-
Notifications
You must be signed in to change notification settings - Fork 296
/
NBSlice.mo
1136 lines (1048 loc) · 45.9 KB
/
NBSlice.mo
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
/*
* This file is part of OpenModelica.
*
* Copyright (c) 1998-2020, Open Source Modelica Consortium (OSMC),
* c/o Linköpings universitet, Department of Computer and Information Science,
* SE-58183 Linköping, Sweden.
*
* All rights reserved.
*
* THIS PROGRAM IS PROVIDED UNDER THE TERMS OF GPL VERSION 3 LICENSE OR
* THIS OSMC PUBLIC LICENSE (OSMC-PL) VERSION 1.2.
* ANY USE, REPRODUCTION OR DISTRIBUTION OF THIS PROGRAM CONSTITUTES
* RECIPIENT'S ACCEPTANCE OF THE OSMC PUBLIC LICENSE OR THE GPL VERSION 3,
* ACCORDING TO RECIPIENTS CHOICE.
*
* The OpenModelica software and the Open Source Modelica
* Consortium (OSMC) Public License (OSMC-PL) are obtained
* from OSMC, either from the above address,
* from the URLs: http://www.ida.liu.se/projects/OpenModelica or
* http://www.openmodelica.org, and in the OpenModelica distribution.
* GNU version 3 is obtained from: http://www.gnu.org/copyleft/gpl.html.
*
* This program is distributed WITHOUT ANY WARRANTY; without
* even the implied warranty of MERCHANTABILITY or FITNESS
* FOR A PARTICULAR PURPOSE, EXCEPT AS EXPRESSLY SET FORTH
* IN THE BY RECIPIENT SELECTED SUBSIDIARY LICENSE CONDITIONS OF OSMC-PL.
*
* See the full OSMC Public License conditions for more details.
*
*/
encapsulated uniontype NBSlice<T>
" file: NBSlice.mo
package: NBSlice
description: This file contains util functions for slicing operations.
"
protected
import Slice = NBSlice;
// NF imports
import ComponentRef = NFComponentRef;
import Dimension = NFDimension;
import Expression = NFExpression;
import Operator = NFOperator;
import SimplifyExp = NFSimplifyExp;
import Subscript = NFSubscript;
import Type = NFType;
import Variable = NFVariable;
// NB imports
import NBAdjacency.Mapping;
import BackendUtil = NBBackendUtil;
import NBEquation.{Equation, Iterator, Frame, FrameLocation, RecollectStatus, FrameOrderingStatus};
import Replacements = NBReplacements;
import BVariable = NBVariable;
import NBVariable.VariablePointers;
// Util imports
import List;
import UnorderedMap;
public
type IntLst = list<Integer>;
record SLICE
T t;
IntLst indices;
end SLICE;
// ############################################################
// Member Functions
// ############################################################
function getT
input Slice<T> slice;
output T t = slice.t;
end getT;
function isEqual
input Slice<T> slice1;
input Slice<T> slice2;
input isEqualT func;
output Boolean b = func(slice1.t, slice2.t) and List.isEqualOnTrue(slice1.indices, slice2.indices, intEq);
end isEqual;
function toString
input Slice<T> slice;
input toStringT func;
input Integer maxLength = 10;
output String str;
protected
String sliceStr;
algorithm
str := func(slice.t);
if maxLength > 0 then
str := str + "\n\t slice: " + List.toString(inList = slice.indices, inPrintFunc = intString, maxLength = 10);
end if;
end toString;
function lstToString
input list<Slice<T>> lst;
input toStringT_ func;
input Integer maxLength = 10;
partial function toStringT_ = toStringT "ugly hack to make type T known to subfunction";
output String str = List.toString(lst, function toString(func = func, maxLength = maxLength), "", "\t", ";\n\t", ";", false);
end lstToString;
function isFull
input Slice<T> slice;
output Boolean b = listEmpty(slice.indices);
end isFull;
function size
input Slice<T> slice;
input sizeT func;
output Integer s;
algorithm
if listEmpty(slice.indices) then
s := func(slice.t);
else
s := listLength(slice.indices);
end if;
end size;
function simplify
"only to be used for unordered purposes!
lists of all indices are meaningful if they are not in the natural ascending order
and can indicate range reversal in for loops."
input output Slice<T> slice;
input sizeT func;
algorithm
if listLength(slice.indices) == func(slice.t) then
slice.indices := {};
else
slice.indices := List.sort(slice.indices, intGt);
end if;
end simplify;
function addToSliceMap
input T t;
input Integer i;
input UnorderedMap<T, IntLst> map;
algorithm
if UnorderedMap.contains(t, map) then
UnorderedMap.add(t, i :: UnorderedMap.getSafe(t, map, sourceInfo()), map);
else
UnorderedMap.addNew(t, {i}, map);
end if;
end addToSliceMap;
function fromTpl
input tuple<T, IntLst> tpl;
output Slice<T> slice;
protected
T t;
IntLst lst;
algorithm
(t, lst) := tpl;
slice := SLICE(t, lst);
end fromTpl;
function fromMap
input UnorderedMap<T, IntLst> map;
output list<Slice<T>> slices = list(fromTpl(tpl) for tpl in UnorderedMap.toList(map));
end fromMap;
function apply
input output Slice<T> slice;
input applyT func;
algorithm
slice.t := func(slice.t);
end apply;
function applyMutable
input Slice<T> slice;
input applyMutableT func;
algorithm
func(slice.t);
end applyMutable;
// ############################################################
// Partial Functions
// ############################################################
partial function toStringT
input T t;
output String str;
end toStringT;
partial function sizeT
input T t;
output Integer s;
end sizeT;
partial function isEqualT
input T t1;
input T t2;
output Boolean b;
end isEqualT;
partial function applyT
input output T t;
end applyT;
partial function applyMutableT
input T t;
end applyMutableT;
partial function filterCref
"partial function that needs to be provided.
decides if the the cref is added to the list pointer."
input output ComponentRef cref;
input UnorderedSet<ComponentRef> acc;
end filterCref;
partial function getDependentCrefIndices
input list<ComponentRef> dependencies "dependent var crefs";
input UnorderedMap<ComponentRef, Integer> map "unordered map to check for relevance";
input Mapping mapping "array <-> scalar index mapping";
input Integer eqn_arr_idx;
output array<list<Integer>> indices;
output array<array<Integer>> mode_to_var;
end getDependentCrefIndices;
// ############################################################
// cref accumulation Functions
// use with:
// Equation.collectCrefs()
// filterExp()
// ############################################################
function filterExp
"wrapper function that applies filter cref to
a cref expression."
input output Expression exp;
input filterCref filter;
input UnorderedSet<ComponentRef> acc;
algorithm
() := match exp
case Expression.CREF() algorithm filter(exp.cref, acc); then ();
else ();
end match;
end filterExp;
function getSliceCandidates
"Used to collect all slices of a certain variable name.
Note: the name has to be stripped of all subscripts for this to work."
extends filterCref;
input ComponentRef name "the name of the variable";
algorithm
if ComponentRef.isEqual(name, ComponentRef.stripSubscriptsAll(cref)) then
UnorderedSet.add(cref, acc);
end if;
end getSliceCandidates;
function getDependentCref
"checks if crefs are relevant in the given context and collects them."
extends filterCref;
input UnorderedMap<ComponentRef, Integer> map "unordered map to check for relevance";
input Boolean pseudo;
protected
ComponentRef checkCref, childCref;
list<Pointer<Variable>> record_children;
algorithm
// if causalized in pseudo array mode, the variables will only have subscript-free variables
checkCref := if pseudo then ComponentRef.stripSubscriptsAll(cref) else cref;
record_children := BVariable.getRecordChildren(BVariable.getVarPointer(checkCref));
if listEmpty(record_children) then
// not a record
if UnorderedMap.contains(checkCref, map) then
UnorderedSet.add(cref, acc);
end if;
else
// its a record, instead parse all the children
for child in record_children loop
childCref := BVariable.getVarName(child);
if UnorderedMap.contains(childCref, map) then
UnorderedSet.add(childCref, acc);
end if;
end for;
end if;
end getDependentCref;
function getDependentCrefCausalized
"checks if crefs are relevant in the given context and collects them.
previously found crefs are replaced by their dependencies! only works on causalized systems."
extends filterCref;
input UnorderedSet<ComponentRef> set "unordered set to check for array crefs for relevance";
protected
ComponentRef checkCref, childCref;
list<Pointer<Variable>> record_children;
algorithm
// always remove subscripts here, this analysis is for sparsity pattern -> currently always scalarized!
checkCref := ComponentRef.stripSubscriptsAll(cref);
record_children := BVariable.getRecordChildren(BVariable.getVarPointer(checkCref));
if listEmpty(record_children) then
// not a record
if UnorderedSet.contains(checkCref, set) then
UnorderedSet.add(cref, acc);
end if;
else
// its a record, instead parse all the children
for child in record_children loop
childCref := BVariable.getVarName(child);
if UnorderedSet.contains(childCref, set) then
UnorderedSet.add(childCref, acc);
end if;
end for;
end if;
end getDependentCrefCausalized;
function getUnsolvableExpCrefs
"finds all unsolvable crefs in an expression."
input output Expression exp "the exp to check for unsolvable crefs";
input UnorderedSet<ComponentRef> acc "accumulator for relevant crefs";
input UnorderedMap<ComponentRef, Integer> map "unordered map to check for relevance";
input Boolean pseudo;
algorithm
// put all unsolvable logic here!
exp := match exp
case Expression.RANGE() then Expression.map(exp, function filterExp(filter = function getDependentCref(map = map, pseudo = pseudo), acc = acc));
case Expression.LBINARY() then Expression.map(exp, function filterExp(filter = function getDependentCref(map = map, pseudo = pseudo), acc = acc));
case Expression.RELATION() then Expression.map(exp, function filterExp(filter = function getDependentCref(map = map, pseudo = pseudo), acc = acc));
else exp;
end match;
end getUnsolvableExpCrefs;
function getDependentCrefIndicesPseudoScalar
"Scalar equations.
Turns cref dependencies into index lists, used for adjacency."
input list<ComponentRef> dependencies "dependent var crefs";
input UnorderedMap<ComponentRef, Integer> map "unordered map to check for relevance";
input Mapping mapping "array <-> scalar index mapping";
output list<Integer> indices = {};
protected
list<ComponentRef> scalarized_dependencies = List.flatten(list(ComponentRef.scalarizeAll(dep) for dep in dependencies));
ComponentRef stripped;
Integer var_arr_idx, var_start, var_scal_idx;
list<Integer> sizes, int_subs;
algorithm
for cref in scalarized_dependencies loop
stripped := ComponentRef.stripSubscriptsAll(cref);
var_arr_idx := UnorderedMap.getSafe(stripped, map, sourceInfo());
(var_start, _) := mapping.var_AtS[var_arr_idx];
sizes := ComponentRef.sizes(stripped);
int_subs := ComponentRef.subscriptsToInteger(cref);
var_scal_idx := locationToIndex(List.zip(sizes, int_subs), var_start);
indices := var_scal_idx :: indices;
end for;
// remove duplicates and sort
if not listEmpty(indices) then
indices := List.sort(List.uniqueIntN(indices, max(i for i in indices)), intLt);
end if;
end getDependentCrefIndicesPseudoScalar;
function getDependentCrefIndicesPseudoFull
"equations that will get full dependency.
Turns cref dependencies into index lists, used for adjacency."
extends getDependentCrefIndices;
protected
list<ComponentRef> scalarized_dependencies = List.flatten(list(ComponentRef.scalarizeAll(dep) for dep in dependencies));
ComponentRef stripped;
Integer eqn_start, eqn_size, var_arr_idx, var_scal_idx, mode = 1;
list<Integer> scal_lst;
Integer idx;
array<Integer> mode_to_var_row;
list<Subscript> subs;
list<Dimension> dims;
Type ty;
algorithm
(eqn_start, eqn_size) := mapping.eqn_AtS[eqn_arr_idx];
indices := arrayCreate(eqn_size, {});
mode_to_var := arrayCreate(eqn_size, arrayCreate(0,0));
// create unique array for each equation
for i in 1:eqn_size loop
mode_to_var[i] := arrayCreate(listLength(scalarized_dependencies),-1);
end for;
for cref in scalarized_dependencies loop
stripped := ComponentRef.stripSubscriptsAll(cref);
var_arr_idx := UnorderedMap.getSafe(stripped, map, sourceInfo());
// build range in reverse, it will be flipped anyway
subs := ComponentRef.subscriptsAllWithWholeFlat(cref);
ty := ComponentRef.getSubscriptedType(stripped, true);
dims := Type.arrayDims(ty);
scal_lst := Mapping.getVarScalIndices(var_arr_idx, mapping, subs, dims, true);
if intMod(eqn_size, listLength(scal_lst)) <> 0 then
Error.addMessage(Error.INTERNAL_ERROR,{getInstanceName()
+ " failed because flattened indices " + intString(listLength(scal_lst))
+ " could not be repeated to fit equation size " + intString(eqn_size) + ". lst: " + List.toString(scal_lst, intString)});
fail();
else
// fill the equation with repeated scalar lists
scal_lst := List.repeat(scal_lst, realInt(eqn_size/listLength(scal_lst)));
end if;
idx := 1;
for var_scal_idx in listReverse(scal_lst) loop
mode_to_var_row := mode_to_var[idx];
arrayUpdate(mode_to_var_row, mode, var_scal_idx);
arrayUpdate(mode_to_var, idx, mode_to_var_row);
indices[idx] := var_scal_idx :: indices[idx];
idx := idx + 1;
end for;
mode := mode + 1;
end for;
// sort
for i in 1:arrayLength(indices) loop
indices[i] := List.sort(UnorderedSet.unique_list(indices[i], Util.id, intEq), intLt);
end for;
end getDependentCrefIndicesPseudoFull;
function getDependentCrefIndicesPseudoFor
"For-Loop equations.
Turns cref dependencies into index lists, used for adjacency."
extends getDependentCrefIndices;
input Iterator iter "iterator frames";
protected
list<ComponentRef> names;
list<Expression> ranges;
list<tuple<ComponentRef, Expression>> frames;
Integer eqn_size, iter_size, body_size, mode = 1;
updateDependencies func;
algorithm
// get iterator size and frames
iter_size := Iterator.size(iter);
(names, ranges) := Iterator.getFrames(iter);
frames := List.zip(names, ranges);
// get eqn size and create the adjacency matrix and causalization mode arrays
(_, eqn_size) := mapping.eqn_AtS[eqn_arr_idx];
indices := arrayCreate(eqn_size, {});
mode_to_var := arrayCreate(eqn_size, arrayCreate(0,0));
// sanity check for eqn size and get size of body equation
if mod(eqn_size, iter_size) == 0 then
body_size := realInt(eqn_size/iter_size);
else
Error.addMessage(Error.INTERNAL_ERROR,{getInstanceName()
+ " failed because the equation size " + intString(eqn_size)
+ " could not be divided by the iterator size " + intString(iter_size) + " without rest."});
end if;
// create unique array for each equation
for i in 1:eqn_size loop
mode_to_var[i] := arrayCreate(listLength(dependencies),-1);
end for;
// create rows
for dep in dependencies loop
func := function updateDependenciesInteger(mode = mode, mode_to_var = mode_to_var, indices = indices);
fillDependencyArray(dep, body_size, frames, mapping, map, func);
// increase mode index
mode := mode + 1;
end for;
// sort (kabdelhak: is this needed? try to FixMe)
for i in 1:arrayLength(indices) loop
indices[i] := List.sort(UnorderedSet.unique_list(indices[i], Util.id, intEq), intLt);
end for;
end getDependentCrefIndicesPseudoFor;
function getDependentCrefsPseudoForCausalized
"(Jacobian) For-Loop equations.
Turns cref dependencies into index lists, used for adjacency."
input ComponentRef row_cref "cref representing the current row";
input list<ComponentRef> dependencies "dependent var crefs";
input VariablePointers var_rep "scalarized variable representatives";
input VariablePointers eqn_rep "scalarized equation representatives";
input Mapping var_rep_mapping "index mapping for variable representatives";
input Mapping eqn_rep_mapping "index mapping for equation representatives";
input Iterator iter "iterator frames";
input Integer eqn_size "full equation size (not considering the slice)";
input list<Integer> slice = {} "optional slice, empty list implies full slice";
input Boolean implicit = false "do not compute row cref indices if implicit";
output list<tuple<ComponentRef, list<ComponentRef>>> tpl_lst "cref -> dependencies for each scalar cref";
protected
list<ComponentRef> names;
list<Expression> ranges;
list<tuple<ComponentRef, Expression>> frames;
Integer num_rows, iter_size, body_size;
list<ComponentRef> row_crefs;
list<Integer> row_scal_lst;
list<list<Integer>> accum_row_lst = {};
array<list<ComponentRef>> accum_dep_arr;
list<list<ComponentRef>> accum_dep_lst;
updateDependencies func_var, func_eqn;
algorithm
// create the array of maximum equation size and slice afterwards
accum_dep_arr := arrayCreate(eqn_size, {});
// get iterator size and frames
iter_size := Iterator.size(iter);
(names, ranges) := Iterator.getFrames(iter);
frames := List.zip(names, ranges);
// sanity check for eqn size and get size of body equation
if mod(eqn_size, iter_size) == 0 then
body_size := realInt(eqn_size/iter_size);
else
Error.addMessage(Error.INTERNAL_ERROR,{getInstanceName()
+ " failed because the equation size " + intString(eqn_size)
+ " could not be divided by the iterator size " + intString(iter_size) + " without rest."});
end if;
// get row cref lst
if implicit then
row_crefs := ComponentRef.scalarizeAll(row_cref);
else
for cref in ComponentRef.scalarizeAll(row_cref) loop
row_scal_lst := getCrefInFrameIndices(cref, frames, eqn_rep_mapping, eqn_rep.map);
accum_row_lst := row_scal_lst :: accum_row_lst;
end for;
row_scal_lst := List.flatten(accum_row_lst);
row_crefs := list(VariablePointers.varSlice(eqn_rep, i, eqn_rep_mapping) for i in row_scal_lst);
end if;
row_crefs := if listEmpty(slice) then row_crefs else List.getAtIndexLst(row_crefs, slice, true);
num_rows := listLength(row_crefs);
// prepare the functions to update dependencies
func_var := function updateDependenciesCref(accum_dep_arr = accum_dep_arr, vars = var_rep, mapping = var_rep_mapping);
func_eqn := function updateDependenciesCref(accum_dep_arr = accum_dep_arr, vars = eqn_rep, mapping = eqn_rep_mapping);
for dep in dependencies loop
if UnorderedMap.contains(dep, var_rep.map) then
fillDependencyArray(dep, body_size, frames, var_rep_mapping, var_rep.map, func_var);
elseif UnorderedMap.contains(dep, eqn_rep.map) then
fillDependencyArray(dep, body_size, frames, eqn_rep_mapping, eqn_rep.map, func_eqn);
end if;
end for;
accum_dep_lst := listReverse(arrayList(accum_dep_arr));
accum_dep_lst := if listEmpty(slice) then accum_dep_lst else List.getAtIndexLst(accum_dep_lst, slice, true);
tpl_lst := List.zip(row_crefs, accum_dep_lst);
end getDependentCrefsPseudoForCausalized;
function fillDependencyArray
"body function of getDependentCrefsPseudoFor and getDependentCrefsPseudoForCausalized
this generates all entries to jacobian or adjacency matrices for a specific dependency.
This dependency might be an array cref, part of a reduction or contain slices."
input ComponentRef dep;
input Integer body_size;
input list<tuple<ComponentRef, Expression>> frames;
input Mapping mapping;
input UnorderedMap<ComponentRef, Integer> map;
input updateDependencies func;
protected
Integer scal_length, body_repeat, eqn_idx;
list<Integer> scal_lst;
list<tuple<ComponentRef, list<Integer>>> scal_tpl_lst = {};
algorithm
// get all dependencies for each scalarized cref
// Note: scalarization does not remove the iterators, therefore it can still yield
// multiple scalar indices when evaluated along the iterator frames
for scal_cref in ComponentRef.scalarizeAll(dep) loop
scal_lst := getCrefInFrameIndices(scal_cref, frames, mapping, map);
scal_tpl_lst := (scal_cref, scal_lst) :: scal_tpl_lst;
end for;
// check wether or not the element has to be repeated to fit the body
scal_length := listLength(scal_tpl_lst);
if mod(scal_length, body_size) == 0 then
body_repeat := realInt(scal_length/body_size);
else
Error.addMessage(Error.INTERNAL_ERROR,{getInstanceName()
+ " failed because number of flattened indices " + intString(scal_length)
+ " for dependency " + ComponentRef.toString(dep)
+ " could not be divided by the body size " + intString(body_size) + " without rest."});
fail();
end if;
eqn_idx := 1;
for tpl in scal_tpl_lst loop
(_, scal_lst) := tpl;
// reverse the scalar index list to traverse it in the correct order
scal_lst := listReverse(scal_lst);
// check if body_repeat > 1 to set the causalization mode to -1 for unsolvable
if body_repeat > 1 then
// reset the counter to 1 if the body is supposed to be repeated
// ToDo: reductions are only tested for body equations of size 1!
eqn_idx := 1;
end if;
for var_idx in scal_lst loop
// we now know that there is a dependency of equation (eqn_idx) to variable (var_idx)
// call the function that adds this specific variable to the correct structure
eqn_idx := func(eqn_idx, var_idx);
end for;
end for;
end fillDependencyArray;
partial function updateDependencies
input output Integer eqn_idx;
input Integer var_idx;
end updateDependencies;
function updateDependenciesCref
"(jacobian) adds the variable of (var_idx) as a dependency to (eqn_idx)
jacobian depencies are stored as component references"
extends updateDependencies;
input array<list<ComponentRef>> accum_dep_arr; //mutable
input VariablePointers vars;
input Mapping mapping;
algorithm
arrayUpdate(accum_dep_arr, eqn_idx, VariablePointers.varSlice(vars, var_idx, mapping) :: accum_dep_arr[eqn_idx]);
eqn_idx := eqn_idx + 1;
end updateDependenciesCref;
function updateDependenciesInteger
"(adjacency) adds the variable of (var_idx) as a dependency to (eqn_idx)
adjacency dependencies are stored as integers
also updates the causalization modes"
extends updateDependencies;
input Integer mode;
input array<array<Integer>> mode_to_var; //mutable
input array<list<Integer>> indices; //mutable
protected
array<Integer> mode_to_var_row;
algorithm
// get the clean pointer to the scalar row to avoid double indexing (meta modelica jank)
mode_to_var_row := mode_to_var[eqn_idx];
// set the dependency mode for this scalar equation to the scalar variable
arrayUpdate(mode_to_var_row, mode, var_idx);
// this is the adjacency matrix row. each dependency cref
// will add exactly one integer to each row belonging to this for-equation
arrayUpdate(indices, eqn_idx, var_idx :: indices[eqn_idx]);
eqn_idx := eqn_idx + 1;
end updateDependenciesInteger;
function getDependentCrefsPseudoArrayCausalized
"Array equations.
Turns cref dependencies into index lists, used for adjacency."
input ComponentRef row_cref "cref representing the current row";
input list<ComponentRef> dependencies "dependent var crefs";
input list<Integer> slice = {} "optional slice, empty least means all";
output list<tuple<ComponentRef, list<ComponentRef>>> tpl_lst "cref -> dependencies for each scalar cref";
protected
list<ComponentRef> row_cref_scal, dep_scal;
list<list<ComponentRef>> dependencies_scal = {};
Boolean sliced = not listEmpty(slice);
algorithm
row_cref_scal := ComponentRef.scalarizeAll(row_cref);
if sliced then
row_cref_scal := List.getAtIndexLst(row_cref_scal, slice, true);
end if;
for dep in listReverse(dependencies) loop
dep_scal := ComponentRef.scalarizeAll(dep);
if sliced then
dep_scal := List.getAtIndexLst(dep_scal, slice, true);
end if;
dependencies_scal := dep_scal :: dependencies_scal;
end for;
dependencies_scal := List.transposeList(dependencies_scal);
tpl_lst := List.zip(row_cref_scal, dependencies_scal);
end getDependentCrefsPseudoArrayCausalized;
function locationToIndex
"reverse function to indexToLocation()
maps a frame location to a scalar index starting from first index (one based!)"
input list<tuple<Integer,Integer>> size_val_tpl_lst;
input output Integer index;
protected
Integer size, val, factor = 1;
algorithm
for tpl in listReverse(size_val_tpl_lst) loop
(size, val) := tpl;
index := index + (val-1) * factor;
factor := factor * size;
end for;
end locationToIndex;
function indexToLocation
"reverse function to locationToIndex()
maps a scalar index to its frame location (zero based!)"
input Integer index;
input list<Integer> sizes;
output list<Integer> vals = {};
protected
Integer iterator = index;
Integer divisor = product(s for s in sizes);
algorithm
for size in sizes loop
divisor := intDiv(divisor, size);
vals := intDiv(iterator, divisor) :: vals;
iterator := mod(iterator, divisor);
end for;
end indexToLocation;
function transposeLocations
"transpose the location indices.
Before: Each inner list of indices represents a scalar equations
location inside all of the dimensions
After: Each inner array of indices represents the location of all
scalar equations for just one of the dimensions.
(still in order from Sorting)"
input list<list<Integer>> locations;
input Integer out_size;
output list<array<Integer>> locations_transposed;
protected
array<list<Integer>> lT_tmp = arrayCreate(out_size, {});
array<array<Integer>> lT_tmp2 = arrayCreate(out_size, arrayCreate(0,0));
Integer idx;
algorithm
for location in locations loop
idx := 1;
for i in location loop
lT_tmp[idx] := i :: lT_tmp[idx];
idx := idx + 1;
end for;
end for;
for j in 1:arrayLength(lT_tmp) loop
lT_tmp2[j] := listArray(listReverse(lT_tmp[j]));
end for;
locations_transposed := listReverse(arrayList(lT_tmp2));
end transposeLocations;
function orderTransposedFrameLocations
"order the frame locations by ascending inertia.
(the longer the chain of equal values at the start, the higher the inertia)
This is done to perform necessary reordering of nested for-loops"
input output list<FrameLocation> frame_locations_transposed;
output UnorderedMap<ComponentRef, Expression> replacements = UnorderedMap.new<Expression>(ComponentRef.hash, ComponentRef.isEqual);
output FrameOrderingStatus status;
protected
list<tuple<Integer, FrameLocation>> frame_inertia_lst;
algorithm
// get inertia for each frame
frame_inertia_lst := list((frameLocationInertia(frame), frame) for frame in frame_locations_transposed);
// sort by inertia (ascending)
frame_inertia_lst := List.sort(frame_inertia_lst, Util.compareTupleIntGt);
// resolve equal inertia (diagonal slices)
(frame_inertia_lst, status) := resolveEqualInertia(frame_inertia_lst, replacements);
frame_locations_transposed := list(Util.tuple22(frame_inertia) for frame_inertia in frame_inertia_lst);
end orderTransposedFrameLocations;
protected function frameLocationInertia
"the longer the chain of equal values at the start, the higher the inertia"
input FrameLocation frameLocation;
output Integer inertia = 1;
protected
array<Integer> dim;
algorithm
dim := Util.tuple21(frameLocation);
while inertia < arrayLength(dim) and dim[inertia] == dim[inertia+1] loop
inertia := inertia + 1;
end while;
end frameLocationInertia;
protected function resolveEqualInertia
"Squashing all equal inertia frames (nested loops) into one.
Equal inertia for frames shows that they 'fire' at the same time.
These frames have to change in one step, therefore they should be merged to
a single one."
input list<tuple<Integer, FrameLocation>> frame_inertia_lst;
input UnorderedMap<ComponentRef, Expression> replacements;
output list<tuple<Integer, FrameLocation>> resolved = {};
output FrameOrderingStatus status = NBEquation.FrameOrderingStatus.UNCHANGED;
protected
tuple<Integer, FrameLocation> tpl1, tpl2;
list<tuple<Integer, FrameLocation>> rest;
algorithm
tpl1 :: rest := frame_inertia_lst;
while not listEmpty(rest) loop
tpl2 :: rest := rest;
tpl1 := match (tpl1, tpl2)
local
Integer inertia1, inertia2, m, b;
array<Integer> loc1, loc2;
ComponentRef name1, name2;
Operator addOp, mulOp;
Expression linMap;
// equal inertia, combine the frames
case ((inertia1, (loc1, (name1, _))), (inertia2, (loc2, (name2, _)))) guard(inertia1 == inertia2) algorithm
addOp := Operator.fromClassification((NFOperator.MathClassification.ADDITION, NFOperator.SizeClassification.SCALAR), Type.INTEGER());
mulOp := Operator.fromClassification((NFOperator.MathClassification.MULTIPLICATION, NFOperator.SizeClassification.SCALAR), Type.INTEGER());
if arrayLength(loc1) <> arrayLength(loc2) then
Error.addMessage(Error.INTERNAL_ERROR,{getInstanceName() + " failed because frames have same inertia but different length.\n"
+ List.toString(arrayList(loc1), intString) + "\n" + List.toString(arrayList(loc2), intString)});
status := NBEquation.FrameOrderingStatus.FAILURE;
elseif arrayLength(loc1) == 1 then
b := loc2[1] - loc1[1];
linMap := Expression.fromCref(name1);
if b <> 0 then
linMap := Expression.MULTARY({Expression.INTEGER(b), linMap}, {}, addOp);
end if;
UnorderedMap.add(name2, linMap, replacements);
status := NBEquation.FrameOrderingStatus.CHANGED;
else
// compute linear map from frame1 to frame2 (y = m*x + b)
// ToDo: integer to real conversion might be wrong?
m := realInt((loc2[1]-loc2[1+inertia2])/(loc1[1]-loc1[1+inertia1]));
b := loc2[1]-m*loc1[1];
// check if linear map holds
for i in 2:arrayLength(loc1) loop
if loc2[i] <> m*loc1[i] + b then
Error.addMessage(Error.INTERNAL_ERROR,{getInstanceName() + " failed because frames have same inertia but the linear map does not hold.\n"
+ "map: y = " + intString(m) + " * x + " + intString(b) + "\n" + List.toString(arrayList(loc1), intString) + "\n" + List.toString(arrayList(loc2), intString)});
status := NBEquation.FrameOrderingStatus.FAILURE;
end if;
end for;
linMap := Expression.fromCref(name1);
if m <> 1 then
linMap := Expression.MULTARY({Expression.INTEGER(m), linMap}, {}, mulOp);
end if;
if b <> 0 then
linMap := Expression.MULTARY({Expression.INTEGER(b), linMap}, {}, addOp);
end if;
UnorderedMap.add(name2, linMap, replacements);
status := NBEquation.FrameOrderingStatus.CHANGED;
end if;
then tpl1;
// different inertia
else algorithm
resolved := tpl1 :: resolved;
then tpl2;
end match;
end while;
resolved := listReverse(tpl1 :: resolved);
end resolveEqualInertia;
public function recollectRangesHeuristic
"consecutively builds up the new frames from frame locations.
Assumes that slicing along the dimensions is possible.
Basic Idea:
1. iterate over each frame location
2. take first (start) and second (stop) element of frame dim to start the search for a pattern (step = stop - start)
3. shift the stop location further until the step changes and safe the start-step-stop pattern
3.1 iterate over the rest of the dim and check if the pattern holds for all of it
3.2 if it not holds search a missing diagonal for this dimension (reconstruct diagonal)
4. increase the shift for the length of the previous pattern and go to next frame location (shifting happens inherently in step 3)"
input list<FrameLocation> frame_locations_transposed;
output list<Frame> frames = {};
output Option<UnorderedMap<ComponentRef, Expression>> removed_diagonal = NONE();
output RecollectStatus status;
protected
array<Integer> dim;
Frame frame;
Integer check_shift, pre_shift, shift = 1;
Integer start, step, stop, max_size, new_step, new_stop, check_stop;
Boolean fail_;
list<Integer> rest;
list<Integer> starts = {}, stops = {}, steps = {}, shifts = {};
list<Boolean> failed = {};
Integer min_dim, max_dim;
list<FrameLocation> diagonal;
UnorderedMap<ComponentRef, Expression> replacements;
FrameOrderingStatus fos;
algorithm
for tpl in frame_locations_transposed loop
// 1. iterate over each frame location
fail_ := false;
(dim, frame) := tpl;
pre_shift := shift;
max_size := arrayLength(dim);
if max_size == 1 then
// if there is only one frame, it is a single equation at that exact point
frames := applyNewFrameRange(frame, (dim[1], 1, dim[1])) :: frames;
starts := dim[1] :: starts;
steps := 0 :: steps;
stops := dim[1] :: stops;
shifts := shift :: shifts;
else
// 2. take first (start) and second (stop) element of frame dim to start the search for a pattern (step = stop - start)
start := dim[1];
stop := dim[1 + shift];
step := stop - start;
if step == 0 then
// if the step size is zero, this range only has a single entry
// this should not happen?
frames := applyNewFrameRange(frame, (start, 1, stop)) :: frames;
starts := start :: starts;
steps := step :: steps;
stops := stop :: stops;
shifts := shift :: shifts;
else
// 3. shift the stop location further until the step changes and safe the start-step-stop pattern
new_step := step;
new_stop := stop;
while (new_step == step) and (shift + pre_shift < max_size) loop
stop := new_stop;
shift := shift + pre_shift;
new_stop := dim[1 + shift];
new_step := new_stop - stop;
end while;
if new_step == step then
// if new_step and step are still equal we hit the end (max_size)
stop := new_stop;
shift := shift + pre_shift; //not necessary but more correct
else
// 3.1 iterate over the rest of the dim and check if the pattern holds for all of it
check_shift := shift;
while (check_shift + pre_shift < max_size) loop
new_step := step;
while (new_step == step) and (check_shift + pre_shift < max_size) loop
check_stop := new_stop;
check_shift := check_shift + pre_shift;
new_stop := dim[1 + check_shift];
new_step := new_stop - check_stop;
end while;
// has to be the same amount of steps after the step size changes
if (check_shift + pre_shift == max_size) then
check_shift := check_shift + pre_shift;
end if;
if not intMod(check_shift, shift) == 0 then
fail_ := true;
break;
end if;
end while;
end if;
// use max/min dim instead of start and stop because the start or end
// could be missing (missing diagonals)
min_dim := min(d for d in dim);
max_dim := max(d for d in dim);
if fail_ then
if step > 0 then
frames := applyNewFrameRange(frame, (min_dim, step, max_dim)) :: frames;
else
frames := applyNewFrameRange(frame, (max_dim, step, min_dim)) :: frames;
end if;
else
frames := applyNewFrameRange(frame, (start, step, stop)) :: frames;
end if;
steps := step :: steps;
starts := if step > 0 then min_dim :: starts else max_dim :: starts;
stops := if step > 0 then max_dim :: stops else min_dim :: stops;
shifts := shift :: shifts;
failed := fail_ :: failed;
end if;
end if;
end for;
// 3.2 if it not holds search a missing diagonal for this dimension (reconstruct diagonal)
// if any dimension was not consistent, try to find a missing diagonal
// it is stored in an unordered map as linear map for the indices
if List.fold(failed, boolOr, false) then
diagonal := reconstructDiagonal(frame_locations_transposed, listReverse(starts), listReverse(steps), listReverse(stops), listReverse(shifts), listReverse(failed));
(diagonal, replacements, fos) := orderTransposedFrameLocations(diagonal);
if fos == NBEquation.FrameOrderingStatus.CHANGED then
removed_diagonal := SOME(replacements);
status := NBEquation.RecollectStatus.SUCCESS;
else
// no equal inertia to resolve or unable to resolve
status := NBEquation.RecollectStatus.FAILURE;
end if;
else
status := NBEquation.RecollectStatus.SUCCESS;
end if;
end recollectRangesHeuristic;
function reconstructDiagonal
"reconstructs a supposed missing diagonal if it exists.
ToDo1: create multiple diagonals if missing indices are found in one go without reset"
input list<FrameLocation> frame_locations_transposed;
input list<Integer> starts;
input list<Integer> steps;
input list<Integer> stops;
input list<Integer> shifts;
input list<Boolean> failed;
output list<FrameLocation> diagonal = {};
protected
Integer start, step, stop, pos, shift = 1;
Boolean fail_;
list<Integer> start_rest = starts, step_rest = steps, stop_rest = stops, shift_rest = shifts;
list<Boolean> fail_rest = failed;
array<Integer> dim;
list<Integer> missing_dims;
Frame frame;
algorithm
// ToDo: all lists have to be of equal length!
// default first shift to 1
for tpl in frame_locations_transposed loop
// get dims and frame from tpl
(dim, frame) := tpl;
// take out start, step, stop, fail
start :: start_rest := start_rest;
step :: step_rest := step_rest;
stop :: stop_rest := stop_rest;
fail_ :: fail_rest := fail_rest;
// initialize missing dims and pos
missing_dims := {};
pos := start;
if fail_ then
for i in 1:shift:arrayLength(dim) loop
while dim[i] <> pos loop
// ToDo1
missing_dims := pos :: missing_dims;
pos := pos + step;
if (sign(step)*pos > sign(step)*stop) then
break;
end if;
end while;
if (sign(step)*(pos+step) > sign(step)*stop) then