yade-dev team mailing list archive

[Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>]

(to yade-dev)
>
> That's true. We need to move metis-library to yade-dem.org/packages
> and compile the next version of packages against it.

I must say the behavior of cmake is not completely consistent at the
moment. I will fix it eventually but I can already clarify a little.
What we do is solving a matrix system, and depending on available libs
this is what should happen:
1/ suitesparse+openblas+metis : this is the optimal situation,
suitesparse will use metis for preconditionning and openblas for
matrix/vector operations.
2/ suitesparse+openblas: functionality should be the same, just the
preconditionning may be a bit less efficient
3/ suitesparse alone (i.e. with some "legacy" blas/lapack): again not
fundamentally different but a bit less efficient

All the above is direct solving (Cholesky factorization). The situation
we have now for yadedaily/12.04 corresponds to:
4/ no suitesparse available, use iterative (Gauss-Seidel) algorithm
implemented directly in yade's code for solving the system. Quite slow.

The thing is we could very much build as in 2, then get acceptable
performance without metis. It's just that the build system is going
straight to 4) as soon as one lib is missing. Since we should try and
stick to metis anyway, I don't see a high priority in fixing that.

Bruno

>
> Anton
>
> _______________________________________________
> Mailing list: https://launchpad.net/~yade-users
> Post to     : yade-users@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~yade-users
> More help   : https://help.launchpad.net/ListHelp
>
>
>


-- 
_______________
Bruno Chareyre
Associate Professor
ENSE³ - Grenoble INP
Lab. 3SR
BP 53
38041 Grenoble cedex 9
Tél : +33 4 56 52 86 21
Fax : +33 4 76 82 70 43
________________