-
-
Notifications
You must be signed in to change notification settings - Fork 78
/
solve.jl
1339 lines (1192 loc) · 48.3 KB
/
solve.jl
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
## Common Interface Solve Functions
function DiffEqBase.__solve(
prob::Union{DiffEqBase.AbstractODEProblem,DiffEqBase.AbstractDAEProblem},
alg::algType,timeseries=[],ts=[],ks=[],
recompile::Type{Val{recompile_flag}}=Val{true};
kwargs...) where {algType<:Union{SundialsODEAlgorithm,SundialsDAEAlgorithm},
recompile_flag}
integrator = DiffEqBase.__init(prob,alg,timeseries,ts,ks;kwargs...)
if integrator.sol.retcode == :Default
solve!(integrator)
end
integrator.sol
end
function DiffEqBase.__init(
prob::DiffEqBase.AbstractODEProblem{uType, tupType, isinplace},
alg::SundialsODEAlgorithm{Method,LinearSolver},
timeseries=[], ts=[], ks=[];
verbose=true,
callback=nothing, abstol=1/10^6, reltol=1/10^3,
saveat=Float64[], tstops=Float64[],
maxiters=Int(1e5),
dt = nothing, dtmin = 0.0, dtmax = 0.0,
timeseries_errors=true,
dense_errors = false,
save_everystep=isempty(saveat),
save_on = true,
save_start = save_everystep || isempty(saveat) || typeof(saveat) <: Number ? true : prob.tspan[1] in saveat,
save_end = save_everystep || isempty(saveat) || typeof(saveat) <: Number ? true : prob.tspan[2] in saveat,
dense = save_everystep && isempty(saveat),
progress=false,progress_name="ODE",
progress_message = DiffEqBase.ODE_DEFAULT_PROG_MESSAGE,
save_timeseries = nothing,
advance_to_tstop = false,stop_at_next_tstop=false,
userdata=nothing,
alias_u0=false,
kwargs...) where {uType, tupType, isinplace, Method, LinearSolver}
tType = eltype(tupType)
if verbose
warned = !isempty(kwargs) && DiffEqBase.check_keywords(alg, kwargs, warnlist)
if !(typeof(prob.f) <: DiffEqBase.AbstractParameterizedFunction) && typeof(alg) <: CVODE_BDF
if DiffEqBase.has_tgrad(prob.f)
@warn("Explicit t-gradient given to this stiff solver is ignored.")
warned = true
end
end
warned && DiffEqBase.warn_compat()
end
if prob.f.mass_matrix != LinearAlgebra.I
error("This solver is not able to use mass matrices.")
end
if typeof(reltol) <: AbstractArray
error("Sundials only allows scalar reltol.")
end
progress && Logging.@logmsg(-1,progress_name,_id=_id = :Sundials,progress=0)
callbacks_internal = DiffEqBase.CallbackSet(callback)
max_len_cb = DiffEqBase.max_vector_callback_length(callbacks_internal)
if max_len_cb isa VectorContinuousCallback
callback_cache = DiffEqBase.CallbackCache(max_len_cb.len,Float64,Float64)
else
callback_cache = nothing
end
tspan = prob.tspan
t0 = tspan[1]
tdir = sign(tspan[2]-tspan[1])
tstops_internal, saveat_internal =
tstop_saveat_disc_handling(tstops,saveat,tdir,tspan,tType)
if typeof(prob.u0) <: Number
u0 = [prob.u0]
else
if alias_u0
u0 = vec(prob.u0)
else
u0 = vec(copy(prob.u0))
end
end
sizeu = size(prob.u0)
### Fix the more general function to Sundials allowed style
if !isinplace && typeof(prob.u0)<:Number
f! = (du, u, p, t) -> (du .= prob.f(first(u), p, t); Cint(0))
elseif !isinplace && typeof(prob.u0)<:Vector{Float64}
f! = (du, u, p, t) -> (du .= prob.f(u, p, t); Cint(0))
elseif !isinplace && typeof(prob.u0)<:AbstractArray
f! = (du, u, p, t) -> (du .= vec(prob.f(reshape(u, sizeu), p, t)); Cint(0))
elseif typeof(prob.u0)<:Vector{Float64}
f! = prob.f
else # Then it's an in-place function on an abstract array
f! = (du, u, p, t) -> (prob.f(reshape(du, sizeu), reshape(u, sizeu), p, t);
du=vec(du); 0)
end
if typeof(alg) <: CVODE_BDF
alg_code = CV_BDF
elseif typeof(alg) <: CVODE_Adams
alg_code = CV_ADAMS
end
#if Method == :Newton
# method_code = CV_NEWTON
#elseif Method == :Functional
# method_code = CV_FUNCTIONAL
#end
mem_ptr = CVodeCreate(alg_code)
(mem_ptr == C_NULL) && error("Failed to allocate CVODE solver object")
mem = Handle(mem_ptr)
!verbose && CVodeSetErrHandlerFn(mem,@cfunction(null_error_handler, Nothing,
(Cint, Char,
Char, Ptr{Cvoid})),C_NULL)
ures = Vector{uType}()
dures = Vector{uType}()
save_start ? ts = [t0] : ts = Float64[]
u0nv = NVector(u0)
_u0 = copy(u0)
utmp = NVector(_u0)
userfun = FunJac(f!,prob.f.jac,prob.p,nothing,prob.f.jac_prototype,alg.prec,alg.psetup,u0,_u0)
function getcfunf(::T) where T
@cfunction(cvodefunjac, Cint, (realtype, N_Vector, N_Vector, Ref{T}))
end
flag = CVodeInit(mem,getcfunf(userfun),
t0, convert(N_Vector, utmp))
dt != nothing && (flag = CVodeSetInitStep(mem, dt))
flag = CVodeSetMinStep(mem, dtmin)
flag = CVodeSetMaxStep(mem, dtmax)
flag = CVodeSetUserData(mem, userfun)
if typeof(abstol) <: Array
flag = CVodeSVtolerances(mem, reltol, abstol)
else
flag = CVodeSStolerances(mem, reltol, abstol)
end
flag = CVodeSetMaxNumSteps(mem, maxiters)
flag = CVodeSetMaxOrd(mem, alg.max_order)
flag = CVodeSetMaxHnilWarns(mem, alg.max_hnil_warns)
flag = CVodeSetStabLimDet(mem, alg.stability_limit_detect)
flag = CVodeSetMaxErrTestFails(mem, alg.max_error_test_failures)
flag = CVodeSetMaxNonlinIters(mem, alg.max_nonlinear_iters)
flag = CVodeSetMaxConvFails(mem, alg.max_convergence_failures)
nojacobian = true
if Method == :Newton # Only use a linear solver if it's a Newton-based method
if LinearSolver in (:Dense, :LapackDense)
nojacobian = false
A = SUNDenseMatrix(length(u0),length(u0))
_A = MatrixHandle(A,DenseMatrix())
if LinearSolver === :Dense
LS = SUNLinSol_Dense(u0,A)
_LS = LinSolHandle(LS,Dense())
else
LS = SUNLinSol_LapackDense(u0,A)
_LS = LinSolHandle(LS,LapackDense())
end
elseif LinearSolver in (:Band, :LapackBand)
nojacobian = false
A = SUNBandMatrix(length(u0), alg.jac_upper, alg.jac_lower)
_A = MatrixHandle(A,BandMatrix())
if LinearSolver === :Band
LS = SUNLinSol_Band(u0,A)
_LS = LinSolHandle(LS,Band())
else
LS = SUNLinSol_LapackBand(u0,A)
_LS = LinSolHandle(LS,LapackBand())
end
elseif LinearSolver == :Diagonal
nojacobian = false
flag = CVDiag(mem)
_A = nothing
_LS = nothing
elseif LinearSolver == :GMRES
LS = SUNLinSol_SPGMR(u0, alg.prec_side, alg.krylov_dim)
_A = nothing
_LS = Sundials.LinSolHandle(LS,Sundials.SPGMR())
elseif LinearSolver == :FGMRES
LS = SUNLinSol_SPFGMR(u0, alg.prec_side, alg.krylov_dim)
_A = nothing
_LS = LinSolHandle(LS,SPFGMR())
elseif LinearSolver == :BCG
LS = SUNLinSol_SPBCGS(u0, alg.prec_side, alg.krylov_dim)
_A = nothing
_LS = LinSolHandle(LS,SPBCGS())
elseif LinearSolver == :PCG
LS = SUNLinSol_PCG(u0, alg.prec_side, alg.krylov_dim)
_A = nothing
_LS = LinSolHandle(LS,PCG())
elseif LinearSolver == :TFQMR
LS = SUNLinSol_SPTFQMR(u0, alg.prec_side, alg.krylov_dim)
_A = nothing
_LS = LinSolHandle(LS,PTFQMR())
elseif LinearSolver == :KLU
nojacobian = false
nnz = length(SparseArrays.nonzeros(prob.f.jac_prototype))
A = SUNSparseMatrix(length(u0),length(u0), nnz, CSC_MAT)
LS = SUNLinSol_KLU(u0, A)
_A = MatrixHandle(A,SparseMatrix())
_LS = LinSolHandle(LS,KLU())
end
if LinearSolver !== :Diagonal
flag = CVodeSetLinearSolver(mem, LS, _A === nothing ? C_NULL : A)
end
NLS = SUNNonlinSol_Newton(u0)
else
_A = nothing
_LS = nothing
# TODO: Anderson Acceleration
anderson_m = 0
NLS = SUNNonlinSol_FixedPoint(u0, anderson_m)
end
CVodeSetNonlinearSolver(mem, NLS)
if DiffEqBase.has_jac(prob.f) && Method == :Newton
function getcfunjac(::T) where T
@cfunction(cvodejac,
Cint,
(realtype,
N_Vector,
N_Vector,
SUNMatrix,
Ref{T},
N_Vector,
N_Vector,
N_Vector))
end
jac = getcfunjac(userfun)
flag = CVodeSetUserData(mem, userfun)
nojacobian || (flag = CVodeSetJacFn(mem, jac))
else
jac = nothing
end
if typeof(prob.f.jac_prototype) <: DiffEqBase.AbstractDiffEqLinearOperator
function getcfunjtimes(::T) where T
@cfunction(jactimes,
Cint,
(N_Vector,
N_Vector,
realtype,
N_Vector,
N_Vector,
Ref{T},
N_Vector))
end
jtimes = getcfunjtimes(userfun)
CVodeSetJacTimes(mem, C_NULL, jtimes)
end
if alg.prec !== nothing
function getpercfun(::T) where T
@cfunction(precsolve,
Cint,
(Float64,
N_Vector,
N_Vector,
N_Vector,
N_Vector,Float64,Float64,Int,
Ref{T}))
end
precfun = getpercfun(userfun)
function getpsetupfun(::T) where T
@cfunction(precsetup,
Cint,
(Float64,
N_Vector,
N_Vector,
Int,
Ptr{Int},Float64,Ref{T}))
end
psetupfun = alg.psetup === nothing ? C_NULL : getpsetupfun(userfun)
CVodeSetPreconditioner(mem, psetupfun, precfun)
end
callbacks_internal == nothing ? tmp = nothing : tmp = similar(u0)
callbacks_internal == nothing ? uprev = nothing : uprev = similar(u0)
tout = [tspan[1]]
if save_start
save_value!(ures,u0,uType,sizeu)
if dense
f!(_u0,u0,prob.p,tspan[1])
save_value!(dures,utmp,uType,sizeu)
end
end
sol = DiffEqBase.build_solution(prob, alg, ts, ures,
dense = dense,
interp = dense ? DiffEqBase.HermiteInterpolation(ts,ures,dures) :
DiffEqBase.LinearInterpolation(ts,ures),
timeseries_errors = timeseries_errors,
destats = DiffEqBase.DEStats(0),
calculate_error = false)
opts = DEOptions(saveat_internal,tstops_internal,save_everystep,dense,
timeseries_errors,dense_errors,save_on,save_end,
callbacks_internal,abstol,reltol,verbose,advance_to_tstop,stop_at_next_tstop,
progress,progress_name,progress_message,maxiters)
integrator = CVODEIntegrator(u0,prob.p,t0,t0,mem,_LS,_A,sol,alg,f!,userfun,jac,opts,
tout,tdir,sizeu,false,tmp,uprev,Cint(flag),false,0,1,callback_cache,0.)
initialize_callbacks!(integrator)
integrator
end # function solve
function DiffEqBase.__init(
prob::DiffEqBase.AbstractODEProblem{uType, tupType, isinplace},
alg::ARKODE{Method,LinearSolver,MassLinearSolver},
timeseries=[], ts=[], ks=[];
verbose=true,
callback=nothing, abstol=1/10^6, reltol=1/10^3,
saveat=Float64[], tstops=Float64[],
maxiters=Int(1e5),
dt = nothing, dtmin = 0.0, dtmax = 0.0,
timeseries_errors=true,
dense_errors = false,
save_everystep=isempty(saveat), dense = save_everystep,
save_on = true, save_start = true, save_end = true,
save_timeseries = nothing,
progress=false,progress_name="ODE",
progress_message = DiffEqBase.ODE_DEFAULT_PROG_MESSAGE,
advance_to_tstop = false,stop_at_next_tstop=false,
userdata=nothing,
alias_u0=false,
kwargs...) where {uType, tupType, isinplace, Method, LinearSolver, MassLinearSolver}
tType = eltype(tupType)
if verbose
warned = !isempty(kwargs) && DiffEqBase.check_keywords(alg, kwargs, warnlist)
if !(typeof(prob.f) <: DiffEqBase.AbstractParameterizedFunction)
if typeof(prob.f) <: SplitFunction ? DiffEqBase.has_tgrad(prob.f.f1) : DiffEqBase.has_tgrad(prob.f)
@warn("Explicit t-gradient given to this stiff solver is ignored.")
warned = true
end
end
warned && DiffEqBase.warn_compat()
end
if typeof(reltol) <: AbstractArray
error("Sundials only allows scalar reltol.")
end
progress && Logging.@logmsg(-1,progress_name,_id=_id = :Sundials,progress=0)
callbacks_internal = DiffEqBase.CallbackSet(callback)
max_len_cb = DiffEqBase.max_vector_callback_length(callbacks_internal)
if max_len_cb isa VectorContinuousCallback
callback_cache = DiffEqBase.CallbackCache(max_len_cb.len,Float64,Float64)
else
callback_cache = nothing
end
tspan = prob.tspan
t0 = tspan[1]
tdir = sign(tspan[2]-tspan[1])
tstops_internal, saveat_internal =
tstop_saveat_disc_handling(tstops,saveat,tdir,tspan,tType)
if typeof(prob.u0) <: Number
u0 = [prob.u0]
else
if alias_u0
u0 = vec(prob.u0)
else
u0 = vec(copy(prob.u0))
end
end
sizeu = size(prob.u0)
ures = Vector{uType}()
dures = Vector{uType}()
save_start ? ts = [t0] : ts = Float64[]
u0nv = NVector(u0)
_u0 = copy(u0)
utmp = NVector(_u0)
function arkodemem(;fe=C_NULL, fi=C_NULL, t0=t0, u0=convert(N_Vector, u0nv))
mem_ptr = ARKStepCreate(fe, fi, t0, u0)
(mem_ptr == C_NULL) && error("Failed to allocate ARKODE solver object")
mem = Handle(mem_ptr)
!verbose && ARKStepSetErrHandlerFn(mem,@cfunction(null_error_handler, Nothing,
(Cint, Char,
Char, Ptr{Cvoid})),C_NULL)
return mem
end
### Fix the more general function to Sundials allowed style
if !isinplace && typeof(prob.u0)<:Number
f! = (du, u, p, t) -> (du .= prob.f(first(u), p, t); Cint(0))
elseif !isinplace && typeof(prob.u0)<:Vector{Float64}
f! = (du, u, p, t) -> (du .= prob.f(u, p, t); Cint(0))
elseif !isinplace && typeof(prob.u0)<:AbstractArray
f! = (du, u, p, t) -> (du .= vec(prob.f(reshape(u, sizeu), p, t)); Cint(0))
elseif typeof(prob.u0)<:Vector{Float64}
f! = prob.f
else # Then it's an in-place function on an abstract array
f! = (du, u, p, t) -> (prob.f(reshape(du, sizeu), reshape(u, sizeu), p, t);
du=vec(du); Cint(0))
end
if typeof(prob.problem_type) <: SplitODEProblem
### Fix the more general function to Sundials allowed style
if !isinplace && typeof(prob.u0)<:Number
f1! = (du, u, p, t) -> (du .= prob.f.f1(first(u), p, t); Cint(0))
f2! = (du, u, p, t) -> (du .= prob.f.f2(first(u), p, t); Cint(0))
elseif !isinplace && typeof(prob.u0)<:Vector{Float64}
f1! = (du, u, p, t) -> (du .= prob.f.f1(u, p, t); Cint(0))
f2! = (du, u, p, t) -> (du .= prob.f.f2(u, p, t); Cint(0))
elseif !isinplace && typeof(prob.u0)<:AbstractArray
f1! = (du, u, p, t) -> (du .= vec(prob.f.f1(reshape(u, sizeu), p, t)); Cint(0))
f2! = (du, u, p, t) -> (du .= vec(prob.f.f2(reshape(u, sizeu), p, t)); Cint(0))
elseif typeof(prob.u0)<:Vector{Float64}
f1! = prob.f.f1
f2! = prob.f.f2
else # Then it's an in-place function on an abstract array
f1! = (du, u, p, t) -> (prob.f.f1(reshape(du, sizeu), reshape(u, sizeu), p, t);
du=vec(du); Cint(0))
f2! = (du, u, p, t) -> (prob.f.f2(reshape(du, sizeu), reshape(u, sizeu), p, t);
du=vec(du); Cint(0))
end
userfun = FunJac(f1!,f2!,prob.f.f1.jac,prob.p,prob.f.mass_matrix,
prob.f.f1.jac_prototype,alg.prec,alg.psetup,u0,_u0,nothing)
function getcfunjac(::T) where T
@cfunction(cvodefunjac, Cint,
(realtype, N_Vector,
N_Vector, Ref{T}))
end
function getcfunjac2(::T) where T
@cfunction(cvodefunjac2, Cint,
(realtype, N_Vector,
N_Vector, Ref{T}))
end
cfj1 = getcfunjac(userfun)
cfj2 = getcfunjac2(userfun)
mem = arkodemem(fi=cfj1, fe=cfj2)
else
userfun = FunJac(f!,prob.f.jac,prob.p,prob.f.mass_matrix,prob.f.jac_prototype,alg.prec,alg.psetup,u0,_u0)
if alg.stiffness == Explicit()
function getcfun1(::T) where T
@cfunction(cvodefunjac, Cint,
(realtype, N_Vector,
N_Vector, Ref{T}))
end
cfj1 = getcfun1(userfun)
mem = arkodemem(fe=cfj1)
elseif alg.stiffness == Implicit()
function getcfun2(::T) where T
@cfunction(cvodefunjac, Cint,
(realtype, N_Vector,
N_Vector, Ref{T}))
end
cfj2 = getcfun2(userfun)
mem = arkodemem(fi=cfj2)
end
end
dt != nothing && (flag = ARKStepSetInitStep(mem, dt))
flag = ARKStepSetMinStep(mem, dtmin)
flag = ARKStepSetMaxStep(mem, dtmax)
flag = ARKStepSetUserData(mem, userfun)
if typeof(abstol) <: Array
flag = ARKStepSVtolerances(mem, reltol, abstol)
else
flag = ARKStepSStolerances(mem, reltol, abstol)
end
flag = ARKStepSetMaxNumSteps(mem, maxiters)
flag = ARKStepSetMaxHnilWarns(mem, alg.max_hnil_warns)
flag = ARKStepSetMaxErrTestFails(mem, alg.max_error_test_failures)
flag = ARKStepSetMaxNonlinIters(mem, alg.max_nonlinear_iters)
flag = ARKStepSetMaxConvFails(mem, alg.max_convergence_failures)
flag = ARKStepSetPredictorMethod(mem, alg.predictor_method)
flag = ARKStepSetNonlinConvCoef(mem, alg.nonlinear_convergence_coefficient)
flag = ARKStepSetDenseOrder(mem,alg.dense_order)
#=
Reference from Manual on ARKODE
To choose an explicit table, set itable to a negative value. This automatically calls ARKStepSetExplicit(). However, if the problem is posed in explicit form, i.e. 𝑦 ̇ = 𝑓 (𝑡, 𝑦), then we recommend that the ERKStep time- stepper module be used instead of ARKStep.
To select an implicit table, set etable to a negative value. This automatically calls ARKStepSetImplicit(). If both itable and etable are non-negative, then these should match an existing implicit/explicit pair, listed in the section Additive Butcher tables. This automatically calls ARKStepSetImEx().
=#
if alg.itable == nothing && alg.etable == nothing
flag = ARKStepSetOrder(mem,alg.order)
elseif alg.itable == nothing && alg.etable != nothing
flag = ARKStepSetTableNum(mem, -1, alg.etable)
elseif alg.itable != nothing && alg.etable == nothing
flag = ARKStepSetTableNum(mem, alg.itable, -1)
else
flag = ARKStepSetTableNum(mem, alg.itable, alg.etable)
end
flag = ARKStepSetNonlinCRDown(mem,alg.crdown)
flag = ARKStepSetNonlinRDiv(mem, alg.rdiv)
flag = ARKStepSetDeltaGammaMax(mem, alg.dgmax)
flag = ARKStepSetMaxStepsBetweenLSet(mem, alg.msbp)
#flag = ARKStepSetAdaptivityMethod(mem,alg.adaptivity_method,1,0)
#flag = ARKStepSetFixedStep(mem,)
alg.set_optimal_params && (flag = ARKStepSetOptimalParams(mem))
if Method == :Newton # Only use a linear solver if it's a Newton-based method
if LinearSolver in (:Dense, :LapackDense)
nojacobian = false
A = SUNDenseMatrix(length(u0),length(u0))
_A = MatrixHandle(A,DenseMatrix())
if LinearSolver === :Dense
LS = SUNLinSol_Dense(u0,A)
_LS = LinSolHandle(LS,Dense())
else
LS = SUNLinSol_LapackDense(u0,A)
_LS = LinSolHandle(LS,LapackDense())
end
elseif LinearSolver in (:Band, :LapackBand)
nojacobian = false
A = SUNBandMatrix(length(u0), alg.jac_upper, alg.jac_lower)
_A = MatrixHandle(A,BandMatrix())
if LinearSolver === :Band
LS = SUNLinSol_Band(u0,A)
_LS = LinSolHandle(LS,Band())
else
LS = SUNLinSol_LapackBand(u0,A)
_LS = LinSolHandle(LS,LapackBand())
end
elseif LinearSolver == :GMRES
LS = SUNLinSol_SPGMR(u0, alg.prec_side, alg.krylov_dim)
_A = nothing
_LS = Sundials.LinSolHandle(LS,Sundials.SPGMR())
elseif LinearSolver == :FGMRES
LS = SUNLinSol_SPFGMR(u0, alg.prec_side, alg.krylov_dim)
_A = nothing
_LS = LinSolHandle(LS,SPFGMR())
elseif LinearSolver == :BCG
LS = SUNLinSol_SPBCGS(u0, alg.prec_side, alg.krylov_dim)
_A = nothing
_LS = LinSolHandle(LS,SPBCGS())
elseif LinearSolver == :PCG
LS = SUNLinSol_PCG(u0, alg.prec_side, alg.krylov_dim)
_A = nothing
_LS = LinSolHandle(LS,PCG())
elseif LinearSolver == :TFQMR
LS = SUNLinSol_SPTFQMR(u0, alg.prec_side, alg.krylov_dim)
_A = nothing
_LS = LinSolHandle(LS,PTFQMR())
elseif LinearSolver == :KLU
nnz = length(SparseArrays.nonzeros(prob.f.jac_prototype))
A = SUNSparseMatrix(length(u0),length(u0), nnz, CSC_MAT)
LS = SUNLinSol_KLU(u0, A)
_A = MatrixHandle(A,SparseMatrix())
_LS = LinSolHandle(LS,KLU())
end
flag = ARKStepSetLinearSolver(mem, LS, _A === nothing ? C_NULL : A)
elseif Method == :Functional
ARKStepSetFixedPoint(mem, Clong(alg.krylov_dim))
else
_A = nothing
_LS = nothing
end
if (typeof(prob.problem_type) <: SplitODEProblem &&
typeof(prob.f.f1.jac_prototype) <: DiffEqBase.AbstractDiffEqLinearOperator) ||
(!(typeof(prob.problem_type) <: SplitODEProblem) &&
typeof(prob.f.jac_prototype) <: DiffEqBase.AbstractDiffEqLinearOperator)
function getcfunjtimes(::T) where T
@cfunction(jactimes,
Cint,
(N_Vector,
N_Vector,
realtype,
N_Vector,
N_Vector,
Ref{T},
N_Vector))
end
jtimes = getcfunjtimes(userfun)
ARKStepSetJacTimes(mem, C_NULL, jtimes)
end
if prob.f.mass_matrix != LinearAlgebra.I
if MassLinearSolver in (:Dense, :LapackDense)
nojacobian = false
M = SUNDenseMatrix(length(u0),length(u0))
_M = MatrixHandle(M,DenseMatrix())
if MassLinearSolver === :Dense
MLS = SUNLinSol_Dense(u0,M)
_MLS = LinSolHandle(MLS,Dense())
else
MLS = SUNLinSol_LapackDense(u0,M)
_MLS = LinSolHandle(MLS,LapackDense())
end
elseif MassLinearSolver in (:Band, :LapackBand)
nojacobian = false
M = SUNBandMatrix(length(u0), alg.jac_upper, alg.jac_lower)
_M = MatrixHandle(M,BandMatrix())
if MassLinearSolver === :Band
MLS = SUNLinSol_Band(u0,M)
_MLS = LinSolHandle(MLS,Band())
else
MLS = SUNLinSol_LapackBand(u0,M)
_MLS = LinSolHandle(MLS,LapackBand())
end
elseif MassLinearSolver == :GMRES
MLS = SUNLinSol_SPGMR(u0, alg.prec_side, alg.mass_krylov_dim)
_M = nothing
_MLS = LinSolHandle(MLS,SPGMR())
elseif MassLinearSolver == :FGMRES
MLS = SUNLinSol_SPGMR(u0, alg.prec_side, alg.mass_krylov_dim)
_M = nothing
_MLS = LinSolHandle(MLS,SPFGMR())
elseif MassLinearSolver == :BCG
MLS = SUNLinSol_SPGMR(u0, alg.prec_side, alg.mass_krylov_dim)
_M = nothing
_MLS = LinSolHandle(MLS,SPBCGS())
elseif MassLinearSolver == :PCG
MLS = SUNLinSol_SPGMR(u0, alg.prec_side, alg.mass_krylov_dim)
_M = nothing
_MLS = LinSolHandle(MLS,PCG())
elseif MassLinearSolver == :TFQMR
MLS = SUNLinSol_SPGMR(u0, alg.prec_side, alg.mass_krylov_dim)
_M = nothing
_MLS = LinSolHandle(MLS,PTFQMR())
elseif MassLinearSolver == :KLU
nnz = length(SparseArrays.nonzeros(prob.f.mass_matrix))
M = SUNSparseMatrix(length(u0),length(u0), nnz, CSC_MAT)
MLS = SUNLinSol_KLU(u0, M)
_M = MatrixHandle(M,SparseMatrix())
_MLS = LinSolHandle(MLS,KLU())
end
flag = ARKStepSetMassLinearSolver(mem, MLS, _M === nothing ? C_NULL : M, false)
function getmatfun(::T) where T
@cfunction(massmat,
Cint,
(realtype,
SUNMatrix,
Ref{T},
N_Vector,
N_Vector,
N_Vector))
end
matfun = getmatfun(userfun)
ARKStepSetMassFn(mem,matfun)
else
_M = nothing
_MLS = nothing
end
if DiffEqBase.has_jac(prob.f)
function getfunjac(::T) where T
@cfunction(cvodejac,
Cint,
(realtype,
N_Vector,
N_Vector,
SUNMatrix,
Ref{T},
N_Vector,
N_Vector,
N_Vector))
end
jac = getfunjac(userfun)
flag = ARKStepSetUserData(mem, userfun)
flag = ARKStepSetJacFn(mem, jac)
else
jac = nothing
end
if alg.prec !== nothing
function getpercfun(::T) where T
@cfunction(precsolve,
Cint,
(Float64,
N_Vector,
N_Vector,
N_Vector,
N_Vector,Float64,Float64,Int,
Ref{T}))
end
precfun = getpercfun(userfun)
function getpsetupfun(::T) where T
@cfunction(precsetup,
Cint,
(Float64,
N_Vector,
N_Vector,
Int,
Ptr{Int},Float64,Ref{T}))
end
psetupfun = alg.psetup === nothing ? C_NULL : getpsetupfun(userfun)
ARKStepSetPreconditioner(mem, psetupfun, precfun)
end
callbacks_internal == nothing ? tmp = nothing : tmp = similar(u0)
callbacks_internal == nothing ? uprev = nothing : uprev = similar(u0)
tout = [tspan[1]]
if save_start
save_value!(ures,u0,uType,sizeu)
if dense
f!(_u0,u0,prob.p,tspan[1])
save_value!(dures,utmp,uType,sizeu)
end
end
sol = DiffEqBase.build_solution(prob, alg, ts, ures,
dense = dense,
interp = dense ? DiffEqBase.HermiteInterpolation(ts,ures,dures) :
DiffEqBase.LinearInterpolation(ts,ures),
timeseries_errors = timeseries_errors,
destats = DiffEqBase.DEStats(0),
calculate_error = false)
opts = DEOptions(saveat_internal,tstops_internal,save_everystep,dense,
timeseries_errors,dense_errors,save_on,save_end,
callbacks_internal,abstol,reltol,verbose,advance_to_tstop,stop_at_next_tstop,
progress,progress_name,progress_message,maxiters)
integrator = ARKODEIntegrator(utmp,prob.p,t0,t0,mem,_LS,_A,_MLS,_M,sol,alg,f!,userfun,jac,opts,
tout,tdir,sizeu,false,tmp,uprev,Cint(flag),false,0,1,callback_cache,0.)
initialize_callbacks!(integrator)
integrator
end # function solve
function tstop_saveat_disc_handling(tstops,saveat,tdir,tspan,tType)
if isempty(tstops) # TODO: Specialize more
tstops_vec = [tspan[2]]
else
tstops_vec = vec(collect(tType,Iterators.filter(x->tdir*tspan[1]<tdir*x≤tdir*tspan[end],Iterators.flatten((tstops,tspan[end])))))
end
if tdir>0
tstops_internal = DataStructures.BinaryMinHeap(tstops_vec)
else
tstops_internal = DataStructures.BinaryMaxHeap(tstops_vec)
end
if typeof(saveat) <: Number
if (tspan[1]:saveat:tspan[end])[end] == tspan[end]
saveat_vec = convert(Vector{tType},collect(tType,tspan[1]+saveat:saveat:tspan[end]))
else
saveat_vec = convert(Vector{tType},collect(tType,tspan[1]+saveat:saveat:(tspan[end]-saveat)))
end
elseif isempty(saveat)
saveat_vec = saveat
else
saveat_vec = vec(collect(tType,Iterators.filter(x->tdir*tspan[1]<tdir*x<tdir*tspan[end],saveat)))
end
if tdir>0
saveat_internal = DataStructures.BinaryMinHeap(saveat_vec)
else
saveat_internal = DataStructures.BinaryMaxHeap(saveat_vec)
end
tstops_internal,saveat_internal
end
## Solve for DAEs uses IDA
function DiffEqBase.__init(
prob::DiffEqBase.AbstractDAEProblem{uType, duType, tupType, isinplace},
alg::SundialsDAEAlgorithm{LinearSolver},
timeseries=[], ts=[], ks=[];
verbose=true,
dt=nothing, dtmax=0.0,
save_on=true, save_start=true,
callback=nothing, abstol=1/10^6, reltol=1/10^3,
saveat=Float64[], tstops=Float64[], maxiters=Int(1e5),
timeseries_errors=true,
dense_errors = false,
save_everystep=isempty(saveat), dense=save_everystep,
save_timeseries=nothing, save_end = true,
progress=false,progress_name="ODE",
progress_message = DiffEqBase.ODE_DEFAULT_PROG_MESSAGE,
advance_to_tstop = false, stop_at_next_tstop = false,
userdata=nothing,
kwargs...) where {uType, duType, tupType, isinplace, LinearSolver}
tType = eltype(tupType)
if verbose
warned = !isempty(kwargs) && DiffEqBase.check_keywords(alg, kwargs, warnida)
if !(typeof(prob.f) <: DiffEqBase.AbstractParameterizedFunction)
if DiffEqBase.has_tgrad(prob.f)
@warn("Explicit t-gradient given to this stiff solver is ignored.")
warned = true
end
end
warned && DiffEqBase.warn_compat()
end
if typeof(reltol) <: AbstractArray
error("Sundials only allows scalar reltol.")
end
progress && Logging.@logmsg(-1,progress_name,_id=_id = :Sundials,progress=0)
callbacks_internal = DiffEqBase.CallbackSet(callback)
max_len_cb = DiffEqBase.max_vector_callback_length(callbacks_internal)
if max_len_cb isa VectorContinuousCallback
callback_cache = DiffEqBase.CallbackCache(max_len_cb.len,Float64,Float64)
else
callback_cache = nothing
end
tspan = prob.tspan
t0 = tspan[1]
tdir = sign(tspan[2]-tspan[1])
tstops_internal, saveat_internal =
tstop_saveat_disc_handling(tstops,saveat,tdir,tspan,tType)
if typeof(prob.u0) <: Number
u0 = [prob.u0]
else
u0 = vec(copy(prob.u0))
end
if typeof(prob.du0) <: Number
du0 = [prob.du0]
else
du0 = vec(copy(prob.du0))
end
sizeu = size(prob.u0)
sizedu = size(prob.du0)
### Fix the more general function to Sundials allowed style
if !isinplace && typeof(prob.u0)<:Number
f! = (out, du, u, p, t) -> (out .= prob.f(first(du),first(u), p, t); Cint(0))
elseif !isinplace && typeof(prob.u0)<:Vector{Float64}
f! = (out, du, u, p, t) -> (out .= prob.f(du, u, p, t); Cint(0))
elseif !isinplace && typeof(prob.u0)<:AbstractArray
f! = (out, du, u, p, t) -> (out .= vec(
prob.f(reshape(du, sizedu), reshape(u, sizeu), p, t)
);Cint(0))
elseif typeof(prob.u0)<:Vector{Float64}
f! = prob.f
else # Then it's an in-place function on an abstract array
f! = (out, du, u, p, t) -> (prob.f(reshape(out, sizeu), reshape(du, sizedu),
reshape(u, sizeu), p, t); Cint(0))
end
mem_ptr = IDACreate()
(mem_ptr == C_NULL) && error("Failed to allocate IDA solver object")
mem = Handle(mem_ptr)
!verbose && IDASetErrHandlerFn(mem,@cfunction(null_error_handler, Nothing,
(Cint, Char,
Char, Ptr{Cvoid})),C_NULL)
ures = Vector{uType}()
dures = Vector{uType}()
ts = [t0]
_u0 = copy(u0)
utmp = NVector(_u0)
_du0 = copy(du0)
dutmp = NVector(_du0)
rtest = zeros(length(u0))
userfun = FunJac(f!,prob.f.jac,prob.p,nothing,prob.f.jac_prototype,alg.prec,alg.psetup,_u0,_du0,rtest)
u0nv = NVector(u0)
function getcfun(::T) where T
@cfunction(idasolfun,
Cint, (realtype, N_Vector, N_Vector,
N_Vector, Ref{T}))
end
cfun = getcfun(userfun)
flag = IDAInit(mem, cfun,
t0, convert(N_Vector, utmp),
convert(N_Vector, dutmp))
dt != nothing && (flag = IDASetInitStep(mem, dt))
flag = IDASetUserData(mem, userfun)
flag = IDASetMaxStep(mem, dtmax)
if typeof(abstol) <: Array
flag = IDASVtolerances(mem, reltol, abstol)
else
flag = IDASStolerances(mem, reltol, abstol)
end
flag = IDASetMaxNumSteps(mem, maxiters)
flag = IDASetMaxOrd(mem,alg.max_order)
flag = IDASetMaxErrTestFails(mem,alg.max_error_test_failures)
flag = IDASetNonlinConvCoef(mem,alg.nonlinear_convergence_coefficient)
flag = IDASetMaxNonlinIters(mem,alg.max_nonlinear_iters)
flag = IDASetMaxConvFails(mem,alg.max_convergence_failures)
flag = IDASetNonlinConvCoefIC(mem,alg.nonlinear_convergence_coefficient_ic)
flag = IDASetMaxNumStepsIC(mem,alg.max_num_steps_ic)
flag = IDASetMaxNumJacsIC(mem,alg.max_num_jacs_ic)
flag = IDASetMaxNumItersIC(mem,alg.max_num_iters_ic)
#flag = IDASetMaxBacksIC(mem,alg.max_num_backs_ic) # Needs newer version?
flag = IDASetLineSearchOffIC(mem,alg.use_linesearch_ic)
if LinearSolver in (:Dense, :LapackDense)
nojacobian = false
A = SUNDenseMatrix(length(u0),length(u0))
_A = MatrixHandle(A,DenseMatrix())
if LinearSolver === :Dense
LS = SUNLinSol_Dense(u0,A)
_LS = LinSolHandle(LS,Dense())
else
LS = SUNLinSol_LapackDense(u0,A)
_LS = LinSolHandle(LS,LapackDense())
end
elseif LinearSolver in (:Band, :LapackBand)
nojacobian = false
A = SUNBandMatrix(length(u0), alg.jac_upper, alg.jac_lower)
_A = MatrixHandle(A,BandMatrix())
if LinearSolver === :Band
LS = SUNLinSol_Band(u0,A)
_LS = LinSolHandle(LS,Band())
else
LS = SUNLinSol_LapackBand(u0,A)
_LS = LinSolHandle(LS,LapackBand())
end
elseif LinearSolver == :GMRES
LS = SUNLinSol_SPGMR(u0, alg.prec_side, alg.krylov_dim)
_A = nothing
_LS = LinSolHandle(LS,SPGMR())
elseif LinearSolver == :FGMRES
LS = SUNLinSol_SPFGMR(u0, alg.prec_side, alg.krylov_dim)
_A = nothing
_LS = LinSolHandle(LS,SPFGMR())
elseif LinearSolver == :BCG
LS = SUNLinSol_SPBCGS(u0, alg.prec_side, alg.krylov_dim)
_A = nothing
_LS = LinSolHandle(LS,SPBCGS())
elseif LinearSolver == :PCG
LS = SUNLinSol_PCG(u0, alg.prec_side, alg.krylov_dim)
_A = nothing
_LS = LinSolHandle(LS,PCG())
elseif LinearSolver == :TFQMR
LS = SUNLinSol_SPTFQMR(u0, alg.prec_side, alg.krylov_dim)
_A = nothing
_LS = LinSolHandle(LS,PTFQMR())
elseif LinearSolver == :KLU
nnz = length(SparseArrays.nonzeros(prob.f.jac_prototype))
A = SUNSparseMatrix(length(u0),length(u0), nnz, Sundials.CSC_MAT)
LS = SUNLinSol_KLU(u0, A)
_A = MatrixHandle(A,SparseMatrix())
_LS = LinSolHandle(LS,KLU())
end
flag = IDASetLinearSolver(mem, LS, _A === nothing ? C_NULL : A)
if typeof(prob.f.jac_prototype) <: DiffEqBase.AbstractDiffEqLinearOperator
function getcfunjtimes(::T) where T
@cfunction(idajactimes,
Cint,
(realtype,
N_Vector,N_Vector,N_Vector,N_Vector,N_Vector,
realtype,
Ref{T},
N_Vector,N_Vector))
end
jtimes = getcfunjtimes(userfun)
IDASetJacTimes(mem, C_NULL, jtimes)
end
if alg.prec !== nothing
function getpercfun(::T) where T
@cfunction(idaprecsolve,
Cint,
(Float64,