[petsc-users] GPU local direct solve of penta-diagonal

[Karl Rupp]

Hi guys,

 > Provided you have a good parallel sparse direct solve for a single SM,
> you could unleash 32 direct solves (or perhaps 16) which run
> concurrently on the K20x. One only needs to set an environment variable
> to use Hypre Q

Thanks for your inputs on HyperQ. I'm afraid this still won't give the 
good performance Marc and Ed are looking for, mostly because there is 
simply not enough parallelism in sparse direct solvers for systems of 
that size (cf. Jed's comment). They might actually work quite well on 
the CPU if a symbolic factorization is first carried out in a 
preprocessing step and then the actual numbers are computed in each 
Picard iteration.

Best regards,
Karli