Schrodinger equation solver doesn't work in the latest sfepy

[GoogleCodeExporter]

Schrodinger equation solver doesn't work in the latest sfepy:

$ ./schroedinger.py -s
sfe: warning: other missing: ['functions', 'modules', 'epbc_[0-9]+|epbcs',
'lcbc_[0-9]+|lcbcs', 'nbc_[0-9]+|nbcs']
sfe: warning: left over: ['funV', 'solver_2']
sfe: reading mesh...
       nodes: 100% |############################################| Time:
00:00:00
       elements: 100% |#########################################| Time:
00:00:00
sfe: ...done in 1.20 s
sfe: setting up domain edges...
sfe: ...done in 0.89 s
sfe: setting up domain faces...
sfe: ...done in 0.66 s
sfe: creating regions...
sfe:     leaf Omega region_1000
sfe:     leaf Surface region_2
sfe: ...done in 1.04 s
sfe: equation "rhs":
sfe: dw_mass_scalar.i1.Omega( v, Psi )
sfe: equation "lhs":
sfe:   dw_laplace.i1.Omega( m.val, v, Psi )
               + dw_mass_scalar_variable.i1.Omega( matV.V, v, Psi )
sfe: describing geometries...
sfe: ...done in 0.16 s
sfe: setting up dof connectivities...
sfe: ...done in 0.00 s
sfe: matrix shape: (7964, 7964)
sfe: assembling matrix graph...
sfe: ...done in 0.10 s
sfe: matrix structural nonzeros: 117024 (1.85e-03% fill)
sfe: updating materials...
sfe:     m
sfe:     matV
sfe: ...done in 0.01 s
sfe: assembling lhs...
sfe:   setting up dof connectivities...
sfe:   ...done in 0.00 s
Traceback (most recent call last):
  File "./schroedinger.py", line 317, in <module>
    main()
  File "./schroedinger.py", line 312, in main
    evp = solveEigenProblem1( conf, options )
  File "./schroedinger.py", line 211, in solveEigenProblem1
    dwMode = 'matrix', tangentMatrix = pb.mtxA )
  File "/home/ondra/sfepy/sfe/fem/evaluate.py", line 186, in evalTermOP
    chunkSize = problem.domain.shape.nEl, **kwargs )
  File "/home/ondra/sfepy/sfe/fem/evaluate.py", line 237, in evalTerm
    chunkSize, groupCanFail = False, **kwargs )
  File "/home/ondra/sfepy/sfe/fem/evaluate.py", line 173, in assembleMatrix
    **args ):
  File "/home/ondra/sfepy/sfe/terms/termsMass.py", line 129, in __call__
    fargs = self.buildCFunArgs( state, ap, vg, **kwargs )
  File "/home/ondra/sfepy/sfe/terms/termsMass.py", line 172, in buildCFunArgs
    modeIn = 'vertex' )
  File "/home/ondra/sfepy/sfe/terms/cache.py", line 178, in __call__
    self.update( key, groupIndx, ih, **kwargs )
  File "/home/ondra/sfepy/sfe/terms/cachesBasic.py", line 350, in update
    matQP = self.data[key][ckey][ih].reshape( vshape )
ValueError: total size of new array must be unchanged

Original issue reported on code.google.com by ondrej.c...@gmail.com on 11 Jun 2008 at 3:21

[GoogleCodeExporter]

Whatever the problem is, let's make sfepy more robust to raise a meaningful 
exception
at the right place. + a test for it.

Original comment by ondrej.c...@gmail.com on 11 Jun 2008 at 3:25

[GoogleCodeExporter]

I used "hg bisect" and here is what I got:

The first bad revision is:
changeset:   180:c73e7db416dd
user:        Robert Cimrman <cimrman3@ntc.zcu.cz>
date:        Tue May 06 18:03:32 2008 +0200
summary:     fixed MatInQPDataCache to work with diffusion tensor in vertices, 
in

well, I think it doesn't need any more comments. :) Do you know how to fix it?

Original comment by ondrej.c...@gmail.com on 11 Jun 2008 at 11:00

• Added labels: Priority-High
• Removed labels: Priority-Medium

[GoogleCodeExporter]

This patch makes it run with the latest sfepy:

diff --git a/sfe/terms/cachesBasic.py b/sfe/terms/cachesBasic.py
--- a/sfe/terms/cachesBasic.py
+++ b/sfe/terms/cachesBasic.py
@@ -341,7 +341,7 @@ class MatInQPDataCache( DataCache ):

             # dq_state_in_qp() works for vectors -> make a view of
             # shape (nEl, nQP, nRow * nCol, 1).
-            vshape = shape[0:2] + (nm.prod( mat.shape[1:] ), 1)
+            vshape = shape[0:2] + (mat.shape[1], 1)
 ##             print self
 ##             print self.shape, ckey
 ##             print vshape

Could you please explain me (or even better in the comments in the sources) 
what is
this supposed to do? And then let's write a test for this, as clearly it can be
broken sometimes.

Original comment by ondrej.c...@gmail.com on 11 Jun 2008 at 11:12

• Changed state: Started

[GoogleCodeExporter]

I will check it. The problem is that the tests do not cover special solvers as
schroedinger.py - I will make a test veryfing if it runs.

Original comment by robert.c...@gmail.com on 12 Jun 2008 at 11:20

[GoogleCodeExporter]

Fixed (the input files).

Explanation: MatInQPDataCache is a class that can, for example, interpolate 
material
data from vertices into Gauss points of the elements. The changeset 180 
broadened the
possible input data shapes, but I forgot to fix also the schreodinger input 
files.

Test it please.

Original comment by robert.c...@gmail.com on 12 Jun 2008 at 11:58

[GoogleCodeExporter]

Works well now. Thanks!

Original comment by ondrej.c...@gmail.com on 12 Jun 2008 at 1:29

• Changed state: Verified

[GoogleCodeExporter]

Migrated to http://github.com/sfepy/sfepy/issues/43

Original comment by robert.c...@gmail.com on 30 Jan 2012 at 10:24

• Changed state: Migrated