EFD parallel computing - from when?
 

when will it possible to start a parallel computing with EFD?

Re: EFD parallel computing - from when?
 

It as been intimated (to me) that Shared Memory (multi processors using the same memory) will be available in version 9 (which might come in September). What the speed advantage will be is going to be interesting but I'm not holding my breath. Distributed Memory (multi computers) will come in later versions I suspect.

Personally: I think that multicore processing will be nice but I spend more time trying to understand the problems than actually on computation, I would prefer to see something like in firefox where extensions can be written and added by users (I am sick of always changing the horizontal axis on the goal monitor from iterations to travels). But I think I should rather start a new thread for this.

Re: EFD parallel computing - from when?
 

in the current day and age it is fairly ridiculous not to have parallel capability for CFD stuff... a quad node pc is fairly cheap and can dramatically improve accuracy everything else being equal....

Re: EFD parallel computing - from when?
 

tell me, why is there an interest in multicore processing when the use of GPUs looks promising. Sure the nVidia Tesla is an expensive piece of hardware but if an article I've recently read suggesting twenty-fold improvements, I'm sure that may be well worth the while (the Engineer, 30/06/08). And with CUDA available is it very difficult for EFD to adopt "hardware acceleration"? Okay, maybe we can have both. /n

Re: EFD parallel computing - from when?
 

I just would like to add the fact that one philosophy of EFD is to keep the level of accuracy of one solution with a coarser mesh. that's why they are improving the inner code quality in each version and that's why, for this kind of software, I totally disagree on the "ridiculous" term you just used. Of course parallel capabilities will help a lot, but I don't think until now it was a huge lack for this software.

Re: EFD parallel computing - from when?
 

yep, and i'm sure that in the course of this "inner quality improvement process" they did away with discretization and resulting dispersion/diffusion errors directly proportional to the mesh size, you know that whole annoying mesh independence concept... am i to believe that people run all those multi-million meshes because they just do not know any better? and btw what do you mean by this kind of software? are the basic pdes somehow different? anyways, just for fun, take a square room, run a coarse mesh, double the mesh and see if anything changes, then imagine a much larger room....

Re: EFD parallel computing - from when?
 

FLoEFD.V9 has the ability of parallel computation.

Re: EFD parallel computing - from when?
 

Parallel, but shared memory I assume?

Re: EFD parallel computing - from when?
 

Yes.at moment.