yade-users team mailing list archive

[Michael Jensen <jenmichael2000@xxxxxxxx>]

"Forget about running serious simulations (given size of simulation) on a
laptop. Get a decent 4core (or more) machine with DDR3 RAM (you need
several GB of RAM for compilation, probably not so much for running)."

hehe.. I might reference you on that, -one more reason for going with experiments :-D

that is what I had begun to suspect, hence my comment about fighting a loosing battle.  I presume that you are refering to the time duration of the simulation being enormous?  or is it possible that you think that a simulation of this size just wouldn't work very well on a laptop?

Also, I would be interested in your thoughts on how much smaller my timestep might have to be to take into account the motion of the system.. 5000 rpm, max radius 5 cm..

and thank's for the numbers on computation time, however rough :-)

-mike

--- On Fri, 7/9/10, Václav Šmilauer <eudoxos@xxxxxxxx> wrote:

From: Václav Šmilauer <eudoxos@xxxxxxxx>
Subject: Re: [Yade-users] hello? -help with a spinning bucket!
To: yade-users@xxxxxxxxxxxxxxxxxxx
Received: Friday, July 9, 2010, 1:07 AM


> I assume it has only one core, and I certainly don't know how to hook
> up other ones, though I know people who surely do.

If there is just one core, you cannot add more (if you know someone who
does, he would be probably doing that all the time). 

Forget about running serious simulations (given size of simulation) on a
laptop. Get a decent 4core (or more) machine with DDR3 RAM (you need
several GB of RAM for compilation, probably not so much for running).

I am usually running with 4 cores (as I have only 4 ;-) ), but there is
not much speedup beyond 4 cores (depending on exact simulation setup; we
pretty suck at scalability on multicore machines).

> am I fighting what must surely be a loosing battle, my goal being to
> simulate up to 1 million particles (1 cubic millimeter glass beads in
> a volume of about 1 liter), for at least 4-10 seconds?  10 000
> particles? 1000?

Depends on the critical Δt (determined roughly by particle diameter and
stiffness). If it were 1e-6, then you need about 1e7 timesteps...

Read https://yade-dem.org/sphinx/formulation.html#cost , that gives an
esstimate of the order of magnitude, about 1e-6s of computation per
particle and per timestep. That gives you 10s with 1 particle :-) So
with 1k particles, you will be somewhere around 3hrs and with 10k around
a day. Keep in mind that it is very very very rough.

Cheers, vaclav


_______________________________________________
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