE_DENSITY
if (p.drhodt[i] != 0) {
temp = SAFETY_FIRST * (fabs(p.rho[i])+rhomin_d)/fabs(p.drhodt[i]);
dtrho = min(temp, dtrho);
}
#endif
#if INTEGRATE_ENERGY
if (p.dedt[i] != 0 && hasEnergy) {
temp = SAFETY_FIRST * (fabs(p.e[i])+emin_d)/fabs(p.dedt[i]);
dte = min(temp, dte);
}
#endif
#if PALPHA_POROSITY
if (p.dalphadt[i] != 0) {
temp = 1.0e-2 / fabs(p.dalphadt[i]);
dtalpha = min(temp, dtalpha);
}
#endif
#if EPSALPHA_POROSITY
if (p.dalpha_epspordt[i] != 0) {
temp = 1.0e-1 / fabs(p.dalpha_epspordt[i]);
dtalpha_epspor = min(temp, dtalpha_epspor);
}
if (p.depsilon_vdt[i] != 0) {
temp = SAFETY_FIRST * (fabs(p.epsilon_v[i])+epsilon_vmin_d)/fabs(p.depsilon_vdt[i]);
dtepsilon_v = min(temp, dtepsilon_v);
}
#endif
#if FRAGMENTATION
if (p.dddt[i] != 0) {
temp = SAFETY_FIRST * (fabs(p.d[i])+damagemin_d)/fabs(p.dddt[i]);
dtdamage = min(temp, dtdamage);
}
#endif
}
i = threadIdx.x;
sharedForces[i] = forces;
sharedCourant[i] = courant;
sharedS[i] = dtS;
sharede[i] = dte;
sharedrho[i] = dtrho;
shareddamage[i] = dtdamage;
sharedalpha[i] = dtalpha;
sharedbeta[i] = dtbeta;
sharedalpha_epspor[i] = dtalpha_epspor;
sharedepsilon_v[i] = dtepsilon_v;
#if ARTIFICIAL_VISCOSITY
sharedArtVisc[i] = dtartvisc;
#endif
for (j = NUM_THREADS_LIMITTIMESTEP / 2; j > 0; j /= 2) {
__syncthreads();
if (i < j) {
k = i + j;
sharedForces[i] = forces = min(forces, sharedForces[k]);
sharedCourant[i] = courant = min(courant, sharedCourant[k]);
sharedS[i] = dtS = min(dtS, sharedS[k]);
sharede[i] = dte = min(dte, sharede[k]);
sharedrho[i] = dtrho = min(dtrho, sharedrho[k]);
shareddamage[i] = dtdamage = min(dtdamage, shareddamage[k]);
sharedalpha[i] = dtalpha = min(dtalpha, sharedalpha[k]);
sharedalpha_epspor[i] = dtalpha_epspor = min(dtalpha_epspor, sharedalpha_epspor[k]);
sharedepsilon_v[i] = dtepsilon_v = min(dtepsilon_v, sharedepsilon_v[k]);
#if ARTIFICIAL_VISCOSITY
sharedArtVisc[i] = dtartvisc = min(dtartvisc, sharedArtVisc[k]);
#endif
#if INVISCID_SPH
sharedbeta[i] = dtbeta = min(dtbeta, sharedbeta[i]);
#endif
}
}
// write block result to global memory
if (i == 0) {
k = blockIdx.x;
forcesPerBlock[k] = forces;
courantPerBlock[k] = courant;
dtSPerBlock[k] = dtS;
dtePerBlock[k] = dte;
dtrhoPerBlock[k] = dtrho;
dtdamagePerBlock[k] = dtdamage;
dtalphaPerBlock[k] = dtalpha;
dtalpha_epsporPerBlock[k] = dtalpha_epspor;
dtepsilon_vPerBlock[k] = dtepsilon_v;
#if ARTIFICIAL_VISCOSITY
dtartviscPerBlock[k] = dtartvisc;
#endif
#if INVISCID_SPH
dtbetaPerBlock[k] = dtbeta;
#endif
m = gridDim.x - 1;
if (m == atomicInc((unsigned int *)&blockCount, m)) {
// last block, so combine all block results
for (j = 0; j <= m; j++) {
forces = min(forces, forcesPerBlock[j]);
courant = min(courant, courantPerBlock[j]);
dtS = min(dtS, dtSPerBlock[j]);
dte = min(dte, dtePerBlock[j]);
dtrho = min(dtrho, dtrhoPerBlock[j]);
dtdamage = min(dtdamage, dtdamagePerBlock[j]);
dtalpha = min(dtalpha, dtalphaPerBlock[j]);
dtalpha_epspor = min(dtalpha_epspor, dtalpha_epsporPerBlock[j]);
dtepsilon_v = min(dtepsilon_v, dtepsilon_vPerBlock[j]);
#if ARTIFICIAL_VISCOSITY
dtartvisc = min(dtartvisc, dtartviscPerBlock[j]);
#endif
#if INVISCID_SPH
dtbeta = min(dtbeta, dtbetaPerBlock[j]);
#endif
}
// set new timestep
dt = dtx = min(COURANT*courant, forces*0.2);
#if SOLID
dt = min(dt, dtS);
#endif
#if INTEGRATE_ENERGY
dt = min(dt, dte);
#endif
#if INTEGRATE_DENSITY
dt = min(dt, dtrho);
#endif
#if FRAGMENTATION
dt = min(dt, dtdamage);
#endif
#if PALPHA_POROSITY
dt = min(dt, dtalpha);
#endif
#if EPSALPHA_POROSITY
dt = min(dt, dtalpha_epspor);
dt = min(dt, dtepsilon_v);
#endif
#if ARTIFICIAL_VISCOSITY
dt = min(dt, dtartvisc);
#endif
#if INVISCID_SPH
dt = min(dt, dtbeta);
#endif
dt = min(dt, endTimeD - currentTimeD);
printf("Time Step Information: dt(v and x): %e dtS: %e dte: %e dtrho: %e dtdamage: %e dtalpha: %e dtalpha_epspor: %e dtepsilon_v: %e\n",
dtx, dtS, dte, dtrho, dtdamage, dtalpha, dtalpha_epspor, dtepsilon_v);
printf("time: %e timestep set to %e, integrating until %e \n", currentTimeD, dt, endTimeD);
#if !PALPHA_POROSITY
currentTimeD += dt;
#endif
// reset block count
blockCount = 0;
}
}
}
void predictor_corrector()
{
double *courantPerBlock, *forcesPerBlock;
double *dtSPerBlock, *dtePerBlock, *dtrhoPerBlock;
double *dtdamagePerBlock;
double *dtalphaPerBlock;
double *dtartviscPerBlock;
double *maxpressureDiffPerBlock;
double *dtbetaPerBlock;
double *dtalpha_epsporPerBlock;
double *dtepsilon_vPerBlock;
int pressureChangeSmallEnough_host;
double maxpressureDiff_host;
double maxpressureDiff_previous;
int maxpressureDiff_cnt;
cudaVerify(cudaMalloc((void**)&courantPerBlock, sizeof(double)*numberOfMultiprocessors));
cudaVerify(cudaMalloc((void**)&forcesPerBlock, sizeof(double)*numberOfMultiprocessors));
cudaVerify(cudaMalloc((void**)&dtSPerBlock, sizeof(double)*numberOfMultiprocessors));
cudaVerify(cudaMalloc((void**)&dtePerBlock, sizeof(double)*numberOfMultiprocessors));
cudaVerify(cudaMalloc((void**)&dtrhoPerBlock, sizeof(double)*numberOfMultiprocessors));
cudaVerify(cudaMalloc((void**)&dtdamagePerBlock, sizeof(double)*numberOfMultiprocessors));
cudaVerify(cudaMalloc((void**)&dtalphaPerBlock, sizeof(double)*numberOfMultiprocessors));
cudaVerify(cudaMalloc((void**)&dtbetaPerBlock, sizeof(double)*numberOfMultiprocessors));
cudaVerify(cudaMalloc((void**)&maxpressureDiffPerBlock, sizeof(double)*numberOfMultiprocessors));
cudaVerify(cudaMalloc((void**)&dtartviscPerBlock, sizeof(double)*numberOfMultiprocessors));
cudaVerify(cudaMalloc((void**)&dtalpha_epsporPerBlock, sizeof(double)*numberOfMultiprocessors));
cudaVerify(cudaMalloc((void**)&dtepsilon_vPerBlock, sizeof(double)*numberOfMultiprocessors));
int lastTimestep = startTimestep + numberOfTimesteps;
int timestep;
double substep_currentTime;
currentTime = startTime;
double endTime = startTime;
int allocate_immutables = 1;
// alloc mem for one rhs
allocate_particles_memory(&predictor_device, allocate_immutables);
copy_particles_immutables_device_to_device(&predictor_device, &p_device);
copy_particles_variables_device_to_device(&predictor_device, &p_device);
/* tell the gpu the current time */
cudaVerify(cudaMemcpyToSymbol(currentTimeD, ¤tTime, sizeof(double)));
cudaVerify(cudaMemcpyToSymbol(predictor, &predictor_device, sizeof(struct Particle)));
#if GRAVITATING_POINT_MASSES
allocate_pointmass_memory(&predictor_pointmass_device, allocate_immutables);
copy_pointmass_immutables_device_to_device(&predictor_pointmass_device, &pointmass_device);
/* tell the gpu the current time */
cudaVerify(cudaMemcpyToSymbol(predictor_pointmass, &predictor_pointmass_device, sizeof(struct Pointmass)));
#endif
for (timestep = startTimestep; timestep < lastTimestep; timestep++) {
fprintf(stdout, "calculating step %d\n", timestep);
printf("\nstep %d / %d\n", timestep, lastTimestep);
endTime += timePerStep;
fprintf(stdout, " currenttime: %e \t endtime: %e\n", currentTime, endTime);
/* tell the gpu the time step */
cudaVerify(cudaMemcpyToSymbol(dt, &timePerStep, sizeof(double)));
/* tell the gpu the end time */
cudaVerify(cudaMemcpyToSymbol(endTimeD, &endTime, sizeof(double)));
// checking for changes in angular momentum
if (param.angular_momentum_check > 0) {
double L_current = calculate_angular_momentum();
double L_change_relative;
if (L_ini > 0) {
L_change_relative = fabs((L_ini - L_current)/L_ini);
}
if (param.verbose) {
fprintf(stdout, "Checking angular momentum conservation.\n");
fprintf(stdout, "Initial angular momentum: %.17e\n", L_ini);
fprintf(stdout, "Current angular momentum: %.17e\n", L_current);
fprintf(stdout, "Relative change: %.17e\n", L_change_relative);
}
if (L_change_relative > param.angular_momentum_check) {
fprintf(stderr, "Conservation of angular momentum violated. Exiting.\n");
exit(111);
}
}
while (currentTime < endTime) {
cudaVerify(cudaDeviceSynchronize());
// calculate first right hand side with p_device
cudaVerify(cudaMemcpyToSymbol(p, &p_device, sizeof(struct Particle)));
#if GRAVITATING_POINT_MASSES
cudaVerify(cudaMemcpyToSymbol(pointmass, &pointmass_device, sizeof(struct Pointmass)));
#endif
cudaVerify(cudaDeviceSynchronize());
cudaVerify(cudaMemcpyFromSymbol(¤tTime, currentTimeD, sizeof(double)));
substep_currentTime = currentTime;
cudaVerify(cudaMemcpyToSymbol(substep_currentTimeD, &substep_currentTime, sizeof(double)));
rightHandSide();
cudaVerify(cudaDeviceSynchronize());
cudaVerifyKernel((setTimestep<<<numberOfMultiprocessors, NUM_THREADS_LIMITTIMESTEP>>>(
forcesPerBlock, courantPerBlock,
dtSPerBlock, dtePerBlock, dtrhoPerBlock, dtdamagePerBlock,
dtalphaPerBlock, dtartviscPerBlock, dtbetaPerBlock, dtalpha_epsporPerBlock, dtepsilon_vPerBlock)));
cudaVerify(cudaDeviceSynchronize());
/* get the time and the time step from the gpu */
cudaVerify(cudaMemcpyFromSymbol(&dt_host, dt, sizeof(double)));
substep_currentTime = currentTime + dt_host * 0.5;
cudaVerify(cudaDeviceSynchronize());
#if PALPHA_POROSITY
maxpressureDiff_cnt = 0;
maxpressureDiff_host = 0;
maxpressureDiff_previous = 0;
pressureChangeSmallEnough_host = FALSE;
while (pressureChangeSmallEnough_host == FALSE) {
#endif
// do the predictor step (writes to predictor)
cudaVerifyKernel((PredictorStep<<<numberOfMultiprocessors, NUM_THREADS_PC_INTEGRATOR>>>()));
cudaVerify(cudaDeviceSynchronize());
// get the derivatives at the predictor locations
cudaVerify(cudaMemcpyToSymbol(p, &predictor_device, sizeof(struct Particle)));
#if GRAVITATING_POINT_MASSES
cudaVerify(cudaMemcpyToSymbol(pointmass, &predictor_pointmass_device, sizeof(struct Pointmass)));
#endif
if (param.selfgravity) {
copy_gravitational_accels_device_to_device(&predictor_device, &p_device);
}
cudaVerify(cudaMemcpyToSymbol(substep_currentTimeD, &substep_currentTime, sizeof(double)));
rightHandSide();
// do the corrector step with the predictor and write it to in predictor
cudaVerify(cudaMemcpyToSymbol(p, &p_device, sizeof(struct Particle)));
#if GRAVITATING_POINT_MASSES
cudaVerify(cudaMemcpyToSymbol(pointmass, &pointmass_device, sizeof(struct Pointmass)));
#endif
#if PALPHA_POROSITY
cudaVerifyKernel((CorrectorStepPorous<<<numberOfMultiprocessors, NUM_THREADS_PC_INTEGRATOR>>>()));
cudaVerify(cudaDeviceSynchronize());
cudaVerify(cudaMemcpyToSymbol(p, &predictor_device, sizeof(struct Particle)));
#if GRAVITATING_POINT_MASSES
cudaVerify(cudaMemcpyToSymbol(pointmass, &pointmass_device, sizeof(struct Pointmass)));
#endif
cudaVerifyKernel((calculatePressure<<<numberOfMultiprocessors * 4, NUM_THREADS_PRESSURE>>>()));
cudaVerify(cudaDeviceSynchronize());
cudaVerifyKernel((pressureChangeCheck<<<numberOfMultiprocessors, NUM_THREADS_PC_INTEGRATOR>>>(maxpressureDiffPerBlock)));
cudaVerify(cudaDeviceSynchronize());
cudaVerify(cudaMemcpyFromSymbol(&pressureChangeSmallEnough_host, pressureChangeSmallEnough, sizeof(int)));
cudaVerify(cudaMemcpyFromSymbol(&maxpressureDiff_host, maxpressureDiff, sizeof(double)));
if (pressureChangeSmallEnough_host == FALSE) {
/* redo predictor step with smaller timestep, derivatives are in p_device */
printf("Reducing timestep due to Pressure Check function to: %.17e\n", dt_host);
if (fabs(maxpressureDiff_host -maxpressureDiff_previous) < 1e-3) {
maxpressureDiff_cnt++;
}
maxpressureDiff_previous = maxpressureDiff_host;
if (maxpressureDiff_cnt > 1) {
printf("Cannot reduce timestep anymore, continuing with dt %.17e and maxpressurediff %.17e", dt_host, maxpressureDiff_host);
pressureChangeSmallEnough_host = TRUE;
}
}
if (pressureChangeSmallEnough_host == FALSE) {
cudaVerify(cudaMemcpyFromSymbol(&dt_host, dt, sizeof(double)));
substep_currentTime = currentTime + dt_host * 0.5;
cudaVerify(cudaMemcpyToSymbol(p, &p_device, sizeof(struct Particle)));
#if GRAVITATING_POINT_MASSES
cudaVerify(cudaMemcpyToSymbol(pointmass, &pointmass_device, sizeof(struct Pointmass)));
#endif
printf("Reducing timestep due to Pressure Check function to: %e\n", dt_host);
} else {
cudaVerify(cudaMemcpyFromSymbol(&dt_host, dt, sizeof(double)));
currentTime += dt_host;
cudaVerify(cudaMemcpyToSymbol(currentTimeD, ¤tTime, sizeof(double)));
copy_particles_variables_device_to_device(&p_device, &predictor_device);
#if GRAVITATING_POINT_MASSES
copy_pointmass_variables_device_to_device(&pointmass_device, &predictor_pointmass_device);
#endif
cudaVerify(cudaDeviceSynchronize());
}
}
#else
cudaVerifyKernel((CorrectorStep<<<numberOfMultiprocessors, NUM_THREADS_PC_INTEGRATOR>>>()));
cudaVerify(cudaDeviceSynchronize());
#endif
/* get the time and the time step from the gpu */
cudaVerify(cudaMemcpyFromSymbol(¤tTime, currentTimeD, sizeof(double)));
} // current time < end time loop
// write results
#if FRAGMENTATION
cudaVerify(cudaDeviceSynchronize());
cudaVerifyKernel((damageLimit<<<numberOfMultiprocessors*4, NUM_THREADS_PC_INTEGRATOR>>>()));
cudaVerify(cudaDeviceSynchronize());
#endif
copyToHostAndWriteToFile(timestep, lastTimestep);
} // timestep loop
// free memory
int free_immutables = 1;
free_particles_memory(&predictor_device, free_immutables);
#if GRAVITATING_POINT_MASSES
free_pointmass_memory(&predictor_pointmass_device, free_immutables);
#endif
cudaVerify(cudaFree(courantPerBlock));
cudaVerify(cudaFree(forcesPerBlock));
cudaVerify(cudaFree(dtSPerBlock));
cudaVerify(cudaFree(dtePerBlock));
cudaVerify(cudaFree(dtrhoPerBlock));
cudaVerify(cudaFree(dtdamagePerBlock));
cudaVerify(cudaFree(dtalphaPerBlock));
cudaVerify(cudaFree(dtbetaPerBlock));
cudaVerify(cudaFree(dtartviscPerBlock));
cudaVerify(cudaFree(maxpressureDiffPerBlock));
cudaVerify(cudaFree(dtalpha_epsporPerBlock));
cudaVerify(cudaFree(dtepsilon_vPerBlock));
}