Re: [Starpu-devel] Questions about StarPU worker recongnization

[Fangli Pi <hpcfapix@hlrs.de>]

Hello,

For example, I have a parallel task (codelet with power model and performance 
model), and I feed, in my application, energy consumption values to the task, 
during calibrating period.
And then when the task is really executed (conf.calibrate = -1;), will the 
scheduler (pheft) consider the consumption value to chose the number of 
worker size? 
In another word, as you said the parallel worker size is chosen by the 
scheduler, I am interested in how the scheduler choose the worker size? Is it 
only the all available workers, or could depend on some other values?

Besides, for a non-parallel task, I can simply get info about which cpu core 
executed the task by looking at starpu_profiling_task_info : workerid.
But for parallel task, how could I get info about the combined workers? What 
is the output of the starpu_profiling_task_info : workerid?

Best regards,
Fangli

----- Original Message -----
From: "Samuel Thibault" <samuel.thibault@inria.fr>
To: "Fangli Pi" <hpcfapix@hlrs.de>
Cc: "starpu-devel" <starpu-devel@lists.gforge.inria.fr>
Sent: Wednesday, January 20, 2016 1:11:54 PM
Subject: Re: [Starpu-devel] Questions about StarPU worker recongnization

Hello,

Fangli Pi, on Wed 20 Jan 2016 13:09:21 +0100, wrote:
> For example of the "vector_scal_spmd.c". Inside the cpu function, number of 
> available cpu cores are get by:
> int nb_workers = starpu_combined_worker_get_size();
> 
> then how could i keep this number for later usage?

Which kind of usage are you thinking about?  The parallel worker size
can be different for different tasks, chosen by the scheduler.

Samuel
-- 
High Performance Computing Center (HLRS)   
University of Stuttgart                    
Nobelstr. 19                               
D-70569 Stuttgart, Germany

phone: ++49-711-685-60442