/[escript]/trunk/escript/src/Data.cpp
ViewVC logotype

Contents of /trunk/escript/src/Data.cpp

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1327 - (show annotations)
Fri Oct 12 07:10:40 2007 UTC (12 years, 5 months ago) by matt
File size: 76014 byte(s)
Initial rewrite of binary escript operations.


1
2 /* $Id$ */
3
4 /*******************************************************
5 *
6 * Copyright 2003-2007 by ACceSS MNRF
7 * Copyright 2007 by University of Queensland
8 *
9 * http://esscc.uq.edu.au
10 * Primary Business: Queensland, Australia
11 * Licensed under the Open Software License version 3.0
12 * http://www.opensource.org/licenses/osl-3.0.php
13 *
14 *******************************************************/
15
16 #include "Data.h"
17
18 #include "DataExpanded.h"
19 #include "DataConstant.h"
20 #include "DataTagged.h"
21 #include "DataEmpty.h"
22 #include "DataArrayView.h"
23 #include "FunctionSpaceFactory.h"
24 #include "AbstractContinuousDomain.h"
25 #include "UnaryFuncs.h"
26 extern "C" {
27 #include "escript/blocktimer.h"
28 }
29
30 #include <fstream>
31 #include <algorithm>
32 #include <vector>
33 #include <functional>
34
35 #include <boost/python/dict.hpp>
36 #include <boost/python/extract.hpp>
37 #include <boost/python/long.hpp>
38
39 using namespace std;
40 using namespace boost::python;
41 using namespace boost;
42 using namespace escript;
43
44 Data::Data()
45 {
46 //
47 // Default data is type DataEmpty
48 DataAbstract* temp=new DataEmpty();
49 shared_ptr<DataAbstract> temp_data(temp);
50 m_data=temp_data;
51 m_protected=false;
52 }
53
54 Data::Data(double value,
55 const tuple& shape,
56 const FunctionSpace& what,
57 bool expanded)
58 {
59 DataArrayView::ShapeType dataPointShape;
60 for (int i = 0; i < shape.attr("__len__")(); ++i) {
61 dataPointShape.push_back(extract<const int>(shape[i]));
62 }
63
64 int len = DataArrayView::noValues(dataPointShape);
65 DataVector temp_data(len,value,len);
66 DataArrayView temp_dataView(temp_data, dataPointShape);
67
68 initialise(temp_dataView, what, expanded);
69
70 m_protected=false;
71 }
72
73 Data::Data(double value,
74 const DataArrayView::ShapeType& dataPointShape,
75 const FunctionSpace& what,
76 bool expanded)
77 {
78 int len = DataArrayView::noValues(dataPointShape);
79
80 DataVector temp_data(len,value,len);
81 DataArrayView temp_dataView(temp_data, dataPointShape);
82
83 initialise(temp_dataView, what, expanded);
84
85 m_protected=false;
86 }
87
88 Data::Data(const Data& inData)
89 {
90 m_data=inData.m_data;
91 m_protected=inData.isProtected();
92 }
93
94 Data::Data(const Data& inData,
95 const DataArrayView::RegionType& region)
96 {
97 //
98 // Create Data which is a slice of another Data
99 DataAbstract* tmp = inData.m_data->getSlice(region);
100 shared_ptr<DataAbstract> temp_data(tmp);
101 m_data=temp_data;
102 m_protected=false;
103 }
104
105 Data::Data(const Data& inData,
106 const FunctionSpace& functionspace)
107 {
108 if (inData.getFunctionSpace()==functionspace) {
109 m_data=inData.m_data;
110 } else {
111 Data tmp(0,inData.getPointDataView().getShape(),functionspace,true);
112 // Note: Must use a reference or pointer to a derived object
113 // in order to get polymorphic behaviour. Shouldn't really
114 // be able to create an instance of AbstractDomain but that was done
115 // as a boost:python work around which may no longer be required.
116 const AbstractDomain& inDataDomain=inData.getDomain();
117 if (inDataDomain==functionspace.getDomain()) {
118 inDataDomain.interpolateOnDomain(tmp,inData);
119 } else {
120 inDataDomain.interpolateACross(tmp,inData);
121 }
122 m_data=tmp.m_data;
123 }
124 m_protected=false;
125 }
126
127 Data::Data(const DataTagged::TagListType& tagKeys,
128 const DataTagged::ValueListType & values,
129 const DataArrayView& defaultValue,
130 const FunctionSpace& what,
131 bool expanded)
132 {
133 DataAbstract* temp=new DataTagged(tagKeys,values,defaultValue,what);
134 shared_ptr<DataAbstract> temp_data(temp);
135 m_data=temp_data;
136 m_protected=false;
137 if (expanded) {
138 expand();
139 }
140 }
141
142 Data::Data(const numeric::array& value,
143 const FunctionSpace& what,
144 bool expanded)
145 {
146 initialise(value,what,expanded);
147 m_protected=false;
148 }
149
150 Data::Data(const DataArrayView& value,
151 const FunctionSpace& what,
152 bool expanded)
153 {
154 initialise(value,what,expanded);
155 m_protected=false;
156 }
157
158 Data::Data(const object& value,
159 const FunctionSpace& what,
160 bool expanded)
161 {
162 numeric::array asNumArray(value);
163 initialise(asNumArray,what,expanded);
164 m_protected=false;
165 }
166
167
168 Data::Data(const object& value,
169 const Data& other)
170 {
171
172 numeric::array asNumArray(value);
173
174
175 // extract the shape of the numarray
176 DataArrayView::ShapeType tempShape;
177 for (int i=0; i < asNumArray.getrank(); i++) {
178 tempShape.push_back(extract<int>(asNumArray.getshape()[i]));
179 }
180 // get the space for the data vector
181 int len = DataArrayView::noValues(tempShape);
182 DataVector temp_data(len, 0.0, len);
183 DataArrayView temp_dataView(temp_data, tempShape);
184 temp_dataView.copy(asNumArray);
185
186 //
187 // Create DataConstant using the given value and all other parameters
188 // copied from other. If value is a rank 0 object this Data
189 // will assume the point data shape of other.
190
191 if (temp_dataView.getRank()==0) {
192 int len = DataArrayView::noValues(other.getPointDataView().getShape());
193
194 DataVector temp2_data(len, temp_dataView(), len);
195 DataArrayView temp2_dataView(temp_data, other.getPointDataView().getShape());
196 initialise(temp2_dataView, other.getFunctionSpace(), false);
197
198 } else {
199 //
200 // Create a DataConstant with the same sample shape as other
201 initialise(temp_dataView, other.getFunctionSpace(), false);
202 }
203 m_protected=false;
204 }
205
206 Data::~Data()
207 {
208
209 }
210
211 escriptDataC
212 Data::getDataC()
213 {
214 escriptDataC temp;
215 temp.m_dataPtr=(void*)this;
216 return temp;
217 }
218
219 escriptDataC
220 Data::getDataC() const
221 {
222 escriptDataC temp;
223 temp.m_dataPtr=(void*)this;
224 return temp;
225 }
226
227 const boost::python::tuple
228 Data::getShapeTuple() const
229 {
230 const DataArrayView::ShapeType& shape=getDataPointShape();
231 switch(getDataPointRank()) {
232 case 0:
233 return make_tuple();
234 case 1:
235 return make_tuple(long_(shape[0]));
236 case 2:
237 return make_tuple(long_(shape[0]),long_(shape[1]));
238 case 3:
239 return make_tuple(long_(shape[0]),long_(shape[1]),long_(shape[2]));
240 case 4:
241 return make_tuple(long_(shape[0]),long_(shape[1]),long_(shape[2]),long_(shape[3]));
242 default:
243 throw DataException("Error - illegal Data rank.");
244 }
245 }
246 void
247 Data::copy(const Data& other)
248 {
249 //
250 // Perform a deep copy
251 {
252 DataExpanded* temp=dynamic_cast<DataExpanded*>(other.m_data.get());
253 if (temp!=0) {
254 //
255 // Construct a DataExpanded copy
256 DataAbstract* newData=new DataExpanded(*temp);
257 shared_ptr<DataAbstract> temp_data(newData);
258 m_data=temp_data;
259 return;
260 }
261 }
262 {
263 DataTagged* temp=dynamic_cast<DataTagged*>(other.m_data.get());
264 if (temp!=0) {
265 //
266 // Construct a DataTagged copy
267 DataAbstract* newData=new DataTagged(*temp);
268 shared_ptr<DataAbstract> temp_data(newData);
269 m_data=temp_data;
270 return;
271 }
272 }
273 {
274 DataConstant* temp=dynamic_cast<DataConstant*>(other.m_data.get());
275 if (temp!=0) {
276 //
277 // Construct a DataConstant copy
278 DataAbstract* newData=new DataConstant(*temp);
279 shared_ptr<DataAbstract> temp_data(newData);
280 m_data=temp_data;
281 return;
282 }
283 }
284 {
285 DataEmpty* temp=dynamic_cast<DataEmpty*>(other.m_data.get());
286 if (temp!=0) {
287 //
288 // Construct a DataEmpty copy
289 DataAbstract* newData=new DataEmpty();
290 shared_ptr<DataAbstract> temp_data(newData);
291 m_data=temp_data;
292 return;
293 }
294 }
295 throw DataException("Error - Copy not implemented for this Data type.");
296 }
297
298
299 void
300 Data::setToZero()
301 {
302 {
303 DataExpanded* temp=dynamic_cast<DataExpanded*>(m_data.get());
304 if (temp!=0) {
305 temp->setToZero();
306 return;
307 }
308 }
309 {
310 DataTagged* temp=dynamic_cast<DataTagged*>(m_data.get());
311 if (temp!=0) {
312 temp->setToZero();
313 return;
314 }
315 }
316 {
317 DataConstant* temp=dynamic_cast<DataConstant*>(m_data.get());
318 if (temp!=0) {
319 temp->setToZero();
320 return;
321 }
322 }
323 throw DataException("Error - Data can not be set to zero.");
324 }
325
326 void
327 Data::copyWithMask(const Data& other,
328 const Data& mask)
329 {
330 Data mask1;
331 Data mask2;
332
333 mask1 = mask.wherePositive();
334 mask2.copy(mask1);
335
336 mask1 *= other;
337 mask2 *= *this;
338 mask2 = *this - mask2;
339
340 *this = mask1 + mask2;
341 }
342
343 bool
344 Data::isExpanded() const
345 {
346 DataExpanded* temp=dynamic_cast<DataExpanded*>(m_data.get());
347 return (temp!=0);
348 }
349
350 bool
351 Data::isTagged() const
352 {
353 DataTagged* temp=dynamic_cast<DataTagged*>(m_data.get());
354 return (temp!=0);
355 }
356
357 bool
358 Data::isEmpty() const
359 {
360 DataEmpty* temp=dynamic_cast<DataEmpty*>(m_data.get());
361 return (temp!=0);
362 }
363
364 bool
365 Data::isConstant() const
366 {
367 DataConstant* temp=dynamic_cast<DataConstant*>(m_data.get());
368 return (temp!=0);
369 }
370
371 void
372 Data::setProtection()
373 {
374 m_protected=true;
375 }
376
377 bool
378 Data::isProtected() const
379 {
380 return m_protected;
381 }
382
383
384
385 void
386 Data::expand()
387 {
388 if (isConstant()) {
389 DataConstant* tempDataConst=dynamic_cast<DataConstant*>(m_data.get());
390 DataAbstract* temp=new DataExpanded(*tempDataConst);
391 shared_ptr<DataAbstract> temp_data(temp);
392 m_data=temp_data;
393 } else if (isTagged()) {
394 DataTagged* tempDataTag=dynamic_cast<DataTagged*>(m_data.get());
395 DataAbstract* temp=new DataExpanded(*tempDataTag);
396 shared_ptr<DataAbstract> temp_data(temp);
397 m_data=temp_data;
398 } else if (isExpanded()) {
399 //
400 // do nothing
401 } else if (isEmpty()) {
402 throw DataException("Error - Expansion of DataEmpty not possible.");
403 } else {
404 throw DataException("Error - Expansion not implemented for this Data type.");
405 }
406 }
407
408 void
409 Data::tag()
410 {
411 if (isConstant()) {
412 DataConstant* tempDataConst=dynamic_cast<DataConstant*>(m_data.get());
413 DataAbstract* temp=new DataTagged(*tempDataConst);
414 shared_ptr<DataAbstract> temp_data(temp);
415 m_data=temp_data;
416 } else if (isTagged()) {
417 // do nothing
418 } else if (isExpanded()) {
419 throw DataException("Error - Creating tag data from DataExpanded not possible.");
420 } else if (isEmpty()) {
421 throw DataException("Error - Creating tag data from DataEmpty not possible.");
422 } else {
423 throw DataException("Error - Tagging not implemented for this Data type.");
424 }
425 }
426
427 Data
428 Data::oneOver() const
429 {
430 return escript::unaryOp(*this,bind1st(divides<double>(),1.));
431 }
432
433 Data
434 Data::wherePositive() const
435 {
436 return escript::unaryOp(*this,bind2nd(greater<double>(),0.0));
437 }
438
439 Data
440 Data::whereNegative() const
441 {
442 return escript::unaryOp(*this,bind2nd(less<double>(),0.0));
443 }
444
445 Data
446 Data::whereNonNegative() const
447 {
448 return escript::unaryOp(*this,bind2nd(greater_equal<double>(),0.0));
449 }
450
451 Data
452 Data::whereNonPositive() const
453 {
454 return escript::unaryOp(*this,bind2nd(less_equal<double>(),0.0));
455 }
456
457 Data
458 Data::whereZero(double tol) const
459 {
460 Data dataAbs=abs();
461 return escript::unaryOp(dataAbs,bind2nd(less_equal<double>(),tol));
462 }
463
464 Data
465 Data::whereNonZero(double tol) const
466 {
467 Data dataAbs=abs();
468 return escript::unaryOp(dataAbs,bind2nd(greater<double>(),tol));
469 }
470
471 Data
472 Data::interpolate(const FunctionSpace& functionspace) const
473 {
474 return Data(*this,functionspace);
475 }
476
477 bool
478 Data::probeInterpolation(const FunctionSpace& functionspace) const
479 {
480 if (getFunctionSpace()==functionspace) {
481 return true;
482 } else {
483 const AbstractDomain& domain=getDomain();
484 if (domain==functionspace.getDomain()) {
485 return domain.probeInterpolationOnDomain(getFunctionSpace().getTypeCode(),functionspace.getTypeCode());
486 } else {
487 return domain.probeInterpolationACross(getFunctionSpace().getTypeCode(),functionspace.getDomain(),functionspace.getTypeCode());
488 }
489 }
490 }
491
492 Data
493 Data::gradOn(const FunctionSpace& functionspace) const
494 {
495 double blocktimer_start = blocktimer_time();
496 if (functionspace.getDomain()!=getDomain())
497 throw DataException("Error - gradient cannot be calculated on different domains.");
498 DataArrayView::ShapeType grad_shape=getPointDataView().getShape();
499 grad_shape.push_back(functionspace.getDim());
500 Data out(0.0,grad_shape,functionspace,true);
501 getDomain().setToGradient(out,*this);
502 blocktimer_increment("grad()", blocktimer_start);
503 return out;
504 }
505
506 Data
507 Data::grad() const
508 {
509 return gradOn(escript::function(getDomain()));
510 }
511
512 int
513 Data::getDataPointSize() const
514 {
515 return getPointDataView().noValues();
516 }
517
518 DataArrayView::ValueType::size_type
519 Data::getLength() const
520 {
521 return m_data->getLength();
522 }
523
524 const DataArrayView::ShapeType&
525 Data::getDataPointShape() const
526 {
527 return getPointDataView().getShape();
528 }
529
530
531
532 const
533 boost::python::numeric::array
534 Data:: getValueOfDataPoint(int dataPointNo)
535 {
536 size_t length=0;
537 int i, j, k, l;
538 //
539 // determine the rank and shape of each data point
540 int dataPointRank = getDataPointRank();
541 DataArrayView::ShapeType dataPointShape = getDataPointShape();
542
543 //
544 // create the numeric array to be returned
545 boost::python::numeric::array numArray(0.0);
546
547 //
548 // the shape of the returned numeric array will be the same
549 // as that of the data point
550 int arrayRank = dataPointRank;
551 DataArrayView::ShapeType arrayShape = dataPointShape;
552
553 //
554 // resize the numeric array to the shape just calculated
555 if (arrayRank==0) {
556 numArray.resize(1);
557 }
558 if (arrayRank==1) {
559 numArray.resize(arrayShape[0]);
560 }
561 if (arrayRank==2) {
562 numArray.resize(arrayShape[0],arrayShape[1]);
563 }
564 if (arrayRank==3) {
565 numArray.resize(arrayShape[0],arrayShape[1],arrayShape[2]);
566 }
567 if (arrayRank==4) {
568 numArray.resize(arrayShape[0],arrayShape[1],arrayShape[2],arrayShape[3]);
569 }
570
571 if (getNumDataPointsPerSample()>0) {
572 int sampleNo = dataPointNo/getNumDataPointsPerSample();
573 int dataPointNoInSample = dataPointNo - sampleNo * getNumDataPointsPerSample();
574 //
575 // Check a valid sample number has been supplied
576 if ((sampleNo >= getNumSamples()) || (sampleNo < 0 )) {
577 throw DataException("Error - Data::convertToNumArray: invalid sampleNo.");
578 }
579
580 //
581 // Check a valid data point number has been supplied
582 if ((dataPointNoInSample >= getNumDataPointsPerSample()) || (dataPointNoInSample < 0)) {
583 throw DataException("Error - Data::convertToNumArray: invalid dataPointNoInSample.");
584 }
585 // TODO: global error handling
586 // create a view of the data if it is stored locally
587 DataArrayView dataPointView = getDataPoint(sampleNo, dataPointNoInSample);
588
589 switch( dataPointRank ){
590 case 0 :
591 numArray[0] = dataPointView();
592 break;
593 case 1 :
594 for( i=0; i<dataPointShape[0]; i++ )
595 numArray[i]=dataPointView(i);
596 break;
597 case 2 :
598 for( i=0; i<dataPointShape[0]; i++ )
599 for( j=0; j<dataPointShape[1]; j++)
600 numArray[make_tuple(i,j)]=dataPointView(i,j);
601 break;
602 case 3 :
603 for( i=0; i<dataPointShape[0]; i++ )
604 for( j=0; j<dataPointShape[1]; j++ )
605 for( k=0; k<dataPointShape[2]; k++)
606 numArray[make_tuple(i,j,k)]=dataPointView(i,j,k);
607 break;
608 case 4 :
609 for( i=0; i<dataPointShape[0]; i++ )
610 for( j=0; j<dataPointShape[1]; j++ )
611 for( k=0; k<dataPointShape[2]; k++ )
612 for( l=0; l<dataPointShape[3]; l++)
613 numArray[make_tuple(i,j,k,l)]=dataPointView(i,j,k,l);
614 break;
615 }
616 }
617 //
618 // return the array
619 return numArray;
620
621 }
622 void
623 Data::setValueOfDataPointToPyObject(int dataPointNo, const boost::python::object& py_object)
624 {
625 // this will throw if the value cannot be represented
626 boost::python::numeric::array num_array(py_object);
627 setValueOfDataPointToArray(dataPointNo,num_array);
628
629
630 }
631
632 void
633 Data::setValueOfDataPointToArray(int dataPointNo, const boost::python::numeric::array& num_array)
634 {
635 if (isProtected()) {
636 throw DataException("Error - attempt to update protected Data object.");
637 }
638 //
639 // check rank
640 if (num_array.getrank()<getDataPointRank())
641 throw DataException("Rank of numarray does not match Data object rank");
642
643 //
644 // check shape of num_array
645 for (int i=0; i<getDataPointRank(); i++) {
646 if (extract<int>(num_array.getshape()[i])!=getDataPointShape()[i])
647 throw DataException("Shape of numarray does not match Data object rank");
648 }
649 //
650 // make sure data is expanded:
651 if (!isExpanded()) {
652 expand();
653 }
654 if (getNumDataPointsPerSample()>0) {
655 int sampleNo = dataPointNo/getNumDataPointsPerSample();
656 int dataPointNoInSample = dataPointNo - sampleNo * getNumDataPointsPerSample();
657 m_data->copyToDataPoint(sampleNo, dataPointNoInSample,num_array);
658 } else {
659 m_data->copyToDataPoint(-1, 0,num_array);
660 }
661 }
662
663 void
664 Data::setValueOfDataPoint(int dataPointNo, const double value)
665 {
666 if (isProtected()) {
667 throw DataException("Error - attempt to update protected Data object.");
668 }
669 //
670 // make sure data is expanded:
671 if (!isExpanded()) {
672 expand();
673 }
674 if (getNumDataPointsPerSample()>0) {
675 int sampleNo = dataPointNo/getNumDataPointsPerSample();
676 int dataPointNoInSample = dataPointNo - sampleNo * getNumDataPointsPerSample();
677 m_data->copyToDataPoint(sampleNo, dataPointNoInSample,value);
678 } else {
679 m_data->copyToDataPoint(-1, 0,value);
680 }
681 }
682
683 const
684 boost::python::numeric::array
685 Data::getValueOfGlobalDataPoint(int procNo, int dataPointNo)
686 {
687 size_t length=0;
688 int i, j, k, l, pos;
689 //
690 // determine the rank and shape of each data point
691 int dataPointRank = getDataPointRank();
692 DataArrayView::ShapeType dataPointShape = getDataPointShape();
693
694 //
695 // create the numeric array to be returned
696 boost::python::numeric::array numArray(0.0);
697
698 //
699 // the shape of the returned numeric array will be the same
700 // as that of the data point
701 int arrayRank = dataPointRank;
702 DataArrayView::ShapeType arrayShape = dataPointShape;
703
704 //
705 // resize the numeric array to the shape just calculated
706 if (arrayRank==0) {
707 numArray.resize(1);
708 }
709 if (arrayRank==1) {
710 numArray.resize(arrayShape[0]);
711 }
712 if (arrayRank==2) {
713 numArray.resize(arrayShape[0],arrayShape[1]);
714 }
715 if (arrayRank==3) {
716 numArray.resize(arrayShape[0],arrayShape[1],arrayShape[2]);
717 }
718 if (arrayRank==4) {
719 numArray.resize(arrayShape[0],arrayShape[1],arrayShape[2],arrayShape[3]);
720 }
721
722 // added for the MPI communication
723 length=1;
724 for( i=0; i<arrayRank; i++ ) length *= arrayShape[i];
725 double *tmpData = new double[length];
726
727 //
728 // load the values for the data point into the numeric array.
729
730 // updated for the MPI case
731 if( get_MPIRank()==procNo ){
732 if (getNumDataPointsPerSample()>0) {
733 int sampleNo = dataPointNo/getNumDataPointsPerSample();
734 int dataPointNoInSample = dataPointNo - sampleNo * getNumDataPointsPerSample();
735 //
736 // Check a valid sample number has been supplied
737 if ((sampleNo >= getNumSamples()) || (sampleNo < 0 )) {
738 throw DataException("Error - Data::convertToNumArray: invalid sampleNo.");
739 }
740
741 //
742 // Check a valid data point number has been supplied
743 if ((dataPointNoInSample >= getNumDataPointsPerSample()) || (dataPointNoInSample < 0)) {
744 throw DataException("Error - Data::convertToNumArray: invalid dataPointNoInSample.");
745 }
746 // TODO: global error handling
747 // create a view of the data if it is stored locally
748 DataArrayView dataPointView = getDataPoint(sampleNo, dataPointNoInSample);
749
750 // pack the data from the view into tmpData for MPI communication
751 pos=0;
752 switch( dataPointRank ){
753 case 0 :
754 tmpData[0] = dataPointView();
755 break;
756 case 1 :
757 for( i=0; i<dataPointShape[0]; i++ )
758 tmpData[i]=dataPointView(i);
759 break;
760 case 2 :
761 for( i=0; i<dataPointShape[0]; i++ )
762 for( j=0; j<dataPointShape[1]; j++, pos++ )
763 tmpData[pos]=dataPointView(i,j);
764 break;
765 case 3 :
766 for( i=0; i<dataPointShape[0]; i++ )
767 for( j=0; j<dataPointShape[1]; j++ )
768 for( k=0; k<dataPointShape[2]; k++, pos++ )
769 tmpData[pos]=dataPointView(i,j,k);
770 break;
771 case 4 :
772 for( i=0; i<dataPointShape[0]; i++ )
773 for( j=0; j<dataPointShape[1]; j++ )
774 for( k=0; k<dataPointShape[2]; k++ )
775 for( l=0; l<dataPointShape[3]; l++, pos++ )
776 tmpData[pos]=dataPointView(i,j,k,l);
777 break;
778 }
779 }
780 }
781 #ifdef PASO_MPI
782 // broadcast the data to all other processes
783 MPI_Bcast( tmpData, length, MPI_DOUBLE, procNo, get_MPIComm() );
784 #endif
785
786 // unpack the data
787 switch( dataPointRank ){
788 case 0 :
789 numArray[0]=tmpData[0];
790 break;
791 case 1 :
792 for( i=0; i<dataPointShape[0]; i++ )
793 numArray[i]=tmpData[i];
794 break;
795 case 2 :
796 for( i=0; i<dataPointShape[0]; i++ )
797 for( j=0; j<dataPointShape[1]; j++ )
798 numArray[make_tuple(i,j)]=tmpData[i+j*dataPointShape[0]];
799 break;
800 case 3 :
801 for( i=0; i<dataPointShape[0]; i++ )
802 for( j=0; j<dataPointShape[1]; j++ )
803 for( k=0; k<dataPointShape[2]; k++ )
804 numArray[make_tuple(i,j,k)]=tmpData[i+dataPointShape[0]*(j*+k*dataPointShape[1])];
805 break;
806 case 4 :
807 for( i=0; i<dataPointShape[0]; i++ )
808 for( j=0; j<dataPointShape[1]; j++ )
809 for( k=0; k<dataPointShape[2]; k++ )
810 for( l=0; l<dataPointShape[3]; l++ )
811 numArray[make_tuple(i,j,k,l)]=tmpData[i+dataPointShape[0]*(j*+dataPointShape[1]*(k+l*dataPointShape[2]))];
812 break;
813 }
814
815 delete [] tmpData;
816 //
817 // return the loaded array
818 return numArray;
819 }
820
821
822
823 boost::python::numeric::array
824 Data::integrate() const
825 {
826 int index;
827 int rank = getDataPointRank();
828 DataArrayView::ShapeType shape = getDataPointShape();
829 int dataPointSize = getDataPointSize();
830
831 //
832 // calculate the integral values
833 vector<double> integrals(dataPointSize);
834 vector<double> integrals_local(dataPointSize);
835 #ifdef PASO_MPI
836 AbstractContinuousDomain::asAbstractContinuousDomain(getDomain()).setToIntegrals(integrals_local,*this);
837 // Global sum: use an array instead of a vector because elements of array are guaranteed to be contiguous in memory
838 double *tmp = new double[dataPointSize];
839 double *tmp_local = new double[dataPointSize];
840 for (int i=0; i<dataPointSize; i++) { tmp_local[i] = integrals_local[i]; }
841 MPI_Allreduce( &tmp_local[0], &tmp[0], dataPointSize, MPI_DOUBLE, MPI_SUM, MPI_COMM_WORLD );
842 for (int i=0; i<dataPointSize; i++) { integrals[i] = tmp[i]; }
843 delete[] tmp;
844 delete[] tmp_local;
845 #else
846 AbstractContinuousDomain::asAbstractContinuousDomain(getDomain()).setToIntegrals(integrals,*this);
847 #endif
848
849 //
850 // create the numeric array to be returned
851 // and load the array with the integral values
852 boost::python::numeric::array bp_array(1.0);
853 if (rank==0) {
854 bp_array.resize(1);
855 index = 0;
856 bp_array[0] = integrals[index];
857 }
858 if (rank==1) {
859 bp_array.resize(shape[0]);
860 for (int i=0; i<shape[0]; i++) {
861 index = i;
862 bp_array[i] = integrals[index];
863 }
864 }
865 if (rank==2) {
866 bp_array.resize(shape[0],shape[1]);
867 for (int i=0; i<shape[0]; i++) {
868 for (int j=0; j<shape[1]; j++) {
869 index = i + shape[0] * j;
870 bp_array[make_tuple(i,j)] = integrals[index];
871 }
872 }
873 }
874 if (rank==3) {
875 bp_array.resize(shape[0],shape[1],shape[2]);
876 for (int i=0; i<shape[0]; i++) {
877 for (int j=0; j<shape[1]; j++) {
878 for (int k=0; k<shape[2]; k++) {
879 index = i + shape[0] * ( j + shape[1] * k );
880 bp_array[make_tuple(i,j,k)] = integrals[index];
881 }
882 }
883 }
884 }
885 if (rank==4) {
886 bp_array.resize(shape[0],shape[1],shape[2],shape[3]);
887 for (int i=0; i<shape[0]; i++) {
888 for (int j=0; j<shape[1]; j++) {
889 for (int k=0; k<shape[2]; k++) {
890 for (int l=0; l<shape[3]; l++) {
891 index = i + shape[0] * ( j + shape[1] * ( k + shape[2] * l ) );
892 bp_array[make_tuple(i,j,k,l)] = integrals[index];
893 }
894 }
895 }
896 }
897 }
898
899 //
900 // return the loaded array
901 return bp_array;
902 }
903
904 Data
905 Data::sin() const
906 {
907 return escript::unaryOp(*this,(Data::UnaryDFunPtr)::sin);
908 }
909
910 Data
911 Data::cos() const
912 {
913 return escript::unaryOp(*this,(Data::UnaryDFunPtr)::cos);
914 }
915
916 Data
917 Data::tan() const
918 {
919 return escript::unaryOp(*this,(Data::UnaryDFunPtr)::tan);
920 }
921
922 Data
923 Data::asin() const
924 {
925 return escript::unaryOp(*this,(Data::UnaryDFunPtr)::asin);
926 }
927
928 Data
929 Data::acos() const
930 {
931 return escript::unaryOp(*this,(Data::UnaryDFunPtr)::acos);
932 }
933
934
935 Data
936 Data::atan() const
937 {
938 return escript::unaryOp(*this,(Data::UnaryDFunPtr)::atan);
939 }
940
941 Data
942 Data::sinh() const
943 {
944 return escript::unaryOp(*this,(Data::UnaryDFunPtr)::sinh);
945 }
946
947 Data
948 Data::cosh() const
949 {
950 return escript::unaryOp(*this,(Data::UnaryDFunPtr)::cosh);
951 }
952
953 Data
954 Data::tanh() const
955 {
956 return escript::unaryOp(*this,(Data::UnaryDFunPtr)::tanh);
957 }
958
959
960 Data
961 Data::erf() const
962 {
963 #ifdef _WIN32
964 throw DataException("Error - Data:: erf function is not supported on _WIN32 platforms.");
965 #else
966 return escript::unaryOp(*this,(Data::UnaryDFunPtr)::erf);
967 #endif
968 }
969
970 Data
971 Data::asinh() const
972 {
973 #ifdef _WIN32
974 return escript::unaryOp(*this,escript::asinh_substitute);
975 #else
976 return escript::unaryOp(*this,(Data::UnaryDFunPtr)::asinh);
977 #endif
978 }
979
980 Data
981 Data::acosh() const
982 {
983 #ifdef _WIN32
984 return escript::unaryOp(*this,escript::acosh_substitute);
985 #else
986 return escript::unaryOp(*this,(Data::UnaryDFunPtr)::acosh);
987 #endif
988 }
989
990 Data
991 Data::atanh() const
992 {
993 #ifdef _WIN32
994 return escript::unaryOp(*this,escript::atanh_substitute);
995 #else
996 return escript::unaryOp(*this,(Data::UnaryDFunPtr)::atanh);
997 #endif
998 }
999
1000 Data
1001 Data::log10() const
1002 {
1003 return escript::unaryOp(*this,(Data::UnaryDFunPtr)::log10);
1004 }
1005
1006 Data
1007 Data::log() const
1008 {
1009 return escript::unaryOp(*this,(Data::UnaryDFunPtr)::log);
1010 }
1011
1012 Data
1013 Data::sign() const
1014 {
1015 return escript::unaryOp(*this,escript::fsign);
1016 }
1017
1018 Data
1019 Data::abs() const
1020 {
1021 return escript::unaryOp(*this,(Data::UnaryDFunPtr)::fabs);
1022 }
1023
1024 Data
1025 Data::neg() const
1026 {
1027 return escript::unaryOp(*this,negate<double>());
1028 }
1029
1030 Data
1031 Data::pos() const
1032 {
1033 Data result;
1034 // perform a deep copy
1035 result.copy(*this);
1036 return result;
1037 }
1038
1039 Data
1040 Data::exp() const
1041 {
1042 return escript::unaryOp(*this,(Data::UnaryDFunPtr)::exp);
1043 }
1044
1045 Data
1046 Data::sqrt() const
1047 {
1048 return escript::unaryOp(*this,(Data::UnaryDFunPtr)::sqrt);
1049 }
1050
1051 double
1052 Data::Lsup() const
1053 {
1054 double localValue, globalValue;
1055 //
1056 // set the initial absolute maximum value to zero
1057
1058 AbsMax abs_max_func;
1059 localValue = algorithm(abs_max_func,0);
1060 #ifdef PASO_MPI
1061 MPI_Allreduce( &localValue, &globalValue, 1, MPI_DOUBLE, MPI_MAX, MPI_COMM_WORLD );
1062 return globalValue;
1063 #else
1064 return localValue;
1065 #endif
1066 }
1067
1068 double
1069 Data::sup() const
1070 {
1071 double localValue, globalValue;
1072 //
1073 // set the initial maximum value to min possible double
1074 FMax fmax_func;
1075 localValue = algorithm(fmax_func,numeric_limits<double>::max()*-1);
1076 #ifdef PASO_MPI
1077 MPI_Allreduce( &localValue, &globalValue, 1, MPI_DOUBLE, MPI_MAX, MPI_COMM_WORLD );
1078 return globalValue;
1079 #else
1080 return localValue;
1081 #endif
1082 }
1083
1084 double
1085 Data::inf() const
1086 {
1087 double localValue, globalValue;
1088 //
1089 // set the initial minimum value to max possible double
1090 FMin fmin_func;
1091 localValue = algorithm(fmin_func,numeric_limits<double>::max());
1092 #ifdef PASO_MPI
1093 MPI_Allreduce( &localValue, &globalValue, 1, MPI_DOUBLE, MPI_MIN, MPI_COMM_WORLD );
1094 return globalValue;
1095 #else
1096 return localValue;
1097 #endif
1098 }
1099
1100 /* TODO */
1101 /* global reduction */
1102 Data
1103 Data::maxval() const
1104 {
1105 //
1106 // set the initial maximum value to min possible double
1107 FMax fmax_func;
1108 return dp_algorithm(fmax_func,numeric_limits<double>::max()*-1);
1109 }
1110
1111 Data
1112 Data::minval() const
1113 {
1114 //
1115 // set the initial minimum value to max possible double
1116 FMin fmin_func;
1117 return dp_algorithm(fmin_func,numeric_limits<double>::max());
1118 }
1119
1120 Data
1121 Data::swapaxes(const int axis0, const int axis1) const
1122 {
1123 int axis0_tmp,axis1_tmp;
1124 DataArrayView::ShapeType s=getDataPointShape();
1125 DataArrayView::ShapeType ev_shape;
1126 // Here's the equivalent of python s_out=s[axis_offset:]+s[:axis_offset]
1127 // which goes thru all shape vector elements starting with axis_offset (at index=rank wrap around to 0)
1128 int rank=getDataPointRank();
1129 if (rank<2) {
1130 throw DataException("Error - Data::swapaxes argument must have at least rank 2.");
1131 }
1132 if (axis0<0 || axis0>rank-1) {
1133 throw DataException("Error - Data::swapaxes: axis0 must be between 0 and rank-1=" + rank-1);
1134 }
1135 if (axis1<0 || axis1>rank-1) {
1136 throw DataException("Error - Data::swapaxes: axis1 must be between 0 and rank-1=" + rank-1);
1137 }
1138 if (axis0 == axis1) {
1139 throw DataException("Error - Data::swapaxes: axis indices must be different.");
1140 }
1141 if (axis0 > axis1) {
1142 axis0_tmp=axis1;
1143 axis1_tmp=axis0;
1144 } else {
1145 axis0_tmp=axis0;
1146 axis1_tmp=axis1;
1147 }
1148 for (int i=0; i<rank; i++) {
1149 if (i == axis0_tmp) {
1150 ev_shape.push_back(s[axis1_tmp]);
1151 } else if (i == axis1_tmp) {
1152 ev_shape.push_back(s[axis0_tmp]);
1153 } else {
1154 ev_shape.push_back(s[i]);
1155 }
1156 }
1157 Data ev(0.,ev_shape,getFunctionSpace());
1158 ev.typeMatchRight(*this);
1159 m_data->swapaxes(ev.m_data.get(), axis0_tmp, axis1_tmp);
1160 return ev;
1161
1162 }
1163
1164 Data
1165 Data::symmetric() const
1166 {
1167 // check input
1168 DataArrayView::ShapeType s=getDataPointShape();
1169 if (getDataPointRank()==2) {
1170 if(s[0] != s[1])
1171 throw DataException("Error - Data::symmetric can only be calculated for rank 2 object with equal first and second dimension.");
1172 }
1173 else if (getDataPointRank()==4) {
1174 if(!(s[0] == s[2] && s[1] == s[3]))
1175 throw DataException("Error - Data::symmetric can only be calculated for rank 4 object with dim0==dim2 and dim1==dim3.");
1176 }
1177 else {
1178 throw DataException("Error - Data::symmetric can only be calculated for rank 2 or 4 object.");
1179 }
1180 Data ev(0.,getDataPointShape(),getFunctionSpace());
1181 ev.typeMatchRight(*this);
1182 m_data->symmetric(ev.m_data.get());
1183 return ev;
1184 }
1185
1186 Data
1187 Data::nonsymmetric() const
1188 {
1189 // check input
1190 DataArrayView::ShapeType s=getDataPointShape();
1191 if (getDataPointRank()==2) {
1192 if(s[0] != s[1])
1193 throw DataException("Error - Data::nonsymmetric can only be calculated for rank 2 object with equal first and second dimension.");
1194 DataArrayView::ShapeType ev_shape;
1195 ev_shape.push_back(s[0]);
1196 ev_shape.push_back(s[1]);
1197 Data ev(0.,ev_shape,getFunctionSpace());
1198 ev.typeMatchRight(*this);
1199 m_data->nonsymmetric(ev.m_data.get());
1200 return ev;
1201 }
1202 else if (getDataPointRank()==4) {
1203 if(!(s[0] == s[2] && s[1] == s[3]))
1204 throw DataException("Error - Data::nonsymmetric can only be calculated for rank 4 object with dim0==dim2 and dim1==dim3.");
1205 DataArrayView::ShapeType ev_shape;
1206 ev_shape.push_back(s[0]);
1207 ev_shape.push_back(s[1]);
1208 ev_shape.push_back(s[2]);
1209 ev_shape.push_back(s[3]);
1210 Data ev(0.,ev_shape,getFunctionSpace());
1211 ev.typeMatchRight(*this);
1212 m_data->nonsymmetric(ev.m_data.get());
1213 return ev;
1214 }
1215 else {
1216 throw DataException("Error - Data::nonsymmetric can only be calculated for rank 2 or 4 object.");
1217 }
1218 }
1219
1220 Data
1221 Data::trace(int axis_offset) const
1222 {
1223 DataArrayView::ShapeType s=getDataPointShape();
1224 if (getDataPointRank()==2) {
1225 DataArrayView::ShapeType ev_shape;
1226 Data ev(0.,ev_shape,getFunctionSpace());
1227 ev.typeMatchRight(*this);
1228 m_data->trace(ev.m_data.get(), axis_offset);
1229 return ev;
1230 }
1231 if (getDataPointRank()==3) {
1232 DataArrayView::ShapeType ev_shape;
1233 if (axis_offset==0) {
1234 int s2=s[2];
1235 ev_shape.push_back(s2);
1236 }
1237 else if (axis_offset==1) {
1238 int s0=s[0];
1239 ev_shape.push_back(s0);
1240 }
1241 Data ev(0.,ev_shape,getFunctionSpace());
1242 ev.typeMatchRight(*this);
1243 m_data->trace(ev.m_data.get(), axis_offset);
1244 return ev;
1245 }
1246 if (getDataPointRank()==4) {
1247 DataArrayView::ShapeType ev_shape;
1248 if (axis_offset==0) {
1249 ev_shape.push_back(s[2]);
1250 ev_shape.push_back(s[3]);
1251 }
1252 else if (axis_offset==1) {
1253 ev_shape.push_back(s[0]);
1254 ev_shape.push_back(s[3]);
1255 }
1256 else if (axis_offset==2) {
1257 ev_shape.push_back(s[0]);
1258 ev_shape.push_back(s[1]);
1259 }
1260 Data ev(0.,ev_shape,getFunctionSpace());
1261 ev.typeMatchRight(*this);
1262 m_data->trace(ev.m_data.get(), axis_offset);
1263 return ev;
1264 }
1265 else {
1266 throw DataException("Error - Data::trace can only be calculated for rank 2, 3 or 4 object.");
1267 }
1268 }
1269
1270 Data
1271 Data::transpose(int axis_offset) const
1272 {
1273 DataArrayView::ShapeType s=getDataPointShape();
1274 DataArrayView::ShapeType ev_shape;
1275 // Here's the equivalent of python s_out=s[axis_offset:]+s[:axis_offset]
1276 // which goes thru all shape vector elements starting with axis_offset (at index=rank wrap around to 0)
1277 int rank=getDataPointRank();
1278 if (axis_offset<0 || axis_offset>rank) {
1279 throw DataException("Error - Data::transpose must have 0 <= axis_offset <= rank=" + rank);
1280 }
1281 for (int i=0; i<rank; i++) {
1282 int index = (axis_offset+i)%rank;
1283 ev_shape.push_back(s[index]); // Append to new shape
1284 }
1285 Data ev(0.,ev_shape,getFunctionSpace());
1286 ev.typeMatchRight(*this);
1287 m_data->transpose(ev.m_data.get(), axis_offset);
1288 return ev;
1289 }
1290
1291 Data
1292 Data::eigenvalues() const
1293 {
1294 // check input
1295 DataArrayView::ShapeType s=getDataPointShape();
1296 if (getDataPointRank()!=2)
1297 throw DataException("Error - Data::eigenvalues can only be calculated for rank 2 object.");
1298 if(s[0] != s[1])
1299 throw DataException("Error - Data::eigenvalues can only be calculated for object with equal first and second dimension.");
1300 // create return
1301 DataArrayView::ShapeType ev_shape(1,s[0]);
1302 Data ev(0.,ev_shape,getFunctionSpace());
1303 ev.typeMatchRight(*this);
1304 m_data->eigenvalues(ev.m_data.get());
1305 return ev;
1306 }
1307
1308 const boost::python::tuple
1309 Data::eigenvalues_and_eigenvectors(const double tol) const
1310 {
1311 DataArrayView::ShapeType s=getDataPointShape();
1312 if (getDataPointRank()!=2)
1313 throw DataException("Error - Data::eigenvalues and eigenvectors can only be calculated for rank 2 object.");
1314 if(s[0] != s[1])
1315 throw DataException("Error - Data::eigenvalues and eigenvectors can only be calculated for object with equal first and second dimension.");
1316 // create return
1317 DataArrayView::ShapeType ev_shape(1,s[0]);
1318 Data ev(0.,ev_shape,getFunctionSpace());
1319 ev.typeMatchRight(*this);
1320 DataArrayView::ShapeType V_shape(2,s[0]);
1321 Data V(0.,V_shape,getFunctionSpace());
1322 V.typeMatchRight(*this);
1323 m_data->eigenvalues_and_eigenvectors(ev.m_data.get(),V.m_data.get(),tol);
1324 return make_tuple(boost::python::object(ev),boost::python::object(V));
1325 }
1326
1327 const boost::python::tuple
1328 Data::minGlobalDataPoint() const
1329 {
1330 // NB: calc_minGlobalDataPoint( had to be split off from minGlobalDataPoint( as boost::make_tuple causes an
1331 // abort (for unknown reasons) if there are openmp directives with it in the
1332 // surrounding function
1333
1334 int DataPointNo;
1335 int ProcNo;
1336 calc_minGlobalDataPoint(ProcNo,DataPointNo);
1337 return make_tuple(ProcNo,DataPointNo);
1338 }
1339
1340 void
1341 Data::calc_minGlobalDataPoint(int& ProcNo,
1342 int& DataPointNo) const
1343 {
1344 int i,j;
1345 int lowi=0,lowj=0;
1346 double min=numeric_limits<double>::max();
1347
1348 Data temp=minval();
1349
1350 int numSamples=temp.getNumSamples();
1351 int numDPPSample=temp.getNumDataPointsPerSample();
1352
1353 double next,local_min;
1354 int local_lowi,local_lowj;
1355
1356 #pragma omp parallel private(next,local_min,local_lowi,local_lowj)
1357 {
1358 local_min=min;
1359 #pragma omp for private(i,j) schedule(static)
1360 for (i=0; i<numSamples; i++) {
1361 for (j=0; j<numDPPSample; j++) {
1362 next=temp.getDataPoint(i,j)();
1363 if (next<local_min) {
1364 local_min=next;
1365 local_lowi=i;
1366 local_lowj=j;
1367 }
1368 }
1369 }
1370 #pragma omp critical
1371 if (local_min<min) {
1372 min=local_min;
1373 lowi=local_lowi;
1374 lowj=local_lowj;
1375 }
1376 }
1377
1378 #ifdef PASO_MPI
1379 // determine the processor on which the minimum occurs
1380 next = temp.getDataPoint(lowi,lowj)();
1381 int lowProc = 0;
1382 double *globalMins = new double[get_MPISize()+1];
1383 int error = MPI_Gather ( &next, 1, MPI_DOUBLE, globalMins, 1, MPI_DOUBLE, 0, get_MPIComm() );
1384
1385 if( get_MPIRank()==0 ){
1386 next = globalMins[lowProc];
1387 for( i=1; i<get_MPISize(); i++ )
1388 if( next>globalMins[i] ){
1389 lowProc = i;
1390 next = globalMins[i];
1391 }
1392 }
1393 MPI_Bcast( &lowProc, 1, MPI_DOUBLE, 0, get_MPIComm() );
1394
1395 delete [] globalMins;
1396 ProcNo = lowProc;
1397 #else
1398 ProcNo = 0;
1399 #endif
1400 DataPointNo = lowj + lowi * numDPPSample;
1401 }
1402
1403 void
1404 Data::saveDX(std::string fileName) const
1405 {
1406 boost::python::dict args;
1407 args["data"]=boost::python::object(this);
1408 getDomain().saveDX(fileName,args);
1409 return;
1410 }
1411
1412 void
1413 Data::saveVTK(std::string fileName) const
1414 {
1415 boost::python::dict args;
1416 args["data"]=boost::python::object(this);
1417 getDomain().saveVTK(fileName,args);
1418 return;
1419 }
1420
1421 Data&
1422 Data::operator+=(const Data& right)
1423 {
1424 if (isProtected()) {
1425 throw DataException("Error - attempt to update protected Data object.");
1426 }
1427 binaryOp(right,plus<double>());
1428 return (*this);
1429 }
1430
1431 Data&
1432 Data::operator+=(const boost::python::object& right)
1433 {
1434 Data tmp(right,getFunctionSpace(),false);
1435 binaryOp(tmp,plus<double>());
1436 return (*this);
1437 }
1438 Data&
1439 Data::operator=(const Data& other)
1440 {
1441 copy(other);
1442 return (*this);
1443 }
1444
1445 Data&
1446 Data::operator-=(const Data& right)
1447 {
1448 if (isProtected()) {
1449 throw DataException("Error - attempt to update protected Data object.");
1450 }
1451 binaryOp(right,minus<double>());
1452 return (*this);
1453 }
1454
1455 Data&
1456 Data::operator-=(const boost::python::object& right)
1457 {
1458 Data tmp(right,getFunctionSpace(),false);
1459 binaryOp(tmp,minus<double>());
1460 return (*this);
1461 }
1462
1463 Data&
1464 Data::operator*=(const Data& right)
1465 {
1466 if (isProtected()) {
1467 throw DataException("Error - attempt to update protected Data object.");
1468 }
1469 binaryOp(right,multiplies<double>());
1470 return (*this);
1471 }
1472
1473 Data&
1474 Data::operator*=(const boost::python::object& right)
1475 {
1476 Data tmp(right,getFunctionSpace(),false);
1477 binaryOp(tmp,multiplies<double>());
1478 return (*this);
1479 }
1480
1481 Data&
1482 Data::operator/=(const Data& right)
1483 {
1484 if (isProtected()) {
1485 throw DataException("Error - attempt to update protected Data object.");
1486 }
1487 binaryOp(right,divides<double>());
1488 return (*this);
1489 }
1490
1491 Data&
1492 Data::operator/=(const boost::python::object& right)
1493 {
1494 Data tmp(right,getFunctionSpace(),false);
1495 binaryOp(tmp,divides<double>());
1496 return (*this);
1497 }
1498
1499 Data
1500 Data::rpowO(const boost::python::object& left) const
1501 {
1502 Data left_d(left,*this);
1503 return left_d.powD(*this);
1504 }
1505
1506 Data
1507 Data::powO(const boost::python::object& right) const
1508 {
1509 Data tmp(right,getFunctionSpace(),false);
1510 return powD(tmp);
1511 }
1512
1513 Data
1514 Data::powD(const Data& right) const
1515 {
1516 Data result;
1517 if (getDataPointRank()<right.getDataPointRank()) {
1518 result.copy(right);
1519 result.binaryOp(*this,escript::rpow);
1520 } else {
1521 result.copy(*this);
1522 result.binaryOp(right,(Data::BinaryDFunPtr)::pow);
1523 }
1524 return result;
1525 }
1526
1527
1528 //
1529 // NOTE: It is essential to specify the namespace this operator belongs to
1530 Data
1531 escript::operator+(const Data& left, const Data& right)
1532 {
1533 return C_TensorBinaryOperation(left, right, plus<double>());
1534 }
1535
1536 //
1537 // NOTE: It is essential to specify the namespace this operator belongs to
1538 Data
1539 escript::operator-(const Data& left, const Data& right)
1540 {
1541 return C_TensorBinaryOperation(left, right, minus<double>());
1542 }
1543
1544 //
1545 // NOTE: It is essential to specify the namespace this operator belongs to
1546 Data
1547 escript::operator*(const Data& left, const Data& right)
1548 {
1549 return C_TensorBinaryOperation(left, right, multiplies<double>());
1550 }
1551
1552 //
1553 // NOTE: It is essential to specify the namespace this operator belongs to
1554 Data
1555 escript::operator/(const Data& left, const Data& right)
1556 {
1557 return C_TensorBinaryOperation(left, right, divides<double>());
1558 }
1559
1560 //
1561 // NOTE: It is essential to specify the namespace this operator belongs to
1562 Data
1563 escript::operator+(const Data& left, const boost::python::object& right)
1564 {
1565 return left+Data(right,left.getFunctionSpace(),false);
1566 }
1567
1568 //
1569 // NOTE: It is essential to specify the namespace this operator belongs to
1570 Data
1571 escript::operator-(const Data& left, const boost::python::object& right)
1572 {
1573 return left-Data(right,left.getFunctionSpace(),false);
1574 }
1575
1576 //
1577 // NOTE: It is essential to specify the namespace this operator belongs to
1578 Data
1579 escript::operator*(const Data& left, const boost::python::object& right)
1580 {
1581 return left*Data(right,left.getFunctionSpace(),false);
1582 }
1583
1584 //
1585 // NOTE: It is essential to specify the namespace this operator belongs to
1586 Data
1587 escript::operator/(const Data& left, const boost::python::object& right)
1588 {
1589 return left/Data(right,left.getFunctionSpace(),false);
1590 }
1591
1592 //
1593 // NOTE: It is essential to specify the namespace this operator belongs to
1594 Data
1595 escript::operator+(const boost::python::object& left, const Data& right)
1596 {
1597 return Data(left,right.getFunctionSpace(),false)+right;
1598 }
1599
1600 //
1601 // NOTE: It is essential to specify the namespace this operator belongs to
1602 Data
1603 escript::operator-(const boost::python::object& left, const Data& right)
1604 {
1605 return Data(left,right.getFunctionSpace(),false)-right;
1606 }
1607
1608 //
1609 // NOTE: It is essential to specify the namespace this operator belongs to
1610 Data
1611 escript::operator*(const boost::python::object& left, const Data& right)
1612 {
1613 return Data(left,right.getFunctionSpace(),false)*right;
1614 }
1615
1616 //
1617 // NOTE: It is essential to specify the namespace this operator belongs to
1618 Data
1619 escript::operator/(const boost::python::object& left, const Data& right)
1620 {
1621 return Data(left,right.getFunctionSpace(),false)/right;
1622 }
1623
1624 //
1625 //bool escript::operator==(const Data& left, const Data& right)
1626 //{
1627 // /*
1628 // NB: this operator does very little at this point, and isn't to
1629 // be relied on. Requires further implementation.
1630 // */
1631 //
1632 // bool ret;
1633 //
1634 // if (left.isEmpty()) {
1635 // if(!right.isEmpty()) {
1636 // ret = false;
1637 // } else {
1638 // ret = true;
1639 // }
1640 // }
1641 //
1642 // if (left.isConstant()) {
1643 // if(!right.isConstant()) {
1644 // ret = false;
1645 // } else {
1646 // ret = true;
1647 // }
1648 // }
1649 //
1650 // if (left.isTagged()) {
1651 // if(!right.isTagged()) {
1652 // ret = false;
1653 // } else {
1654 // ret = true;
1655 // }
1656 // }
1657 //
1658 // if (left.isExpanded()) {
1659 // if(!right.isExpanded()) {
1660 // ret = false;
1661 // } else {
1662 // ret = true;
1663 // }
1664 // }
1665 //
1666 // return ret;
1667 //}
1668
1669 /* TODO */
1670 /* global reduction */
1671 Data
1672 Data::getItem(const boost::python::object& key) const
1673 {
1674 const DataArrayView& view=getPointDataView();
1675
1676 DataArrayView::RegionType slice_region=view.getSliceRegion(key);
1677
1678 if (slice_region.size()!=view.getRank()) {
1679 throw DataException("Error - slice size does not match Data rank.");
1680 }
1681
1682 return getSlice(slice_region);
1683 }
1684
1685 /* TODO */
1686 /* global reduction */
1687 Data
1688 Data::getSlice(const DataArrayView::RegionType& region) const
1689 {
1690 return Data(*this,region);
1691 }
1692
1693 /* TODO */
1694 /* global reduction */
1695 void
1696 Data::setItemO(const boost::python::object& key,
1697 const boost::python::object& value)
1698 {
1699 Data tempData(value,getFunctionSpace());
1700 setItemD(key,tempData);
1701 }
1702
1703 void
1704 Data::setItemD(const boost::python::object& key,
1705 const Data& value)
1706 {
1707 const DataArrayView& view=getPointDataView();
1708
1709 DataArrayView::RegionType slice_region=view.getSliceRegion(key);
1710 if (slice_region.size()!=view.getRank()) {
1711 throw DataException("Error - slice size does not match Data rank.");
1712 }
1713 if (getFunctionSpace()!=value.getFunctionSpace()) {
1714 setSlice(Data(value,getFunctionSpace()),slice_region);
1715 } else {
1716 setSlice(value,slice_region);
1717 }
1718 }
1719
1720 void
1721 Data::setSlice(const Data& value,
1722 const DataArrayView::RegionType& region)
1723 {
1724 if (isProtected()) {
1725 throw DataException("Error - attempt to update protected Data object.");
1726 }
1727 Data tempValue(value);
1728 typeMatchLeft(tempValue);
1729 typeMatchRight(tempValue);
1730 m_data->setSlice(tempValue.m_data.get(),region);
1731 }
1732
1733 void
1734 Data::typeMatchLeft(Data& right) const
1735 {
1736 if (isExpanded()){
1737 right.expand();
1738 } else if (isTagged()) {
1739 if (right.isConstant()) {
1740 right.tag();
1741 }
1742 }
1743 }
1744
1745 void
1746 Data::typeMatchRight(const Data& right)
1747 {
1748 if (isTagged()) {
1749 if (right.isExpanded()) {
1750 expand();
1751 }
1752 } else if (isConstant()) {
1753 if (right.isExpanded()) {
1754 expand();
1755 } else if (right.isTagged()) {
1756 tag();
1757 }
1758 }
1759 }
1760
1761 void
1762 Data::setTaggedValueByName(std::string name,
1763 const boost::python::object& value)
1764 {
1765 if (getFunctionSpace().getDomain().isValidTagName(name)) {
1766 int tagKey=getFunctionSpace().getDomain().getTag(name);
1767 setTaggedValue(tagKey,value);
1768 }
1769 }
1770 void
1771 Data::setTaggedValue(int tagKey,
1772 const boost::python::object& value)
1773 {
1774 if (isProtected()) {
1775 throw DataException("Error - attempt to update protected Data object.");
1776 }
1777 //
1778 // Ensure underlying data object is of type DataTagged
1779 tag();
1780
1781 if (!isTagged()) {
1782 throw DataException("Error - DataTagged conversion failed!!");
1783 }
1784
1785 numeric::array asNumArray(value);
1786
1787
1788 // extract the shape of the numarray
1789 DataArrayView::ShapeType tempShape;
1790 for (int i=0; i < asNumArray.getrank(); i++) {
1791 tempShape.push_back(extract<int>(asNumArray.getshape()[i]));
1792 }
1793
1794 // get the space for the data vector
1795 int len = DataArrayView::noValues(tempShape);
1796 DataVector temp_data(len, 0.0, len);
1797 DataArrayView temp_dataView(temp_data, tempShape);
1798 temp_dataView.copy(asNumArray);
1799
1800 //
1801 // Call DataAbstract::setTaggedValue
1802 m_data->setTaggedValue(tagKey,temp_dataView);
1803 }
1804
1805 void
1806 Data::setTaggedValueFromCPP(int tagKey,
1807 const DataArrayView& value)
1808 {
1809 if (isProtected()) {
1810 throw DataException("Error - attempt to update protected Data object.");
1811 }
1812 //
1813 // Ensure underlying data object is of type DataTagged
1814 tag();
1815
1816 if (!isTagged()) {
1817 throw DataException("Error - DataTagged conversion failed!!");
1818 }
1819
1820 //
1821 // Call DataAbstract::setTaggedValue
1822 m_data->setTaggedValue(tagKey,value);
1823 }
1824
1825 int
1826 Data::getTagNumber(int dpno)
1827 {
1828 return m_data->getTagNumber(dpno);
1829 }
1830
1831 void
1832 Data::archiveData(const std::string fileName)
1833 {
1834 cout << "Archiving Data object to: " << fileName << endl;
1835
1836 //
1837 // Determine type of this Data object
1838 int dataType = -1;
1839
1840 if (isEmpty()) {
1841 dataType = 0;
1842 cout << "\tdataType: DataEmpty" << endl;
1843 }
1844 if (isConstant()) {
1845 dataType = 1;
1846 cout << "\tdataType: DataConstant" << endl;
1847 }
1848 if (isTagged()) {
1849 dataType = 2;
1850 cout << "\tdataType: DataTagged" << endl;
1851 }
1852 if (isExpanded()) {
1853 dataType = 3;
1854 cout << "\tdataType: DataExpanded" << endl;
1855 }
1856
1857 if (dataType == -1) {
1858 throw DataException("archiveData Error: undefined dataType");
1859 }
1860
1861 //
1862 // Collect data items common to all Data types
1863 int noSamples = getNumSamples();
1864 int noDPPSample = getNumDataPointsPerSample();
1865 int functionSpaceType = getFunctionSpace().getTypeCode();
1866 int dataPointRank = getDataPointRank();
1867 int dataPointSize = getDataPointSize();
1868 int dataLength = getLength();
1869 DataArrayView::ShapeType dataPointShape = getDataPointShape();
1870 vector<int> referenceNumbers(noSamples);
1871 for (int sampleNo=0; sampleNo<noSamples; sampleNo++) {
1872 referenceNumbers[sampleNo] = getFunctionSpace().getReferenceIDOfSample(sampleNo);
1873 }
1874 vector<int> tagNumbers(noSamples);
1875 if (isTagged()) {
1876 for (int sampleNo=0; sampleNo<noSamples; sampleNo++) {
1877 tagNumbers[sampleNo] = getFunctionSpace().getTagFromSampleNo(sampleNo);
1878 }
1879 }
1880
1881 cout << "\tnoSamples: " << noSamples << " noDPPSample: " << noDPPSample << endl;
1882 cout << "\tfunctionSpaceType: " << functionSpaceType << endl;
1883 cout << "\trank: " << dataPointRank << " size: " << dataPointSize << " length: " << dataLength << endl;
1884
1885 //
1886 // Flatten Shape to an array of integers suitable for writing to file
1887 int flatShape[4] = {0,0,0,0};
1888 cout << "\tshape: < ";
1889 for (int dim=0; dim<dataPointRank; dim++) {
1890 flatShape[dim] = dataPointShape[dim];
1891 cout << dataPointShape[dim] << " ";
1892 }
1893 cout << ">" << endl;
1894
1895 //
1896 // Open archive file
1897 ofstream archiveFile;
1898 archiveFile.open(fileName.data(), ios::out);
1899
1900 if (!archiveFile.good()) {
1901 throw DataException("archiveData Error: problem opening archive file");
1902 }
1903
1904 //
1905 // Write common data items to archive file
1906 archiveFile.write(reinterpret_cast<char *>(&dataType),sizeof(int));
1907 archiveFile.write(reinterpret_cast<char *>(&noSamples),sizeof(int));
1908 archiveFile.write(reinterpret_cast<char *>(&noDPPSample),sizeof(int));
1909 archiveFile.write(reinterpret_cast<char *>(&functionSpaceType),sizeof(int));
1910 archiveFile.write(reinterpret_cast<char *>(&dataPointRank),sizeof(int));
1911 archiveFile.write(reinterpret_cast<char *>(&dataPointSize),sizeof(int));
1912 archiveFile.write(reinterpret_cast<char *>(&dataLength),sizeof(int));
1913 for (int dim = 0; dim < 4; dim++) {
1914 archiveFile.write(reinterpret_cast<char *>(&flatShape[dim]),sizeof(int));
1915 }
1916 for (int sampleNo=0; sampleNo<noSamples; sampleNo++) {
1917 archiveFile.write(reinterpret_cast<char *>(&referenceNumbers[sampleNo]),sizeof(int));
1918 }
1919 if (isTagged()) {
1920 for (int sampleNo=0; sampleNo<noSamples; sampleNo++) {
1921 archiveFile.write(reinterpret_cast<char *>(&tagNumbers[sampleNo]),sizeof(int));
1922 }
1923 }
1924
1925 if (!archiveFile.good()) {
1926 throw DataException("archiveData Error: problem writing to archive file");
1927 }
1928
1929 //
1930 // Archive underlying data values for each Data type
1931 int noValues;
1932 switch (dataType) {
1933 case 0:
1934 // DataEmpty
1935 noValues = 0;
1936 archiveFile.write(reinterpret_cast<char *>(&noValues),sizeof(int));
1937 cout << "\tnoValues: " << noValues << endl;
1938 break;
1939 case 1:
1940 // DataConstant
1941 noValues = m_data->getLength();
1942 archiveFile.write(reinterpret_cast<char *>(&noValues),sizeof(int));
1943 cout << "\tnoValues: " << noValues << endl;
1944 if (m_data->archiveData(archiveFile,noValues)) {
1945 throw DataException("archiveData Error: problem writing data to archive file");
1946 }
1947 break;
1948 case 2:
1949 // DataTagged
1950 noValues = m_data->getLength();
1951 archiveFile.write(reinterpret_cast<char *>(&noValues),sizeof(int));
1952 cout << "\tnoValues: " << noValues << endl;
1953 if (m_data->archiveData(archiveFile,noValues)) {
1954 throw DataException("archiveData Error: problem writing data to archive file");
1955 }
1956 break;
1957 case 3:
1958 // DataExpanded
1959 noValues = m_data->getLength();
1960 archiveFile.write(reinterpret_cast<char *>(&noValues),sizeof(int));
1961 cout << "\tnoValues: " << noValues << endl;
1962 if (m_data->archiveData(archiveFile,noValues)) {
1963 throw DataException("archiveData Error: problem writing data to archive file");
1964 }
1965 break;
1966 }
1967
1968 if (!archiveFile.good()) {
1969 throw DataException("archiveData Error: problem writing data to archive file");
1970 }
1971
1972 //
1973 // Close archive file
1974 archiveFile.close();
1975
1976 if (!archiveFile.good()) {
1977 throw DataException("archiveData Error: problem closing archive file");
1978 }
1979
1980 }
1981
1982 void
1983 Data::extractData(const std::string fileName,
1984 const FunctionSpace& fspace)
1985 {
1986 //
1987 // Can only extract Data to an object which is initially DataEmpty
1988 if (!isEmpty()) {
1989 throw DataException("extractData Error: can only extract to DataEmpty object");
1990 }
1991
1992 cout << "Extracting Data object from: " << fileName << endl;
1993
1994 int dataType;
1995 int noSamples;
1996 int noDPPSample;
1997 int functionSpaceType;
1998 int dataPointRank;
1999 int dataPointSize;
2000 int dataLength;
2001 DataArrayView::ShapeType dataPointShape;
2002 int flatShape[4];
2003
2004 //
2005 // Open the archive file
2006 ifstream archiveFile;
2007 archiveFile.open(fileName.data(), ios::in);
2008
2009 if (!archiveFile.good()) {
2010 throw DataException("extractData Error: problem opening archive file");
2011 }
2012
2013 //
2014 // Read common data items from archive file
2015 archiveFile.read(reinterpret_cast<char *>(&dataType),sizeof(int));
2016 archiveFile.read(reinterpret_cast<char *>(&noSamples),sizeof(int));
2017 archiveFile.read(reinterpret_cast<char *>(&noDPPSample),sizeof(int));
2018 archiveFile.read(reinterpret_cast<char *>(&functionSpaceType),sizeof(int));
2019 archiveFile.read(reinterpret_cast<char *>(&dataPointRank),sizeof(int));
2020 archiveFile.read(reinterpret_cast<char *>(&dataPointSize),sizeof(int));
2021 archiveFile.read(reinterpret_cast<char *>(&dataLength),sizeof(int));
2022 for (int dim = 0; dim < 4; dim++) {
2023 archiveFile.read(reinterpret_cast<char *>(&flatShape[dim]),sizeof(int));
2024 if (flatShape[dim]>0) {
2025 dataPointShape.push_back(flatShape[dim]);
2026 }
2027 }
2028 vector<int> referenceNumbers(noSamples);
2029 for (int sampleNo=0; sampleNo<noSamples; sampleNo++) {
2030 archiveFile.read(reinterpret_cast<char *>(&referenceNumbers[sampleNo]),sizeof(int));
2031 }
2032 vector<int> tagNumbers(noSamples);
2033 if (dataType==2) {
2034 for (int sampleNo=0; sampleNo<noSamples; sampleNo++) {
2035 archiveFile.read(reinterpret_cast<char *>(&tagNumbers[sampleNo]),sizeof(int));
2036 }
2037 }
2038
2039 if (!archiveFile.good()) {
2040 throw DataException("extractData Error: problem reading from archive file");
2041 }
2042
2043 //
2044 // Verify the values just read from the archive file
2045 switch (dataType) {
2046 case 0:
2047 cout << "\tdataType: DataEmpty" << endl;
2048 break;
2049 case 1:
2050 cout << "\tdataType: DataConstant" << endl;
2051 break;
2052 case 2:
2053 cout << "\tdataType: DataTagged" << endl;
2054 break;
2055 case 3:
2056 cout << "\tdataType: DataExpanded" << endl;
2057 break;
2058 default:
2059 throw DataException("extractData Error: undefined dataType read from archive file");
2060 break;
2061 }
2062
2063 cout << "\tnoSamples: " << noSamples << " noDPPSample: " << noDPPSample << endl;
2064 cout << "\tfunctionSpaceType: " << functionSpaceType << endl;
2065 cout << "\trank: " << dataPointRank << " size: " << dataPointSize << " length: " << dataLength << endl;
2066 cout << "\tshape: < ";
2067 for (int dim = 0; dim < dataPointRank; dim++) {
2068 cout << dataPointShape[dim] << " ";
2069 }
2070 cout << ">" << endl;
2071
2072 //
2073 // Verify that supplied FunctionSpace object is compatible with this Data object.
2074 if ( (fspace.getTypeCode()!=functionSpaceType) ||
2075 (fspace.getNumSamples()!=noSamples) ||
2076 (fspace.getNumDPPSample()!=noDPPSample)
2077 ) {
2078 throw DataException("extractData Error: incompatible FunctionSpace");
2079 }
2080 for (int sampleNo=0; sampleNo<noSamples; sampleNo++) {
2081 if (referenceNumbers[sampleNo] != fspace.getReferenceIDOfSample(sampleNo)) {
2082 throw DataException("extractData Error: incompatible FunctionSpace");
2083 }
2084 }
2085 if (dataType==2) {
2086 for (int sampleNo=0; sampleNo<noSamples; sampleNo++) {
2087 if (tagNumbers[sampleNo] != fspace.getTagFromSampleNo(sampleNo)) {
2088 throw DataException("extractData Error: incompatible FunctionSpace");
2089 }
2090 }
2091 }
2092
2093 //
2094 // Construct a DataVector to hold underlying data values
2095 DataVector dataVec(dataLength);
2096
2097 //
2098 // Load this DataVector with the appropriate values
2099 int noValues;
2100 archiveFile.read(reinterpret_cast<char *>(&noValues),sizeof(int));
2101 cout << "\tnoValues: " << noValues << endl;
2102 switch (dataType) {
2103 case 0:
2104 // DataEmpty
2105 if (noValues != 0) {
2106 throw DataException("extractData Error: problem reading data from archive file");
2107 }
2108 break;
2109 case 1:
2110 // DataConstant
2111 if (dataVec.extractData(archiveFile,noValues)) {
2112 throw DataException("extractData Error: problem reading data from archive file");
2113 }
2114 break;
2115 case 2:
2116 // DataTagged
2117 if (dataVec.extractData(archiveFile,noValues)) {
2118 throw DataException("extractData Error: problem reading data from archive file");
2119 }
2120 break;
2121 case 3:
2122 // DataExpanded
2123 if (dataVec.extractData(archiveFile,noValues)) {
2124 throw DataException("extractData Error: problem reading data from archive file");
2125 }
2126 break;
2127 }
2128
2129 if (!archiveFile.good()) {
2130 throw DataException("extractData Error: problem reading from archive file");
2131 }
2132
2133 //
2134 // Close archive file
2135 archiveFile.close();
2136
2137 if (!archiveFile.good()) {
2138 throw DataException("extractData Error: problem closing archive file");
2139 }
2140
2141 //
2142 // Construct an appropriate Data object
2143 DataAbstract* tempData;
2144 switch (dataType) {
2145 case 0:
2146 // DataEmpty
2147 tempData=new DataEmpty();
2148 break;
2149 case 1:
2150 // DataConstant
2151 tempData=new DataConstant(fspace,dataPointShape,dataVec);
2152 break;
2153 case 2:
2154 // DataTagged
2155 tempData=new DataTagged(fspace,dataPointShape,tagNumbers,dataVec);
2156 break;
2157 case 3:
2158 // DataExpanded
2159 tempData=new DataExpanded(fspace,dataPointShape,dataVec);
2160 break;
2161 }
2162 shared_ptr<DataAbstract> temp_data(tempData);
2163 m_data=temp_data;
2164 }
2165
2166 ostream& escript::operator<<(ostream& o, const Data& data)
2167 {
2168 o << data.toString();
2169 return o;
2170 }
2171
2172 Data
2173 escript::C_GeneralTensorProduct(Data& arg_0,
2174 Data& arg_1,
2175 int axis_offset,
2176 int transpose)
2177 {
2178 // General tensor product: res(SL x SR) = arg_0(SL x SM) * arg_1(SM x SR)
2179 // SM is the product of the last axis_offset entries in arg_0.getShape().
2180
2181 // Interpolate if necessary and find an appropriate function space
2182 Data arg_0_Z, arg_1_Z;
2183 if (arg_0.getFunctionSpace()!=arg_1.getFunctionSpace()) {
2184 if (arg_0.probeInterpolation(arg_1.getFunctionSpace())) {
2185 arg_0_Z = arg_0.interpolate(arg_1.getFunctionSpace());
2186 arg_1_Z = Data(arg_1);
2187 }
2188 else if (arg_1.probeInterpolation(arg_0.getFunctionSpace())) {
2189 arg_1_Z=arg_1.interpolate(arg_0.getFunctionSpace());
2190 arg_0_Z =Data(arg_0);
2191 }
2192 else {
2193 throw DataException("Error - C_GeneralTensorProduct: arguments have incompatible function spaces.");
2194 }
2195 } else {
2196 arg_0_Z = Data(arg_0);
2197 arg_1_Z = Data(arg_1);
2198 }
2199 // Get rank and shape of inputs
2200 int rank0 = arg_0_Z.getDataPointRank();
2201 int rank1 = arg_1_Z.getDataPointRank();
2202 DataArrayView::ShapeType shape0 = arg_0_Z.getDataPointShape();
2203 DataArrayView::ShapeType shape1 = arg_1_Z.getDataPointShape();
2204
2205 // Prepare for the loops of the product and verify compatibility of shapes
2206 int start0=0, start1=0;
2207 if (transpose == 0) {}
2208 else if (transpose == 1) { start0 = axis_offset; }
2209 else if (transpose == 2) { start1 = rank1-axis_offset; }
2210 else { throw DataException("C_GeneralTensorProduct: Error - transpose should be 0, 1 or 2"); }
2211
2212 // Adjust the shapes for transpose
2213 DataArrayView::ShapeType tmpShape0;
2214 DataArrayView::ShapeType tmpShape1;
2215 for (int i=0; i<rank0; i++) { tmpShape0.push_back( shape0[(i+start0)%rank0] ); }
2216 for (int i=0; i<rank1; i++) { tmpShape1.push_back( shape1[(i+start1)%rank1] ); }
2217
2218 #if 0
2219 // For debugging: show shape after transpose
2220 char tmp[100];
2221 std::string shapeStr;
2222 shapeStr = "(";
2223 for (int i=0; i<rank0; i++) { sprintf(tmp, "%d,", tmpShape0[i]); shapeStr += tmp; }
2224 shapeStr += ")";
2225 cout << "C_GeneralTensorProduct: Shape of arg0 is " << shapeStr << endl;
2226 shapeStr = "(";
2227 for (int i=0; i<rank1; i++) { sprintf(tmp, "%d,", tmpShape1[i]); shapeStr += tmp; }
2228 shapeStr += ")";
2229 cout << "C_GeneralTensorProduct: Shape of arg1 is " << shapeStr << endl;
2230 #endif
2231
2232 // Prepare for the loops of the product
2233 int SL=1, SM=1, SR=1;
2234 for (int i=0; i<rank0-axis_offset; i++) {
2235 SL *= tmpShape0[i];
2236 }
2237 for (int i=rank0-axis_offset; i<rank0; i++) {
2238 if (tmpShape0[i] != tmpShape1[i-(rank0-axis_offset)]) {
2239 throw DataException("C_GeneralTensorProduct: Error - incompatible shapes");
2240 }
2241 SM *= tmpShape0[i];
2242 }
2243 for (int i=axis_offset; i<rank1; i++) {
2244 SR *= tmpShape1[i];
2245 }
2246
2247 // Define the shape of the output
2248 DataArrayView::ShapeType shape2;
2249 for (int i=0; i<rank0-axis_offset; i++) { shape2.push_back(tmpShape0[i]); } // First part of arg_0_Z
2250 for (int i=axis_offset; i<rank1; i++) { shape2.push_back(tmpShape1[i]); } // Last part of arg_1_Z
2251
2252 // Declare output Data object
2253 Data res;
2254
2255 if (arg_0_Z.isConstant() && arg_1_Z.isConstant()) {
2256 res = Data(0.0, shape2, arg_1_Z.getFunctionSpace()); // DataConstant output
2257 double *ptr_0 = &((arg_0_Z.getPointDataView().getData())[0]);
2258 double *ptr_1 = &((arg_1_Z.getPointDataView().getData())[0]);
2259 double *ptr_2 = &((res.getPointDataView().getData())[0]);
2260 matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2261 }
2262 else if (arg_0_Z.isConstant() && arg_1_Z.isTagged()) {
2263
2264 // Prepare the DataConstant input
2265 DataConstant* tmp_0=dynamic_cast<DataConstant*>(arg_0_Z.borrowData());
2266 if (tmp_0==0) { throw DataException("GTP Programming error - casting to DataConstant."); }
2267
2268 // Borrow DataTagged input from Data object
2269 DataTagged* tmp_1=dynamic_cast<DataTagged*>(arg_1_Z.borrowData());
2270 if (tmp_1==0) { throw DataException("GTP_1 Programming error - casting to DataTagged."); }
2271
2272 // Prepare a DataTagged output 2
2273 res = Data(0.0, shape2, arg_1_Z.getFunctionSpace()); // DataTagged output
2274 res.tag();
2275 DataTagged* tmp_2=dynamic_cast<DataTagged*>(res.borrowData());
2276 if (tmp_2==0) { throw DataException("GTP Programming error - casting to DataTagged."); }
2277
2278 // Prepare offset into DataConstant
2279 int offset_0 = tmp_0->getPointOffset(0,0);
2280 double *ptr_0 = &((arg_0_Z.getPointDataView().getData())[offset_0]);
2281 // Get the views
2282 DataArrayView view_1 = tmp_1->getDefaultValue();
2283 DataArrayView view_2 = tmp_2->getDefaultValue();
2284 // Get the pointers to the actual data
2285 double *ptr_1 = &((view_1.getData())[0]);
2286 double *ptr_2 = &((view_2.getData())[0]);
2287 // Compute an MVP for the default
2288 matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2289 // Compute an MVP for each tag
2290 const DataTagged::DataMapType& lookup_1=tmp_1->getTagLookup();
2291 DataTagged::DataMapType::const_iterator i; // i->first is a tag, i->second is an offset into memory
2292 for (i=lookup_1.begin();i!=lookup_1.end();i++) {
2293 tmp_2->addTaggedValue(i->first,tmp_2->getDefaultValue());
2294 DataArrayView view_1 = tmp_1->getDataPointByTag(i->first);
2295 DataArrayView view_2 = tmp_2->getDataPointByTag(i->first);
2296 double *ptr_1 = &view_1.getData(0);
2297 double *ptr_2 = &view_2.getData(0);
2298 matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2299 }
2300
2301 }
2302 else if (arg_0_Z.isConstant() && arg_1_Z.isExpanded()) {
2303
2304 res = Data(0.0, shape2, arg_1_Z.getFunctionSpace(),true); // DataExpanded output
2305 DataConstant* tmp_0=dynamic_cast<DataConstant*>(arg_0_Z.borrowData());
2306 DataExpanded* tmp_1=dynamic_cast<DataExpanded*>(arg_1_Z.borrowData());
2307 DataExpanded* tmp_2=dynamic_cast<DataExpanded*>(res.borrowData());
2308 if (tmp_0==0) { throw DataException("GTP Programming error - casting to DataConstant."); }
2309 if (tmp_1==0) { throw DataException("GTP Programming error - casting to DataExpanded."); }
2310 if (tmp_2==0) { throw DataException("GTP Programming error - casting to DataExpanded."); }
2311 int sampleNo_1,dataPointNo_1;
2312 int numSamples_1 = arg_1_Z.getNumSamples();
2313 int numDataPointsPerSample_1 = arg_1_Z.getNumDataPointsPerSample();
2314 int offset_0 = tmp_0->getPointOffset(0,0);
2315 #pragma omp parallel for private(sampleNo_1,dataPointNo_1) schedule(static)
2316 for (sampleNo_1 = 0; sampleNo_1 < numSamples_1; sampleNo_1++) {
2317 for (dataPointNo_1 = 0; dataPointNo_1 < numDataPointsPerSample_1; dataPointNo_1++) {
2318 int offset_1 = tmp_1->getPointOffset(sampleNo_1,dataPointNo_1);
2319 int offset_2 = tmp_2->getPointOffset(sampleNo_1,dataPointNo_1);
2320 double *ptr_0 = &((arg_0_Z.getPointDataView().getData())[offset_0]);
2321 double *ptr_1 = &((arg_1_Z.getPointDataView().getData())[offset_1]);
2322 double *ptr_2 = &((res.getPointDataView().getData())[offset_2]);
2323 matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2324 }
2325 }
2326
2327 }
2328 else if (arg_0_Z.isTagged() && arg_1_Z.isConstant()) {
2329
2330 // Borrow DataTagged input from Data object
2331 DataTagged* tmp_0=dynamic_cast<DataTagged*>(arg_0_Z.borrowData());
2332 if (tmp_0==0) { throw DataException("GTP_0 Programming error - casting to DataTagged."); }
2333
2334 // Prepare the DataConstant input
2335 DataConstant* tmp_1=dynamic_cast<DataConstant*>(arg_1_Z.borrowData());
2336 if (tmp_1==0) { throw DataException("GTP Programming error - casting to DataConstant."); }
2337
2338 // Prepare a DataTagged output 2
2339 res = Data(0.0, shape2, arg_0_Z.getFunctionSpace()); // DataTagged output
2340 res.tag();
2341 DataTagged* tmp_2=dynamic_cast<DataTagged*>(res.borrowData());
2342 if (tmp_2==0) { throw DataException("GTP Programming error - casting to DataTagged."); }
2343
2344 // Prepare offset into DataConstant
2345 int offset_1 = tmp_1->getPointOffset(0,0);
2346 double *ptr_1 = &((arg_1_Z.getPointDataView().getData())[offset_1]);
2347 // Get the views
2348 DataArrayView view_0 = tmp_0->getDefaultValue();
2349 DataArrayView view_2 = tmp_2->getDefaultValue();
2350 // Get the pointers to the actual data
2351 double *ptr_0 = &((view_0.getData())[0]);
2352 double *ptr_2 = &((view_2.getData())[0]);
2353 // Compute an MVP for the default
2354 matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2355 // Compute an MVP for each tag
2356 const DataTagged::DataMapType& lookup_0=tmp_0->getTagLookup();
2357 DataTagged::DataMapType::const_iterator i; // i->first is a tag, i->second is an offset into memory
2358 for (i=lookup_0.begin();i!=lookup_0.end();i++) {
2359 tmp_2->addTaggedValue(i->first,tmp_2->getDefaultValue());
2360 DataArrayView view_0 = tmp_0->getDataPointByTag(i->first);
2361 DataArrayView view_2 = tmp_2->getDataPointByTag(i->first);
2362 double *ptr_0 = &view_0.getData(0);
2363 double *ptr_2 = &view_2.getData(0);
2364 matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2365 }
2366
2367 }
2368 else if (arg_0_Z.isTagged() && arg_1_Z.isTagged()) {
2369
2370 // Borrow DataTagged input from Data object
2371 DataTagged* tmp_0=dynamic_cast<DataTagged*>(arg_0_Z.borrowData());
2372 if (tmp_0==0) { throw DataException("GTP Programming error - casting to DataTagged."); }
2373
2374 // Borrow DataTagged input from Data object
2375 DataTagged* tmp_1=dynamic_cast<DataTagged*>(arg_1_Z.borrowData());
2376 if (tmp_1==0) { throw DataException("GTP Programming error - casting to DataTagged."); }
2377
2378 // Prepare a DataTagged output 2
2379 res = Data(0.0, shape2, arg_1_Z.getFunctionSpace());
2380 res.tag(); // DataTagged output
2381 DataTagged* tmp_2=dynamic_cast<DataTagged*>(res.borrowData());
2382 if (tmp_2==0) { throw DataException("GTP Programming error - casting to DataTagged."); }
2383
2384 // Get the views
2385 DataArrayView view_0 = tmp_0->getDefaultValue();
2386 DataArrayView view_1 = tmp_1->getDefaultValue();
2387 DataArrayView view_2 = tmp_2->getDefaultValue();
2388 // Get the pointers to the actual data
2389 double *ptr_0 = &((view_0.getData())[0]);
2390 double *ptr_1 = &((view_1.getData())[0]);
2391 double *ptr_2 = &((view_2.getData())[0]);
2392 // Compute an MVP for the default
2393 matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2394 // Merge the tags
2395 DataTagged::DataMapType::const_iterator i; // i->first is a tag, i->second is an offset into memory
2396 const DataTagged::DataMapType& lookup_0=tmp_0->getTagLookup();
2397 const DataTagged::DataMapType& lookup_1=tmp_1->getTagLookup();
2398 for (i=lookup_0.begin();i!=lookup_0.end();i++) {
2399 tmp_2->addTaggedValue(i->first,tmp_2->getDefaultValue()); // use tmp_2 to get correct shape
2400 }
2401 for (i=lookup_1.begin();i!=lookup_1.end();i++) {
2402 tmp_2->addTaggedValue(i->first,tmp_2->getDefaultValue());
2403 }
2404 // Compute an MVP for each tag
2405 const DataTagged::DataMapType& lookup_2=tmp_2->getTagLookup();
2406 for (i=lookup_2.begin();i!=lookup_2.end();i++) {
2407 DataArrayView view_0 = tmp_0->getDataPointByTag(i->first);
2408 DataArrayView view_1 = tmp_1->getDataPointByTag(i->first);
2409 DataArrayView view_2 = tmp_2->getDataPointByTag(i->first);
2410 double *ptr_0 = &view_0.getData(0);
2411 double *ptr_1 = &view_1.getData(0);
2412 double *ptr_2 = &view_2.getData(0);
2413 matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2414 }
2415
2416 }
2417 else if (arg_0_Z.isTagged() && arg_1_Z.isExpanded()) {
2418
2419 // After finding a common function space above the two inputs have the same numSamples and num DPPS
2420 res = Data(0.0, shape2, arg_1_Z.getFunctionSpace(),true); // DataExpanded output
2421 DataTagged* tmp_0=dynamic_cast<DataTagged*>(arg_0_Z.borrowData());
2422 DataExpanded* tmp_1=dynamic_cast<DataExpanded*>(arg_1_Z.borrowData());
2423 DataExpanded* tmp_2=dynamic_cast<DataExpanded*>(res.borrowData());
2424 if (tmp_0==0) { throw DataException("GTP Programming error - casting to DataTagged."); }
2425 if (tmp_1==0) { throw DataException("GTP Programming error - casting to DataExpanded."); }
2426 if (tmp_2==0) { throw DataException("GTP Programming error - casting to DataExpanded."); }
2427 int sampleNo_0,dataPointNo_0;
2428 int numSamples_0 = arg_0_Z.getNumSamples();
2429 int numDataPointsPerSample_0 = arg_0_Z.getNumDataPointsPerSample();
2430 #pragma omp parallel for private(sampleNo_0,dataPointNo_0) schedule(static)
2431 for (sampleNo_0 = 0; sampleNo_0 < numSamples_0; sampleNo_0++) {
2432 int offset_0 = tmp_0->getPointOffset(sampleNo_0,0); // They're all the same, so just use #0
2433 double *ptr_0 = &((arg_0_Z.getPointDataView().getData())[offset_0]);
2434 for (dataPointNo_0 = 0; dataPointNo_0 < numDataPointsPerSample_0; dataPointNo_0++) {
2435 int offset_1 = tmp_1->getPointOffset(sampleNo_0,dataPointNo_0);
2436 int offset_2 = tmp_2->getPointOffset(sampleNo_0,dataPointNo_0);
2437 double *ptr_1 = &((arg_1_Z.getPointDataView().getData())[offset_1]);
2438 double *ptr_2 = &((res.getPointDataView().getData())[offset_2]);
2439 matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2440 }
2441 }
2442
2443 }
2444 else if (arg_0_Z.isExpanded() && arg_1_Z.isConstant()) {
2445
2446 res = Data(0.0, shape2, arg_1_Z.getFunctionSpace(),true); // DataExpanded output
2447 DataExpanded* tmp_0=dynamic_cast<DataExpanded*>(arg_0_Z.borrowData());
2448 DataConstant* tmp_1=dynamic_cast<DataConstant*>(arg_1_Z.borrowData());
2449 DataExpanded* tmp_2=dynamic_cast<DataExpanded*>(res.borrowData());
2450 if (tmp_0==0) { throw DataException("GTP Programming error - casting to DataExpanded."); }
2451 if (tmp_1==0) { throw DataException("GTP Programming error - casting to DataConstant."); }
2452 if (tmp_2==0) { throw DataException("GTP Programming error - casting to DataExpanded."); }
2453 int sampleNo_0,dataPointNo_0;
2454 int numSamples_0 = arg_0_Z.getNumSamples();
2455 int numDataPointsPerSample_0 = arg_0_Z.getNumDataPointsPerSample();
2456 int offset_1 = tmp_1->getPointOffset(0,0);
2457 #pragma omp parallel for private(sampleNo_0,dataPointNo_0) schedule(static)
2458 for (sampleNo_0 = 0; sampleNo_0 < numSamples_0; sampleNo_0++) {
2459 for (dataPointNo_0 = 0; dataPointNo_0 < numDataPointsPerSample_0; dataPointNo_0++) {
2460 int offset_0 = tmp_0->getPointOffset(sampleNo_0,dataPointNo_0);
2461 int offset_2 = tmp_2->getPointOffset(sampleNo_0,dataPointNo_0);
2462 double *ptr_0 = &((arg_0_Z.getPointDataView().getData())[offset_0]);
2463 double *ptr_1 = &((arg_1_Z.getPointDataView().getData())[offset_1]);
2464 double *ptr_2 = &((res.getPointDataView().getData())[offset_2]);
2465 matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2466 }
2467 }
2468
2469
2470 }
2471 else if (arg_0_Z.isExpanded() && arg_1_Z.isTagged()) {
2472
2473 // After finding a common function space above the two inputs have the same numSamples and num DPPS
2474 res = Data(0.0, shape2, arg_1_Z.getFunctionSpace(),true); // DataExpanded output
2475 DataExpanded* tmp_0=dynamic_cast<DataExpanded*>(arg_0_Z.borrowData());
2476 DataTagged* tmp_1=dynamic_cast<DataTagged*>(arg_1_Z.borrowData());
2477 DataExpanded* tmp_2=dynamic_cast<DataExpanded*>(res.borrowData());
2478 if (tmp_0==0) { throw DataException("GTP Programming error - casting to DataExpanded."); }
2479 if (tmp_1==0) { throw DataException("GTP Programming error - casting to DataTagged."); }
2480 if (tmp_2==0) { throw DataException("GTP Programming error - casting to DataExpanded."); }
2481 int sampleNo_0,dataPointNo_0;
2482 int numSamples_0 = arg_0_Z.getNumSamples();
2483 int numDataPointsPerSample_0 = arg_0_Z.getNumDataPointsPerSample();
2484 #pragma omp parallel for private(sampleNo_0,dataPointNo_0) schedule(static)
2485 for (sampleNo_0 = 0; sampleNo_0 < numSamples_0; sampleNo_0++) {
2486 int offset_1 = tmp_1->getPointOffset(sampleNo_0,0);
2487 double *ptr_1 = &((arg_1_Z.getPointDataView().getData())[offset_1]);
2488 for (dataPointNo_0 = 0; dataPointNo_0 < numDataPointsPerSample_0; dataPointNo_0++) {
2489 int offset_0 = tmp_0->getPointOffset(sampleNo_0,dataPointNo_0);
2490 int offset_2 = tmp_2->getPointOffset(sampleNo_0,dataPointNo_0);
2491 double *ptr_0 = &((arg_0_Z.getPointDataView().getData())[offset_0]);
2492 double *ptr_2 = &((res.getPointDataView().getData())[offset_2]);
2493 matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2494 }
2495 }
2496
2497 }
2498 else if (arg_0_Z.isExpanded() && arg_1_Z.isExpanded()) {
2499
2500 // After finding a common function space above the two inputs have the same numSamples and num DPPS
2501 res = Data(0.0, shape2, arg_1_Z.getFunctionSpace(),true); // DataExpanded output
2502 DataExpanded* tmp_0=dynamic_cast<DataExpanded*>(arg_0_Z.borrowData());
2503 DataExpanded* tmp_1=dynamic_cast<DataExpanded*>(arg_1_Z.borrowData());
2504 DataExpanded* tmp_2=dynamic_cast<DataExpanded*>(res.borrowData());
2505 if (tmp_0==0) { throw DataException("GTP Programming error - casting to DataExpanded."); }
2506 if (tmp_1==0) { throw DataException("GTP Programming error - casting to DataExpanded."); }
2507 if (tmp_2==0) { throw DataException("GTP Programming error - casting to DataExpanded."); }
2508 int sampleNo_0,dataPointNo_0;
2509 int numSamples_0 = arg_0_Z.getNumSamples();
2510 int numDataPointsPerSample_0 = arg_0_Z.getNumDataPointsPerSample();
2511 #pragma omp parallel for private(sampleNo_0,dataPointNo_0) schedule(static)
2512 for (sampleNo_0 = 0; sampleNo_0 < numSamples_0; sampleNo_0++) {
2513 for (dataPointNo_0 = 0; dataPointNo_0 < numDataPointsPerSample_0; dataPointNo_0++) {
2514 int offset_0 = tmp_0->getPointOffset(sampleNo_0,dataPointNo_0);
2515 int offset_1 = tmp_1->getPointOffset(sampleNo_0,dataPointNo_0);
2516 int offset_2 = tmp_2->getPointOffset(sampleNo_0,dataPointNo_0);
2517 double *ptr_0 = &((arg_0_Z.getPointDataView().getData())[offset_0]);
2518 double *ptr_1 = &((arg_1_Z.getPointDataView().getData())[offset_1]);
2519 double *ptr_2 = &((res.getPointDataView().getData())[offset_2]);
2520 matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2521 }
2522 }
2523
2524 }
2525 else {
2526 throw DataException("Error - C_GeneralTensorProduct: unknown combination of inputs");
2527 }
2528
2529 return res;
2530 }
2531
2532 DataAbstract*
2533 Data::borrowData() const
2534 {
2535 return m_data.get();
2536 }
2537
2538 /* Member functions specific to the MPI implementation */
2539
2540 void
2541 Data::print()
2542 {
2543 int i,j;
2544
2545 printf( "Data is %dX%d\n", getNumSamples(), getNumDataPointsPerSample() );
2546 for( i=0; i<getNumSamples(); i++ )
2547 {
2548 printf( "[%6d]", i );
2549 for( j=0; j<getNumDataPointsPerSample(); j++ )
2550 printf( "\t%10.7g", (getSampleData(i))[j] );
2551 printf( "\n" );
2552 }
2553 }
2554 void
2555 Data::dump(const std::string fileName) const
2556 {
2557 try
2558 {
2559 return m_data->dump(fileName);
2560 }
2561 catch (exception& e)
2562 {
2563 cout << e.what() << endl;
2564 }
2565 }
2566
2567 int
2568 Data::get_MPISize() const
2569 {
2570 int error, size;
2571 #ifdef PASO_MPI
2572 error = MPI_Comm_size( get_MPIComm(), &size );
2573 #else
2574 size = 1;
2575 #endif
2576 return size;
2577 }
2578
2579 int
2580 Data::get_MPIRank() const
2581 {
2582 int error, rank;
2583 #ifdef PASO_MPI
2584 error = MPI_Comm_rank( get_MPIComm(), &rank );
2585 #else
2586 rank = 0;
2587 #endif
2588 return rank;
2589 }
2590
2591 MPI_Comm
2592 Data::get_MPIComm() const
2593 {
2594 #ifdef PASO_MPI
2595 return MPI_COMM_WORLD;
2596 #else
2597 return -1;
2598 #endif
2599 }
2600

Properties

Name Value
svn:eol-style native
svn:keywords Author Date Id Revision

  ViewVC Help
Powered by ViewVC 1.1.26