Re: [ESPResSo-users] different outcomes for CPU and GPU

[Georg Rempfer]

I am trying to debug a similar problem right now. I ran you script and only got a difference of about 0.5%. Is that correct?

On Mon, Jun 22, 2015 at 10:02 AM, Renáta Tóthová <address@hidden> wrote:

Hello, Espresso developers,

I am a member of the Cell-in-fluid research group in Slovakia and I have encountered this problem:
I have run the simulation in periodic channel with fluid only. There were only bottom and top wall and fluid flowed due to external force in x direction in the channel.
In first case, the fluid was computed using CPU - command lbfluid cpu. In second case, the fluid was computed using GPU - command lbfluid gpu. In both cases, there is Poiseuille/parabolic flow in the channel, but the values of velocity differ in the two outputs. The values are from a stable flow, after several thousand integration steps. The GPU calculations run on Nvidia Tesla machine. I have used current espresso master branch from github.

In the attachment, I am sending you the tcl script I have used.
Do you know what could be the issue?

Please, let me know. Thank you.
Have a nice day,

Renata Tothova