history_before_2005

来自「FreeFem++可以生成高质量的有限元网格。可以用于流体力学」· 代码 · 共 639 行 · 第 1/2 页

 

21/08/2003:
 -  Bug in optmization in case on non constant robin boundary condition
    correct 
 -  add grey=1  ion plot command to make  grey plot, and //\index{grey=}\index{plot!grey=}
   correct small mistake ( on mac and X11 , ...)
 
 
Make version 1.34 (22/08/2003)
------------------------------

16/09/2003
 - bug in local variable stack if the size is larger the 8 ( complex)
  complex a;real b;
  a=0;b=1; 
  // here a is 0+i; 
  correction is:
  voila la correction dans AFunction.hpp vers la ligne 1486: 
  template<class T>
  inline Type_Expr  NewVariable(aType t,size_t &off) 
   { 
     size_t o= align8(off);//  align    
   //  off += t->un\_ptr\_type->size;
   // bug    off += t->size;
    off += t->un_ptr_type->size; // correction 
    return  Type_Expr(t,new T(o,t));
   }

02/11/2003
  -correct probleme of renumbering the traingles of a  mesh 
   when reading in the bamg software (with readmesh command).
   This implies error when we restore mesh and non P1 finite element
   solution, see example saveandrestore.edp. 
  
28/11/2003
  -correction of bug in window version 
   add the umfpack sparce linear solver (this solver have some problem
   in some case), so I do'nt put this one in default linearsolver. 
   remark, I think is due to the way of taking Dirichlet boundary condition
   Huge value on diagonal (tgv=1e30) and Stokes matrice.//\index{problem!tgv=}\index{solve!tgv=}

  -put this file on the web
  -put on the web 
  -the codewarroir projet to build arpack ands umfpack lib.
  
Make version 1.36   
-------------------

 - Change the graphic window in Window xx version

8/12/2003
 - repare in mpi version  get mesh via mpi
 - forget build quadtree soo the interpolation bugs
 - add 4 line in file parallelempi.cpp in mesh serialization (recivied
   //  add 3 line FH 08/12/2003  forget build quadtree sorry      
       Fem2D::R2 Pn,Px;
       a->BoundingBox(Pn,Px);
       a->quadtree=new FQuadTree(m,Pn,Px,m->nv);
      
18/12/2003
  - add in mpi version://\index{broadcast}
    broadcast(processor(1),th); //  broadcast  th from proc 1 to all other.
    see the mpi exemple
    change the precision to 6 to 12 is savemesh.
  - add tool to change the default precision on ostream or ofstream 
    with the C++ syntaxe
    see saverestore.edp example.
    cout.precision(12); //\index{precision}
  - add UMFPACK linear solver ( not well test)   //\index{UMFPACK}
    http://www.cise.ufl.edu/research/sparse/umfpack
  - add full matrix with few operator //\index{matrix!full}
     real [int,int] A(10,10); 
     A= 2;
     A(5,5) = 2;
     cout << A << endl;
  -add array of mesh //\index{array!mesh}

08/01/2004
   - merge all mesh example in one file call mesh.edp
   - add discontinuous Galerkin method (see LapDG2.edp example)
   - add syntaxe to get mesh information  (see end of mesh.edp exemple)
   - add meshsplit  function to make conformal recusive locat mesh spliting //\index{meshsplit}
     (see mesh.edp example)
   - add dynamic load via dlopen  see load.edp in example++-load  //\index{dynamic load}

Make version 1.37 
-----------------

15/01/2004
  - change metrix= in metric= the in adapted mesh metric=[m11,m12,m22]//  \index{meshadapt!metric=}
    (change the ordre to compatible with the fonction order with IsMetric=1)

02/02/2004
  - change printing in gmres algorithme
  - print the size of matrix 
06/02/2004
  - change intalledges  in change of discontinuous Galerkin loop also boundary edge//\index{intalledges}
	 jump(u) is external - internal value of u with normal go to internal to external 
	 jump(u)  on boundary is -internal value of u //\index{jump}
	 average(u)  on boundary is internal value of u  //\index{average}
  - add nTonEdge to get the number of Triangles which see the  current edge//\index{nTonEdge}
  so see the new LapDG2.edp example for full detail
07/02/2004
  - correct bug in real[string] map; // map array  \index{array!string}
        change in get_element and operator <  of String 
  - add int array 
Make Version 1.38
----------------
  - remove (void) line 14 of throwassert.hpp file (erreur compile g++3.3.5 )
  make change in macro expansion of shell (no perfect to day)
     
 16/04/2004 
  - add matrix tools 
    bluid interpolation interpolation matric from a FEspace VH  to an other FEspace Vh
       matrix Ih=matrix B= interplotematrix(VH,Vh); 
	// where  Vh correspond to  line and Vh to column 
       the named parameters can be 
          t= true or false  (to bluid the transpose or not matrix)
          op=0 ,1 or 2 (to build the interpolation of value, dx,dy )
          inside=true or false  (to remove or not  all ouside interpolation  quadrature point)
     build diagonal  sparce  matrix (type is morse)
     do the matrice product of to sparce matrix (type is morse)
   add build a sparce matrix from a full matrix so if you have install UMFPACK solver
     it is possible to solve systeme with full matrix.
   add set function to change the solver 
   see exemple  sparce-matrix.edp 
	   
  - correct bug flag ouside of mesh 
    add line in file fem.cpp line 864
       outside=true; 
     so some time the P0 interpolatation is no zero
     add new king of mesh of multiplicator data :
     mesh emptyTh=emptymesh(Th); // to build a mesh with no internal point 
     see mesh.edp example

Make Version 1.40
-----------------

  - add les linear combination of sparce matrice
  - set a sparce solver to a sparce matrice (to day just UMFPACK)
  - transforme a full matrix in sparce matrix
  - add istream f("toto");
     f.good() or f.EOF to test state the file 
     remark: the state is set after the read so the previous value is wrong

02/06/2004
----------
  -  correction in call macro with string "..." parameter  
24/06/2004
----------
   add tools to bluid periodic adapmesh  see sphere.edp of exemple
   add HaveUMFPACK global variable see sparse-matrix.edp exemple   
   Change all the structure of the software 
   Use ./configure to build all the make file see (README)
   you can build the Window version with g++ under cygwin 
   ( http: cygwin.... )

Make Version 1.41 
-----------------

07/07/2004 
----------
an anonymous CVS server is avialable to get FreeFem++ source
cvs -d :pserver:[e-mail]:/Users/pubcvs/cvs login
(password = freefem++)
cvs -d :pserver:[e-mail]:/Users/pubcvs/cvs freefem++ co
the TAG release_1_41_before_packaging is created

30/08/04
--------
 Correct  divide by 0 in plot instruction  when the bounding box is flat.
 Set default solver to LU because UMFPACK sometime by bad result.
 Correct all g++-3.4 error
 Optimize Choleski and Crout solver (divide cpu time by two on pentium)
 Add stuff for adding new finite element due to the new software achitecture.

31/08/04
--------
 Add string[string] array
 correct bug in  macro expansion 
     macro parameter will be not expanded 
  example of bug:
     macro a(i) i //
     macro b  a(x)  //  bug here 
     b = 1 ; // we get "i=1" or we want "x =1"

Make Version 1.42
-----------------

11/10/04
--------
 Add Finite element with complex value, and matrix<complex> 
 see sparse-cmatrix.edp  and  FEComplex.edp files
 for a full examples

 Vh<complex> ur; 
 varf av(u,v)=int(Th)(u*v*(1i+1));
 Matrix<complex> A = vav(Vh,Vh); 
 // warning don't mixte Matrix<complex> and Matrix<real> in a expression     
 
19/10/04
--------
  correct UMFPACK driver problem, ands set this solver as a default 
    solver.
  add finite element connectivite inquire ice FE.epd exemple 
  fespace Wh(Th,P2);
   Wh.nt //  given the number of element 
   Wh.ndof //  given the number of degree of freedom
   Wh.ndofK //  given the number of degree of freedom on one element 
   Wh(k,i)  // given the number of i  degree of freedom of element k.
   //  with $0 <= i <  Wh.ndofK $ and  $0 <= k <  Wh.nt $ 

}
  build the source and compile version 1.43 
01/11/04
--------
   correct problem on period adapted mesh if orientation are reverse.
   add cvs tag release_1_43
   Now  the current version is 1.44 (CVS)

04/11/004
---------
   correct problem in LinearGMRES and LinearGC 
   in LinearCG algorithme for solving A*x = b0+b
   the matrix function can compute A*x or A*x-b0 ,
   and the other part b of the right end side can  exist or not.
   But in LinearGMRES must the function must compute A*x and 
   b0 can't exist. 
   see algo.edp  exist for  a full example

06/12/2004
__________
  change  function name interplotematrix to interpolate
  correct  bug introduce in 04/12 change
  a new graphic FreeFem++-cs: integrated development environment (located in src/ide)
  see REAME and maunal-full.pdf
  add new examples++-tutorial
  mat_interpol.edp  # to test interpolate function
  shur-comp.edp  # a shur complement domaine decomposition

11/12/2004
 TAG release_1_44
 now the current version is 1.45

29/12/2004  
   add concat operator # in macro generator.

04/01/2005
   add new quadrature formular on edge (a,b) :qf1pElump  : 1/2(f(a)+f(b))*length(ab)

05/01/2005
  correction bug  unset FE variable dans le .edp   =>
  use a null pointeur (in sol) to set the *X  array in  problem.cpp
   (*X)[K(df)]= (*sol[kfe])[SK(kdf)] ;  
   make some time make a seg trap error. 
06/01/2005
   rewrite the macro generation tool to be more to suppress some bug
   in argument subtition
   correct error in line numbering in case of newline in string
   cosmetic change in output 

20/01/2005
   count () in macro parameter to handle   couple of () in macro parameter.
   macro sum(a,b) (a+b); 
   sum((1),2) // now works and given (1)+2
   sum(atan2(2,1),2) // now works and given atan2(2,1)+2
   
   correct bug in linear combination of sparce matrix:
   the correction is:
    change line 1069, in file MatriceCreuse_tpl.hpp
           mij[make_pair(i,j)] = coef*a[k];
    to
           mij[make_pair(i,j)] += coef*a[k];
    
    Correct a segfault if second derivative of test function exist in 
    problem definition.  The Change in problem.cpp:
    resize : KN<double> buf(Vh.MaximalNbOfDF()*3*Vh.N);
    to     : KN<double> buf(Vh.MaximalNbOfDF()*last_operatortype*Vh.N);

   
24/01/2005
   add  named parameter nbtx=..  in buildmesh  function to change the default 
   maximal number of vertex in the build mesh.

26/01/2005
   correct pb with  --disable-pdf in configure.ac
   correct pdflatex and latex compilation 
   correct  FreeFem++.app problem when usind in a shell script FreeFem++-CoCoa
   correct web site and distrib file  freefem++-1.45-2.tar.gz

02/02/2005
   correct problem with socket in FLTK in freefem++-cs  (not wet finish) ALH
   add nbvx=  named parameter to change the default maximum number vertices in creted mesh
   in builmesh (see mesh.edp exemple)  
   
10/02/2005
   add 2 new quadrature formular qf7pT and qf9pT with respectively 15 and 21 points
   add  P2b a new finite element  : P2 + bubble 
   make drag and drop on ide version + copy paste 

14/02/2005
   create a error  when you do   uh = ..  when uh is a  componant of a vectorial fe 
     element function (for exemple fespace Vh(Th,TR), Vh [uh,vh]; uh = 1 (error here ;)
     because it's impossible to build   uh and vh independantly.

17/02/2005
   FreeFem++-cs: Communication speed between client and server has been
   increased ten-fold (ALH)

   
23/02/2005
     add jump and mean of complex function  (FH)

23/02/2005
    FreeFem++-cs: grammar-driven syntax highlighting works on
    (examples++-tutorial/...) Laplace.edp, adapt.edp and
    adaptindicatorP1.edp (ALH)