In SLEPc we have an example (ex34.c, contributed by Fande Kong) that uses 
DMPlex. In the nightly tests we see a difference that did not ocurr yesterday.

The problem may be caused by the merge of branch tisaac/feature-space-tensor

The difference appears only in complex scalars. To reproduce the problem:

REAL:~/soft/slepc/src/eps/examples/tutorials (master)$ ./ex34 
-petscspace_degree 1 -petscspace_poly_tensor
 Solution method: power (nonlinear)

 Number of requested eigenvalues: 1
k: 64.392 error: 6.91725e-07

COMPLEX:~/soft/slepc/src/eps/examples/tutorials (master)$ ./ex34 
-petscspace_degree 1 -petscspace_poly_tensor
 Solution method: power (nonlinear)

 Number of requested eigenvalues: 1
k: 165.007 error: 199.037



Jose

Reply via email to