Re: [Starpu-devel] Scheduling accelerators

[Muhammad Usama Asad <muhammad-usama.asad@s2012.tu-chemnitz.de>]

Hey Samuel,

Thanks for your quick kind response. I appreciate ladies working in  
technical engineering research area but got surprised ;)

I need to partition vision algorithms or computational workload and  
then will be able to easily map. Can you please recommend or refer to  
some research group which can help me out in partitioning vision  
algorithms.

Secondly I need to know how mature StarPU is for scheduling and  
mapping. I mean is it in use in industries ? I am supposed to use this  
scheduler for a TI device (SoC).

Thanks for your assistance.

Regards,
Muhammad





Zitat von Samuel Thibault <samuel.thibault@inria.fr>:

> Hello,
>
> Muhammad Usama Asad, on Tue 12 Jan 2016 16:51:39 +0100, wrote:
> > Respected Sir,
>
> There are ladies in here too ;)
>
> > Can you propose some optimized scheduling algorithm (SW framework) for
> > vision applications tailored for using computer vision accelerators?
>
> Well, I don't know much how such kind of computation is typically
> divided. Perhaps the current StarPU scheduling strategies will be fine
> enough already, or you may have to write one, but that's made easy in
> StarPU.
>
> > And strategies for partitioning the algorithm and strategies for mapping
> > those partitioned algorithms into hardware resources available on chipset?
>
> StarPU itself does not decide how to partition data, it only does the
> mapping.
>
> Samuel