[Trilinos-Users] FW: question about anasazi tests

[Baker, Christopher G.]

Jason, could you attached just a list of the failing tests so I can get a better idea where the problem might be?

Chris


On 12/15/09 2:57 PM, "Jason Slemons" <slemons at cray.com> wrote:


Hello,

I've run a few Anasazi tests (with 10.0 release, using GNU) and they all seem to get similar errors; the test is essentially to run  Anasazi_IRTRThyra_test.exe:

252/320 Test: Anasazi_IRTRThyra_test_MPI_4_3
Command: "/usr/bin/aprun" "-n" "4" "/home/users/slemons/trilinos-10.0-Source/gnu.big.BUILD/packages/anasazi/thyra/test/IRTR/Anasazi_IRTRThyra_test.exe" "--verbose" "--skinny" "--sort=LR"

Many of them have this error about unused parameters which then, it seems to me, are causing all the rest of the problems:

Failure! Unused parameters:
WARNING: Parameter "Block Size" 5   [unused] is unused
WARNING: Parameter "Convergence Tolerance" 1e-06   [unused] is unused
WARNING: Parameter "Maximum Iterations" 450   [unused] is unused
WARNING: Parameter "Skinny Solver" 1  [unused] is unused
WARNING: Parameter "Verbosity" 89  [unused] is unused
WARNING: Parameter "Which" LR   [unused] is unused

Anasazi::SolverUtils::directSolver(): In HEGV, DPOTRF or DHEEV returned an error code (Anasazi::SolverUtils::directSolver(): In HEGV, DPOTRF or DHEEV returned an error code (44Anasazi::SolverUtils::directSolver(): In HEGV, DPOTRF or DHEEV returned an error code ().
).
4
).

Anasazi::SolverUtils::directSolver(): In HEGV, DPOTRF or DHEEV returned an error code (4).
Exception: /opt/cray/trilinos/10.0.0/GNU/include/AnasaziSIRTR.hpp:848:
Throw number = 1
Throw test that evaluated to true: ret != 0
Anasazi::SIRTR::iterate(): failure solving projected eigenproblem after retraction. ret == -20AA: Teuchos::SerialDenseMatrix
Values_copied : yes
Rows : 5
Columns : 5
LDA : 5
nan nan nan nan nan
nan nan nan nan nan
...

I have tested compiling with GNU and PGI. Only GNU fails though. PGIs output is :

===============================================================================                 =

                        SIRTR Solver Status

The solver is initialized.
The number of iterations performed is 1
The current block size is             5
The number of auxiliary vectors is    0
The number of operations A*x    is 15
The number of operations B*x    is 15
The number of operations B*x by the orthomanager is 40
The number of operations Prec*x is 0
Parameter rho_prime is  5.000000e-01
Inner stopping condition was negative curvature
Number of inner iterations was 1
Total number of inner iterations is 1
f(x) is -nan

CURRENT EIGENVALUE ESTIMATES
         Eigenvalue        Residual(B)         Residual(2)
-------------------------------------------------------------------------------
               -nan         not current                nan
               -nan         not current                nan
               -nan         not current                nan
               -nan         not current                nan
               -nan         not current                nan
====================================================
...
Which right away seems like somethings going wrong but the 'TEST PASSED' according to the output.
Has anyone seen this before?

-Jason Slemons