[Trilinos-Users] glibc detected in IFPack ILU compute()

[Erik Boman]

Try |"ILU stand-alone" or ||"ILUT stand-alone" as the string to the 
Ifpack factory to skip AdditiveSchwarz in serial.| I am not sure this 
fixes your problem, but it's easy to try.

Erik

james williams wrote:
> Hi All,
>
> I'm using the Ifpack factory class to build an ILU preconditioner for a 
> serial code.  After creating and initializing the preconditioner from my 
> FillCompleted EpetraCrs matrix, I call compute() and the program crashes 
> with:
>
> *** glibc detected *** ./zoupa: free(): invalid next size (normal): 
> 0x00000000069df2e0 ***
> *** glibc detected *** ./zoupa: malloc(): memory corruption: 
> 0x00000000069eeb10 ***
>
> When I run the code with valgrind, I get:
>
> ==9743== Invalid write of size 4
> ==9743==    at 0xBBE0F0: Ifpack_ILU::Compute() (in 
> /storage/jww/SIM_VERSIONS/IMEX_ML_SIM/zoupa)
> ==9743==    by 0xB39F0C: Ifpack_AdditiveSchwarz<Ifpack_ILU>::Compute() 
> (in /storage/jww/SIM_VERSIONS/IMEX_ML_SIM/zoupa)
> ==9743==    by 0x4B09B2: compif_ (pass_A.cpp:62)
> ==9743==    by 0x4C593D: stepper_ (stepper.f90:268)
> ==9743==    by 0x486356: MAIN__ (ice.f90:220)
> ==9743==    by 0x4DD639: main (in 
> /storage/jww/SIM_VERSIONS/IMEX_ML_SIM/zoupa)
> ==9743==  Address 0x744285c is 0 bytes after a block of size 63,516 alloc'd
>
> This looks like writing off the end of the of the memory block, any 
> ideas of why this could be happening?
>
> Additional notes:
>
> 1) The matrix is sparse, square, non-symmetric.
>
> 2) Using an ML preconditioner with the same matrix, and an analogous 
> build process runs with no problems.
>
> Any insight into what could be going wrong is much appreciated,
> James
>
> _______________________________________________
> Trilinos-Users mailing list
> Trilinos-Users at software.sandia.gov
> http://software.sandia.gov/mailman/listinfo/trilinos-users
>