forked from sanshar/Dice
-
Notifications
You must be signed in to change notification settings - Fork 0
/
SHCIbasics.cpp
1591 lines (1403 loc) · 55.5 KB
/
SHCIbasics.cpp
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
/*
Developed by Sandeep Sharma with contributions from James E. T. Smith and Adam
A. Holmes, 2017 Copyright (c) 2017, Sandeep Sharma
This file is part of DICE.
This program 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.
This program 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
this program. If not, see <http://www.gnu.org/licenses/>.
*/
#include "SHCIbasics.h"
#include <boost/archive/binary_iarchive.hpp>
#include <boost/archive/binary_oarchive.hpp>
#include <boost/serialization/map.hpp>
#include <boost/serialization/serialization.hpp>
#include <boost/serialization/set.hpp>
#include <boost/serialization/shared_ptr.hpp>
#include <boost/serialization/vector.hpp>
#include <fstream>
#include <map>
#include <tuple>
#include <vector>
#include "Davidson.h"
#include "Determinants.h"
#include "Hmult.h"
#include "SHCISortMpiUtils.h"
#include "SHCIgetdeterminants.h"
#include "SHCImakeHamiltonian.h"
#include "SHCIrdm.h"
#include "SHCIsampledeterminants.h"
#include "SHCIshm.h"
#include "SHCItime.h"
#include "boost/format.hpp"
#include "input.h"
#include "integral.h"
#include "math.h"
#include "communicate.h"
using namespace std;
using namespace Eigen;
using namespace boost;
using namespace SHCISortMpiUtils;
double SHCIbasics::DoPerturbativeStochastic2SingleListDoubleEpsilon2AllTogether(
Determinant *Dets, CItype *ci, int DetsSize, double &E0, oneInt &I1,
twoInt &I2, twoIntHeatBathSHM &I2HB, vector<int> &irrep, schedule &schd,
double coreE, int nelec, int root) {
if (schd.nPTiter == 0)
return 0;
pout << format("Performing (semi)stochastic PT for state: %3i") % (root)
<< endl;
double epsilon2 = schd.epsilon2;
schd.epsilon2 = schd.epsilon2Large;
vector<MatrixXx> vdVector;
double Psi1Norm;
double EptLarge = 0.0;
if (schd.epsilon2 < 999)
pout << endl
<< "1/ Deterministic calculation with epsilon2=" << schd.epsilon2
<< endl;
EptLarge = DoPerturbativeDeterministic(Dets, ci, DetsSize, E0, I1, I2, I2HB,
irrep, schd, coreE, nelec, root,
vdVector, Psi1Norm);
schd.epsilon2 = epsilon2;
pout << endl
<< "2/ Stochastic calculation with epsilon2=" << schd.epsilon2 << endl;
int norbs = Determinant::norbs;
Determinant *SortedDets;
std::vector<Determinant> SortedDetsvec;
if (commrank == 0) {
for (int i = 0; i < DetsSize; i++) SortedDetsvec.push_back(Dets[i]);
std::sort(SortedDetsvec.begin(), SortedDetsvec.end());
}
#ifndef SERIAL
MPI_Barrier(MPI_COMM_WORLD);
#endif
SHMVecFromVecs(SortedDetsvec, SortedDets, shciSortedDets, SortedDetsSegment,
regionSortedDets);
SortedDetsvec.clear();
int niter = schd.nPTiter;
// double eps = 0.001;
int Nsample = schd.SampleN;
double AvgenergyEN = 0.0, AvgenergyEN2 = 0.0, stddev = 0.0;
double AverageDen = 0.0;
int currentIter = 0;
int sampleSize = 0;
int num_thrds = 1;
double cumulative = 0.0;
for (int i = 0; i < DetsSize; i++) cumulative += abs(ci[i]);
std::vector<int> alias;
std::vector<double> prob;
if (commrank == 0)
SHCIsampledeterminants::setUpAliasMethod(ci, DetsSize, cumulative, alias,
prob);
StitchDEH uniqueDEH;
double totalPT = 0.0;
double totalPTLargeEps = 0;
size_t ntries = 0;
int AllDistinctSample = 0;
size_t Nmc = commsize * num_thrds * Nsample;
std::vector<int> allSample(Nmc, -1);
std::vector<CItype> allwts(Nmc, 0.);
pout << format("%6s %18s %5s %18s %10s %10s") % ("Iter") % ("EPTcurrent") %
("State") % ("EPTavg") % ("Error") % ("Time(s)")
<< endl;
int size = commsize, rank = commrank;
for (int iter = 0; iter < niter; iter++) {
std::vector<CItype> wts1(Nsample, 0.0);
std::vector<int> Sample1(Nsample, -1);
int ithrd = 0;
vector<size_t> all_to_all(size * size, 0);
if (commrank == 0) {
std::fill(allSample.begin(), allSample.end(), -1);
AllDistinctSample = SHCIsampledeterminants::sample_N2_alias(
ci, DetsSize, cumulative, allSample, allwts, alias, prob);
}
#ifndef SERIAL
MPI_Bcast(&allSample[0], allSample.size(), MPI_INT, 0, MPI_COMM_WORLD);
MPI_Bcast(&AllDistinctSample, 1, MPI_INT, 0, MPI_COMM_WORLD);
#ifndef Complex
MPI_Bcast(&allwts[0], allwts.size(), MPI_DOUBLE, 0, MPI_COMM_WORLD);
#else
MPI_Bcast(&allwts[0], 2 * allwts.size(), MPI_DOUBLE, 0, MPI_COMM_WORLD);
#endif
#endif
int distinctSample = 0;
for (int i = 0; i < AllDistinctSample; i++) {
if (i % size != rank) continue;
wts1[distinctSample] = allwts[i];
Sample1[distinctSample] = allSample[i];
distinctSample++;
}
double norm = 0.0;
for (int i = 0; i < distinctSample; i++) {
int I = Sample1[i];
SHCIgetdeterminants::getDeterminantsStochastic2Epsilon(
Dets[I], schd.epsilon2 / abs(ci[I]), schd.epsilon2Large / abs(ci[I]),
wts1[i], ci[I], I1, I2, I2HB, irrep, coreE, E0, *uniqueDEH.Det,
*uniqueDEH.Num, *uniqueDEH.Num2, *uniqueDEH.present,
*uniqueDEH.Energy, schd, Nmc, nelec);
}
if (commsize > 1) {
boost::shared_ptr<vector<Determinant>> &Det = uniqueDEH.Det;
boost::shared_ptr<vector<CItype>> &Num = uniqueDEH.Num;
boost::shared_ptr<vector<CItype>> &Num2 = uniqueDEH.Num2;
boost::shared_ptr<vector<double>> &Energy = uniqueDEH.Energy;
boost::shared_ptr<vector<char>> &present = uniqueDEH.present;
std::vector<size_t> hashValues(Det->size());
std::vector<size_t> all_to_all_cumulative(size, 0);
for (int i = 0; i < Det->size(); i++) {
hashValues[i] = Det->at(i).getHash();
all_to_all[rank * size + hashValues[i] % size]++;
}
for (int i = 0; i < size; i++)
all_to_all_cumulative[i] =
i == 0 ? all_to_all[rank * size + i]
: all_to_all_cumulative[i - 1] + all_to_all[rank * size + i];
size_t dsize = Det->size() == 0 ? 1 : Det->size();
vector<Determinant> atoaDets(dsize);
vector<CItype> atoaNum(dsize);
vector<CItype> atoaNum2(dsize);
vector<double> atoaE(dsize);
vector<char> atoaPresent(dsize);
#ifndef SERIAL
vector<size_t> all_to_allCopy = all_to_all;
MPI_Allreduce(&all_to_allCopy[0], &all_to_all[0], 2 * size * size,
MPI_INT, MPI_SUM, MPI_COMM_WORLD);
#endif
vector<size_t> counter(size, 0);
for (int i = 0; i < Det->size(); i++) {
int toProc = hashValues[i] % size;
size_t index =
toProc == 0 ? counter[0]
: counter[toProc] + all_to_all_cumulative[toProc - 1];
atoaDets[index] = Det->at(i);
atoaNum[index] = Num->at(i);
atoaNum2[index] = Num2->at(i);
atoaE[index] = Energy->at(i);
atoaPresent[index] = present->at(i);
counter[toProc]++;
}
vector<int> sendcts(size, 0), senddisp(size, 0), recvcts(size, 0),
recvdisp(size, 0);
vector<int> sendctsDets(size, 0), senddispDets(size, 0),
recvctsDets(size, 0), recvdispDets(size, 0);
vector<int> sendctsPresent(size, 0), senddispPresent(size, 0),
recvctsPresent(size, 0), recvdispPresent(size, 0);
size_t recvSize = 0;
for (int i = 0; i < size; i++) {
sendcts[i] =
all_to_all[rank * size + i] * sizeof(CItype) / sizeof(double);
senddisp[i] = i == 0 ? 0 : senddisp[i - 1] + sendcts[i - 1];
recvcts[i] =
all_to_all[i * size + rank] * sizeof(CItype) / sizeof(double);
recvdisp[i] = i == 0 ? 0 : recvdisp[i - 1] + recvcts[i - 1];
sendctsDets[i] =
all_to_all[rank * size + i] * sizeof(Determinant) / sizeof(double);
senddispDets[i] = i == 0 ? 0 : senddispDets[i - 1] + sendctsDets[i - 1];
recvctsDets[i] =
all_to_all[i * size + rank] * sizeof(Determinant) / sizeof(double);
recvdispDets[i] = i == 0 ? 0 : recvdispDets[i - 1] + recvctsDets[i - 1];
sendctsPresent[i] = all_to_all[rank * size + i];
senddispPresent[i] =
i == 0 ? 0 : senddispPresent[i - 1] + sendctsPresent[i - 1];
recvctsPresent[i] = all_to_all[i * size + rank];
recvdispPresent[i] =
i == 0 ? 0 : recvdispPresent[i - 1] + recvctsPresent[i - 1];
recvSize += all_to_all[i * size + rank];
}
recvSize = recvSize == 0 ? 1 : recvSize;
Det->resize(recvSize), Num->resize(recvSize), Energy->resize(recvSize);
Num2->resize(recvSize), present->resize(recvSize);
#ifndef SERIAL
MPI_Alltoallv(&atoaNum.at(0), &sendcts[0], &senddisp[0], MPI_DOUBLE,
&Num->at(0), &recvcts[0], &recvdisp[0], MPI_DOUBLE,
MPI_COMM_WORLD);
MPI_Alltoallv(&atoaNum2.at(0), &sendcts[0], &senddisp[0], MPI_DOUBLE,
&Num2->at(0), &recvcts[0], &recvdisp[0], MPI_DOUBLE,
MPI_COMM_WORLD);
MPI_Alltoallv(&atoaE.at(0), &sendctsPresent[0], &senddispPresent[0],
MPI_DOUBLE, &Energy->at(0), &recvctsPresent[0],
&recvdispPresent[0], MPI_DOUBLE, MPI_COMM_WORLD);
MPI_Alltoallv(&atoaPresent.at(0), &sendctsPresent[0], &senddispPresent[0],
MPI_CHAR, &present->at(0), &recvctsPresent[0],
&recvdispPresent[0], MPI_CHAR, MPI_COMM_WORLD);
MPI_Alltoallv(&atoaDets.at(0).repr[0], &sendctsDets[0], &senddispDets[0],
MPI_DOUBLE, &(Det->at(0).repr[0]), &recvctsDets[0],
&recvdispDets[0], MPI_DOUBLE, MPI_COMM_WORLD);
#endif
}
uniqueDEH.MergeSort();
double energyEN = 0.0, energyENLargeEps = 0.0;
vector<Determinant> &Psi1 = *uniqueDEH.Det;
vector<CItype> &numerator1A = *uniqueDEH.Num;
vector<CItype> &numerator2A = *uniqueDEH.Num2;
vector<char> &present = *uniqueDEH.present;
vector<double> &det_energy = *uniqueDEH.Energy;
CItype currentNum1A = 0.;
CItype currentNum2A = 0.;
CItype currentNum1B = 0.;
CItype currentNum2B = 0.;
size_t vec_it = 0, i = 0;
while (i < Psi1.size() && vec_it < DetsSize) {
if (Psi1[i] < SortedDets[vec_it]) {
currentNum1A += numerator1A[i];
currentNum2A += numerator2A[i];
if (present[i]) {
currentNum1B += numerator1A[i];
currentNum2B += numerator2A[i];
}
if (i == Psi1.size() - 1) {
#ifndef Complex
energyEN +=
(pow(abs(currentNum1A), 2) * Nmc / (Nmc - 1) - currentNum2A) /
(det_energy[i] - E0);
energyENLargeEps +=
(pow(abs(currentNum1B), 2) * Nmc / (Nmc - 1) - currentNum2B) /
(det_energy[i] - E0);
#else
energyEN += (pow(abs(currentNum1A), 2) * Nmc / (Nmc - 1) -
currentNum2A.real()) /
(det_energy[i] - E0);
energyENLargeEps += (pow(abs(currentNum1B), 2) * Nmc / (Nmc - 1) -
currentNum2B.real()) /
(det_energy[i] - E0);
#endif
} else if (!(Psi1[i] == Psi1[i + 1])) {
#ifndef Complex
energyEN +=
(pow(abs(currentNum1A), 2) * Nmc / (Nmc - 1) - currentNum2A) /
(det_energy[i] - E0);
energyENLargeEps +=
(pow(abs(currentNum1B), 2) * Nmc / (Nmc - 1) - currentNum2B) /
(det_energy[i] - E0);
#else
energyEN += (pow(abs(currentNum1A), 2) * Nmc / (Nmc - 1) -
currentNum2A.real()) /
(det_energy[i] - E0);
energyENLargeEps += (pow(abs(currentNum1B), 2) * Nmc / (Nmc - 1) -
currentNum2B.real()) /
(det_energy[i] - E0);
#endif
currentNum1A = 0.;
currentNum2A = 0.;
currentNum1B = 0.;
currentNum2B = 0.;
}
i++;
} else if (SortedDets[vec_it] < Psi1[i] && vec_it != DetsSize) {
vec_it++;
} else {
if (i == Psi1.size() - 1 || Psi1[i] == Psi1[i + 1]) {
i++;
} else {
vec_it++;
i++;
}
}
}
while (i < Psi1.size()) {
currentNum1A += numerator1A[i];
currentNum2A += numerator2A[i];
if (present[i]) {
currentNum1B += numerator1A[i];
currentNum2B += numerator2A[i];
}
if (i == Psi1.size() - 1) {
#ifndef Complex
energyEN +=
(pow(abs(currentNum1A), 2) * Nmc / (Nmc - 1) - currentNum2A) /
(det_energy[i] - E0);
energyENLargeEps +=
(pow(abs(currentNum1B), 2) * Nmc / (Nmc - 1) - currentNum2B) /
(det_energy[i] - E0);
#else
energyEN += (pow(abs(currentNum1A), 2) * Nmc / (Nmc - 1) -
currentNum2A.real()) /
(det_energy[i] - E0);
energyENLargeEps += (pow(abs(currentNum1B), 2) * Nmc / (Nmc - 1) -
currentNum2B.real()) /
(det_energy[i] - E0);
#endif
} else if (!(Psi1[i] == Psi1[i + 1])) {
#ifndef Complex
energyEN +=
(pow(abs(currentNum1A), 2) * Nmc / (Nmc - 1) - currentNum2A) /
(det_energy[i] - E0);
energyENLargeEps +=
(pow(abs(currentNum1B), 2) * Nmc / (Nmc - 1) - currentNum2B) /
(det_energy[i] - E0);
#else
energyEN += (pow(abs(currentNum1A), 2) * Nmc / (Nmc - 1) -
currentNum2A.real()) /
(det_energy[i] - E0);
energyENLargeEps += (pow(abs(currentNum1B), 2) * Nmc / (Nmc - 1) -
currentNum2B.real()) /
(det_energy[i] - E0);
#endif
currentNum1A = 0.;
currentNum2A = 0.;
currentNum1B = 0.;
currentNum2B = 0.;
}
i++;
}
totalPT = 0;
totalPTLargeEps = 0;
totalPT += energyEN;
totalPTLargeEps += energyENLargeEps;
double finalE = 0., finalELargeEps = 0;
#ifndef SERIAL
MPI_Allreduce(&totalPT, &finalE, 1, MPI_DOUBLE, MPI_SUM, MPI_COMM_WORLD);
MPI_Allreduce(&totalPTLargeEps, &finalELargeEps, 1, MPI_DOUBLE, MPI_SUM,
MPI_COMM_WORLD);
#else
finalE = totalPT;
finalELargeEps = totalPTLargeEps;
#endif
if (commrank == 0) {
currentIter++;
AvgenergyEN += -finalE + finalELargeEps + EptLarge;
AvgenergyEN2 += pow(-finalE + finalELargeEps + EptLarge, 2);
stddev = currentIter < 5
? 1e4
: pow((currentIter * AvgenergyEN2 - pow(AvgenergyEN, 2)) /
currentIter / (currentIter - 1) / currentIter,
0.5);
if (currentIter < 5)
std::cout << format("%6i %18.10f %5i %18.10f %10s %10.2f") %
(currentIter) %
(E0 - finalE + finalELargeEps + EptLarge) % (root) %
(E0 + AvgenergyEN / currentIter) % "--" %
(getTime() - startofCalc);
else
std::cout << format("%6i %18.10f %5i %18.10f %10.2e %10.2f") %
(currentIter) %
(E0 - finalE + finalELargeEps + EptLarge) % (root) %
(E0 + AvgenergyEN / currentIter) % stddev %
(getTime() - startofCalc);
pout << endl;
}
#ifndef SERIAL
MPI_Bcast(¤tIter, 1, MPI_INT, 0, MPI_COMM_WORLD);
MPI_Bcast(&stddev, 1, MPI_DOUBLE, 0, MPI_COMM_WORLD);
MPI_Bcast(&AvgenergyEN, 1, MPI_DOUBLE, 0, MPI_COMM_WORLD);
#endif
uniqueDEH.clear();
if (stddev < schd.targetError) {
AvgenergyEN /= currentIter;
// pout << "Standard Error : " << stddev << " less than " <<
// schd.targetError << endl;
pout << "Semistochastic PT calculation converged" << endl;
pout << format("PTEnergy: %18.10f") % (E0 + AvgenergyEN) << " +/- ";
pout << format("%8.2e") % (stddev) << endl;
pout << format("Time(s): %10.2f") % (getTime() - startofCalc) << endl;
break;
}
}
return AvgenergyEN;
}
double SHCIbasics::DoPerturbativeDeterministic(
Determinant *Dets, CItype *ci, int DetsSize, double &E0, oneInt &I1,
twoInt &I2, twoIntHeatBathSHM &I2HB, vector<int> &irrep, schedule &schd,
double coreE, int nelec, int root, vector<MatrixXx> &vdVector,
double &Psi1Norm, bool appendPsi1ToPsi0) {
#ifndef SERIAL
boost::mpi::communicator world;
#endif
int norbs = Determinant::norbs;
Determinant *SortedDets;
std::vector<Determinant> SortedDetsvec;
if (commrank == 0) {
for (int i = 0; i < DetsSize; i++) SortedDetsvec.push_back(Dets[i]);
std::sort(SortedDetsvec.begin(), SortedDetsvec.end());
}
#ifndef SERIAL
MPI_Barrier(MPI_COMM_WORLD);
#endif
SHMVecFromVecs(SortedDetsvec, SortedDets, shciSortedDets, SortedDetsSegment,
regionSortedDets);
SortedDetsvec.clear();
double energyEN = 0.0;
double Psi1NormProc = 0.0;
StitchDEH uniqueDEH;
double totalPT = 0.0;
int ntries = 0;
int size = commsize, rank = commrank;
vector<size_t> all_to_all(size * size, 0);
if (schd.DoRDM || schd.doResponse) {
uniqueDEH.extra_info = true;
for (int i = 0; i < DetsSize; i++) {
if (i % size != rank) continue;
SHCIgetdeterminants::getDeterminantsDeterministicPTKeepRefDets(
Dets[i], i, abs(schd.epsilon2 / ci[i]), ci[i], I1, I2, I2HB, irrep,
coreE, E0, *uniqueDEH.Det, *uniqueDEH.Num, *uniqueDEH.Energy,
*uniqueDEH.var_indices_beforeMerge,
*uniqueDEH.orbDifference_beforeMerge, schd, nelec);
}
} else {
for (int i = 0; i < DetsSize; i++) {
if ((i % size != rank)) continue;
SHCIgetdeterminants::getDeterminantsDeterministicPT(
Dets[i], abs(schd.epsilon2 / ci[i]), ci[i], 0.0, I1, I2, I2HB, irrep,
coreE, E0, *uniqueDEH.Det, *uniqueDEH.Num, *uniqueDEH.Energy, schd, 0,
nelec);
// if (i%100000 == 0 && omp_get_thread_num()==0 && commrank == 0) pout <<
// "# " << i << endl;
}
}
if (commsize > 1) {
boost::shared_ptr<vector<Determinant>> &Det = uniqueDEH.Det;
boost::shared_ptr<vector<CItype>> &Num = uniqueDEH.Num;
boost::shared_ptr<vector<double>> &Energy = uniqueDEH.Energy;
boost::shared_ptr<vector<int>> &var_indices =
uniqueDEH.var_indices_beforeMerge;
boost::shared_ptr<vector<size_t>> &orbDifference =
uniqueDEH.orbDifference_beforeMerge;
std::vector<size_t> hashValues(Det->size());
std::vector<size_t> all_to_all_cumulative(size, 0);
for (int i = 0; i < Det->size(); i++) {
hashValues[i] = Det->at(i).getHash();
all_to_all[rank * size + hashValues[i] % size]++;
}
for (int i = 0; i < size; i++)
all_to_all_cumulative[i] =
i == 0 ? all_to_all[rank * size + i]
: all_to_all_cumulative[i - 1] + all_to_all[rank * size + i];
size_t dsize = Det->size() == 0 ? 1 : Det->size();
vector<Determinant> atoaDets(dsize);
vector<CItype> atoaNum(dsize);
vector<double> atoaE(dsize);
vector<int> atoaVarIndices;
vector<size_t> atoaOrbDiff;
if (schd.DoRDM || schd.doResponse) {
atoaVarIndices.resize(dsize);
atoaOrbDiff.resize(dsize);
}
#ifndef SERIAL
vector<size_t> all_to_allCopy = all_to_all;
MPI_Allreduce(&all_to_allCopy[0], &all_to_all[0], 2 * size * size, MPI_INT,
MPI_SUM, MPI_COMM_WORLD);
#endif
vector<size_t> counter(size, 0);
for (int i = 0; i < Det->size(); i++) {
int toProc = hashValues[i] % size;
size_t index = toProc == 0
? counter[0]
: counter[toProc] + all_to_all_cumulative[toProc - 1];
atoaDets[index] = Det->at(i);
atoaNum[index] = Num->at(i);
atoaE[index] = Energy->at(i);
if (schd.DoRDM || schd.doResponse) {
atoaVarIndices[index] = var_indices->at(i);
atoaOrbDiff[index] = orbDifference->at(i);
}
counter[toProc]++;
}
vector<int> sendcts(size, 0), senddisp(size, 0), recvcts(size, 0),
recvdisp(size, 0);
vector<int> sendctsDets(size, 0), senddispDets(size, 0),
recvctsDets(size, 0), recvdispDets(size, 0);
vector<int> sendctsVarDiff(size, 0), senddispVarDiff(size, 0),
recvctsVarDiff(size, 0), recvdispVarDiff(size, 0);
size_t recvSize = 0;
for (int i = 0; i < size; i++) {
sendcts[i] =
all_to_all[rank * size + i] * sizeof(CItype) / sizeof(double);
senddisp[i] = i == 0 ? 0 : senddisp[i - 1] + sendcts[i - 1];
recvcts[i] =
all_to_all[i * size + rank] * sizeof(CItype) / sizeof(double);
recvdisp[i] = i == 0 ? 0 : recvdisp[i - 1] + recvcts[i - 1];
sendctsDets[i] =
all_to_all[rank * size + i] * sizeof(Determinant) / sizeof(double);
senddispDets[i] = i == 0 ? 0 : senddispDets[i - 1] + sendctsDets[i - 1];
recvctsDets[i] =
all_to_all[i * size + rank] * sizeof(Determinant) / sizeof(double);
recvdispDets[i] = i == 0 ? 0 : recvdispDets[i - 1] + recvctsDets[i - 1];
sendctsVarDiff[i] = all_to_all[rank * size + i];
senddispVarDiff[i] =
i == 0 ? 0 : senddispVarDiff[i - 1] + sendctsVarDiff[i - 1];
recvctsVarDiff[i] = all_to_all[i * size + rank];
recvdispVarDiff[i] =
i == 0 ? 0 : recvdispVarDiff[i - 1] + recvctsVarDiff[i - 1];
recvSize += all_to_all[i * size + rank];
}
recvSize = recvSize == 0 ? 1 : recvSize;
Det->resize(recvSize), Num->resize(recvSize), Energy->resize(recvSize);
if (schd.DoRDM || schd.doResponse) {
var_indices->resize(recvSize);
orbDifference->resize(recvSize);
}
#ifndef SERIAL
MPI_Alltoallv(&atoaNum.at(0), &sendcts[0], &senddisp[0], MPI_DOUBLE,
&Num->at(0), &recvcts[0], &recvdisp[0], MPI_DOUBLE,
MPI_COMM_WORLD);
MPI_Alltoallv(&atoaE.at(0), &sendctsVarDiff[0], &senddispVarDiff[0],
MPI_DOUBLE, &Energy->at(0), &recvctsVarDiff[0],
&recvdispVarDiff[0], MPI_DOUBLE, MPI_COMM_WORLD);
MPI_Alltoallv(&atoaDets.at(0).repr[0], &sendctsDets[0], &senddispDets[0],
MPI_DOUBLE, &(Det->at(0).repr[0]), &recvctsDets[0],
&recvdispDets[0], MPI_DOUBLE, MPI_COMM_WORLD);
if (schd.DoRDM || schd.doResponse) {
MPI_Alltoallv(&atoaVarIndices.at(0), &sendctsVarDiff[0],
&senddispVarDiff[0], MPI_INT, &(var_indices->at(0)),
&recvctsVarDiff[0], &recvdispVarDiff[0], MPI_INT,
MPI_COMM_WORLD);
MPI_Alltoallv(&atoaOrbDiff.at(0), &sendctsVarDiff[0], &senddispVarDiff[0],
MPI_DOUBLE, &(orbDifference->at(0)), &recvctsVarDiff[0],
&recvdispVarDiff[0], MPI_DOUBLE, MPI_COMM_WORLD);
}
#endif
uniqueDEH.Num2->clear();
}
uniqueDEH.MergeSortAndRemoveDuplicates();
uniqueDEH.RemoveDetsPresentIn(SortedDets, DetsSize);
vector<Determinant> &hasHEDDets = *uniqueDEH.Det;
vector<CItype> &hasHEDNumerator = *uniqueDEH.Num;
vector<double> &hasHEDEnergy = *uniqueDEH.Energy;
double PTEnergy = 0.0;
double psi1normthrd = 0.0;
for (size_t i = 0; i < hasHEDDets.size(); i++) {
psi1normthrd += pow(abs(hasHEDNumerator[i] / (E0 - hasHEDEnergy[i])), 2);
PTEnergy += pow(abs(hasHEDNumerator[i]), 2) / (E0 - hasHEDEnergy[i]);
}
Psi1NormProc += psi1normthrd;
totalPT += PTEnergy;
/*
//PROBABLY SHOULD DELETE IT
{
char file [5000];
sprintf (file, "%s/%d-PTdata.bkp" , schd.prefix[0].c_str(), commrank );
std::ofstream ofs(file, std::ios::binary);
boost::archive::binary_oarchive save(ofs);
printf("writing file %s\n", file);
save << uniqueDEH;
ofs.close();
}
*/
double finalE = 0.;
#ifndef SERIAL
mpi::all_reduce(world, totalPT, finalE, std::plus<double>());
mpi::all_reduce(world, Psi1NormProc, Psi1Norm, std::plus<double>());
#else
finalE = totalPT;
#endif
if (commrank == 0) {
pout << "Deterministic PT calculation converged" << endl;
pout << format("PTEnergy: %18.10f") % (E0 + finalE) << endl;
pout << format("Time(s): %10.2f") % (getTime() - startofCalc) << endl;
}
if (schd.doResponse || schd.DoRDM) { // build RHS for the lambda equation
pout << endl << "Now calculating PT RDM" << endl;
MatrixXx s2RDM, twoRDM;
SHCIrdm::loadRDM(schd, s2RDM, twoRDM, root);
#ifndef SERIAL
mpi::broadcast(world, s2RDM, 0);
if (schd.DoSpinRDM) mpi::broadcast(world, twoRDM, 0);
#endif
if (commrank != 0) {
s2RDM = 0. * s2RDM;
twoRDM = 0. * twoRDM;
}
// SHCIrdm::ComputeEnergyFromSpatialRDM(norbs/2, nelec, I1, I2, coreE,
// s2RDM);
SHCIrdm::UpdateRDMResponsePerturbativeDeterministic(
Dets, DetsSize, ci, E0, I1, I2, schd, coreE, nelec, norbs, uniqueDEH,
root, Psi1Norm, s2RDM, twoRDM);
// SHCIrdm::ComputeEnergyFromSpatialRDM(norbs/2, nelec, I1, I2, coreE,
// s2RDM);
SHCIrdm::saveRDM(schd, s2RDM, twoRDM, root);
if (schd.RdmType == RELAXED) {
// construct the vector Via x da
// where Via is the perturbation matrix element
// da are the elements of the PT wavefunctions
vdVector[root] = MatrixXx::Zero(DetsSize, 1);
vector<Determinant> &uniqueDets = *uniqueDEH.Det;
vector<double> &uniqueEnergy = *uniqueDEH.Energy;
vector<CItype> &uniqueNumerator = *uniqueDEH.Num;
vector<vector<int>> &uniqueVarIndices = *uniqueDEH.var_indices;
vector<vector<size_t>> &uniqueOrbDiff = *uniqueDEH.orbDifference;
for (int a = 0; a < uniqueDets.size(); a++) {
CItype da = uniqueNumerator[a] /
(E0 - uniqueEnergy[a]); // coefficient for det a
for (int i = 0; i < uniqueVarIndices[a].size(); i++) {
int I = uniqueVarIndices[a][i]; // index of the Var determinant
size_t orbDiff;
#ifndef Complex
vdVector[root](I, 0) -=
da * Hij(Dets[I], uniqueDets[a], I1, I2, coreE, orbDiff);
#else
vdVector[root](I, 0) -=
conj(da) * Hij(Dets[I], uniqueDets[a], I1, I2, coreE, orbDiff);
#endif
}
}
#ifndef SERIAL
MPI_Allreduce(MPI_IN_PLACE, &vdVector[root](0, 0), vdVector[root].rows(),
MPI_DOUBLE, MPI_SUM, MPI_COMM_WORLD);
#endif
}
}
return finalE;
}
void unpackTrevState(vector<Determinant> &Dets, int &DetsSize,
vector<MatrixXx> &ci, SparseHam &sparseHam, bool DoRDM,
oneInt &I1, twoInt &I2, double &coreE) {
#ifndef SERIAL
boost::mpi::communicator world;
#endif
int oldLen = Dets.size();
vector<int> partnerLocation(oldLen, -1);
if (commrank == 0) {
if (Determinant::Trev != 0) {
int numDets = 0;
for (int i = 0; i < DetsSize; i++) {
if (Dets[i].hasUnpairedElectrons())
numDets += 2;
else
numDets += 1;
}
Dets.resize(numDets);
vector<MatrixXx> cibkp = ci;
for (int i = 0; i < ci.size(); i++) {
ci[i].resize(Dets.size(), 1);
ci[i].block(0, 0, cibkp[i].rows(), 1) = 1. * cibkp[i];
}
int newIndex = 0, oldLen = cibkp[0].rows();
for (int i = 0; i < oldLen; i++) {
if (Dets[i].hasUnpairedElectrons()) {
partnerLocation[i] = newIndex;
Dets[newIndex + oldLen] = Dets[i];
Dets[newIndex + oldLen].flipAlphaBeta();
for (int j = 0; j < ci.size(); j++) {
ci[j](i, 0) = cibkp[j](i, 0) / sqrt(2.0);
double parity = Dets[i].parityOfFlipAlphaBeta();
ci[j](newIndex + oldLen, 0) =
Determinant::Trev * parity * cibkp[j](i, 0) / sqrt(2.0);
}
newIndex++;
}
}
}
}
DetsSize = Dets.size();
#ifndef SERIAL
MPI_Bcast(&DetsSize, 1, MPI_INT, 0, MPI_COMM_WORLD);
MPI_Bcast(&oldLen, 1, MPI_INT, 0, MPI_COMM_WORLD);
#endif
}
// this takes in a ci vector for determinants placed in Dets
// it then does a SHCI varitional calculation and the resulting
// ci and dets are returned here
// At input usually the Dets will just have a HF or some such determinant
// and ci will be just 1.0
vector<double> SHCIbasics::DoVariational(vector<MatrixXx> &ci,
vector<Determinant> &Dets,
schedule &schd, twoInt &I2,
twoIntHeatBathSHM &I2HB,
vector<int> &irrep, oneInt &I1,
double &coreE, int nelec, bool DoRDM) {
int proc = 0, nprocs = 1;
#ifndef SERIAL
MPI_Comm_rank(MPI_COMM_WORLD, &proc);
MPI_Comm_size(MPI_COMM_WORLD, &nprocs);
boost::mpi::communicator world;
#endif
// Put determinants on the shared memory
Determinant *SHMDets, *SortedDets;
SHMVecFromVecs(Dets, SHMDets, shciDetsCI, DetsCISegment, regionDetsCI);
if (proc != 0) Dets.resize(0);
std::vector<Determinant> SortedDetsvec; // only proc 1 has it
if (commrank == 0) {
SortedDetsvec = Dets;
std::sort(SortedDetsvec.begin(), SortedDetsvec.end());
}
int SortedDetsSize = SortedDetsvec.size(), DetsSize = Dets.size();
SHMVecFromVecs(SortedDetsvec, SortedDets, shciSortedDets, SortedDetsSegment,
regionSortedDets);
Dets.clear();
SortedDetsvec.clear();
#ifndef SERIAL
mpi::broadcast(world, SortedDetsSize, 0);
mpi::broadcast(world, DetsSize, 0);
#endif
// sometimes coreenergy is huge and it kills the stability of davidson solver
// so make it zero and just add it to the converged energy
double coreEbkp = coreE;
coreE = 0.0;
if (schd.outputlevel > 0 && commrank == 0)
Time::print_time("start variation");
// initialize the sparse Hamiltonian and helpers
int nroots = ci.size();
SHCImakeHamiltonian::HamHelpers2 helper2;
SHCImakeHamiltonian::SparseHam sparseHam;
if (schd.DavidsonType == DISK) {
sparseHam.diskio = true;
sparseHam.BatchSize = 100000;
// sparseHam.BatchSize = 10;
sparseHam.setNbatches(DetsSize);
sparseHam.prefix = schd.prefix[0];
}
MatrixXx diag;
size_t norbs = 2. * I2.Direct.rows();
int Norbs = norbs;
CItype e0 = SHMDets[0].Energy(I1, I2, coreE);
size_t orbDiff;
if (Determinant::Trev != 0)
updateHijForTReversal(e0, SHMDets[0], SHMDets[0], I1, I2, coreE, orbDiff);
vector<double> E0(nroots, abs(e0));
// make helpers and then put them on the shared memory
if (proc == 0) {
helper2.PopulateHelpers(SHMDets, DetsSize, 0);
}
helper2.MakeSHMHelpers();
// if it is not direct Hamiltonian then generate it
if (schd.DavidsonType != DIRECT) {
sparseHam.makeFromHelper(helper2, SHMDets, 0, DetsSize, Norbs, I1, I2,
coreE, schd.DoRDM || schd.DoOneRDM);
}
// update the Hamiltonian with SOC terms
#ifdef Complex
SHCImakeHamiltonian::updateSOCconnections(
SHMDets, 0, DetsSize, SortedDets, sparseHam.connections,
sparseHam.orbDifference, sparseHam.Helements, norbs, I1, nelec, false);
#endif
pout << format("%4s %4s %10s %10.2e %18s %9s %10s\n") % ("Iter") %
("Root") % ("Eps1 ") % ("#Var. Det.") % ("Energy") %
("#Davidson") % ("Time(s)");
int prevSize = 0;
// If this is a restart calculation then read from disk
int iterstart = 0;
if (schd.restart || schd.fullrestart) {
bool converged;
readVariationalResult(iterstart, ci, Dets, sparseHam, E0, converged, schd,
helper2);
// after reading restart put dets on shared memory
SHMVecFromVecs(Dets, SHMDets, shciDetsCI, DetsCISegment, regionDetsCI);
DetsSize = Dets.size();
SortedDetsSize = DetsSize;
// put sorted dets on shared memory as well
SHMVecFromVecs(Dets, SortedDets, shciSortedDets, SortedDetsSegment,
regionSortedDets);
#ifndef SERIAL
mpi::broadcast(world, SortedDetsSize, 0);
mpi::broadcast(world, DetsSize, 0);
#endif
if (localrank == 0) std::sort(SortedDets, SortedDets + SortedDetsSize);
#ifndef SERIAL
MPI_Barrier(MPI_COMM_WORLD);
#endif
Dets.clear();
// Make helpers and make sparse hamiltonian if full restart and not direct
helper2.MakeSHMHelpers();
if (!(schd.DavidsonType == DIRECT ||
(!schd.fullrestart && converged &&
iterstart >= schd.epsilon1.size() - 1))) {
sparseHam.clear();
sparseHam.makeFromHelper(helper2, SHMDets, 0, DetsSize, Norbs, I1, I2,
coreE, schd.DoRDM || schd.DoOneRDM);
}
for (int i = 0; i < E0.size(); i++)
pout << format("%4i %4i %10.2e %10.2e - %18.10f %10.2f\n") %
(iterstart) % (i) % schd.epsilon1[iterstart] % DetsSize %
(E0[i] + coreEbkp) % (getTime() - startofCalc);
if (!schd.fullrestart)
iterstart++;
else
iterstart = 0;
// if the calculation is converged then exit
if (schd.outputlevel > 0)
pout << "Converged: " << converged
<< " loaded iter: " << iterstart
<< " max iter: " << schd.epsilon1.size()
<< " loaded eps1: " << schd.epsilon1[iterstart-1]
<< " min eps1: " << schd.epsilon1[schd.epsilon1.size()-1]
<< endl;
if (converged && iterstart-1 <= schd.epsilon1.size() &&
schd.epsilon1[iterstart-1] <= schd.epsilon1[schd.epsilon1.size()-1]) {
for (int i = 0; i < E0.size(); i++)
E0[i] += coreEbkp;
coreE = coreEbkp;
pout << "# restarting from a converged calculation, moving to "
"perturbative part.!!"
<< endl;
Dets.resize(DetsSize);
for (int i = 0; i < DetsSize; i++) Dets[i] = SHMDets[i];
#ifndef SERIAL
MPI_Barrier(MPI_COMM_WORLD);
mpi::broadcast(world, E0, 0);
#endif
unpackTrevState(Dets, DetsSize, ci, sparseHam, false, I1, I2, coreE);
return E0;
}
}
for (int iter = iterstart; iter < schd.epsilon1.size(); iter++) {
double epsilon1 = schd.epsilon1[iter];
StitchDEH uniqueDEH;
// for multiple states, use the sum of squares of states
// to do the seclection process
if (schd.outputlevel > 0)
pout << format("#-------------Iter=%4i---------------") % iter << endl;
CItype *cMaxSHM;
vector<CItype> cMax;
if (proc == 0) {
cMax.resize(ci[0].rows(), 0);
for (int j = 0; j < ci[0].rows(); j++) {
for (int i = 0; i < ci.size(); i++) cMax[j] += pow(abs(ci[i](j, 0)), 2);
cMax[j] = pow(cMax[j], 0.5);
}
}
SHMVecFromVecs(cMax, cMaxSHM, shcicMax, cMaxSegment, regioncMax);
cMax.clear();
CItype zero = 0.0;
for (int i = 0; i < SortedDetsSize; i++) {
if (i % (commsize) != commrank) continue;
#ifndef Complex
SHCIgetdeterminants::getDeterminantsVariationalApprox(
SHMDets[i], epsilon1 / abs(cMaxSHM[i]), cMaxSHM[i], zero, I1, I2,
I2HB, irrep, coreE, E0[0], *uniqueDEH.Det, schd, 0, nelec, SortedDets,
SortedDetsSize);
#else
SHCIgetdeterminants::getDeterminantsVariational(
SHMDets[i], epsilon1 / abs(cMaxSHM[i]), cMaxSHM[i], zero, I1, I2,
I2HB, irrep, coreE, E0[0], *uniqueDEH.Det, schd, 0, nelec);
#endif
}
if (Determinant::Trev != 0) {
for (int i = 0; i < uniqueDEH.Det->size(); i++)
uniqueDEH.Det->at(i).makeStandard();
}
//*********
// Remove duplicates and put all the dets on all the nodes
sort(uniqueDEH.Det->begin(), uniqueDEH.Det->end());
uniqueDEH.Det->erase(unique(uniqueDEH.Det->begin(), uniqueDEH.Det->end()),
uniqueDEH.Det->end());
if (Determinant::Trev != 0)
uniqueDEH.RemoveOnlyDetsPresentIn(SortedDets, SortedDetsSize);
#ifdef Complex
uniqueDEH.RemoveOnlyDetsPresentIn(SortedDets, SortedDetsSize);
#endif