Accéder au contenu.
Menu Sympa

starpu-devel - [Starpu-devel] Asking collaboration in PhD Research about Scheduling

Objet : Developers list for StarPU

Archives de la liste

[Starpu-devel] Asking collaboration in PhD Research about Scheduling


Chronologique Discussions 
  • From: Marcos Amarís González <marcos.amaris-gonzalez@inria.fr>
  • To: starpu-devel <starpu-devel@lists.gforge.inria.fr>
  • Subject: [Starpu-devel] Asking collaboration in PhD Research about Scheduling
  • Date: Fri, 15 Jan 2016 12:33:31 +0000
  • Authentication-results: mail2-smtp-roc.national.inria.fr; spf=None smtp.pra=marcos.amaris-gonzalez@inria.fr; spf=Pass smtp.mailfrom=marcos.amaris@gmail.com; spf=None smtp.helo=postmaster@mail-wm0-f49.google.com
  • Ironport-phdr: 9a23:n51+khMRWY6EVsUqSVgl6mtUPXoX/o7sNwtQ0KIMzox0KPTyrarrMEGX3/hxlliBBdydsKIazbuP+PuwEUU7or+/81k6OKRWUBEEjchE1ycBO+WiTXPBEfjxciYhF95DXlI2t1uyMExSBdqsLwaK+i760zceF13FOBZvIaytQ8iJ35rxhrn5oM2bSj4LrQT+SIs6FA+xowTVu5teqqpZAYF19CH0pGBVcf9d32JiKAHbtR/94sCt4MwrqHwI6Lpyv/JHBL7mdr4gUPlUASorNyhh/9HirwHeCAeC+HYYemEXiQZTRQfL6wv1U9HwtDH7v6xzwn/JE9fxSOURUC6p6LwjYRrhkqIcfxE+9nNejsFixPZeqQiqoAc5yovdeoCOcfV5eLHUed4IbWtHRMdYESJbVNDvJ7ATBvYMaL4L57L2oEED+EOz
  • List-archive: <http://lists.gforge.inria.fr/pipermail/starpu-devel/>
  • List-id: "Developers list. For discussion of new features, code changes, etc." <starpu-devel.lists.gforge.inria.fr>

Hello good people,

My name is Marcos Amarís, I am a PhD Candidate of Computer Sciences at the University of São Paulo in Brazil (http://www.ime.usp.br/~amaris/). Two months ago I began a Research internship at INRIA Grenoble, over the supervision of Denis Trystram (http://moais.imag.fr/membres/denis.trystram/).

My new topic of research is about Scheduling of Task-based Applications (fine grain) executed in heterogeneous platforms (CPU + GPU).

Our first idea was to profile Benchmarks as Rodinia (https://www.cs.virginia.edu/~skadron/wiki/rodinia), Pairboil (http://impact.crhc.illinois.edu/parboil/parboil.aspx), SHOC (https://github.com/vetter/shoc/wiki), among other. This idea was discarded because, all the applications have one job and one task, different kernels executed in a Coarse Grained approach, I mean the applications are not partitioned in Tasks, and we want to create our Scheduler with a set of applications with multiples or many tasks.

Our second idea was to port some applications of the Benchmark mentioned above to StarPU, this is a hard work and it would consume many time, even more difficult is to partition those applications in task! Really with a great effort some applications can be port to task-based approach (fine grain) and other simply is not possible, for the nature of the problem.

For those reasons that I have mentioned above I decided to ask the collaborations of the StarPU community, and to request developed applications with a Task-based approach in StarPU to analyze them and to try to elaborate my research about Scheduling of task-based applications over Heterogeneous application (CPU + GPU).

I really would appreciate any help, and obviously I would have to put you in the acknowledgments of any Article, or in the bibliography references in case of to be a published Work.

I hope that everybody have a very good year, many peace, harmony, successes and love!

Best Regards,

-Marcos



Archives gérées par MHonArc 2.6.19+.

Haut de le page