-
-
Notifications
You must be signed in to change notification settings - Fork 393
/
nlp.jl
1465 lines (1285 loc) · 55.7 KB
/
nlp.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
# Copyright 2017, Iain Dunning, Joey Huchette, Miles Lubin, and contributors
# This Source Code Form is subject to the terms of the Mozilla Public
# License, v. 2.0. If a copy of the MPL was not distributed with this
# file, You can obtain one at http://mozilla.org/MPL/2.0/.
mutable struct NonlinearExprData
nd::Vector{NodeData}
const_values::Vector{Float64}
end
include("parsenlp.jl")
const NonlinearConstraint = GenericRangeConstraint{NonlinearExprData}
mutable struct NLPData
nlobj
nlconstr::Vector{NonlinearConstraint}
nlexpr::Vector{NonlinearExprData}
nlconstrDuals::Vector{Float64}
nlparamvalues::Vector{Float64}
user_operators::ReverseDiffSparse.UserOperatorRegistry
largest_user_input_dimension::Int
evaluator
end
function NonlinearExpression(m::Model,ex::NonlinearExprData)
initNLP(m)
nldata::NLPData = m.nlpdata
push!(nldata.nlexpr, ex)
return NonlinearExpression(m, length(nldata.nlexpr))
end
function newparameter(m::Model,value::Number)
initNLP(m)
nldata::NLPData = m.nlpdata
push!(nldata.nlparamvalues, value)
return NonlinearParameter(m, length(nldata.nlparamvalues))
end
getvalue(p::NonlinearParameter) = p.m.nlpdata.nlparamvalues[p.index]::Float64
setvalue(p::NonlinearParameter,v::Number) = (p.m.nlpdata.nlparamvalues[p.index] = v)
NLPData() = NLPData(nothing, NonlinearConstraint[], NonlinearExprData[], Float64[], Float64[], ReverseDiffSparse.UserOperatorRegistry(), 0, nothing)
Base.copy(::NLPData) = error("Copying nonlinear problems not yet implemented")
function initNLP(m::Model)
if m.nlpdata === nothing
m.nlpdata = NLPData()
end
end
function getdual(c::ConstraintRef{Model,NonlinearConstraint})
initNLP(c.m)
nldata::NLPData = c.m.nlpdata
if length(nldata.nlconstrDuals) != length(nldata.nlconstr)
getdualwarn(c)
NaN
else
nldata.nlconstrDuals[c.idx]
end
end
mutable struct FunctionStorage
nd::Vector{NodeData}
adj::SparseMatrixCSC{Bool,Int}
const_values::Vector{Float64}
forward_storage::Vector{Float64}
partials_storage::Vector{Float64}
reverse_storage::Vector{Float64}
grad_sparsity::Vector{Int}
hess_I::Vector{Int} # nonzero pattern of hessian
hess_J::Vector{Int}
rinfo::Coloring.RecoveryInfo # coloring info for hessians
seed_matrix::Matrix{Float64}
linearity::Linearity
dependent_subexpressions::Vector{Int} # subexpressions which this function depends on, ordered for forward pass
end
mutable struct SubexpressionStorage
nd::Vector{NodeData}
adj::SparseMatrixCSC{Bool,Int}
const_values::Vector{Float64}
forward_storage::Vector{Float64}
partials_storage::Vector{Float64}
reverse_storage::Vector{Float64}
forward_storage_ϵ::Vector{Float64}
partials_storage_ϵ::Vector{Float64}
reverse_storage_ϵ::Vector{Float64}
linearity::Linearity
end
mutable struct NLPEvaluator <: MathProgBase.AbstractNLPEvaluator
m::Model
A::SparseMatrixCSC{Float64,Int} # linear constraint matrix
parameter_values::Vector{Float64}
has_nlobj::Bool
linobj::Vector{Float64}
objective::FunctionStorage
constraints::Vector{FunctionStorage}
subexpressions::Vector{SubexpressionStorage}
subexpression_order::Vector{Int}
subexpression_forward_values::Vector{Float64}
subexpression_reverse_values::Vector{Float64}
subexpression_linearity::Vector{ReverseDiffSparse.Linearity}
subexpressions_as_julia_expressions::Vector{Any}
last_x::Vector{Float64}
jac_storage::Vector{Float64} # temporary storage for computing jacobians
user_output_buffer::Vector{Float64} # temporary storage for user-defined functions
# storage for computing hessians
# these Float64 vectors are reinterpreted to hold multiple epsilon components
# so the length should be multiplied by the maximum number of epsilon components
disable_2ndorder::Bool # don't offer Hess or HessVec
want_hess::Bool
forward_storage_ϵ::Vector{Float64} # (longest expression)
partials_storage_ϵ::Vector{Float64} # (longest expression)
reverse_storage_ϵ::Vector{Float64} # (longest expression)
input_ϵ::Vector{Float64} # (number of variables)
output_ϵ::Vector{Float64}# (number of variables)
subexpression_forward_values_ϵ::Vector{Float64} # (number of subexpressions)
subexpression_reverse_values_ϵ::Vector{Float64} # (number of subexpressions)
# hessian sparsity pattern
hess_I::Vector{Int}
hess_J::Vector{Int}
max_chunk::Int # chunk size for which we've allocated storage
# timers
eval_f_timer::Float64
eval_g_timer::Float64
eval_grad_f_timer::Float64
eval_jac_g_timer::Float64
eval_hesslag_timer::Float64
function NLPEvaluator(m::Model)
d = new(m)
numVar = m.numCols
d.A = prepConstrMatrix(m)
# check if we have any user-defined operators, in which case we need to
# disable hessians.
if isa(m.nlpdata,NLPData)
nldata::NLPData = m.nlpdata
has_nlobj = isa(nldata.nlobj, NonlinearExprData)
has_user_mv_operator = false
for nlexpr in nldata.nlexpr
has_user_mv_operator |= ReverseDiffSparse.has_user_multivariate_operators(nlexpr.nd)
end
if has_nlobj
has_user_mv_operator |= ReverseDiffSparse.has_user_multivariate_operators(nldata.nlobj.nd)
end
for nlconstr in nldata.nlconstr
has_user_mv_operator |= ReverseDiffSparse.has_user_multivariate_operators(nlconstr.terms.nd)
end
d.disable_2ndorder = has_user_mv_operator
d.user_output_buffer = Array{Float64}(m.nlpdata.largest_user_input_dimension)
d.jac_storage = Array{Float64}(max(numVar,m.nlpdata.largest_user_input_dimension))
else
d.disable_2ndorder = false
d.user_output_buffer = Array{Float64}(0)
d.jac_storage = Array{Float64}(numVar)
end
d.eval_f_timer = 0
d.eval_g_timer = 0
d.eval_grad_f_timer = 0
d.eval_jac_g_timer = 0
d.eval_hesslag_timer = 0
return d
end
end
function simplify_expression(nd::Vector{NodeData}, const_values, subexpression_linearity, fixed_variables, parameter_values, x_values, subexpression_values)
adj = adjmat(nd)
forward_storage = zeros(length(nd))
partials_storage = zeros(length(nd))
linearity = classify_linearity(nd, adj, subexpression_linearity, fixed_variables)
forward_eval(forward_storage, partials_storage, nd, adj, const_values, parameter_values, x_values, subexpression_values)
nd_new = simplify_constants(forward_storage, nd, adj, const_values, linearity)
return nd_new, forward_storage[1], linearity[1]
end
function FunctionStorage(nd::Vector{NodeData}, const_values,numVar, coloring_storage, want_hess::Bool, subexpr::Vector{Vector{NodeData}}, dependent_subexpressions, subexpression_linearity, subexpression_edgelist, subexpression_variables, fixed_variables)
adj = adjmat(nd)
forward_storage = zeros(length(nd))
partials_storage = zeros(length(nd))
reverse_storage = zeros(length(nd))
empty!(coloring_storage)
compute_gradient_sparsity!(coloring_storage, nd)
for k in dependent_subexpressions
compute_gradient_sparsity!(coloring_storage,subexpr[k])
end
grad_sparsity = sort!(collect(coloring_storage))
empty!(coloring_storage)
if want_hess
# compute hessian sparsity
linearity = classify_linearity(nd, adj, subexpression_linearity, fixed_variables)
edgelist = compute_hessian_sparsity(nd, adj, linearity, coloring_storage, subexpression_edgelist, subexpression_variables)
hess_I, hess_J, rinfo = Coloring.hessian_color_preprocess(edgelist, numVar, coloring_storage)
seed_matrix = Coloring.seed_matrix(rinfo)
if linearity[1] == NONLINEAR
@assert length(hess_I) > 0
end
else
hess_I = hess_J = Int[]
rinfo = Coloring.RecoveryInfo()
seed_matrix = Array{Float64}(0,0)
linearity = [NONLINEAR]
end
return FunctionStorage(nd, adj, const_values, forward_storage, partials_storage, reverse_storage, grad_sparsity, hess_I, hess_J, rinfo, seed_matrix, linearity[1],dependent_subexpressions)
end
function SubexpressionStorage(nd::Vector{NodeData}, const_values,numVar, fixed_variables,subexpression_linearity)
adj = adjmat(nd)
forward_storage = zeros(length(nd))
partials_storage = zeros(length(nd))
reverse_storage = zeros(length(nd))
linearity = classify_linearity(nd, adj, subexpression_linearity, fixed_variables)
empty_arr = Array{Float64}(0)
return SubexpressionStorage(nd, adj, const_values, forward_storage, partials_storage, reverse_storage, empty_arr, empty_arr, empty_arr, linearity[1])
end
function MathProgBase.initialize(d::NLPEvaluator, requested_features::Vector{Symbol})
for feat in requested_features
if !(feat in MathProgBase.features_available(d))
error("Unsupported feature $feat")
# TODO: implement Jac-vec products
# for solvers that need them
end
end
if d.eval_f_timer != 0
# we've already been initialized
# assume no new features are being requested.
return
end
initNLP(d.m) #in case the problem is purely linear/quadratic thus far
nldata::NLPData = d.m.nlpdata
d.constraints = FunctionStorage[]
d.last_x = fill(NaN, d.m.numCols)
SIMPLIFY = d.m.simplify_nonlinear_expressions
fixed_variables = SIMPLIFY ? d.m.colLower .== d.m.colUpper : d.m.colLower .== NaN
d.m.colVal[fixed_variables] = d.m.colLower[fixed_variables]
d.parameter_values = nldata.nlparamvalues
tic()
d.linobj = prepAffObjective(d.m)
linrowlb, linrowub = prepConstrBounds(d.m)
numVar = length(d.linobj)
d.want_hess = (:Hess in requested_features)
want_hess_storage = (:HessVec in requested_features) || d.want_hess
coloring_storage = ReverseDiffSparse.Coloring.IndexedSet(numVar)
d.has_nlobj = isa(nldata.nlobj, NonlinearExprData)
max_expr_length = 0
main_expressions = Array{Vector{NodeData}}(0)
subexpr = Array{Vector{NodeData}}(0)
for nlexpr in nldata.nlexpr
push!(subexpr, nlexpr.nd)
end
if d.has_nlobj
push!(main_expressions,nldata.nlobj.nd)
end
for nlconstr in nldata.nlconstr
push!(main_expressions,nlconstr.terms.nd)
end
d.subexpression_order, individual_order = order_subexpressions(main_expressions,subexpr)
d.subexpression_linearity = Array{Linearity}(length(nldata.nlexpr))
subexpression_variables = Array{Vector{Int}}(length(nldata.nlexpr))
subexpression_edgelist = Array{Set{Tuple{Int,Int}}}(length(nldata.nlexpr))
d.subexpressions = Array{SubexpressionStorage}(length(nldata.nlexpr))
d.subexpression_forward_values = Array{Float64}(length(d.subexpressions))
d.subexpression_reverse_values = Array{Float64}(length(d.subexpressions))
empty_edgelist = Set{Tuple{Int,Int}}()
for k in d.subexpression_order # only load expressions which actually are used
if SIMPLIFY
nd_new, forward_value, simplified_linearity = simplify_expression(nldata.nlexpr[k].nd, nldata.nlexpr[k].const_values,d.subexpression_linearity, fixed_variables, d.parameter_values, d.m.colVal, d.subexpression_forward_values)
else
nd_new = nldata.nlexpr[k].nd
forward_value = NaN
simplified_linearity = NONLINEAR
end
d.subexpression_forward_values[k] = forward_value
if simplified_linearity != CONSTANT
d.subexpressions[k] = SubexpressionStorage(nd_new, nldata.nlexpr[k].const_values, numVar, fixed_variables, d.subexpression_linearity)
subex = d.subexpressions[k]
d.subexpression_linearity[k] = subex.linearity
@assert !SIMPLIFY || subex.linearity != CONSTANT
if d.want_hess
empty!(coloring_storage)
compute_gradient_sparsity!(coloring_storage,subex.nd)
# union with all dependent expressions
for idx in list_subexpressions(subex.nd)
union!(coloring_storage, subexpression_variables[idx])
end
subexpression_variables[k] = collect(coloring_storage)
empty!(coloring_storage)
linearity = classify_linearity(subex.nd, subex.adj, d.subexpression_linearity, fixed_variables)
edgelist = compute_hessian_sparsity(subex.nd, subex.adj, linearity,coloring_storage,subexpression_edgelist, subexpression_variables)
subexpression_edgelist[k] = edgelist
end
else
d.subexpression_linearity[k] = simplified_linearity
subexpression_edgelist[k] = empty_edgelist
end
end
if :ExprGraph in requested_features
d.subexpressions_as_julia_expressions = Array{Any}(length(subexpr))
for k in d.subexpression_order
if d.subexpression_linearity[k] != CONSTANT || !SIMPLIFY
ex = d.subexpressions[k]
d.subexpressions_as_julia_expressions[k] = tapeToExpr(d.m, 1, ex.nd, ex.adj, ex.const_values, d.parameter_values, d.subexpressions_as_julia_expressions, nldata.user_operators, true, true)
else
d.subexpressions_as_julia_expressions[k] = d.subexpression_forward_values[k]
end
end
end
if SIMPLIFY
main_expressions = Array{Vector{NodeData}}(0)
# simplify objective and constraint expressions
if d.has_nlobj
nd_new, forward_value = simplify_expression(nldata.nlobj.nd, nldata.nlobj.const_values,d.subexpression_linearity, fixed_variables, d.parameter_values, d.m.colVal, d.subexpression_forward_values)
push!(main_expressions,nd_new)
end
for k in 1:length(nldata.nlconstr)
nd_new, forward_value = simplify_expression(nldata.nlconstr[k].terms.nd, nldata.nlconstr[k].terms.const_values,d.subexpression_linearity, fixed_variables, d.parameter_values, d.m.colVal, d.subexpression_forward_values)
push!(main_expressions,nd_new)
end
# recompute dependencies after simplification
d.subexpression_order, individual_order = order_subexpressions(main_expressions,subexpr)
subexpr = Array{Vector{NodeData}}(length(d.subexpressions))
for k in d.subexpression_order
subexpr[k] = d.subexpressions[k].nd
end
end
max_chunk = 1
if d.has_nlobj
@assert length(d.m.obj.qvars1) == 0 && length(d.m.obj.aff.vars) == 0
nd = main_expressions[1]
d.objective = FunctionStorage(nd, nldata.nlobj.const_values, numVar, coloring_storage, d.want_hess, subexpr, individual_order[1], d.subexpression_linearity, subexpression_edgelist, subexpression_variables, fixed_variables)
max_expr_length = max(max_expr_length, length(d.objective.nd))
max_chunk = max(max_chunk, size(d.objective.seed_matrix,2))
end
for k in 1:length(nldata.nlconstr)
nlconstr = nldata.nlconstr[k]
idx = (d.has_nlobj) ? k+1 : k
nd = main_expressions[idx]
push!(d.constraints, FunctionStorage(nd, nlconstr.terms.const_values, numVar, coloring_storage, d.want_hess, subexpr, individual_order[idx], d.subexpression_linearity, subexpression_edgelist, subexpression_variables, fixed_variables))
max_expr_length = max(max_expr_length, length(d.constraints[end].nd))
max_chunk = max(max_chunk, size(d.constraints[end].seed_matrix,2))
end
max_chunk = min(max_chunk, 10) # 10 is hardcoded upper bound to avoid excess memory allocation
if d.want_hess || want_hess_storage # storage for Hess or HessVec
d.input_ϵ = Array{Float64}(max_chunk*d.m.numCols)
d.output_ϵ = Array{Float64}(max_chunk*d.m.numCols)
d.forward_storage_ϵ = Array{Float64}(max_chunk*max_expr_length)
d.partials_storage_ϵ = Array{Float64}(max_chunk*max_expr_length)
d.reverse_storage_ϵ = Array{Float64}(max_chunk*max_expr_length)
d.subexpression_forward_values_ϵ = Array{Float64}(max_chunk*length(d.subexpressions))
d.subexpression_reverse_values_ϵ = Array{Float64}(max_chunk*length(d.subexpressions))
for k in d.subexpression_order
subex = d.subexpressions[k]
subex.forward_storage_ϵ = zeros(Float64,max_chunk*length(subex.nd))
subex.partials_storage_ϵ = zeros(Float64,max_chunk*length(subex.nd))
subex.reverse_storage_ϵ = zeros(Float64,max_chunk*length(subex.nd))
end
d.max_chunk = max_chunk
if d.want_hess
d.hess_I, d.hess_J = _hesslag_structure(d)
# JIT warm-up
MathProgBase.eval_hesslag(d, Array{Float64}(length(d.hess_I)), d.m.colVal, 1.0, ones(MathProgBase.numconstr(d.m)))
end
end
# JIT warm-up
if :Grad in requested_features
MathProgBase.eval_grad_f(d, zeros(numVar), d.m.colVal)
MathProgBase.eval_g(d, zeros(MathProgBase.numconstr(d.m)), d.m.colVal)
end
tprep = toq()
#println("Prep time: $tprep")
# reset timers
d.eval_f_timer = 0
d.eval_grad_f_timer = 0
d.eval_g_timer = 0
d.eval_jac_g_timer = 0
d.eval_hesslag_timer = 0
nothing
end
function MathProgBase.features_available(d::NLPEvaluator)
features = [:Grad, :Jac, :ExprGraph]
if !d.disable_2ndorder
push!(features,:Hess)
push!(features,:HessVec)
end
return features
end
function forward_eval_all(d::NLPEvaluator,x)
# do a forward pass on all expressions at x
subexpr_values = d.subexpression_forward_values
user_operators = d.m.nlpdata.user_operators::ReverseDiffSparse.UserOperatorRegistry
user_input_buffer = d.jac_storage
user_output_buffer = d.user_output_buffer
SIMPLIFY = d.m.simplify_nonlinear_expressions
for k in d.subexpression_order
if SIMPLIFY && d.subexpression_linearity[k] == CONSTANT
continue
end
ex = d.subexpressions[k]
subexpr_values[k] = forward_eval(ex.forward_storage,ex.partials_storage,ex.nd,ex.adj,ex.const_values,d.parameter_values,x,subexpr_values,user_input_buffer,user_output_buffer, user_operators=user_operators)
end
if d.has_nlobj
ex = d.objective
forward_eval(ex.forward_storage,ex.partials_storage,ex.nd,ex.adj,ex.const_values,d.parameter_values,x,subexpr_values,user_input_buffer,user_output_buffer,user_operators=user_operators)
end
for ex in d.constraints
forward_eval(ex.forward_storage,ex.partials_storage,ex.nd,ex.adj,ex.const_values,d.parameter_values,x,subexpr_values,user_input_buffer,user_output_buffer,user_operators=user_operators)
end
end
function reverse_eval_all(d::NLPEvaluator,x)
# do a reverse pass on all expressions at x
subexpr_reverse_values = d.subexpression_reverse_values
subexpr_values = d.subexpression_forward_values
grad_storage = d.jac_storage
SIMPLIFY = d.m.simplify_nonlinear_expressions
for k in d.subexpression_order
if SIMPLIFY && d.subexpression_linearity[k] == CONSTANT
continue
end
ex = d.subexpressions[k]
reverse_eval(ex.reverse_storage,ex.partials_storage,ex.nd,ex.adj)
end
if d.has_nlobj
ex = d.objective
reverse_eval(ex.reverse_storage,ex.partials_storage,ex.nd,ex.adj)
end
for ex in d.constraints
reverse_eval(ex.reverse_storage,ex.partials_storage,ex.nd,ex.adj)
end
copy!(d.last_x,x)
end
function MathProgBase.eval_f(d::NLPEvaluator, x)
tic()
if d.last_x != x
forward_eval_all(d,x)
reverse_eval_all(d,x)
end
val = zero(eltype(x))
if d.has_nlobj
val = d.objective.forward_storage[1]
else
qobj = d.m.obj::QuadExpr
val = dot(x,d.linobj) + qobj.aff.constant
for k in 1:length(qobj.qvars1)
val += qobj.qcoeffs[k]*x[qobj.qvars1[k].col]*x[qobj.qvars2[k].col]
end
end
d.eval_f_timer += toq()
return val
end
function MathProgBase.eval_grad_f(d::NLPEvaluator, g, x)
tic()
if d.last_x != x
forward_eval_all(d,x)
reverse_eval_all(d,x)
end
SIMPLIFY = d.m.simplify_nonlinear_expressions
if d.has_nlobj
fill!(g,0.0)
ex = d.objective
subexpr_reverse_values = d.subexpression_reverse_values
subexpr_reverse_values[ex.dependent_subexpressions] = 0.0
reverse_extract(g,ex.reverse_storage,ex.nd,ex.adj,subexpr_reverse_values,1.0)
for i in length(ex.dependent_subexpressions):-1:1
k = ex.dependent_subexpressions[i]
if SIMPLIFY && d.subexpression_linearity[k] == CONSTANT
continue
end
subexpr = d.subexpressions[k]
reverse_extract(g,subexpr.reverse_storage,subexpr.nd,subexpr.adj,subexpr_reverse_values,subexpr_reverse_values[k])
end
else
copy!(g,d.linobj)
qobj::QuadExpr = d.m.obj
for k in 1:length(qobj.qvars1)
coef = qobj.qcoeffs[k]
g[qobj.qvars1[k].col] += coef*x[qobj.qvars2[k].col]
g[qobj.qvars2[k].col] += coef*x[qobj.qvars1[k].col]
end
end
d.eval_grad_f_timer += toq()
return
end
function MathProgBase.eval_g(d::NLPEvaluator, g, x)
tic()
if d.last_x != x
forward_eval_all(d,x)
reverse_eval_all(d,x)
end
A = d.A
for i in 1:size(A,1); g[i] = 0.0; end
#fill!(view(g,1:size(A,1)), 0.0)
A_mul_B!(view(g,1:size(A,1)),A,x)
idx = size(A,1)+1
quadconstr = d.m.quadconstr::Vector{QuadConstraint}
for c::QuadConstraint in quadconstr
aff = c.terms.aff
v = aff.constant
for k in 1:length(aff.vars)
v += aff.coeffs[k]*x[aff.vars[k].col]
end
for k in 1:length(c.terms.qvars1)
v += c.terms.qcoeffs[k]*x[c.terms.qvars1[k].col]*x[c.terms.qvars2[k].col]
end
g[idx] = v
idx += 1
end
for ex in d.constraints
g[idx] = ex.forward_storage[1]
idx += 1
end
d.eval_g_timer += toq()
#print("x = ");show(x);println()
#println(size(A,1), " g(x) = ");show(g);println()
return
end
function MathProgBase.eval_jac_g(d::NLPEvaluator, J, x)
tic()
if d.last_x != x
forward_eval_all(d,x)
reverse_eval_all(d,x)
end
fill!(J,0.0)
idx = 1
A = d.A
for col = 1:size(A,2)
for pos = nzrange(A,col)
J[idx] = A.nzval[pos]
idx += 1
end
end
quadconstr = d.m.quadconstr::Vector{QuadConstraint}
for c::QuadConstraint in quadconstr
aff = c.terms.aff
for k in 1:length(aff.vars)
J[idx] = aff.coeffs[k]
idx += 1
end
for k in 1:length(c.terms.qvars1)
coef = c.terms.qcoeffs[k]
qidx1 = c.terms.qvars1[k].col
qidx2 = c.terms.qvars2[k].col
J[idx] = coef*x[qidx2]
J[idx+1] = coef*x[qidx1]
idx += 2
end
end
grad_storage = d.jac_storage
subexpr_reverse_values = d.subexpression_reverse_values
SIMPLIFY = d.m.simplify_nonlinear_expressions
for ex in d.constraints
nzidx = ex.grad_sparsity
grad_storage[nzidx] = 0.0
subexpr_reverse_values[ex.dependent_subexpressions] = 0.0
reverse_extract(grad_storage,ex.reverse_storage,ex.nd,ex.adj,subexpr_reverse_values,1.0)
for i in length(ex.dependent_subexpressions):-1:1
k = ex.dependent_subexpressions[i]
if SIMPLIFY && d.subexpression_linearity[k] == CONSTANT
continue
end
subexpr = d.subexpressions[k]
reverse_extract(grad_storage,subexpr.reverse_storage,subexpr.nd,subexpr.adj,subexpr_reverse_values,subexpr_reverse_values[k])
end
for k in 1:length(nzidx)
J[idx+k-1] = grad_storage[nzidx[k]]
end
idx += length(nzidx)
end
d.eval_jac_g_timer += toq()
#print("x = ");show(x);println()
#print("V ");show(J);println()
return
end
function MathProgBase.eval_hesslag_prod(
d::NLPEvaluator,
h::Vector{Float64}, # output vector
x::Vector{Float64}, # current solution
v::Vector{Float64}, # rhs vector
σ::Float64, # multiplier for objective
μ::Vector{Float64}) # multipliers for each constraint
nldata = d.m.nlpdata::NLPData
if d.last_x != x
forward_eval_all(d,x)
reverse_eval_all(d,x)
end
fill!(h, 0.0)
# quadratic objective
qobj::QuadExpr = d.m.obj
for k in 1:length(qobj.qvars1)
col1 = qobj.qvars1[k].col
col2 = qobj.qvars2[k].col
coef = qobj.qcoeffs[k]
if col1 == col2
h[col1] += σ*2*coef*v[col1]
else
h[col1] += σ*coef*v[col2]
h[col2] += σ*coef*v[col1]
end
end
# quadratic constraints
row = size(d.A,1)+1
quadconstr = d.m.quadconstr::Vector{QuadConstraint}
for c in quadconstr
l = μ[row]
for k in 1:length(c.terms.qvars1)
col1 = c.terms.qvars1[k].col
col2 = c.terms.qvars2[k].col
coef = c.terms.qcoeffs[k]
if col1 == col2
h[col1] += l*2*coef*v[col1]
else
h[col1] += l*coef*v[col2]
h[col2] += l*coef*v[col1]
end
end
row += 1
end
input_ϵ = reinterpret(ForwardDiff.Partials{1,Float64}, d.input_ϵ)
output_ϵ = reinterpret(ForwardDiff.Partials{1,Float64}, d.output_ϵ)
for i in 1:length(x)
input_ϵ[i] = ForwardDiff.Partials((v[i],))
end
# forward evaluate all subexpressions once
subexpr_forward_values_ϵ = reinterpret(ForwardDiff.Partials{1,Float64},d.subexpression_forward_values_ϵ)
subexpr_reverse_values_ϵ = reinterpret(ForwardDiff.Partials{1,Float64},d.subexpression_reverse_values_ϵ)
forward_storage_ϵ = reinterpret(ForwardDiff.Partials{1,Float64},d.forward_storage_ϵ)
reverse_storage_ϵ = reinterpret(ForwardDiff.Partials{1,Float64},d.reverse_storage_ϵ)
partials_storage_ϵ = reinterpret(ForwardDiff.Partials{1,Float64},d.partials_storage_ϵ)
SIMPLIFY = d.m.simplify_nonlinear_expressions
for expridx in d.subexpression_order
if SIMPLIFY && d.subexpression_linearity[expridx] == CONSTANT
continue
end
subexpr = d.subexpressions[expridx]
sub_forward_storage_ϵ = reinterpret(ForwardDiff.Partials{1,Float64},subexpr.forward_storage_ϵ)
sub_partials_storage_ϵ = reinterpret(ForwardDiff.Partials{1,Float64},subexpr.partials_storage_ϵ)
subexpr_forward_values_ϵ[expridx] = forward_eval_ϵ(subexpr.forward_storage,sub_forward_storage_ϵ, subexpr.partials_storage, sub_partials_storage_ϵ, subexpr.nd, subexpr.adj, input_ϵ, subexpr_forward_values_ϵ,user_operators=nldata.user_operators)
end
# we only need to do one reverse pass through the subexpressions as well
zero_ϵ = zero(ForwardDiff.Partials{1,Float64})
fill!(subexpr_reverse_values_ϵ,zero_ϵ)
fill!(d.subexpression_reverse_values,0.0)
fill!(reverse_storage_ϵ,zero_ϵ)
fill!(output_ϵ,zero_ϵ)
if d.has_nlobj
ex = d.objective
forward_eval_ϵ(ex.forward_storage, forward_storage_ϵ, ex.partials_storage, partials_storage_ϵ, ex.nd,ex.adj,input_ϵ, subexpr_forward_values_ϵ,user_operators=nldata.user_operators)
reverse_eval_ϵ(output_ϵ,ex.reverse_storage, reverse_storage_ϵ,ex.partials_storage, partials_storage_ϵ,ex.nd,ex.adj,d.subexpression_reverse_values,subexpr_reverse_values_ϵ, σ, zero_ϵ)
end
for i in 1:length(d.constraints)
ex = d.constraints[i]
l = μ[row]
forward_eval_ϵ(ex.forward_storage, forward_storage_ϵ, ex.partials_storage, partials_storage_ϵ, ex.nd,ex.adj, input_ϵ,subexpr_forward_values_ϵ,user_operators=nldata.user_operators)
reverse_eval_ϵ(output_ϵ, ex.reverse_storage, reverse_storage_ϵ, ex.partials_storage, partials_storage_ϵ, ex.nd,ex.adj,d.subexpression_reverse_values,subexpr_reverse_values_ϵ, l, zero_ϵ)
row += 1
end
for i in length(ex.dependent_subexpressions):-1:1
expridx = ex.dependent_subexpressions[i]
if SIMPLIFY && d.subexpression_linearity[expridx] == CONSTANT
continue
end
subexpr = d.subexpressions[expridx]
sub_reverse_storage_ϵ = reinterpret(ForwardDiff.Partials{1,Float64},subexpr.reverse_storage_ϵ)
sub_partials_storage_ϵ = reinterpret(ForwardDiff.Partials{1,Float64},subexpr.partials_storage_ϵ)
reverse_eval_ϵ(output_ϵ,subexpr.reverse_storage,sub_reverse_storage_ϵ, subexpr.partials_storage, sub_partials_storage_ϵ,subexpr.nd,subexpr.adj,d.subexpression_reverse_values,subexpr_reverse_values_ϵ,d.subexpression_reverse_values[expridx],subexpr_reverse_values_ϵ[expridx])
end
for i in 1:length(x)
h[i] += output_ϵ[i].values[1]
end
end
function MathProgBase.eval_hesslag(
d::NLPEvaluator,
H::Vector{Float64}, # Sparse hessian entry vector
x::Vector{Float64}, # Current solution
obj_factor::Float64, # Lagrangian multiplier for objective
lambda::Vector{Float64}) # Multipliers for each constraint
qobj = d.m.obj::QuadExpr
nldata = d.m.nlpdata::NLPData
d.want_hess || error("Hessian computations were not requested on the call to MathProgBase.initialize.")
if d.last_x != x
forward_eval_all(d,x)
reverse_eval_all(d,x)
end
tic()
# quadratic objective
nzcount = 1
for k in 1:length(qobj.qvars1)
if qobj.qvars1[k].col == qobj.qvars2[k].col
H[nzcount] = obj_factor*2*qobj.qcoeffs[k]
else
H[nzcount] = obj_factor*qobj.qcoeffs[k]
end
nzcount += 1
end
# quadratic constraints
quadconstr = d.m.quadconstr::Vector{QuadConstraint}
for i in 1:length(quadconstr)
c = quadconstr[i]
l = lambda[length(d.m.linconstr)+i]
for k in 1:length(c.terms.qvars1)
if c.terms.qvars1[k].col == c.terms.qvars2[k].col
H[nzcount] = l*2*c.terms.qcoeffs[k]
else
H[nzcount] = l*c.terms.qcoeffs[k]
end
nzcount += 1
end
end
fill!(d.input_ϵ,0.0)
recovery_tmp_storage = d.output_ϵ
nzcount -= 1
if d.has_nlobj
ex = d.objective
chunk = min(size(ex.seed_matrix,2),d.max_chunk)
if chunk == 1
# skip dynamic dispatch
nzthis = hessian_slice(d, ex, x, H, obj_factor, nzcount, recovery_tmp_storage, Val{1})::Int
else
nzthis = hessian_slice(d, ex, x, H, obj_factor, nzcount, recovery_tmp_storage, Val{chunk})::Int
end
nzcount += nzthis
end
for i in 1:length(d.constraints)
ex = d.constraints[i]
chunk = min(size(ex.seed_matrix,2),d.max_chunk)
if chunk == 1
nzthis = hessian_slice(d, ex, x, H, lambda[i+length(quadconstr)+length(d.m.linconstr)], nzcount, recovery_tmp_storage, Val{1})::Int
else
nzthis = hessian_slice(d, ex, x, H, lambda[i+length(quadconstr)+length(d.m.linconstr)], nzcount, recovery_tmp_storage, Val{chunk})::Int
end
nzcount += nzthis
end
d.eval_hesslag_timer += toq()
return
end
function hessian_slice_inner(d, ex, R, input_ϵ, output_ϵ, ::Type{Val{CHUNK}}) where CHUNK
subexpr_forward_values_ϵ = reinterpret_unsafe(ForwardDiff.Partials{CHUNK,Float64},d.subexpression_forward_values_ϵ)
subexpr_reverse_values_ϵ = reinterpret_unsafe(ForwardDiff.Partials{CHUNK,Float64},d.subexpression_reverse_values_ϵ)
forward_storage_ϵ = reinterpret_unsafe(ForwardDiff.Partials{CHUNK,Float64},d.forward_storage_ϵ)
reverse_storage_ϵ = reinterpret_unsafe(ForwardDiff.Partials{CHUNK,Float64},d.reverse_storage_ϵ)
partials_storage_ϵ = reinterpret_unsafe(ForwardDiff.Partials{CHUNK,Float64},d.partials_storage_ϵ)
zero_ϵ = zero(ForwardDiff.Partials{CHUNK,Float64})
user_operators = d.m.nlpdata.user_operators::ReverseDiffSparse.UserOperatorRegistry
SIMPLIFY = d.m.simplify_nonlinear_expressions
# do a forward pass
for expridx in ex.dependent_subexpressions
if SIMPLIFY && d.subexpression_linearity[expridx] == CONSTANT
continue
end
subexpr = d.subexpressions[expridx]
sub_forward_storage_ϵ = reinterpret_unsafe(ForwardDiff.Partials{CHUNK,Float64},subexpr.forward_storage_ϵ)
sub_partials_storage_ϵ = reinterpret_unsafe(ForwardDiff.Partials{CHUNK,Float64},subexpr.partials_storage_ϵ)
subexpr_forward_values_ϵ[expridx] = forward_eval_ϵ(subexpr.forward_storage,sub_forward_storage_ϵ,subexpr.partials_storage,sub_partials_storage_ϵ, subexpr.nd, subexpr.adj, input_ϵ, subexpr_forward_values_ϵ,user_operators=user_operators)
end
forward_eval_ϵ(ex.forward_storage,forward_storage_ϵ,ex.partials_storage, partials_storage_ϵ,ex.nd,ex.adj,input_ϵ, subexpr_forward_values_ϵ,user_operators=user_operators)
# do a reverse pass
subexpr_reverse_values_ϵ[ex.dependent_subexpressions] = zero_ϵ
d.subexpression_reverse_values[ex.dependent_subexpressions] = 0.0
reverse_eval_ϵ(output_ϵ, ex.reverse_storage, reverse_storage_ϵ,ex.partials_storage, partials_storage_ϵ,ex.nd,ex.adj,d.subexpression_reverse_values,subexpr_reverse_values_ϵ, 1.0, zero_ϵ)
for i in length(ex.dependent_subexpressions):-1:1
expridx = ex.dependent_subexpressions[i]
if SIMPLIFY && d.subexpression_linearity[expridx] == CONSTANT
continue
end
subexpr = d.subexpressions[expridx]
sub_reverse_storage_ϵ = reinterpret_unsafe(ForwardDiff.Partials{CHUNK,Float64},subexpr.reverse_storage_ϵ)
sub_partials_storage_ϵ = reinterpret_unsafe(ForwardDiff.Partials{CHUNK,Float64},subexpr.partials_storage_ϵ)
reverse_eval_ϵ(output_ϵ, subexpr.reverse_storage, sub_reverse_storage_ϵ,subexpr.partials_storage,sub_partials_storage_ϵ,subexpr.nd,subexpr.adj,d.subexpression_reverse_values,subexpr_reverse_values_ϵ,d.subexpression_reverse_values[expridx],subexpr_reverse_values_ϵ[expridx])
end
end
function hessian_slice(d, ex, x, H, scale, nzcount, recovery_tmp_storage,::Type{Val{CHUNK}}) where CHUNK
nzthis = length(ex.hess_I)
if ex.linearity == LINEAR
@assert nzthis == 0
return 0
end
R = ex.seed_matrix
Coloring.prepare_seed_matrix!(R,ex.rinfo)
local_to_global_idx = ex.rinfo.local_indices
zero_ϵ = zero(ForwardDiff.Partials{CHUNK,Float64})
input_ϵ_raw = d.input_ϵ
output_ϵ_raw = d.output_ϵ
input_ϵ = reinterpret_unsafe(ForwardDiff.Partials{CHUNK,Float64}, input_ϵ_raw)
output_ϵ = reinterpret_unsafe(ForwardDiff.Partials{CHUNK,Float64}, output_ϵ_raw)
# compute hessian-vector products
num_products = size(R,2) # number of hessian-vector products
num_chunks = div(num_products, CHUNK)
@assert size(R,1) == length(local_to_global_idx)
numVar = length(x)
for k in 1:CHUNK:CHUNK*num_chunks
for r in 1:length(local_to_global_idx)
# set up directional derivatives
@inbounds idx = local_to_global_idx[r]
# load up R[r,k,k+1,...,k+CHUNK-1] into input_ϵ
for s in 1:CHUNK
input_ϵ_raw[(idx-1)*CHUNK + s] = R[r,k+s-1]
end
@inbounds output_ϵ[idx] = zero_ϵ
end
hessian_slice_inner(d, ex, R, input_ϵ, output_ϵ, Val{CHUNK})
# collect directional derivatives
for r in 1:length(local_to_global_idx)
idx = local_to_global_idx[r]
# load output_ϵ into R[r,k,k+1,...,k+CHUNK-1]
for s in 1:CHUNK
R[r,k+s-1] = output_ϵ_raw[(idx-1)*CHUNK + s]
end
@inbounds input_ϵ[idx] = zero_ϵ
end
end
# leftover chunk
remaining = num_products - CHUNK*num_chunks
if remaining > 0
k = CHUNK*num_chunks+1
for r in 1:length(local_to_global_idx)
# set up directional derivatives
@inbounds idx = local_to_global_idx[r]
# load up R[r,k,k+1,...,k+remaining-1] into input_ϵ
for s in 1:remaining
# leave junk in the unused components
input_ϵ_raw[(idx-1)*CHUNK + s] = R[r,k+s-1]
end
@inbounds output_ϵ[idx] = zero_ϵ
end
hessian_slice_inner(d, ex, R, input_ϵ, output_ϵ, Val{CHUNK})
# collect directional derivatives
for r in 1:length(local_to_global_idx)
idx = local_to_global_idx[r]
# load output_ϵ into R[r,k,k+1,...,k+remaining-1]
for s in 1:remaining
R[r,k+s-1] = output_ϵ_raw[(idx-1)*CHUNK + s]
end
@inbounds input_ϵ[idx] = zero_ϵ
end
end
# Output is in R, now recover
#output_slice = view(H, (nzcount+1):(nzcount+nzthis))
output_slice = VectorView(nzcount, nzthis, pointer(H))
Coloring.recover_from_matmat!(output_slice, R, ex.rinfo, recovery_tmp_storage)
scale!(output_slice, scale)
return nzthis
end
MathProgBase.isobjlinear(d::NLPEvaluator) = !d.has_nlobj && (length(d.m.obj.qvars1) == 0)
# interpret quadratic to include purely linear
MathProgBase.isobjquadratic(d::NLPEvaluator) = !d.has_nlobj
MathProgBase.isconstrlinear(d::NLPEvaluator, i::Integer) = (i <= length(d.m.linconstr))
function MathProgBase.jac_structure(d::NLPEvaluator)
# Jacobian structure
jac_I = Int[]
jac_J = Int[]
A = d.A
for col = 1:size(A,2)
for pos = nzrange(A,col)
push!(jac_I, A.rowval[pos])
push!(jac_J, col)
end
end
rowoffset = size(A,1)+1
for c::QuadConstraint in d.m.quadconstr
aff = c.terms.aff
for k in 1:length(aff.vars)
push!(jac_I, rowoffset)
push!(jac_J, aff.vars[k].col)
end
for k in 1:length(c.terms.qvars1)
push!(jac_I, rowoffset)
push!(jac_I, rowoffset)
push!(jac_J, c.terms.qvars1[k].col)
push!(jac_J, c.terms.qvars2[k].col)
end
rowoffset += 1
end
for ex in d.constraints
idx = ex.grad_sparsity
for i in 1:length(idx)
push!(jac_I, rowoffset)
push!(jac_J, idx[i])
end
rowoffset += 1
end
return jac_I, jac_J
end
function MathProgBase.hesslag_structure(d::NLPEvaluator)
d.want_hess || error("Hessian computations were not requested on the call to MathProgBase.initialize.")
return d.hess_I,d.hess_J
end
function _hesslag_structure(d::NLPEvaluator)
hess_I = Int[]
hess_J = Int[]
qobj::QuadExpr = d.m.obj
for k in 1:length(qobj.qvars1)
qidx1 = qobj.qvars1[k].col
qidx2 = qobj.qvars2[k].col
if qidx2 > qidx1