[petsc-users] optimizing repeated calls to KSPsolve?

[Jed Brown]

What will you do with the 500000 responses?

Jed

On Dec 11, 2010 7:49 AM, "Luke Bloy" <luke.bloy at gmail.com> wrote:



On 12/10/2010 06:30 PM, Jed Brown wrote:
>
> On Sat, Dec 11, 2010 at 00:03, Luke Bloy <luke.bloy at g...
Yes I'd like do do this for roughly 500,000 d_i. I'm solving a diffusion
problem, I only have local measures of the diffusion process which is what i
use to determine that matrix M. Now the  500,000 d_i are the boundaries to
my diffusion problem, what i need to know is who much of what gets pumped in
though a given boundary state exits the system through the other boundaries.
Do i really need to do all 500,000 Probably not. this is the highest
resolution mesh of the boundary i can compute from my data. A lower res mesh
would probably be sufficient. But I wont know until i do it either way I'd
like to use the highest res mesh that I can. If you can suggest an
alternative approach I am all ears.

Luke


>>> 2. If you can afford the memory, a direct solve probably makes sense.
>>
>>
>> My understanding...
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20101211/114ccf5b/attachment.htm>