[Trilinos-Users] Trilinos-mkl coupling, info.

Heroux, Michael A maherou at sandia.gov
Thu Aug 9 18:41:07 MDT 2012


It looks like you are using ILP64, which interprets "int" as a 64-bit quantity.  Our bindings to the BLAS/LAPACK assume "int" is 32-bit.  Unless you need to solve dense problems with size greater than 2^31 (about 2 billion), you should probably use LP64, which treats "long" and pointer types as 64-bit but leaves "int" as 32-bit.

Let us know if this is the issue, and whether the proposed fix works.  We are just learning ourselves about ILP64.

Mike


On Aug 9, 2012, at 12:00 PM, "Natarajan Cs" <csnataraj at gmail.com> wrote:

> Hello All,
>             I recently came across a rather odd error which I thought I should report. I have a project that uses trilinos and base mkl. 
> I noticed that when trilinos is compiled with 64-bit mkl (with ilp support), my project would spew out errors during some convolution routines (done with mkl). Trilinos itself was fine, but the mkl portion of my project went sour. The errors weren't also easy to trace back to compiling trilinos with 64-bit mkl. (The error I received was "bad stride"). Once I recompile trilinos with 32-bit mkl everything was golden again. Luckily I had done this mistake once before and could trace back to the source to the error.
> 
> My situation is probably uncommon but thought I should put it out there, in case someone runs into a similar issue.
> 
> Cheers,
> C.S.N
> _______________________________________________
> Trilinos-Users mailing list
> Trilinos-Users at software.sandia.gov
> http://software.sandia.gov/mailman/listinfo/trilinos-users



More information about the Trilinos-Users mailing list