[petsc-dev] Handling ESSL + Lapack and PESSL

[Barry Smith]

On Sep 13, 2010, at 1:29 PM, Jed Brown wrote:

> On Mon, 13 Sep 2010 13:18:54 -0500, Barry Smith <bsmith at mcs.anl.gov> wrote:
>> Does PESSL have some of the same symbols in it as ESSL? If it has
>> 
> 
>> You really don't want to link LAPACK first because then basically
>> you get the LAPACK versions of the codes which (one would presume)
>> are slower then the same interfaces in ESSL (otherwise why bother
>> using ESSL).
> 
> For the fast BLAS?  I agree with your point, but I suspect lots of users
> would prefer to tolerate a somewhat slower GEEV in exchange for not
> having to write two versions of their code.

My point is not only do they get a slower GEEV but they get a slower everything in LAPACK.

If they are not willing to suffer IBM's arrogance and Jack's total  incompetence that let IBM provide a different interface why link with ESSL at all? Why not just use LAPACK?

But you are correct a more specific configure check should be added and then a more specific ifdef used in the PETSc code. Anyone is welcome to add this support but I sure as hell am not going to do it myself.

> 
>> When is KAUST going to grow a brain and order a IBM Blue Waters
>> machine as the next machine and not a IBM Blue Gene?
> 
> Presumably as soon as IBM is ready to install it in 2009.

Bill Gropp managed to get his Blue Waters in 2009 so why couldn't Kaust :-).

Barry

> 
> Jed