/[escript]/branches/clazy/escriptcore/src/DataLazy.cpp
ViewVC logotype

Diff of /branches/clazy/escriptcore/src/DataLazy.cpp

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

branches/schroedinger/escript/src/DataLazy.cpp revision 1888 by jfenwick, Wed Oct 15 04:00:21 2008 UTC trunk/escriptcore/src/DataLazy.cpp revision 6057 by jfenwick, Thu Mar 10 06:00:58 2016 UTC
# Line 1  Line 1 
1    
2  /*******************************************************  /*****************************************************************************
3  *  *
4  * Copyright (c) 2003-2008 by University of Queensland  * Copyright (c) 2003-2016 by The University of Queensland
5  * Earth Systems Science Computational Center (ESSCC)  * http://www.uq.edu.au
 * http://www.uq.edu.au/esscc  
6  *  *
7  * Primary Business: Queensland, Australia  * Primary Business: Queensland, Australia
8  * Licensed under the Open Software License version 3.0  * Licensed under the Open Software License version 3.0
9  * http://www.opensource.org/licenses/osl-3.0.php  * http://www.opensource.org/licenses/osl-3.0.php
10  *  *
11  *******************************************************/  * Development until 2012 by Earth Systems Science Computational Center (ESSCC)
12    * Development 2012-2013 by School of Earth Sciences
13    * Development from 2014 by Centre for Geoscience Computing (GeoComp)
14    *
15    *****************************************************************************/
16    
17  #include "DataLazy.h"  #include "DataLazy.h"
18    #include "Data.h"
19    #include "DataTypes.h"
20    #include "EscriptParams.h"
21    #include "FunctionSpace.h"
22    #include "UnaryFuncs.h"    // for escript::fsign
23    #include "Utils.h"
24    #include "DataMaths.h"
25    
26  #ifdef USE_NETCDF  #ifdef USE_NETCDF
27  #include <netcdfcpp.h>  #include <netcdfcpp.h>
28  #endif  #endif
29  #ifdef PASO_MPI  
30  #include <mpi.h>  #include <iomanip> // for some fancy formatting in debug
31  #endif  
32  #ifdef _OPENMP  using namespace escript::DataTypes;
33  #include <omp.h>  
34  #endif  #define NO_ARG
35  #include "FunctionSpace.h"  
36  #include "DataTypes.h"  // #define LAZYDEBUG(X) if (privdebug){X;}
37  #include "Data.h"  #define LAZYDEBUG(X)
38  #include "UnaryFuncs.h"     // for escript::fsign  namespace
39    {
40    bool privdebug=false;
41    
42    #define ENABLEDEBUG privdebug=true;
43    #define DISABLEDEBUG privdebug=false;
44    }
45    
46    // #define SIZELIMIT if ((m_height>escript::escriptParams.getTOO_MANY_LEVELS()) || (m_children>escript::escriptParams.getTOO_MANY_NODES())) {cerr << "\n!!!!!!! SIZE LIMIT EXCEEDED " << m_children << ";" << m_height << endl << toString() << endl;resolveToIdentity();}
47    
48    // #define SIZELIMIT if ((m_height>escript::escriptParams.getTOO_MANY_LEVELS()) || (m_children>escript::escriptParams.getTOO_MANY_NODES())) {cerr << "SIZE LIMIT EXCEEDED " << m_height << endl;resolveToIdentity();}
49    
50    
51    #define SIZELIMIT if (m_height>escript::escriptParams.getTOO_MANY_LEVELS())  {if (escript::escriptParams.getLAZY_VERBOSE()){cerr << "SIZE LIMIT EXCEEDED height=" << m_height << endl;}resolveToIdentity();}
52    
53    /*
54    How does DataLazy work?
55    ~~~~~~~~~~~~~~~~~~~~~~~
56    
57    Each instance represents a single operation on one or two other DataLazy instances. These arguments are normally
58    denoted left and right.
59    
60    A special operation, IDENTITY, stores an instance of DataReady in the m_id member.
61    This means that all "internal" nodes in the structure are instances of DataLazy.
62    
63    Each operation has a string representation as well as an opgroup - eg G_IDENTITY, G_BINARY, ...
64    Note that IDENTITY is not considered a unary operation.
65    
66    I am avoiding calling the structure formed a tree because it is not guaranteed to be one (eg c=a+a).
67    It must however form a DAG (directed acyclic graph).
68    I will refer to individual DataLazy objects with the structure as nodes.
69    
70    Each node also stores:
71    - m_readytype \in {'E','T','C','?'} ~ indicates what sort of DataReady would be produced if the expression was
72            evaluated.
73    - m_buffsrequired ~ the large number of samples which would need to be kept simultaneously in order to
74            evaluate the expression.
75    - m_samplesize ~ the number of doubles stored in a sample.
76    
77    When a new node is created, the above values are computed based on the values in the child nodes.
78    Eg: if left requires 4 samples and right requires 6 then left+right requires 7 samples.
79    
80    The resolve method, which produces a DataReady from a DataLazy, does the following:
81    1) Create a DataReady to hold the new result.
82    2) Allocate a vector (v) big enough to hold m_buffsrequired samples.
83    3) For each sample, call resolveSample with v, to get its values and copy them into the result object.
84    
85    (In the case of OMP, multiple samples are resolved in parallel so the vector needs to be larger.)
86    
87    resolveSample returns a Vector* and an offset within that vector where the result is stored.
88    Normally, this would be v, but for identity nodes their internal vector is returned instead.
89    
90    The convention that I use, is that the resolve methods should store their results starting at the offset they are passed.
91    
92    For expressions which evaluate to Constant or Tagged, there is a different evaluation method.
93    The collapse method invokes the (non-lazy) operations on the Data class to evaluate the expression.
94    
95    To add a new operator you need to do the following (plus anything I might have forgotten - adding a new group for example):
96    1) Add to the ES_optype.
97    2) determine what opgroup your operation belongs to (X)
98    3) add a string for the op to the end of ES_opstrings
99    4) increase ES_opcount
100    5) add an entry (X) to opgroups
101    6) add an entry to the switch in collapseToReady
102    7) add an entry to resolveX
103    */
104    
105    
106  using namespace std;  using namespace std;
107  using namespace boost;  using namespace boost;
# Line 33  using namespace boost; Line 109  using namespace boost;
109  namespace escript  namespace escript
110  {  {
111    
 const std::string&  
 opToString(ES_optype op);  
   
112  namespace  namespace
113  {  {
114    
115    
116    // enabling this will print out when ever the maximum stacksize used by resolve increases
117    // it assumes _OPENMP is also in use
118    //#define LAZY_STACK_PROF
119    
120    
121    
122    #ifndef _OPENMP
123      #ifdef LAZY_STACK_PROF
124      #undef LAZY_STACK_PROF
125      #endif
126    #endif
127    
128    
129    #ifdef LAZY_STACK_PROF
130    std::vector<void*> stackstart(getNumberOfThreads());
131    std::vector<void*> stackend(getNumberOfThreads());
132    size_t maxstackuse=0;
133    #endif
134    
135  enum ES_opgroup  enum ES_opgroup
136  {  {
137     G_UNKNOWN,     G_UNKNOWN,
138     G_IDENTITY,     G_IDENTITY,
139     G_BINARY,     G_BINARY,            // pointwise operations with two arguments
140     G_UNARY     G_UNARY,             // pointwise operations with one argument
141       G_UNARY_P,           // pointwise operations with one argument, requiring a parameter
142       G_NP1OUT,            // non-pointwise op with one output
143       G_NP1OUT_P,          // non-pointwise op with one output requiring a parameter
144       G_TENSORPROD,        // general tensor product
145       G_NP1OUT_2P,         // non-pointwise op with one output requiring two params
146       G_REDUCTION,         // non-pointwise unary op with a scalar output
147       G_CONDEVAL
148  };  };
149    
150    
151    
152    
153  string ES_opstrings[]={"UNKNOWN","IDENTITY","+","-","*","/","sin","cos","tan",  string ES_opstrings[]={"UNKNOWN","IDENTITY","+","-","*","/","^",
154              "asin","acos","atan","sinh","cosh","tanh","erf",                          "sin","cos","tan",
155              "asinh","acosh","atanh",                          "asin","acos","atan","sinh","cosh","tanh","erf",
156              "log10","log","sign","abs","neg","pos","exp","sqrt",                          "asinh","acosh","atanh",
157              "1/","where>0","where<0","where>=0","where<=0"};                          "log10","log","sign","abs","neg","pos","exp","sqrt",
158  int ES_opcount=32;                          "1/","where>0","where<0","where>=0","where<=0", "where<>0","where=0",
159  ES_opgroup opgroups[]={G_UNKNOWN,G_IDENTITY,G_BINARY,G_BINARY,G_BINARY,G_BINARY,G_UNARY,G_UNARY,G_UNARY, //9                          "symmetric","nonsymmetric",
160              G_UNARY,G_UNARY,G_UNARY,G_UNARY,G_UNARY,G_UNARY,G_UNARY,    // 16                          "prod",
161              G_UNARY,G_UNARY,G_UNARY,                    // 19                          "transpose", "trace",
162              G_UNARY,G_UNARY,G_UNARY,G_UNARY,G_UNARY,G_UNARY,G_UNARY,G_UNARY,        // 27                          "swapaxes",
163              G_UNARY,G_UNARY,G_UNARY,G_UNARY,G_UNARY};                          "minval", "maxval",
164                            "condEval"};
165    int ES_opcount=44;
166    ES_opgroup opgroups[]={G_UNKNOWN,G_IDENTITY,G_BINARY,G_BINARY,G_BINARY,G_BINARY, G_BINARY,
167                            G_UNARY,G_UNARY,G_UNARY, //10
168                            G_UNARY,G_UNARY,G_UNARY,G_UNARY,G_UNARY,G_UNARY,G_UNARY,        // 17
169                            G_UNARY,G_UNARY,G_UNARY,                                        // 20
170                            G_UNARY,G_UNARY,G_UNARY,G_UNARY,G_UNARY,G_UNARY,G_UNARY,G_UNARY,        // 28
171                            G_UNARY,G_UNARY,G_UNARY,G_UNARY,G_UNARY, G_UNARY_P, G_UNARY_P,          // 35
172                            G_NP1OUT,G_NP1OUT,
173                            G_TENSORPROD,
174                            G_NP1OUT_P, G_NP1OUT_P,
175                            G_NP1OUT_2P,
176                            G_REDUCTION, G_REDUCTION,
177                            G_CONDEVAL};
178  inline  inline
179  ES_opgroup  ES_opgroup
180  getOpgroup(ES_optype op)  getOpgroup(ES_optype op)
# Line 74  getOpgroup(ES_optype op) Line 186  getOpgroup(ES_optype op)
186  FunctionSpace  FunctionSpace
187  resultFS(DataAbstract_ptr left, DataAbstract_ptr right, ES_optype op)  resultFS(DataAbstract_ptr left, DataAbstract_ptr right, ES_optype op)
188  {  {
189      // perhaps this should call interpolate and throw or something?          // perhaps this should call interpolate and throw or something?
190      // maybe we need an interpolate node -          // maybe we need an interpolate node -
191      // that way, if interpolate is required in any other op we can just throw a          // that way, if interpolate is required in any other op we can just throw a
192      // programming error exception.          // programming error exception.
193    
194      FunctionSpace l=left->getFunctionSpace();
195      if (left->getFunctionSpace()!=right->getFunctionSpace())    FunctionSpace r=right->getFunctionSpace();
196      {    if (l!=r)
197          throw DataException("FunctionSpaces not equal - interpolation not supported on lazy data.");    {
198      }      signed char res=r.getDomain()->preferredInterpolationOnDomain(r.getTypeCode(), l.getTypeCode());
199      return left->getFunctionSpace();      if (res==1)
200        {
201            return l;
202        }
203        if (res==-1)
204        {
205            return r;
206        }
207        throw DataException("Cannot interpolate between the FunctionSpaces given for operation "+opToString(op)+".");
208      }
209      return l;
210  }  }
211    
212  // return the shape of the result of "left op right"  // return the shape of the result of "left op right"
213    // the shapes resulting from tensor product are more complex to compute so are worked out elsewhere
214  DataTypes::ShapeType  DataTypes::ShapeType
215  resultShape(DataAbstract_ptr left, DataAbstract_ptr right, ES_optype op)  resultShape(DataAbstract_ptr left, DataAbstract_ptr right, ES_optype op)
216  {  {
217      if (left->getShape()!=right->getShape())          if (left->getShape()!=right->getShape())
218      {          {
219          throw DataException("Shapes not the same - shapes must match for lazy data.");            if ((getOpgroup(op)!=G_BINARY) && (getOpgroup(op)!=G_NP1OUT))
220      }            {
221      return left->getShape();                  throw DataException("Shapes not the name - shapes must match for (point)binary operations.");
222              }
223    
224              if (left->getRank()==0)       // we need to allow scalar * anything
225              {
226                    return right->getShape();
227              }
228              if (right->getRank()==0)
229              {
230                    return left->getShape();
231              }
232              throw DataException("Shapes not the same - arguments must have matching shapes (or be scalars) for (point)binary operations on lazy data.");
233            }
234            return left->getShape();
235  }  }
236    
237  size_t  // return the shape for "op left"
238  resultLength(DataAbstract_ptr left, DataAbstract_ptr right, ES_optype op)  
239    DataTypes::ShapeType
240    resultShape(DataAbstract_ptr left, ES_optype op, int axis_offset)
241  {  {
242     switch (getOpgroup(op))          switch(op)
243     {          {
244     case G_BINARY: return left->getLength();          case TRANS:
245     case G_UNARY: return left->getLength();             {                    // for the scoping of variables
246     default:                  const DataTypes::ShapeType& s=left->getShape();
247      throw DataException("Programmer Error - attempt to getLength() for operator "+opToString(op)+".");                  DataTypes::ShapeType sh;
248     }                  int rank=left->getRank();
249                    if (axis_offset<0 || axis_offset>rank)
250                    {
251                stringstream e;
252                e << "Error - Data::transpose must have 0 <= axis_offset <= rank=" << rank;
253                throw DataException(e.str());
254            }
255            for (int i=0; i<rank; i++)
256                    {
257                       int index = (axis_offset+i)%rank;
258               sh.push_back(s[index]); // Append to new shape
259            }
260                    return sh;
261               }
262            break;
263            case TRACE:
264               {
265                    int rank=left->getRank();
266                    if (rank<2)
267                    {
268                       throw DataException("Trace can only be computed for objects with rank 2 or greater.");
269                    }
270                    if ((axis_offset>rank-2) || (axis_offset<0))
271                    {
272                       throw DataException("Trace: axis offset must lie between 0 and rank-2 inclusive.");
273                    }
274                    if (rank==2)
275                    {
276                       return DataTypes::scalarShape;
277                    }
278                    else if (rank==3)
279                    {
280                       DataTypes::ShapeType sh;
281                       if (axis_offset==0)
282                       {
283                            sh.push_back(left->getShape()[2]);
284                       }
285                       else         // offset==1
286                       {
287                            sh.push_back(left->getShape()[0]);
288                       }
289                       return sh;
290                    }
291                    else if (rank==4)
292                    {
293                       DataTypes::ShapeType sh;
294                       const DataTypes::ShapeType& s=left->getShape();
295                       if (axis_offset==0)
296                       {
297                            sh.push_back(s[2]);
298                            sh.push_back(s[3]);
299                       }
300                       else if (axis_offset==1)
301                       {
302                            sh.push_back(s[0]);
303                            sh.push_back(s[3]);
304                       }
305                       else         // offset==2
306                       {
307                            sh.push_back(s[0]);
308                            sh.push_back(s[1]);
309                       }
310                       return sh;
311                    }
312                    else            // unknown rank
313                    {
314                       throw DataException("Error - Data::trace can only be calculated for rank 2, 3 or 4 object.");
315                    }
316               }
317            break;
318            default:
319            throw DataException("Programmer error - resultShape(left,op) can't compute shapes for operator "+opToString(op)+".");
320            }
321  }  }
322    
323  int  DataTypes::ShapeType
324  calcBuffs(const DataLazy_ptr& left, const DataLazy_ptr& right, ES_optype op)  SwapShape(DataAbstract_ptr left, const int axis0, const int axis1)
325  {  {
326     switch(getOpgroup(op))       // This code taken from the Data.cpp swapaxes() method
327     {       // Some of the checks are probably redundant here
328     case G_IDENTITY: return 1;       int axis0_tmp,axis1_tmp;
329     case G_BINARY: return max(left->getBuffsRequired(),right->getBuffsRequired()+1);       const DataTypes::ShapeType& s=left->getShape();
330     case G_UNARY: return max(left->getBuffsRequired(),1);       DataTypes::ShapeType out_shape;
331     default:       // Here's the equivalent of python s_out=s[axis_offset:]+s[:axis_offset]
332      throw DataException("Programmer Error - attempt to calcBuffs() for operator "+opToString(op)+".");       // which goes thru all shape vector elements starting with axis_offset (at index=rank wrap around to 0)
333     }       int rank=left->getRank();
334         if (rank<2) {
335            throw DataException("Error - Data::swapaxes argument must have at least rank 2.");
336         }
337         if (axis0<0 || axis0>rank-1) {
338            stringstream e;
339            e << "Error - Data::swapaxes: axis0 must be between 0 and rank-1=" << (rank-1);
340            throw DataException(e.str());
341         }
342         if (axis1<0 || axis1>rank-1) {
343            stringstream e;
344            e << "Error - Data::swapaxes: axis1 must be between 0 and rank-1=" << (rank-1);
345            throw DataException(e.str());
346         }
347         if (axis0 == axis1) {
348             throw DataException("Error - Data::swapaxes: axis indices must be different.");
349         }
350         if (axis0 > axis1) {
351             axis0_tmp=axis1;
352             axis1_tmp=axis0;
353         } else {
354             axis0_tmp=axis0;
355             axis1_tmp=axis1;
356         }
357         for (int i=0; i<rank; i++) {
358           if (i == axis0_tmp) {
359              out_shape.push_back(s[axis1_tmp]);
360           } else if (i == axis1_tmp) {
361              out_shape.push_back(s[axis0_tmp]);
362           } else {
363              out_shape.push_back(s[i]);
364           }
365         }
366        return out_shape;
367  }  }
368    
369    
370    // determine the output shape for the general tensor product operation
371    // the additional parameters return information required later for the product
372    // the majority of this code is copy pasted from C_General_Tensor_Product
373    DataTypes::ShapeType
374    GTPShape(DataAbstract_ptr left, DataAbstract_ptr right, int axis_offset, int transpose, int& SL, int& SM, int& SR)
375    {
376            
377      // Get rank and shape of inputs
378      int rank0 = left->getRank();
379      int rank1 = right->getRank();
380      const DataTypes::ShapeType& shape0 = left->getShape();
381      const DataTypes::ShapeType& shape1 = right->getShape();
382    
383      // Prepare for the loops of the product and verify compatibility of shapes
384      int start0=0, start1=0;
385      if (transpose == 0)           {}
386      else if (transpose == 1)      { start0 = axis_offset; }
387      else if (transpose == 2)      { start1 = rank1-axis_offset; }
388      else                          { throw DataException("DataLazy GeneralTensorProduct Constructor: Error - transpose should be 0, 1 or 2"); }
389    
390  }   // end anonymous namespace    if (rank0<axis_offset)
391      {
392            throw DataException("DataLazy GeneralTensorProduct Constructor: Error - rank of left < axisoffset");
393      }
394    
395      // Adjust the shapes for transpose
396      DataTypes::ShapeType tmpShape0(rank0);        // pre-sizing the vectors rather
397      DataTypes::ShapeType tmpShape1(rank1);        // than using push_back
398      for (int i=0; i<rank0; i++)   { tmpShape0[i]=shape0[(i+start0)%rank0]; }
399      for (int i=0; i<rank1; i++)   { tmpShape1[i]=shape1[(i+start1)%rank1]; }
400    
401      // Prepare for the loops of the product
402      SL=1, SM=1, SR=1;
403      for (int i=0; i<rank0-axis_offset; i++)       {
404        SL *= tmpShape0[i];
405      }
406      for (int i=rank0-axis_offset; i<rank0; i++)   {
407        if (tmpShape0[i] != tmpShape1[i-(rank0-axis_offset)]) {
408          throw DataException("C_GeneralTensorProduct: Error - incompatible shapes");
409        }
410        SM *= tmpShape0[i];
411      }
412      for (int i=axis_offset; i<rank1; i++)         {
413        SR *= tmpShape1[i];
414      }
415    
416      // Define the shape of the output (rank of shape is the sum of the loop ranges below)
417      DataTypes::ShapeType shape2(rank0+rank1-2*axis_offset);      
418      {                     // block to limit the scope of out_index
419         int out_index=0;
420         for (int i=0; i<rank0-axis_offset; i++, ++out_index) { shape2[out_index]=tmpShape0[i]; } // First part of arg_0_Z
421         for (int i=axis_offset; i<rank1; i++, ++out_index)   { shape2[out_index]=tmpShape1[i]; } // Last part of arg_1_Z
422      }
423    
424      if (shape2.size()>ESCRIPT_MAX_DATA_RANK)
425      {
426         ostringstream os;
427         os << "C_GeneralTensorProduct: Error - Attempt to create a rank " << shape2.size() << " object. The maximum rank is " << ESCRIPT_MAX_DATA_RANK << ".";
428         throw DataException(os.str());
429      }
430    
431      return shape2;
432    }
433    
434    }       // end anonymous namespace
435    
436    
437    
438    // Return a string representing the operation
439  const std::string&  const std::string&
440  opToString(ES_optype op)  opToString(ES_optype op)
441  {  {
# Line 138  opToString(ES_optype op) Line 446  opToString(ES_optype op)
446    return ES_opstrings[op];    return ES_opstrings[op];
447  }  }
448    
449    void DataLazy::LazyNodeSetup()
450    {
451    #ifdef _OPENMP
452        int numthreads=omp_get_max_threads();
453        m_samples.resize(numthreads*m_samplesize);
454        m_sampleids=new int[numthreads];
455        for (int i=0;i<numthreads;++i)
456        {
457            m_sampleids[i]=-1;  
458        }
459    #else
460        m_samples.resize(m_samplesize);
461        m_sampleids=new int[1];
462        m_sampleids[0]=-1;
463    #endif  // _OPENMP
464    }
465    
466    
467    // Creates an identity node
468  DataLazy::DataLazy(DataAbstract_ptr p)  DataLazy::DataLazy(DataAbstract_ptr p)
469      : parent(p->getFunctionSpace(),p->getShape()),          : parent(p->getFunctionSpace(),p->getShape())
470      m_op(IDENTITY)          ,m_sampleids(0),
471            m_samples(1)
472  {  {
473     if (p->isLazy())     if (p->isLazy())
474     {     {
475      // TODO: fix this.   We could make the new node a copy of p?          // I don't want identity of Lazy.
476      // I don't want identity of Lazy.          // Question: Why would that be so bad?
477      // Question: Why would that be so bad?          // Answer: We assume that the child of ID is something we can call getVector on
478      // Answer: We assume that the child of ID is something we can call getVector on          throw DataException("Programmer error - attempt to create identity from a DataLazy.");
     throw DataException("Programmer error - attempt to create identity from a DataLazy.");  
479     }     }
480     else     else
481     {     {
482      m_id=dynamic_pointer_cast<DataReady>(p);          p->makeLazyShared();
483            DataReady_ptr dr=dynamic_pointer_cast<DataReady>(p);
484            makeIdentity(dr);
485    LAZYDEBUG(cout << "Wrapping " << dr.get() << " id=" << m_id.get() << endl;)
486     }     }
487     m_length=p->getLength();  LAZYDEBUG(cout << "(1)Lazy created with " << m_samplesize << endl;)
    m_buffsRequired=1;  
    m_samplesize=getNumDPPSample()*getNoValues();  
 cout << "(1)Lazy created with " << m_samplesize << endl;  
488  }  }
489    
490  DataLazy::DataLazy(DataAbstract_ptr left, ES_optype op)  DataLazy::DataLazy(DataAbstract_ptr left, ES_optype op)
491      : parent(left->getFunctionSpace(),left->getShape()),          : parent(left->getFunctionSpace(),(getOpgroup(op)!=G_REDUCTION)?left->getShape():DataTypes::scalarShape),
492      m_op(op)          m_op(op),
493            m_axis_offset(0),
494            m_transpose(0),
495            m_SL(0), m_SM(0), m_SR(0)
496  {  {
497     if (getOpgroup(op)!=G_UNARY)     if ((getOpgroup(op)!=G_UNARY) && (getOpgroup(op)!=G_NP1OUT) && (getOpgroup(op)!=G_REDUCTION))
498     {     {
499      throw DataException("Programmer error - constructor DataLazy(left, op) will only process UNARY operations.");          throw DataException("Programmer error - constructor DataLazy(left, op) will only process UNARY operations.");
500     }     }
501    
502     DataLazy_ptr lleft;     DataLazy_ptr lleft;
503     if (!left->isLazy())     if (!left->isLazy())
504     {     {
505      lleft=DataLazy_ptr(new DataLazy(left));          lleft=DataLazy_ptr(new DataLazy(left));
506     }     }
507     else     else
508     {     {
509      lleft=dynamic_pointer_cast<DataLazy>(left);          lleft=dynamic_pointer_cast<DataLazy>(left);
510     }     }
511     m_length=left->getLength();     m_readytype=lleft->m_readytype;
512     m_left=lleft;     m_left=lleft;
    m_buffsRequired=1;  
513     m_samplesize=getNumDPPSample()*getNoValues();     m_samplesize=getNumDPPSample()*getNoValues();
514       m_children=m_left->m_children+1;
515       m_height=m_left->m_height+1;
516       LazyNodeSetup();
517       SIZELIMIT
518  }  }
519    
520    
521  DataLazy::DataLazy(DataLazy_ptr left, DataLazy_ptr right, ES_optype op)  // In this constructor we need to consider interpolation
522      : parent(resultFS(left,right,op), resultShape(left,right,op)),  DataLazy::DataLazy(DataAbstract_ptr left, DataAbstract_ptr right, ES_optype op)
523      m_left(left),          : parent(resultFS(left,right,op), resultShape(left,right,op)),
524      m_right(right),          m_op(op),
525      m_op(op)          m_SL(0), m_SM(0), m_SR(0)
526  {  {
527     if (getOpgroup(op)!=G_BINARY)  LAZYDEBUG(cout << "Forming operator with " << left.get() << " " << right.get() << endl;)
528       if ((getOpgroup(op)!=G_BINARY))
529       {
530            throw DataException("Programmer error - constructor DataLazy(left, right, op) will only process BINARY operations.");
531       }
532    
533       if (getFunctionSpace()!=left->getFunctionSpace())    // left needs to be interpolated
534       {
535            FunctionSpace fs=getFunctionSpace();
536            Data ltemp(left);
537            Data tmp(ltemp,fs);
538            left=tmp.borrowDataPtr();
539       }
540       if (getFunctionSpace()!=right->getFunctionSpace())   // right needs to be interpolated
541       {
542            Data tmp(Data(right),getFunctionSpace());
543            right=tmp.borrowDataPtr();
544    LAZYDEBUG(cout << "Right interpolation required " << right.get() << endl;)
545       }
546       left->operandCheck(*right);
547    
548       if (left->isLazy())                  // the children need to be DataLazy. Wrap them in IDENTITY if required
549       {
550            m_left=dynamic_pointer_cast<DataLazy>(left);
551    LAZYDEBUG(cout << "Left is " << m_left->toString() << endl;)
552       }
553       else
554       {
555            m_left=DataLazy_ptr(new DataLazy(left));
556    LAZYDEBUG(cout << "Left " << left.get() << " wrapped " << m_left->m_id.get() << endl;)
557       }
558       if (right->isLazy())
559     {     {
560      throw DataException("Programmer error - constructor DataLazy(left, right, op) will only process BINARY operations.");          m_right=dynamic_pointer_cast<DataLazy>(right);
561    LAZYDEBUG(cout << "Right is " << m_right->toString() << endl;)
562       }
563       else
564       {
565            m_right=DataLazy_ptr(new DataLazy(right));
566    LAZYDEBUG(cout << "Right " << right.get() << " wrapped " << m_right->m_id.get() << endl;)
567       }
568       char lt=m_left->m_readytype;
569       char rt=m_right->m_readytype;
570       if (lt=='E' || rt=='E')
571       {
572            m_readytype='E';
573       }
574       else if (lt=='T' || rt=='T')
575       {
576            m_readytype='T';
577       }
578       else
579       {
580            m_readytype='C';
581     }     }
    m_length=resultLength(m_left,m_right,m_op);  
582     m_samplesize=getNumDPPSample()*getNoValues();     m_samplesize=getNumDPPSample()*getNoValues();
583     m_buffsRequired=calcBuffs(m_left, m_right, m_op);     m_children=m_left->m_children+m_right->m_children+2;
584  cout << "(2)Lazy created with " << m_samplesize << endl;     m_height=max(m_left->m_height,m_right->m_height)+1;
585       LazyNodeSetup();
586       SIZELIMIT
587    LAZYDEBUG(cout << "(3)Lazy created with " << m_samplesize << endl;)
588  }  }
589    
590  DataLazy::DataLazy(DataAbstract_ptr left, DataAbstract_ptr right, ES_optype op)  DataLazy::DataLazy(DataAbstract_ptr left, DataAbstract_ptr right, ES_optype op, int axis_offset, int transpose)
591      : parent(resultFS(left,right,op), resultShape(left,right,op)),          : parent(resultFS(left,right,op), GTPShape(left,right, axis_offset, transpose, m_SL,m_SM, m_SR)),
592      m_op(op)          m_op(op),
593            m_axis_offset(axis_offset),
594            m_transpose(transpose)
595  {  {
596     if (getOpgroup(op)!=G_BINARY)     if ((getOpgroup(op)!=G_TENSORPROD))
597     {     {
598      throw DataException("Programmer error - constructor DataLazy(left, op) will only process BINARY operations.");          throw DataException("Programmer error - constructor DataLazy(left, right, op, ax, tr) will only process BINARY operations which require parameters.");
599     }     }
600     if (left->isLazy())     if ((transpose>2) || (transpose<0))
601     {     {
602      m_left=dynamic_pointer_cast<DataLazy>(left);          throw DataException("DataLazy GeneralTensorProduct constructor: Error - transpose should be 0, 1 or 2");
603       }
604       if (getFunctionSpace()!=left->getFunctionSpace())    // left needs to be interpolated
605       {
606            FunctionSpace fs=getFunctionSpace();
607            Data ltemp(left);
608            Data tmp(ltemp,fs);
609            left=tmp.borrowDataPtr();
610       }
611       if (getFunctionSpace()!=right->getFunctionSpace())   // right needs to be interpolated
612       {
613            Data tmp(Data(right),getFunctionSpace());
614            right=tmp.borrowDataPtr();
615       }
616    //    left->operandCheck(*right);
617    
618       if (left->isLazy())                  // the children need to be DataLazy. Wrap them in IDENTITY if required
619       {
620            m_left=dynamic_pointer_cast<DataLazy>(left);
621     }     }
622     else     else
623     {     {
624      m_left=DataLazy_ptr(new DataLazy(left));          m_left=DataLazy_ptr(new DataLazy(left));
625     }     }
626     if (right->isLazy())     if (right->isLazy())
627     {     {
628      m_right=dynamic_pointer_cast<DataLazy>(right);          m_right=dynamic_pointer_cast<DataLazy>(right);
629       }
630       else
631       {
632            m_right=DataLazy_ptr(new DataLazy(right));
633       }
634       char lt=m_left->m_readytype;
635       char rt=m_right->m_readytype;
636       if (lt=='E' || rt=='E')
637       {
638            m_readytype='E';
639       }
640       else if (lt=='T' || rt=='T')
641       {
642            m_readytype='T';
643       }
644       else
645       {
646            m_readytype='C';
647       }
648       m_samplesize=getNumDPPSample()*getNoValues();
649       m_children=m_left->m_children+m_right->m_children+2;
650       m_height=max(m_left->m_height,m_right->m_height)+1;
651       LazyNodeSetup();
652       SIZELIMIT
653    LAZYDEBUG(cout << "(4)Lazy created with " << m_samplesize << endl;)
654    }
655    
656    
657    DataLazy::DataLazy(DataAbstract_ptr left, ES_optype op, int axis_offset)
658            : parent(left->getFunctionSpace(), resultShape(left,op, axis_offset)),
659            m_op(op),
660            m_axis_offset(axis_offset),
661            m_transpose(0),
662            m_tol(0)
663    {
664       if ((getOpgroup(op)!=G_NP1OUT_P))
665       {
666            throw DataException("Programmer error - constructor DataLazy(left, op, ax) will only process UNARY operations which require parameters.");
667       }
668       DataLazy_ptr lleft;
669       if (!left->isLazy())
670       {
671            lleft=DataLazy_ptr(new DataLazy(left));
672       }
673       else
674       {
675            lleft=dynamic_pointer_cast<DataLazy>(left);
676       }
677       m_readytype=lleft->m_readytype;
678       m_left=lleft;
679       m_samplesize=getNumDPPSample()*getNoValues();
680       m_children=m_left->m_children+1;
681       m_height=m_left->m_height+1;
682       LazyNodeSetup();
683       SIZELIMIT
684    LAZYDEBUG(cout << "(5)Lazy created with " << m_samplesize << endl;)
685    }
686    
687    DataLazy::DataLazy(DataAbstract_ptr left, ES_optype op, double tol)
688            : parent(left->getFunctionSpace(), left->getShape()),
689            m_op(op),
690            m_axis_offset(0),
691            m_transpose(0),
692            m_tol(tol)
693    {
694       if ((getOpgroup(op)!=G_UNARY_P))
695       {
696            throw DataException("Programmer error - constructor DataLazy(left, op, tol) will only process UNARY operations which require parameters.");
697       }
698       DataLazy_ptr lleft;
699       if (!left->isLazy())
700       {
701            lleft=DataLazy_ptr(new DataLazy(left));
702     }     }
703     else     else
704     {     {
705      m_right=DataLazy_ptr(new DataLazy(right));          lleft=dynamic_pointer_cast<DataLazy>(left);
706     }     }
707       m_readytype=lleft->m_readytype;
708       m_left=lleft;
709       m_samplesize=getNumDPPSample()*getNoValues();
710       m_children=m_left->m_children+1;
711       m_height=m_left->m_height+1;
712       LazyNodeSetup();
713       SIZELIMIT
714    LAZYDEBUG(cout << "(6)Lazy created with " << m_samplesize << endl;)
715    }
716    
717    
718     m_length=resultLength(m_left,m_right,m_op);  DataLazy::DataLazy(DataAbstract_ptr left, ES_optype op, const int axis0, const int axis1)
719            : parent(left->getFunctionSpace(), SwapShape(left,axis0,axis1)),
720            m_op(op),
721            m_axis_offset(axis0),
722            m_transpose(axis1),
723            m_tol(0)
724    {
725       if ((getOpgroup(op)!=G_NP1OUT_2P))
726       {
727            throw DataException("Programmer error - constructor DataLazy(left, op, tol) will only process UNARY operations which require two integer parameters.");
728       }
729       DataLazy_ptr lleft;
730       if (!left->isLazy())
731       {
732            lleft=DataLazy_ptr(new DataLazy(left));
733       }
734       else
735       {
736            lleft=dynamic_pointer_cast<DataLazy>(left);
737       }
738       m_readytype=lleft->m_readytype;
739       m_left=lleft;
740     m_samplesize=getNumDPPSample()*getNoValues();     m_samplesize=getNumDPPSample()*getNoValues();
741     m_buffsRequired=calcBuffs(m_left, m_right,m_op);     m_children=m_left->m_children+1;
742  cout << "(3)Lazy created with " << m_samplesize << endl;     m_height=m_left->m_height+1;
743       LazyNodeSetup();
744       SIZELIMIT
745    LAZYDEBUG(cout << "(7)Lazy created with " << m_samplesize << endl;)
746  }  }
747    
748    
749  DataLazy::~DataLazy()  namespace
750    {
751    
752        inline int max3(int a, int b, int c)
753        {
754            int t=(a>b?a:b);
755            return (t>c?t:c);
756    
757        }
758    }
759    
760    DataLazy::DataLazy(DataAbstract_ptr mask, DataAbstract_ptr left, DataAbstract_ptr right/*, double tol*/)
761            : parent(left->getFunctionSpace(), left->getShape()),
762            m_op(CONDEVAL),
763            m_axis_offset(0),
764            m_transpose(0),
765            m_tol(0)
766  {  {
767    
768       DataLazy_ptr lmask;
769       DataLazy_ptr lleft;
770       DataLazy_ptr lright;
771       if (!mask->isLazy())
772       {
773            lmask=DataLazy_ptr(new DataLazy(mask));
774       }
775       else
776       {
777            lmask=dynamic_pointer_cast<DataLazy>(mask);
778       }
779       if (!left->isLazy())
780       {
781            lleft=DataLazy_ptr(new DataLazy(left));
782       }
783       else
784       {
785            lleft=dynamic_pointer_cast<DataLazy>(left);
786       }
787       if (!right->isLazy())
788       {
789            lright=DataLazy_ptr(new DataLazy(right));
790       }
791       else
792       {
793            lright=dynamic_pointer_cast<DataLazy>(right);
794       }
795       m_readytype=lmask->m_readytype;
796       if ((lleft->m_readytype!=lright->m_readytype) || (lmask->m_readytype!=lleft->m_readytype))
797       {
798            throw DataException("Programmer Error - condEval arguments must have the same readytype");
799       }
800       m_left=lleft;
801       m_right=lright;
802       m_mask=lmask;
803       m_samplesize=getNumDPPSample()*getNoValues();
804       m_children=m_left->m_children+m_right->m_children+m_mask->m_children+1;
805       m_height=max3(m_left->m_height,m_right->m_height,m_mask->m_height)+1;
806       LazyNodeSetup();
807       SIZELIMIT
808    LAZYDEBUG(cout << "(8)Lazy created with " << m_samplesize << endl;)
809  }  }
810    
811    
812  int  
813  DataLazy::getBuffsRequired() const  DataLazy::~DataLazy()
814  {  {
815      return m_buffsRequired;     delete[] m_sampleids;
816  }  }
817    
818    
819  // the vector and the offset are a place where the method could write its data if it wishes  /*
820  // it is not obligated to do so. For example, if it has its own storage already, it can use that.    \brief Evaluates the expression using methods on Data.
821  // Hence the return value to indicate where the data is actually stored.    This does the work for the collapse method.
822  // Regardless, the storage should be assumed to be used, even if it isn't.    For reasons of efficiency do not call this method on DataExpanded nodes.
823  const double*  */
824  DataLazy::resolveSample(ValueType& v,int sampleNo,  size_t offset ) const  DataReady_ptr
825    DataLazy::collapseToReady() const
826  {  {
827    if (m_op==IDENTITY)      if (m_readytype=='E')
828      {     // this is more an efficiency concern than anything else
829        throw DataException("Programmer Error - do not use collapse on Expanded data.");
830      }
831      if (m_op==IDENTITY)
832    {    {
833      const ValueType& vec=m_id->getVector();      return m_id;
     return &(vec[m_id->getPointOffset(sampleNo, 0)]);  
834    }    }
835    size_t rightoffset=offset+m_samplesize;    DataReady_ptr pleft=m_left->collapseToReady();
836    const double* left=m_left->resolveSample(v,sampleNo,offset);    Data left(pleft);
837    const double* right=0;    Data right;
838    if (getOpgroup(m_op)==G_BINARY)    if ((getOpgroup(m_op)==G_BINARY) || (getOpgroup(m_op)==G_TENSORPROD))
839    {    {
840      right=m_right->resolveSample(v,sampleNo,rightoffset);      right=Data(m_right->collapseToReady());
841    }    }
842    double* result=&(v[offset]);    Data result;
843      switch(m_op)
844    {    {
845      switch(m_op)      case ADD:
846      {          result=left+right;
847      case ADD:       // since these are pointwise ops, pretend each sample is one point          break;
848      tensor_binary_operation(m_samplesize, left, right, result, plus<double>());      case SUB:          
849      break;          result=left-right;
850      case SUB:                break;
851      tensor_binary_operation(m_samplesize, left, right, result, minus<double>());      case MUL:          
852      break;          result=left*right;
853      case MUL:                break;
854      tensor_binary_operation(m_samplesize, left, right, result, multiplies<double>());      case DIV:          
855      break;          result=left/right;
856      case DIV:                break;
     tensor_binary_operation(m_samplesize, left, right, result, divides<double>());  
     break;  
 // unary ops  
857      case SIN:      case SIN:
858      tensor_unary_operation(m_samplesize, left, result, ::sin);          result=left.sin();      
859      break;          break;
860        case COS:
861            result=left.cos();
862            break;
863        case TAN:
864            result=left.tan();
865            break;
866        case ASIN:
867            result=left.asin();
868            break;
869        case ACOS:
870            result=left.acos();
871            break;
872        case ATAN:
873            result=left.atan();
874            break;
875        case SINH:
876            result=left.sinh();
877            break;
878        case COSH:
879            result=left.cosh();
880            break;
881        case TANH:
882            result=left.tanh();
883            break;
884        case ERF:
885            result=left.erf();
886            break;
887       case ASINH:
888            result=left.asinh();
889            break;
890       case ACOSH:
891            result=left.acosh();
892            break;
893       case ATANH:
894            result=left.atanh();
895            break;
896        case LOG10:
897            result=left.log10();
898            break;
899        case LOG:
900            result=left.log();
901            break;
902        case SIGN:
903            result=left.sign();
904            break;
905        case ABS:
906            result=left.abs();
907            break;
908        case NEG:
909            result=left.neg();
910            break;
911        case POS:
912            // it doesn't mean anything for delayed.
913            // it will just trigger a deep copy of the lazy object
914            throw DataException("Programmer error - POS not supported for lazy data.");
915            break;
916        case EXP:
917            result=left.exp();
918            break;
919        case SQRT:
920            result=left.sqrt();
921            break;
922        case RECIP:
923            result=left.oneOver();
924            break;
925        case GZ:
926            result=left.wherePositive();
927            break;
928        case LZ:
929            result=left.whereNegative();
930            break;
931        case GEZ:
932            result=left.whereNonNegative();
933            break;
934        case LEZ:
935            result=left.whereNonPositive();
936            break;
937        case NEZ:
938            result=left.whereNonZero(m_tol);
939            break;
940        case EZ:
941            result=left.whereZero(m_tol);
942            break;
943        case SYM:
944            result=left.symmetric();
945            break;
946        case NSYM:
947            result=left.nonsymmetric();
948            break;
949        case PROD:
950            result=C_GeneralTensorProduct(left,right,m_axis_offset, m_transpose);
951            break;
952        case TRANS:
953            result=left.transpose(m_axis_offset);
954            break;
955        case TRACE:
956            result=left.trace(m_axis_offset);
957            break;
958        case SWAP:
959            result=left.swapaxes(m_axis_offset, m_transpose);
960            break;
961        case MINVAL:
962            result=left.minval();
963            break;
964        case MAXVAL:
965            result=left.minval();
966            break;
967        default:
968            throw DataException("Programmer error - collapseToReady does not know how to resolve operator "+opToString(m_op)+".");
969      }
970      return result.borrowReadyPtr();
971    }
972    
973    /*
974       \brief Converts the DataLazy into an IDENTITY storing the value of the expression.
975       This method uses the original methods on the Data class to evaluate the expressions.
976       For this reason, it should not be used on DataExpanded instances. (To do so would defeat
977       the purpose of using DataLazy in the first place).
978    */
979    void
980    DataLazy::collapse() const
981    {
982      if (m_op==IDENTITY)
983      {
984            return;
985      }
986      if (m_readytype=='E')
987      {     // this is more an efficiency concern than anything else
988        throw DataException("Programmer Error - do not use collapse on Expanded data.");
989      }
990      m_id=collapseToReady();
991      m_op=IDENTITY;
992    }
993    
994    // The result will be stored in m_samples
995    // The return value is a pointer to the DataVector, offset is the offset within the return value
996    const DataTypes::RealVectorType*
997    DataLazy::resolveNodeSample(int tid, int sampleNo, size_t& roffset) const
998    {
999    LAZYDEBUG(cout << "Resolve sample " << toString() << endl;)
1000            // collapse so we have a 'E' node or an IDENTITY for some other type
1001      if (m_readytype!='E' && m_op!=IDENTITY)
1002      {
1003            collapse();
1004      }
1005      if (m_op==IDENTITY)  
1006      {
1007        const RealVectorType& vec=m_id->getVectorRO();
1008        roffset=m_id->getPointOffset(sampleNo, 0);
1009    #ifdef LAZY_STACK_PROF
1010    int x;
1011    if (&x<stackend[omp_get_thread_num()])
1012    {
1013           stackend[omp_get_thread_num()]=&x;
1014    }
1015    #endif
1016        return &(vec);
1017      }
1018      if (m_readytype!='E')
1019      {
1020        throw DataException("Programmer Error - Collapse did not produce an expanded node.");
1021      }
1022      if (m_sampleids[tid]==sampleNo)
1023      {
1024            roffset=tid*m_samplesize;
1025            return &(m_samples);            // sample is already resolved
1026      }
1027      m_sampleids[tid]=sampleNo;
1028    
1029      switch (getOpgroup(m_op))
1030      {
1031      case G_UNARY:
1032      case G_UNARY_P: return resolveNodeUnary(tid, sampleNo, roffset);
1033      case G_BINARY: return resolveNodeBinary(tid, sampleNo, roffset);
1034      case G_NP1OUT: return resolveNodeNP1OUT(tid, sampleNo, roffset);
1035      case G_NP1OUT_P: return resolveNodeNP1OUT_P(tid, sampleNo, roffset);
1036      case G_TENSORPROD: return resolveNodeTProd(tid, sampleNo, roffset);
1037      case G_NP1OUT_2P: return resolveNodeNP1OUT_2P(tid, sampleNo, roffset);
1038      case G_REDUCTION: return resolveNodeReduction(tid, sampleNo, roffset);
1039      case G_CONDEVAL: return resolveNodeCondEval(tid, sampleNo, roffset);
1040      default:
1041        throw DataException("Programmer Error - resolveSample does not know how to process "+opToString(m_op)+".");
1042      }
1043    }
1044    
1045    const DataTypes::RealVectorType*
1046    DataLazy::resolveNodeUnary(int tid, int sampleNo, size_t& roffset) const
1047    {
1048            // we assume that any collapsing has been done before we get here
1049            // since we only have one argument we don't need to think about only
1050            // processing single points.
1051            // we will also know we won't get identity nodes
1052      if (m_readytype!='E')
1053      {
1054        throw DataException("Programmer error - resolveUnary should only be called on expanded Data.");
1055      }
1056      if (m_op==IDENTITY)
1057      {
1058        throw DataException("Programmer error - resolveNodeUnary should not be called on identity nodes.");
1059      }
1060      const DataTypes::RealVectorType* leftres=m_left->resolveNodeSample(tid, sampleNo, roffset);
1061      const double* left=&((*leftres)[roffset]);
1062      roffset=m_samplesize*tid;
1063      double* result=&(m_samples[roffset]);
1064      switch (m_op)
1065      {
1066        case SIN:  
1067            tensor_unary_operation<double (*)(double)>(m_samplesize, left, result, ::sin);
1068            break;
1069      case COS:      case COS:
1070      tensor_unary_operation(m_samplesize, left, result, ::cos);          tensor_unary_operation<double (*)(double)>(m_samplesize, left, result, ::cos);
1071      break;          break;
1072      case TAN:      case TAN:
1073      tensor_unary_operation(m_samplesize, left, result, ::tan);          tensor_unary_operation<double (*)(double)>(m_samplesize, left, result, ::tan);
1074      break;          break;
1075      case ASIN:      case ASIN:
1076      tensor_unary_operation(m_samplesize, left, result, ::asin);          tensor_unary_operation<double (*)(double)>(m_samplesize, left, result, ::asin);
1077      break;          break;
1078      case ACOS:      case ACOS:
1079      tensor_unary_operation(m_samplesize, left, result, ::acos);          tensor_unary_operation<double (*)(double)>(m_samplesize, left, result, ::acos);
1080      break;          break;
1081      case ATAN:      case ATAN:
1082      tensor_unary_operation(m_samplesize, left, result, ::atan);          tensor_unary_operation<double (*)(double)>(m_samplesize, left, result, ::atan);
1083      break;          break;
1084      case SINH:      case SINH:
1085      tensor_unary_operation(m_samplesize, left, result, ::sinh);          tensor_unary_operation<double (*)(double)>(m_samplesize, left, result, ::sinh);
1086      break;          break;
1087      case COSH:      case COSH:
1088      tensor_unary_operation(m_samplesize, left, result, ::cosh);          tensor_unary_operation<double (*)(double)>(m_samplesize, left, result, ::cosh);
1089      break;          break;
1090      case TANH:      case TANH:
1091      tensor_unary_operation(m_samplesize, left, result, ::tanh);          tensor_unary_operation<double (*)(double)>(m_samplesize, left, result, ::tanh);
1092      break;          break;
1093      case ERF:      case ERF:
1094  #ifdef _WIN32  #if defined (_WIN32) && !defined(__INTEL_COMPILER)
1095      throw DataException("Error - Data:: erf function is not supported on _WIN32 platforms.");          throw DataException("Error - Data:: erf function is not supported on _WIN32 platforms.");
1096  #else  #else
1097      tensor_unary_operation(m_samplesize, left, result, ::erf);          tensor_unary_operation(m_samplesize, left, result, ::erf);
1098      break;          break;
1099  #endif  #endif
1100     case ASINH:     case ASINH:
1101  #ifdef _WIN32  #if defined (_WIN32) && !defined(__INTEL_COMPILER)
1102      tensor_unary_operation(m_samplesize, left, result, escript::asinh_substitute);          tensor_unary_operation(m_samplesize, left, result, escript::asinh_substitute);
1103  #else  #else
1104      tensor_unary_operation(m_samplesize, left, result, ::asinh);          tensor_unary_operation(m_samplesize, left, result, ::asinh);
1105  #endif    #endif  
1106      break;          break;
1107     case ACOSH:     case ACOSH:
1108  #ifdef _WIN32  #if defined (_WIN32) && !defined(__INTEL_COMPILER)
1109      tensor_unary_operation(m_samplesize, left, result, escript::acosh_substitute);          tensor_unary_operation(m_samplesize, left, result, escript::acosh_substitute);
1110  #else  #else
1111      tensor_unary_operation(m_samplesize, left, result, ::acosh);          tensor_unary_operation(m_samplesize, left, result, ::acosh);
1112  #endif    #endif  
1113      break;          break;
1114     case ATANH:     case ATANH:
1115  #ifdef _WIN32  #if defined (_WIN32) && !defined(__INTEL_COMPILER)
1116      tensor_unary_operation(m_samplesize, left, result, escript::atanh_substitute);          tensor_unary_operation(m_samplesize, left, result, escript::atanh_substitute);
1117  #else  #else
1118      tensor_unary_operation(m_samplesize, left, result, ::atanh);          tensor_unary_operation(m_samplesize, left, result, ::atanh);
1119  #endif    #endif  
1120      break;          break;
1121      case LOG10:      case LOG10:
1122      tensor_unary_operation(m_samplesize, left, result, ::log10);          tensor_unary_operation<double (*)(double)>(m_samplesize, left, result, ::log10);
1123      break;          break;
1124      case LOG:      case LOG:
1125      tensor_unary_operation(m_samplesize, left, result, ::log);          tensor_unary_operation<double (*)(double)>(m_samplesize, left, result, ::log);
1126      break;          break;
1127      case SIGN:      case SIGN:
1128      tensor_unary_operation(m_samplesize, left, result, escript::fsign);          tensor_unary_operation(m_samplesize, left, result, escript::fsign);
1129      break;          break;
1130      case ABS:      case ABS:
1131      tensor_unary_operation(m_samplesize, left, result, ::fabs);          tensor_unary_operation<double (*)(double)>(m_samplesize, left, result, ::fabs);
1132      break;          break;
1133      case NEG:      case NEG:
1134      tensor_unary_operation(m_samplesize, left, result, negate<double>());          tensor_unary_operation(m_samplesize, left, result, negate<double>());
1135      break;          break;
1136      case POS:      case POS:
1137      // it doesn't mean anything for delayed.          // it doesn't mean anything for delayed.
1138      // it will just trigger a deep copy of the lazy object          // it will just trigger a deep copy of the lazy object
1139      throw DataException("Programmer error - POS not supported for lazy data.");          throw DataException("Programmer error - POS not supported for lazy data.");
1140      break;          break;
1141      case EXP:      case EXP:
1142      tensor_unary_operation(m_samplesize, left, result, ::exp);          tensor_unary_operation<double (*)(double)>(m_samplesize, left, result, ::exp);
1143      break;          break;
1144      case SQRT:      case SQRT:
1145      tensor_unary_operation(m_samplesize, left, result, ::sqrt);          tensor_unary_operation<double (*)(double)>(m_samplesize, left, result, ::sqrt);
1146      break;          break;
1147      case RECIP:      case RECIP:
1148      tensor_unary_operation(m_samplesize, left, result, bind1st(divides<double>(),1.));          tensor_unary_operation(m_samplesize, left, result, bind1st(divides<double>(),1.));
1149      break;          break;
1150      case GZ:      case GZ:
1151      tensor_unary_operation(m_samplesize, left, result, bind2nd(greater<double>(),0.0));          tensor_unary_operation(m_samplesize, left, result, bind2nd(greater<double>(),0.0));
1152      break;          break;
1153      case LZ:      case LZ:
1154      tensor_unary_operation(m_samplesize, left, result, bind2nd(less<double>(),0.0));          tensor_unary_operation(m_samplesize, left, result, bind2nd(less<double>(),0.0));
1155      break;          break;
1156      case GEZ:      case GEZ:
1157      tensor_unary_operation(m_samplesize, left, result, bind2nd(greater_equal<double>(),0.0));          tensor_unary_operation(m_samplesize, left, result, bind2nd(greater_equal<double>(),0.0));
1158      break;          break;
1159      case LEZ:      case LEZ:
1160      tensor_unary_operation(m_samplesize, left, result, bind2nd(less_equal<double>(),0.0));          tensor_unary_operation(m_samplesize, left, result, bind2nd(less_equal<double>(),0.0));
1161      break;          break;
1162    // There are actually G_UNARY_P but I don't see a compelling reason to treat them differently
1163        case NEZ:
1164            tensor_unary_operation(m_samplesize, left, result, bind2nd(AbsGT(),m_tol));
1165            break;
1166        case EZ:
1167            tensor_unary_operation(m_samplesize, left, result, bind2nd(AbsLTE(),m_tol));
1168            break;
1169    
1170      default:      default:
1171      throw DataException("Programmer error - do not know how to resolve operator "+opToString(m_op)+".");          throw DataException("Programmer error - resolveUnary can not resolve operator "+opToString(m_op)+".");
     }  
1172    }    }
1173    return result;    return &(m_samples);
1174  }  }
1175    
1176    
1177    const DataTypes::RealVectorType*
1178    DataLazy::resolveNodeReduction(int tid, int sampleNo, size_t& roffset) const
1179    {
1180            // we assume that any collapsing has been done before we get here
1181            // since we only have one argument we don't need to think about only
1182            // processing single points.
1183            // we will also know we won't get identity nodes
1184      if (m_readytype!='E')
1185      {
1186        throw DataException("Programmer error - resolveUnary should only be called on expanded Data.");
1187      }
1188      if (m_op==IDENTITY)
1189      {
1190        throw DataException("Programmer error - resolveNodeUnary should not be called on identity nodes.");
1191      }
1192      size_t loffset=0;
1193      const DataTypes::RealVectorType* leftres=m_left->resolveNodeSample(tid, sampleNo, loffset);
1194    
1195      roffset=m_samplesize*tid;
1196      unsigned int ndpps=getNumDPPSample();
1197      unsigned int psize=DataTypes::noValues(m_left->getShape());
1198      double* result=&(m_samples[roffset]);
1199      switch (m_op)
1200      {
1201        case MINVAL:
1202            {
1203              for (unsigned int z=0;z<ndpps;++z)
1204              {
1205                FMin op;
1206                *result=DataMaths::reductionOp(*leftres, m_left->getShape(), loffset, op, numeric_limits<double>::max());
1207                loffset+=psize;
1208                result++;
1209              }
1210            }
1211            break;
1212        case MAXVAL:
1213            {
1214              for (unsigned int z=0;z<ndpps;++z)
1215              {
1216              FMax op;
1217              *result=DataMaths::reductionOp(*leftres, m_left->getShape(), loffset, op, numeric_limits<double>::max()*-1);
1218              loffset+=psize;
1219              result++;
1220              }
1221            }
1222            break;
1223        default:
1224            throw DataException("Programmer error - resolveUnary can not resolve operator "+opToString(m_op)+".");
1225      }
1226      return &(m_samples);
1227    }
1228    
1229    const DataTypes::RealVectorType*
1230    DataLazy::resolveNodeNP1OUT(int tid, int sampleNo, size_t& roffset) const
1231    {
1232            // we assume that any collapsing has been done before we get here
1233            // since we only have one argument we don't need to think about only
1234            // processing single points.
1235      if (m_readytype!='E')
1236      {
1237        throw DataException("Programmer error - resolveNodeNP1OUT should only be called on expanded Data.");
1238      }
1239      if (m_op==IDENTITY)
1240      {
1241        throw DataException("Programmer error - resolveNodeNP1OUT should not be called on identity nodes.");
1242      }
1243      size_t subroffset;
1244      const RealVectorType* leftres=m_left->resolveNodeSample(tid, sampleNo, subroffset);
1245      roffset=m_samplesize*tid;
1246      size_t loop=0;
1247      size_t numsteps=(m_readytype=='E')?getNumDPPSample():1;
1248      size_t step=getNoValues();
1249      size_t offset=roffset;
1250      switch (m_op)
1251      {
1252        case SYM:
1253            for (loop=0;loop<numsteps;++loop)
1254            {
1255                DataMaths::symmetric(*leftres,m_left->getShape(),subroffset, m_samples, getShape(), offset);
1256                subroffset+=step;
1257                offset+=step;
1258            }
1259            break;
1260        case NSYM:
1261            for (loop=0;loop<numsteps;++loop)
1262            {
1263                DataMaths::nonsymmetric(*leftres,m_left->getShape(),subroffset, m_samples, getShape(), offset);
1264                subroffset+=step;
1265                offset+=step;
1266            }
1267            break;
1268        default:
1269            throw DataException("Programmer error - resolveNP1OUT can not resolve operator "+opToString(m_op)+".");
1270      }
1271      return &m_samples;
1272    }
1273    
1274    const DataTypes::RealVectorType*
1275    DataLazy::resolveNodeNP1OUT_P(int tid, int sampleNo, size_t& roffset) const
1276    {
1277            // we assume that any collapsing has been done before we get here
1278            // since we only have one argument we don't need to think about only
1279            // processing single points.
1280      if (m_readytype!='E')
1281      {
1282        throw DataException("Programmer error - resolveNodeNP1OUT_P should only be called on expanded Data.");
1283      }
1284      if (m_op==IDENTITY)
1285      {
1286        throw DataException("Programmer error - resolveNodeNP1OUT_P should not be called on identity nodes.");
1287      }
1288      size_t subroffset;
1289      size_t offset;
1290      const RealVectorType* leftres=m_left->resolveNodeSample(tid, sampleNo, subroffset);
1291      roffset=m_samplesize*tid;
1292      offset=roffset;
1293      size_t loop=0;
1294      size_t numsteps=(m_readytype=='E')?getNumDPPSample():1;
1295      size_t outstep=getNoValues();
1296      size_t instep=m_left->getNoValues();
1297      switch (m_op)
1298      {
1299        case TRACE:
1300            for (loop=0;loop<numsteps;++loop)
1301            {
1302                DataMaths::trace(*leftres,m_left->getShape(),subroffset, m_samples ,getShape(),offset,m_axis_offset);
1303                subroffset+=instep;
1304                offset+=outstep;
1305            }
1306            break;
1307        case TRANS:
1308            for (loop=0;loop<numsteps;++loop)
1309            {
1310                DataMaths::transpose(*leftres,m_left->getShape(),subroffset, m_samples, getShape(),offset,m_axis_offset);
1311                subroffset+=instep;
1312                offset+=outstep;
1313            }
1314            break;
1315        default:
1316            throw DataException("Programmer error - resolveNP1OUTP can not resolve operator "+opToString(m_op)+".");
1317      }
1318      return &m_samples;
1319    }
1320    
1321    
1322    const DataTypes::RealVectorType*
1323    DataLazy::resolveNodeNP1OUT_2P(int tid, int sampleNo, size_t& roffset) const
1324    {
1325      if (m_readytype!='E')
1326      {
1327        throw DataException("Programmer error - resolveNodeNP1OUT_2P should only be called on expanded Data.");
1328      }
1329      if (m_op==IDENTITY)
1330      {
1331        throw DataException("Programmer error - resolveNodeNP1OUT_2P should not be called on identity nodes.");
1332      }
1333      size_t subroffset;
1334      size_t offset;
1335      const RealVectorType* leftres=m_left->resolveNodeSample(tid, sampleNo, subroffset);
1336      roffset=m_samplesize*tid;
1337      offset=roffset;
1338      size_t loop=0;
1339      size_t numsteps=(m_readytype=='E')?getNumDPPSample():1;
1340      size_t outstep=getNoValues();
1341      size_t instep=m_left->getNoValues();
1342      switch (m_op)
1343      {
1344        case SWAP:
1345            for (loop=0;loop<numsteps;++loop)
1346            {
1347                DataMaths::swapaxes(*leftres,m_left->getShape(),subroffset, m_samples, getShape(),offset, m_axis_offset, m_transpose);
1348                subroffset+=instep;
1349                offset+=outstep;
1350            }
1351            break;
1352        default:
1353            throw DataException("Programmer error - resolveNodeNP1OUT2P can not resolve operator "+opToString(m_op)+".");
1354      }
1355      return &m_samples;
1356    }
1357    
1358    const DataTypes::RealVectorType*
1359    DataLazy::resolveNodeCondEval(int tid, int sampleNo, size_t& roffset) const
1360    {
1361      if (m_readytype!='E')
1362      {
1363        throw DataException("Programmer error - resolveNodeCondEval should only be called on expanded Data.");
1364      }
1365      if (m_op!=CONDEVAL)
1366      {
1367        throw DataException("Programmer error - resolveNodeCondEval should only be called on CONDEVAL nodes.");
1368      }
1369      size_t subroffset;
1370    
1371      const RealVectorType* maskres=m_mask->resolveNodeSample(tid, sampleNo, subroffset);
1372      const RealVectorType* srcres=0;
1373      if ((*maskres)[subroffset]>0)
1374      {
1375            srcres=m_left->resolveNodeSample(tid, sampleNo, subroffset);
1376      }
1377      else
1378      {
1379            srcres=m_right->resolveNodeSample(tid, sampleNo, subroffset);
1380      }
1381    
1382      // Now we need to copy the result
1383    
1384      roffset=m_samplesize*tid;
1385      for (int i=0;i<m_samplesize;++i)
1386      {
1387            m_samples[roffset+i]=(*srcres)[subroffset+i];  
1388      }
1389    
1390      return &m_samples;
1391    }
1392    
1393    // This method assumes that any subexpressions which evaluate to Constant or Tagged Data
1394    // have already been collapsed to IDENTITY. So we must have at least one expanded child.
1395    // If both children are expanded, then we can process them in a single operation (we treat
1396    // the whole sample as one big datapoint.
1397    // If one of the children is not expanded, then we need to treat each point in the sample
1398    // individually.
1399    // There is an additional complication when scalar operations are considered.
1400    // For example, 2+Vector.
1401    // In this case each double within the point is treated individually
1402    const DataTypes::RealVectorType*
1403    DataLazy::resolveNodeBinary(int tid, int sampleNo, size_t& roffset) const
1404    {
1405    LAZYDEBUG(cout << "Resolve binary: " << toString() << endl;)
1406    
1407      size_t lroffset=0, rroffset=0;        // offsets in the left and right result vectors
1408            // first work out which of the children are expanded
1409      bool leftExp=(m_left->m_readytype=='E');
1410      bool rightExp=(m_right->m_readytype=='E');
1411      if (!leftExp && !rightExp)
1412      {
1413            throw DataException("Programmer Error - please use collapse if neither argument has type 'E'.");
1414      }
1415      bool leftScalar=(m_left->getRank()==0);
1416      bool rightScalar=(m_right->getRank()==0);
1417      if ((m_left->getRank()!=m_right->getRank()) && (!leftScalar && !rightScalar))
1418      {
1419            throw DataException("resolveBinary - ranks of arguments must match unless one of them is scalar.");
1420      }
1421      size_t leftsize=m_left->getNoValues();
1422      size_t rightsize=m_right->getNoValues();
1423      size_t chunksize=1;                   // how many doubles will be processed in one go
1424      int leftstep=0;               // how far should the left offset advance after each step
1425      int rightstep=0;
1426      int numsteps=0;               // total number of steps for the inner loop
1427      int oleftstep=0;      // the o variables refer to the outer loop
1428      int orightstep=0;     // The outer loop is only required in cases where there is an extended scalar
1429      int onumsteps=1;
1430      
1431      bool LES=(leftExp && leftScalar);     // Left is an expanded scalar
1432      bool RES=(rightExp && rightScalar);
1433      bool LS=(!leftExp && leftScalar);     // left is a single scalar
1434      bool RS=(!rightExp && rightScalar);
1435      bool LN=(!leftExp && !leftScalar);    // left is a single non-scalar
1436      bool RN=(!rightExp && !rightScalar);
1437      bool LEN=(leftExp && !leftScalar);    // left is an expanded non-scalar
1438      bool REN=(rightExp && !rightScalar);
1439    
1440      if ((LES && RES) || (LEN && REN))     // both are Expanded scalars or both are expanded non-scalars
1441      {
1442            chunksize=m_left->getNumDPPSample()*leftsize;
1443            leftstep=0;
1444            rightstep=0;
1445            numsteps=1;
1446      }
1447      else if (LES || RES)
1448      {
1449            chunksize=1;
1450            if (LES)                // left is an expanded scalar
1451            {
1452                    if (RS)
1453                    {
1454                       leftstep=1;
1455                       rightstep=0;
1456                       numsteps=m_left->getNumDPPSample();
1457                    }
1458                    else            // RN or REN
1459                    {
1460                       leftstep=0;
1461                       oleftstep=1;
1462                       rightstep=1;
1463                       orightstep=(RN ? -(int)rightsize : 0);
1464                       numsteps=rightsize;
1465                       onumsteps=m_left->getNumDPPSample();
1466                    }
1467            }
1468            else            // right is an expanded scalar
1469            {
1470                    if (LS)
1471                    {
1472                       rightstep=1;
1473                       leftstep=0;
1474                       numsteps=m_right->getNumDPPSample();
1475                    }
1476                    else
1477                    {
1478                       rightstep=0;
1479                       orightstep=1;
1480                       leftstep=1;
1481                       oleftstep=(LN ? -(int)leftsize : 0);
1482                       numsteps=leftsize;
1483                       onumsteps=m_right->getNumDPPSample();
1484                    }
1485            }
1486      }
1487      else  // this leaves (LEN, RS), (LEN, RN) and their transposes
1488      {
1489            if (LEN)        // and Right will be a single value
1490            {
1491                    chunksize=rightsize;
1492                    leftstep=rightsize;
1493                    rightstep=0;
1494                    numsteps=m_left->getNumDPPSample();
1495                    if (RS)
1496                    {
1497                       numsteps*=leftsize;
1498                    }
1499            }
1500            else    // REN
1501            {
1502                    chunksize=leftsize;
1503                    rightstep=leftsize;
1504                    leftstep=0;
1505                    numsteps=m_right->getNumDPPSample();
1506                    if (LS)
1507                    {
1508                       numsteps*=rightsize;
1509                    }
1510            }
1511      }
1512    
1513      int resultStep=max(leftstep,rightstep);       // only one (at most) should be !=0
1514            // Get the values of sub-expressions
1515      const RealVectorType* left=m_left->resolveNodeSample(tid,sampleNo,lroffset);      
1516      const RealVectorType* right=m_right->resolveNodeSample(tid,sampleNo,rroffset);
1517    LAZYDEBUG(cout << "Post sub calls in " << toString() << endl;)
1518    LAZYDEBUG(cout << "shapes=" << DataTypes::shapeToString(m_left->getShape()) << "," << DataTypes::shapeToString(m_right->getShape()) << endl;)
1519    LAZYDEBUG(cout << "chunksize=" << chunksize << endl << "leftstep=" << leftstep << " rightstep=" << rightstep;)
1520    LAZYDEBUG(cout << " numsteps=" << numsteps << endl << "oleftstep=" << oleftstep << " orightstep=" << orightstep;)
1521    LAZYDEBUG(cout << "onumsteps=" << onumsteps << endl;)
1522    LAZYDEBUG(cout << " DPPS=" << m_left->getNumDPPSample() << "," <<m_right->getNumDPPSample() << endl;)
1523    LAZYDEBUG(cout << "" << LS << RS << LN << RN << LES << RES <<LEN << REN <<   endl;)
1524    
1525    LAZYDEBUG(cout << "Left res["<< lroffset<< "]=" << (*left)[lroffset] << endl;)
1526    LAZYDEBUG(cout << "Right res["<< rroffset<< "]=" << (*right)[rroffset] << endl;)
1527    
1528    
1529      roffset=m_samplesize*tid;
1530      double* resultp=&(m_samples[roffset]);                // results are stored at the vector offset we received
1531      switch(m_op)
1532      {
1533        case ADD:
1534            //PROC_OP(NO_ARG,plus<double>());
1535          DataMaths::binaryOpVectorLazyHelper<real_t, real_t, real_t>(resultp,
1536                 &(*left)[0],
1537                 &(*right)[0],
1538                 chunksize,
1539                 onumsteps,
1540                 numsteps,
1541                 resultStep,
1542                 leftstep,
1543                 rightstep,
1544                 oleftstep,
1545                 orightstep,
1546                 lroffset,
1547                 rroffset,
1548                 escript::ESFunction::PLUSF);  
1549            break;
1550        case SUB:
1551          DataMaths::binaryOpVectorLazyHelper<real_t, real_t, real_t>(resultp,
1552                 &(*left)[0],
1553                 &(*right)[0],
1554                 chunksize,
1555                 onumsteps,
1556                 numsteps,
1557                 resultStep,
1558                 leftstep,
1559                 rightstep,
1560                 oleftstep,
1561                 orightstep,
1562                 lroffset,
1563                 rroffset,
1564                 escript::ESFunction::MINUSF);        
1565            //PROC_OP(NO_ARG,minus<double>());
1566            break;
1567        case MUL:
1568            //PROC_OP(NO_ARG,multiplies<double>());
1569          DataMaths::binaryOpVectorLazyHelper<real_t, real_t, real_t>(resultp,
1570                 &(*left)[0],
1571                 &(*right)[0],
1572                 chunksize,
1573                 onumsteps,
1574                 numsteps,
1575                 resultStep,
1576                 leftstep,
1577                 rightstep,
1578                 oleftstep,
1579                 orightstep,
1580                 lroffset,
1581                 rroffset,
1582                 escript::ESFunction::MULTIPLIESF);      
1583            break;
1584        case DIV:
1585            //PROC_OP(NO_ARG,divides<double>());
1586          DataMaths::binaryOpVectorLazyHelper<real_t, real_t, real_t>(resultp,
1587                 &(*left)[0],
1588                 &(*right)[0],
1589                 chunksize,
1590                 onumsteps,
1591                 numsteps,
1592                 resultStep,
1593                 leftstep,
1594                 rightstep,
1595                 oleftstep,
1596                 orightstep,
1597                 lroffset,
1598                 rroffset,
1599                 escript::ESFunction::DIVIDESF);          
1600            break;
1601        case POW:
1602           //PROC_OP(double (double,double),::pow);
1603          DataMaths::binaryOpVectorLazyHelper<real_t, real_t, real_t>(resultp,
1604                 &(*left)[0],
1605                 &(*right)[0],
1606                 chunksize,
1607                 onumsteps,
1608                 numsteps,
1609                 resultStep,
1610                 leftstep,
1611                 rightstep,
1612                 oleftstep,
1613                 orightstep,
1614                 lroffset,
1615                 rroffset,
1616                 escript::ESFunction::POWF);          
1617            break;
1618        default:
1619            throw DataException("Programmer error - resolveBinary can not resolve operator "+opToString(m_op)+".");
1620      }
1621    LAZYDEBUG(cout << "Result res[" << roffset<< "]" << m_samples[roffset] << endl;)
1622      return &m_samples;
1623    }
1624    
1625    
1626    // This method assumes that any subexpressions which evaluate to Constant or Tagged Data
1627    // have already been collapsed to IDENTITY. So we must have at least one expanded child.
1628    // unlike the other resolve helpers, we must treat these datapoints separately.
1629    const DataTypes::RealVectorType*
1630    DataLazy::resolveNodeTProd(int tid, int sampleNo, size_t& roffset) const
1631    {
1632    LAZYDEBUG(cout << "Resolve TensorProduct: " << toString() << endl;)
1633    
1634      size_t lroffset=0, rroffset=0;        // offsets in the left and right result vectors
1635            // first work out which of the children are expanded
1636      bool leftExp=(m_left->m_readytype=='E');
1637      bool rightExp=(m_right->m_readytype=='E');
1638      int steps=getNumDPPSample();
1639      int leftStep=(leftExp? m_left->getNoValues() : 0);            // do not have scalars as input to this method
1640      int rightStep=(rightExp?m_right->getNoValues() : 0);
1641    
1642      int resultStep=getNoValues();
1643      roffset=m_samplesize*tid;
1644      size_t offset=roffset;
1645    
1646      const RealVectorType* left=m_left->resolveNodeSample(tid, sampleNo, lroffset);
1647    
1648      const RealVectorType* right=m_right->resolveNodeSample(tid, sampleNo, rroffset);
1649    
1650    LAZYDEBUG(cerr << "[Left shape]=" << DataTypes::shapeToString(m_left->getShape()) << "\n[Right shape]=" << DataTypes::shapeToString(m_right->getShape()) << " result=" <<DataTypes::shapeToString(getShape()) <<  endl;
1651    cout << getNoValues() << endl;)
1652    
1653    
1654    LAZYDEBUG(cerr << "Post sub calls: " << toString() << endl;)
1655    LAZYDEBUG(cout << "LeftExp=" << leftExp << " rightExp=" << rightExp << endl;)
1656    LAZYDEBUG(cout << "LeftR=" << m_left->getRank() << " rightExp=" << m_right->getRank() << endl;)
1657    LAZYDEBUG(cout << "LeftSize=" << m_left->getNoValues() << " RightSize=" << m_right->getNoValues() << endl;)
1658    LAZYDEBUG(cout << "m_samplesize=" << m_samplesize << endl;)
1659    LAZYDEBUG(cout << "outputshape=" << DataTypes::shapeToString(getShape()) << endl;)
1660    LAZYDEBUG(cout << "DPPS=" << m_right->getNumDPPSample() <<"."<<endl;)
1661    
1662      double* resultp=&(m_samples[offset]);         // results are stored at the vector offset we received
1663      switch(m_op)
1664      {
1665        case PROD:
1666            for (int i=0;i<steps;++i,resultp+=resultStep)
1667            {
1668              const double *ptr_0 = &((*left)[lroffset]);
1669              const double *ptr_1 = &((*right)[rroffset]);
1670    
1671    LAZYDEBUG(cout << DataTypes::pointToString(*left, m_left->getShape(),lroffset,"LEFT") << endl;)
1672    LAZYDEBUG(cout << DataTypes::pointToString(*right,m_right->getShape(),rroffset, "RIGHT") << endl;)
1673    
1674              matrix_matrix_product(m_SL, m_SM, m_SR, ptr_0, ptr_1, resultp, m_transpose);
1675    
1676              lroffset+=leftStep;
1677              rroffset+=rightStep;
1678            }
1679            break;
1680        default:
1681            throw DataException("Programmer error - resolveTProduct can not resolve operator "+opToString(m_op)+".");
1682      }
1683      roffset=offset;
1684      return &m_samples;
1685    }
1686    
1687    
1688    const DataTypes::RealVectorType*
1689    DataLazy::resolveSample(int sampleNo, size_t& roffset) const
1690    {
1691    #ifdef _OPENMP
1692            int tid=omp_get_thread_num();
1693    #else
1694            int tid=0;
1695    #endif
1696    
1697    #ifdef LAZY_STACK_PROF
1698            stackstart[tid]=&tid;
1699            stackend[tid]=&tid;
1700            const DataTypes::RealVectorType* r=resolveNodeSample(tid, sampleNo, roffset);
1701            size_t d=(size_t)stackstart[tid]-(size_t)stackend[tid];
1702            #pragma omp critical
1703            if (d>maxstackuse)
1704            {
1705    cout << "Max resolve Stack use " << d << endl;
1706                    maxstackuse=d;
1707            }
1708            return r;
1709    #else
1710            return resolveNodeSample(tid, sampleNo, roffset);
1711    #endif
1712    }
1713    
1714    
1715    // This needs to do the work of the identity constructor
1716    void
1717    DataLazy::resolveToIdentity()
1718    {
1719       if (m_op==IDENTITY)
1720            return;
1721       DataReady_ptr p=resolveNodeWorker();
1722       makeIdentity(p);
1723    }
1724    
1725    void DataLazy::makeIdentity(const DataReady_ptr& p)
1726    {
1727       m_op=IDENTITY;
1728       m_axis_offset=0;
1729       m_transpose=0;
1730       m_SL=m_SM=m_SR=0;
1731       m_children=m_height=0;
1732       m_id=p;
1733       if(p->isConstant()) {m_readytype='C';}
1734       else if(p->isExpanded()) {m_readytype='E';}
1735       else if (p->isTagged()) {m_readytype='T';}
1736       else {throw DataException("Unknown DataReady instance in convertToIdentity constructor.");}
1737       m_samplesize=p->getNumDPPSample()*p->getNoValues();
1738       m_left.reset();
1739       m_right.reset();
1740    }
1741    
1742    
1743  DataReady_ptr  DataReady_ptr
1744  DataLazy::resolve()  DataLazy::resolve()
1745  {  {
1746    // This is broken!     We need to have a buffer per thread!      resolveToIdentity();
1747    // so the allocation of v needs to move inside the loop somehow      return m_id;
1748    }
1749    
 cout << "Sample size=" << m_samplesize << endl;  
 cout << "Buffers=" << m_buffsRequired << endl;  
1750    
1751    size_t threadbuffersize=m_samplesize*(max(1,m_buffsRequired)+1);  /* This is really a static method but I think that caused problems in windows */
1752    int numthreads=1;  void
1753    DataLazy::resolveGroupWorker(std::vector<DataLazy*>& dats)
1754    {
1755      if (dats.empty())
1756      {
1757            return;
1758      }
1759      vector<DataLazy*> work;
1760      FunctionSpace fs=dats[0]->getFunctionSpace();
1761      bool match=true;
1762      for (int i=dats.size()-1;i>=0;--i)
1763      {
1764            if (dats[i]->m_readytype!='E')
1765            {
1766                    dats[i]->collapse();
1767            }
1768            if (dats[i]->m_op!=IDENTITY)
1769            {
1770                    work.push_back(dats[i]);
1771                    if (fs!=dats[i]->getFunctionSpace())
1772                    {
1773                            match=false;
1774                    }
1775            }
1776      }
1777      if (work.empty())
1778      {
1779            return;         // no work to do
1780      }
1781      if (match)    // all functionspaces match.  Yes I realise this is overly strict
1782      {             // it is possible that dats[0] is one of the objects which we discarded and
1783                    // all the other functionspaces match.
1784            vector<DataExpanded*> dep;
1785            vector<RealVectorType*> vecs;
1786            for (int i=0;i<work.size();++i)
1787            {
1788                    dep.push_back(new DataExpanded(fs,work[i]->getShape(), RealVectorType(work[i]->getNoValues())));
1789                    vecs.push_back(&(dep[i]->getVectorRW()));
1790            }
1791            int totalsamples=work[0]->getNumSamples();
1792            const RealVectorType* res=0; // Storage for answer
1793            int sample;
1794            #pragma omp parallel private(sample, res)
1795            {
1796                size_t roffset=0;
1797                #pragma omp for schedule(static)
1798                for (sample=0;sample<totalsamples;++sample)
1799                {
1800                    roffset=0;
1801                    int j;
1802                    for (j=work.size()-1;j>=0;--j)
1803                    {
1804  #ifdef _OPENMP  #ifdef _OPENMP
1805    numthreads=omp_get_max_threads();                      res=work[j]->resolveNodeSample(omp_get_thread_num(),sample,roffset);
1806    int threadnum=0;  #else
1807  #endif                      res=work[j]->resolveNodeSample(0,sample,roffset);
1808    ValueType v(numthreads*threadbuffersize);  #endif
1809  cout << "Buffer created with size=" << v.size() << endl;                      RealVectorType::size_type outoffset=dep[j]->getPointOffset(sample,0);
1810    ValueType dummy(getNoValues());                      memcpy(&((*vecs[j])[outoffset]),&((*res)[roffset]),work[j]->m_samplesize*sizeof(RealVectorType::ElementType));
1811    DataExpanded* result=new DataExpanded(getFunctionSpace(),getShape(),dummy);                  }
1812    ValueType& resvec=result->getVector();              }
1813            }
1814            // Now we need to load the new results as identity ops into the lazy nodes
1815            for (int i=work.size()-1;i>=0;--i)
1816            {
1817                work[i]->makeIdentity(REFCOUNTNS::dynamic_pointer_cast<DataReady>(dep[i]->getPtr()));
1818            }
1819      }
1820      else  // functionspaces do not match
1821      {
1822            for (int i=0;i<work.size();++i)
1823            {
1824                    work[i]->resolveToIdentity();
1825            }
1826      }
1827    }
1828    
1829    
1830    
1831    // This version of resolve uses storage in each node to hold results
1832    DataReady_ptr
1833    DataLazy::resolveNodeWorker()
1834    {
1835      if (m_readytype!='E')         // if the whole sub-expression is Constant or Tagged, then evaluate it normally
1836      {
1837        collapse();
1838      }
1839      if (m_op==IDENTITY)           // So a lazy expression of Constant or Tagged data will be returned here.
1840      {
1841        return m_id;
1842      }
1843            // from this point on we must have m_op!=IDENTITY and m_readytype=='E'
1844      DataExpanded* result=new DataExpanded(getFunctionSpace(),getShape(),  RealVectorType(getNoValues()));
1845      RealVectorType& resvec=result->getVectorRW();
1846    DataReady_ptr resptr=DataReady_ptr(result);    DataReady_ptr resptr=DataReady_ptr(result);
1847    
1848    int sample;    int sample;
   int resoffset;  
1849    int totalsamples=getNumSamples();    int totalsamples=getNumSamples();
1850    #pragma omp parallel for private(sample,resoffset,threadnum) schedule(static)    const RealVectorType* res=0;       // Storage for answer
1851    for (sample=0;sample<totalsamples;++sample)  LAZYDEBUG(cout << "Total number of samples=" <<totalsamples << endl;)
1852      #pragma omp parallel private(sample,res)
1853    {    {
1854            size_t roffset=0;
1855    #ifdef LAZY_STACK_PROF
1856            stackstart[omp_get_thread_num()]=&roffset;
1857            stackend[omp_get_thread_num()]=&roffset;
1858    #endif
1859            #pragma omp for schedule(static)
1860            for (sample=0;sample<totalsamples;++sample)
1861            {
1862                    roffset=0;
1863  #ifdef _OPENMP  #ifdef _OPENMP
1864      const double* res=resolveSample(v,sample,threadbuffersize*omp_get_thread_num());                  res=resolveNodeSample(omp_get_thread_num(),sample,roffset);
1865  #else  #else
1866      const double* res=resolveSample(v,sample,0);   // this would normally be v, but not if its a single IDENTITY op.                  res=resolveNodeSample(0,sample,roffset);
1867  #endif  #endif
1868      resoffset=result->getPointOffset(sample,0);  LAZYDEBUG(cout << "Sample #" << sample << endl;)
1869      for (int i=0;i<m_samplesize;++i,++resoffset)    // copy values into the output vector  LAZYDEBUG(cout << "Final res[" << roffset<< "]=" << (*res)[roffset] << (*res)[roffset]<< endl; )
1870      {                  RealVectorType::size_type outoffset=result->getPointOffset(sample,0);
1871      resvec[resoffset]=res[i];                  memcpy(&(resvec[outoffset]),&((*res)[roffset]),m_samplesize*sizeof(RealVectorType::ElementType));
1872      }          }
1873    }    }
1874    #ifdef LAZY_STACK_PROF
1875      for (int i=0;i<getNumberOfThreads();++i)
1876      {
1877            size_t r=((size_t)stackstart[i] - (size_t)stackend[i]);
1878    //      cout << i << " " << stackstart[i] << " .. " << stackend[i] << " = " <<  r << endl;
1879            if (r>maxstackuse)
1880            {
1881                    maxstackuse=r;
1882            }
1883      }
1884      cout << "Max resolve Stack use=" << maxstackuse << endl;
1885    #endif
1886    return resptr;    return resptr;
1887  }  }
1888    
# Line 430  std::string Line 1890  std::string
1890  DataLazy::toString() const  DataLazy::toString() const
1891  {  {
1892    ostringstream oss;    ostringstream oss;
1893    oss << "Lazy Data:";    oss << "Lazy Data: [depth=" << m_height<< "] ";
1894    intoString(oss);    switch (escriptParams.getLAZY_STR_FMT())
1895      {
1896      case 1:       // tree format
1897            oss << endl;
1898            intoTreeString(oss,"");
1899            break;
1900      case 2:       // just the depth
1901            break;
1902      default:
1903            intoString(oss);
1904            break;
1905      }
1906    return oss.str();    return oss.str();
1907  }  }
1908    
1909    
1910  void  void
1911  DataLazy::intoString(ostringstream& oss) const  DataLazy::intoString(ostringstream& oss) const
1912  {  {
1913    //    oss << "[" << m_children <<";"<<m_height <<"]";
1914    switch (getOpgroup(m_op))    switch (getOpgroup(m_op))
1915    {    {
1916    case G_IDENTITY:    case G_IDENTITY:
1917      oss << '@' << m_id.get();          if (m_id->isExpanded())
1918      break;          {
1919               oss << "E";
1920            }
1921            else if (m_id->isTagged())
1922            {
1923              oss << "T";
1924            }
1925            else if (m_id->isConstant())
1926            {
1927              oss << "C";
1928            }
1929            else
1930            {
1931              oss << "?";
1932            }
1933            oss << '@' << m_id.get();
1934            break;
1935    case G_BINARY:    case G_BINARY:
1936      oss << '(';          oss << '(';
1937      m_left->intoString(oss);          m_left->intoString(oss);
1938      oss << ' ' << opToString(m_op) << ' ';          oss << ' ' << opToString(m_op) << ' ';
1939      m_right->intoString(oss);          m_right->intoString(oss);
1940      oss << ')';          oss << ')';
1941      break;          break;
1942    case G_UNARY:    case G_UNARY:
1943      oss << opToString(m_op) << '(';    case G_UNARY_P:
1944      m_left->intoString(oss);    case G_NP1OUT:
1945      oss << ')';    case G_NP1OUT_P:
1946      break;    case G_REDUCTION:
1947            oss << opToString(m_op) << '(';
1948            m_left->intoString(oss);
1949            oss << ')';
1950            break;
1951      case G_TENSORPROD:
1952            oss << opToString(m_op) << '(';
1953            m_left->intoString(oss);
1954            oss << ", ";
1955            m_right->intoString(oss);
1956            oss << ')';
1957            break;
1958      case G_NP1OUT_2P:
1959            oss << opToString(m_op) << '(';
1960            m_left->intoString(oss);
1961            oss << ", " << m_axis_offset << ", " << m_transpose;
1962            oss << ')';
1963            break;
1964      case G_CONDEVAL:
1965            oss << opToString(m_op)<< '(' ;
1966            m_mask->intoString(oss);
1967            oss << " ? ";
1968            m_left->intoString(oss);
1969            oss << " : ";
1970            m_right->intoString(oss);
1971            oss << ')';
1972            break;
1973    default:    default:
1974      oss << "UNKNOWN";          oss << "UNKNOWN";
1975    }    }
1976  }  }
1977    
1978  // Note that in this case, deepCopy does not make copies of the leaves.  
1979  // Hopefully copy on write (or whatever we end up using) will take care of this.  void
1980    DataLazy::intoTreeString(ostringstream& oss, string indent) const
1981    {
1982      oss << '[' << m_rank << ':' << setw(3) << m_samplesize << "] " << indent;
1983      switch (getOpgroup(m_op))
1984      {
1985      case G_IDENTITY:
1986            if (m_id->isExpanded())
1987            {
1988               oss << "E";
1989            }
1990            else if (m_id->isTagged())
1991            {
1992              oss << "T";
1993            }
1994            else if (m_id->isConstant())
1995            {
1996              oss << "C";
1997            }
1998            else
1999            {
2000              oss << "?";
2001            }
2002            oss << '@' << m_id.get() << endl;
2003            break;
2004      case G_BINARY:
2005            oss << opToString(m_op) << endl;
2006            indent+='.';
2007            m_left->intoTreeString(oss, indent);
2008            m_right->intoTreeString(oss, indent);
2009            break;
2010      case G_UNARY:
2011      case G_UNARY_P:
2012      case G_NP1OUT:
2013      case G_NP1OUT_P:
2014      case G_REDUCTION:
2015            oss << opToString(m_op) << endl;
2016            indent+='.';
2017            m_left->intoTreeString(oss, indent);
2018            break;
2019      case G_TENSORPROD:
2020            oss << opToString(m_op) << endl;
2021            indent+='.';
2022            m_left->intoTreeString(oss, indent);
2023            m_right->intoTreeString(oss, indent);
2024            break;
2025      case G_NP1OUT_2P:
2026            oss << opToString(m_op) << ", " << m_axis_offset << ", " << m_transpose<< endl;
2027            indent+='.';
2028            m_left->intoTreeString(oss, indent);
2029            break;
2030      default:
2031            oss << "UNKNOWN";
2032      }
2033    }
2034    
2035    
2036  DataAbstract*  DataAbstract*
2037  DataLazy::deepCopy()  DataLazy::deepCopy() const
2038  {  {
2039    if (m_op==IDENTITY)    switch (getOpgroup(m_op))
2040    {    {
2041      return new DataLazy(m_left);    // we don't need to copy the child here    case G_IDENTITY:  return new DataLazy(m_id->deepCopy()->getPtr());
2042      case G_UNARY:
2043      case G_REDUCTION:      return new DataLazy(m_left->deepCopy()->getPtr(),m_op);
2044      case G_UNARY_P:       return new DataLazy(m_left->deepCopy()->getPtr(), m_op, m_tol);
2045      case G_BINARY:        return new DataLazy(m_left->deepCopy()->getPtr(),m_right->deepCopy()->getPtr(),m_op);
2046      case G_NP1OUT: return new DataLazy(m_left->deepCopy()->getPtr(), m_right->deepCopy()->getPtr(),m_op);
2047      case G_TENSORPROD: return new DataLazy(m_left->deepCopy()->getPtr(), m_right->deepCopy()->getPtr(), m_op, m_axis_offset, m_transpose);
2048      case G_NP1OUT_P:   return new DataLazy(m_left->deepCopy()->getPtr(),m_op,  m_axis_offset);
2049      case G_NP1OUT_2P:  return new DataLazy(m_left->deepCopy()->getPtr(), m_op, m_axis_offset, m_transpose);
2050      default:
2051            throw DataException("Programmer error - do not know how to deepcopy operator "+opToString(m_op)+".");
2052    }    }
   return new DataLazy(m_left->deepCopy()->getPtr(),m_right->deepCopy()->getPtr(),m_op);  
2053  }  }
2054    
2055    
2056  DataTypes::ValueType::size_type  
2057    // There is no single, natural interpretation of getLength on DataLazy.
2058    // Instances of DataReady can look at the size of their vectors.
2059    // For lazy though, it could be the size the data would be if it were resolved;
2060    // or it could be some function of the lengths of the DataReady instances which
2061    // form part of the expression.
2062    // Rather than have people making assumptions, I have disabled the method.
2063    DataTypes::RealVectorType::size_type
2064  DataLazy::getLength() const  DataLazy::getLength() const
2065  {  {
2066    return m_length;    throw DataException("getLength() does not make sense for lazy data.");
2067  }  }
2068    
2069    
# Line 486  DataLazy::getSlice(const DataTypes::Regi Line 2073  DataLazy::getSlice(const DataTypes::Regi
2073    throw DataException("getSlice - not implemented for Lazy objects.");    throw DataException("getSlice - not implemented for Lazy objects.");
2074  }  }
2075    
2076  DataTypes::ValueType::size_type  
2077    // To do this we need to rely on our child nodes
2078    DataTypes::RealVectorType::size_type
2079    DataLazy::getPointOffset(int sampleNo,
2080                     int dataPointNo)
2081    {
2082      if (m_op==IDENTITY)
2083      {
2084            return m_id->getPointOffset(sampleNo,dataPointNo);
2085      }
2086      if (m_readytype!='E')
2087      {
2088            collapse();
2089            return m_id->getPointOffset(sampleNo,dataPointNo);
2090      }
2091      // at this point we do not have an identity node and the expression will be Expanded
2092      // so we only need to know which child to ask
2093      if (m_left->m_readytype=='E')
2094      {
2095            return m_left->getPointOffset(sampleNo,dataPointNo);
2096      }
2097      else
2098      {
2099            return m_right->getPointOffset(sampleNo,dataPointNo);
2100      }
2101    }
2102    
2103    // To do this we need to rely on our child nodes
2104    DataTypes::RealVectorType::size_type
2105  DataLazy::getPointOffset(int sampleNo,  DataLazy::getPointOffset(int sampleNo,
2106                   int dataPointNo) const                   int dataPointNo) const
2107  {  {
2108    throw DataException("getPointOffset - not implemented for Lazy objects - yet.");    if (m_op==IDENTITY)
2109      {
2110            return m_id->getPointOffset(sampleNo,dataPointNo);
2111      }
2112      if (m_readytype=='E')
2113      {
2114        // at this point we do not have an identity node and the expression will be Expanded
2115        // so we only need to know which child to ask
2116        if (m_left->m_readytype=='E')
2117        {
2118            return m_left->getPointOffset(sampleNo,dataPointNo);
2119        }
2120        else
2121        {
2122            return m_right->getPointOffset(sampleNo,dataPointNo);
2123        }
2124      }
2125      if (m_readytype=='C')
2126      {
2127            return m_left->getPointOffset(sampleNo,dataPointNo); // which child doesn't matter
2128      }
2129      throw DataException("Programmer error - getPointOffset on lazy data may require collapsing (but this object is marked const).");
2130    }
2131    
2132    
2133    // I have decided to let Data:: handle this issue.
2134    void
2135    DataLazy::setToZero()
2136    {
2137    //   DataTypes::RealVectorType v(getNoValues(),0);
2138    //   m_id=DataReady_ptr(new DataConstant(getFunctionSpace(),getShape(),v));
2139    //   m_op=IDENTITY;
2140    //   m_right.reset();  
2141    //   m_left.reset();
2142    //   m_readytype='C';
2143    //   m_buffsRequired=1;
2144    
2145      (void)privdebug;  // to stop the compiler complaining about unused privdebug
2146      throw DataException("Programmer error - setToZero not supported for DataLazy (DataLazy objects should be read only).");
2147  }  }
2148    
2149  }   // end namespace  bool
2150    DataLazy::actsExpanded() const
2151    {
2152            return (m_readytype=='E');
2153    }
2154    
2155    } // end namespace
2156    

Legend:
Removed from v.1888  
changed lines
  Added in v.6057

  ViewVC Help
Powered by ViewVC 1.1.26