Re: [moose-devel] Using axonal delay for multi-step execution

[U.S.Bhalla <bh...@nc...>]

Dear Saeed,
     Did you have a look at the branch that Vivek Sagar made for his GPU 
project? That worked under the current clock structure. I gather he 
launched GPU threads from within his section of the code.

The clock itself executes in a single-thread manner.

Best,
     Upi
On Wednesday 15 October 2014 01:09 AM, Saeed Shariati wrote:
> Dear All,
>
> The current version of the GPU version, executes in synchronous and single
> thread mode and by each step, executes the GPU kernels and comes back.
> Therefore, the performance significantly reduce. Beside of multithread
> execution, using axonal delay can boost the performance by reducing number
> of kernel call which is quite expensive in GPU execution.
>
> Would you please guide me how to use Clocking mechanism to obtain this
> functionality? For Example I see in Clock::buildTicks() function, the
> variable "stride_" is responsible for this parameter, but it seems that
> this value hard-coded to one (1).
>
> Any clue or guidance would be appreciated.
>
> Thank you,
> Saeed
> ------------------------------------------------------------------------------
> Comprehensive Server Monitoring with Site24x7.
> Monitor 10 servers for $9/Month.
> Get alerted through email, SMS, voice calls or mobile push notifications.
> Take corrective actions from your mobile device.
> http://p.sf.net/sfu/Zoho
> _______________________________________________
> moose-devel mailing list
> moo...@li...
> https://lists.sourceforge.net/lists/listinfo/moose-devel