Luka Stanisic
Table of Contents
Contact
- Professional email:
luka.stanisic@inria.fr
- Personal email:
stanisic.luka@gmail.com
- Skype:
stanisic_luka
- Most of the projects I work on are on gforge.inria, but still some of them can be found on github.
Address at Inria Bordeaux/Talence
Office B236 Inria 200 Avenue de la Vieille Tour 33405 Talence cedex FRANCE
Current Position
I am a postdoctoral researcher at Inria Bordeaux Sud-Ouest in STORM and HiePACS teams. My main objectives are to experiment with dynamic HPC applications based on StarPU-MPI and to evaluate their performance in simulation. In a first step, medium-size clusters of hybrid nodes will be used with dense linear algebra solver Chameleon. This will then be generalized to larger clusters and assessed in the context of other existing numerical applications, sparse linear algebra solver qr_mumps, and more complex QDWH polar decomposition (in collaboration with KAUST) and ScalFMM library.
I am also involved in research on visualization techniques for evaluating performance of dynamic task-based applications, in collaboration with researchers from POLARIS team in Grenoble and UFRGS university in Brazil. Additionally, a lot of my time is dedicated to enhancing reproducibility of the experimental research in the community.
In collaboration with physicist Marina Krstic Marinkovic, I am working on performance simulation and modeling of openQCD, a lattice QCD code mostly developed at CERN (poster).
In the autumn semester of 2016/2017, I was teaching "Operating Systems" course to Master 1 students. More information about the course is available here.
Previously
I was a Phd student supervised by Arnaud Legrand and Jean-François Méhaut at the LIG (formerly ID Laboratory) laboratory in Grenoble. I worked at the Inria Grenoble Rhone-Alpes in MESCAL and CORSE (previously NANOSIM) teams in the domain of High Performance Computing. I defended my Phd thesis on 30.10.2015 (slides).
My thesis, financed by the French ministry grant, is about modeling and simulation of heterogeneous platforms and was partially done in the context of ANR SONGS project. We have ported a dynamic task-based runtime StarPU on top of SimGrid simulator that gave us a tool for accurate performance prediction of dense and sparse linear algebra applications, such as Cholesky factorization or qr_mumps. More details about this study as well as all the experiment results can be found at the starpu-simgrid website.
In the very beginning of the thesis, as well as during the Master internship, I was working on finding a coarse-grain models of CPUs. To that purpose, I was doing extensive benchmarking, performance evaluation and comparison of CPU caches of ARM and Intel processors. This research was a part of the European Mont-blanc project that investigates energy-efficient hardware solutions for the future super-computer platforms.
To validate our approaches I performed a large experimental studies on various prototype machines. In order to do this properly, with my colleagues we developed an unique methodology based on Git and Org-mode tools.
Research Interests
My research activities are concentrated on two tightly connected topics. First one is performance evaluation of parallel HPC applications, but also computer systems in general, relying on statistical analysis, modeling, simulation, visualization and trace analysis. My second interest is about experimental methodology and reproducible research, and how they can help us do a better science.
Talks
Here are the slides of the several presentations I gave:
- MPCDF seminar in Garching, February 2017 (slides)
- VPA workshop at SuperComputing in Salt Lake City, November 2016 (slides)
- Reproducible Research webinars, session on Git usage, June 2016 (slides,video)
- SIAM PPSC conference in Paris, April 2016 (slides)
- PhD defense, October 2015 (slides)
- ANR SOLHAR plenary meeting in Lyon, June 2015 (slides); in Toulouse, December 2016 (slides)
- EuroPar conference in Porto, August 2014 (slides)
- REPPAR workshop on reproducibility in parallel computing in Porto, August 2014 (slides)
- JointLab for Petascale Computing (JLPC) workshop in Nice, June 2014 (slides)
- AEP workshop on performance evaluation in Nice, June 2014 (slides)
- COMPAS conference in Neuchâtel, April 2014 (slides)
- ANR SONGS plenary meetings in Frejus, January 2016 (slides); in Nancy, January 2015 (slides); in Nice, January 2014 (slides); and in Lyon, June 2013 (slides)
Publications
Here is my publication list as extracted from HAL.
2017
Conference papers
- ref_biblio
- Luka Stanisic, Lucas Mello Schnorr, Augustin Degomme, Franz Heinrich, Arnaud Legrand, et al.. Characterizing the Performance of Modern Architectures Through Opaque Benchmarks: Pitfalls Learned the Hard Way. 31st IEEE International Parallel & Distributed Processing Symposium (RepPar workshop), Jun 2017, Orlando, United States. <hal-01470399v2>
- Accès au texte intégral et bibtex
Reports
- ref_biblio
- Emmanuel Agullo, Bérenger Bramas, Olivier Coulaud, Martin Khannouz, Luka Stanisic. Task-based fast multipole method for clusters of multicore processors. [Research Report] RR-8970, Inria Bordeaux Sud-Ouest. 2017, pp.15. <hal-01387482v4>
- Accès au texte intégral et bibtex
- ref_biblio
- Emmanuel Agullo, Bérenger Bramas, Olivier Coulaud, Luka Stanisic, Samuel Thibault. Modeling Irregular Kernels of Task-based codes: Illustration with the Fast Multipole Method. [Research Report] RR-9036, INRIA Bordeaux. 2017, pp.35. <hal-01474556>
- Accès au texte intégral et bibtex
2016
Conference papers
- ref_biblio
- Vinicius Garcia Pinto, Luka Stanisic, Arnaud Legrand, Lucas Mello Schnorr, Samuel Thibault, et al.. Analyzing Dynamic Task-Based Applications on Hybrid Platforms: An Agile Scripting Approach. 3rd Workshop on Visual Performance Analysis (VPA), Nov 2016, Salt Lake City, United States. <http://www.cedmav.org/events/vpa-2016.html>. <hal-01353962>
- Accès au texte intégral et bibtex
- ref_biblio
- Marina Krstic Marinkovic, Luka Stanisic. Platform independent profiling of a QCD code. 34th annual International Symposium on Lattice Field Theory, Jul 2016, Southampton, United Kingdom. 2017, PoS. <hal-01474342>
- Accès au texte intégral et bibtex
2015
Journal articles
- ref_biblio
- Luka Stanisic, Samuel Thibault, Arnaud Legrand, Brice Videau, Jean-François Méhaut. Faithful Performance Prediction of a Dynamic Task-Based Runtime System for Heterogeneous Multi-Core Architectures. Concurrency and Computation: Practice and Experience, Wiley, 2015, pp.16. <10.1002/cpe>. <hal-01147997>
- Accès au texte intégral et bibtex
- ref_biblio
- Luka Stanisic, Arnaud Legrand, Vincent Danjean. An Effective Git And Org-Mode Based Workflow For Reproducible Research. Operating Systems Review, Association for Computing Machinery, 2015, 49, pp.61 - 70. <10.1145/2723872.2723881>. <hal-01112795>
- Accès au texte intégral et bibtex
Conference papers
- ref_biblio
- Luka Stanisic, Emmanuel Agullo, Alfredo Buttari, Abdou Guermouche, Arnaud Legrand, et al.. Fast and Accurate Simulation of Multithreaded Sparse Linear Algebra Solvers. The 21st IEEE International Conference on Parallel and Distributed Systems, Dec 2015, Melbourne, Australia. <http://www.2015.icpads.org/>. <hal-01180272v2>
- Accès au texte intégral et bibtex
Theses
- ref_biblio
- Luka Stanisic. A Reproducible Research Methodology for Designing and Conducting Faithful Simulations of Dynamic HPC Applications. Distributed, Parallel, and Cluster Computing [cs.DC]. Université Grenoble Alpes, 2015. English. <NNT : 2015GREAM035>. <tel-01248109v2>
- Accès au texte intégral et bibtex
2014
Conference papers
- ref_biblio
- Luka Stanisic, Samuel Thibault, Arnaud Legrand, Brice Videau, Jean-François Méhaut. Modeling and Simulation of a Dynamic Task-Based Runtime System for Heterogeneous Multi-Core Architectures. Euro-par - 20th International Conference on Parallel Processing, Aug 2014, Porto, Portugal. Springer International Publishing Switzerland, pp.50-62, 2014, Euro-Par 2014, LNCS 8632. <hal-01011633>
- Accès au texte intégral et bibtex
- ref_biblio
- Luka Stanisic, Arnaud Legrand. Effective Reproducible Research with Org-Mode and Git. 1st International Workshop on Reproducibility in Parallel Computing, Aug 2014, Porto, Portugal. <http://reppar.org/>. <hal-01083205>
- Accès au texte intégral et bibtex
Documents associated with scientific events
- ref_biblio
- Luka Stanisic, Arnaud Legrand. Writing a Reproducible Article. Pascal Felber; Laurent Philippe; Etienne Riviere; Arnaud Tisserand. ComPAS 2014 : conférence en parallélisme, architecture et systèmes, Apr 2014, Neuchâtel, Switzerland. 2014. <hal-00994575>
- Accès au texte intégral et bibtex
Reports
- ref_biblio
- Luka Stanisic, Samuel Thibault, Arnaud Legrand, Brice Videau, Jean-François Méhaut. Modeling and Simulation of a Dynamic Task-Based Runtime System for Heterogeneous Multi-Core Architectures. [Research Report] RR-8509, INRIA. 2014. <hal-00966862>
- Accès au texte intégral et bibtex
2013
Conference papers
- ref_biblio
- Luka Stanisic, Brice Videau, Johan Cronsioe, Augustin Degomme, Vania Marangozova-Martin, et al.. Performance Analysis of HPC Applications on Low-Power Embedded Platforms. DATE - Design, Automation & Test in Europe, Mar 2013, Grenoble, France. pp.475-480, 2013, <10.7873/DATE.2013.106>. <hal-00872482>
- Accès au texte intégral et bibtex
2012
Master thesis
- ref_biblio
- Luka Stanisic. Towards Modeling and Simulation of Exascale Computing Platforms. Distributed, Parallel, and Cluster Computing [cs.DC]. 2012. <hal-01158585>
- Accès au texte intégral et bibtex
Hobbies
Since the early childhood, I was always passionate about the sports. While in Grenoble, I was playing an amateur "Foot A 7" football league for the Team Paquito and skiing/snowboarding in the mountains. Formerly, I used to play futsal for faculty team Elektricar in Belgrade and before that basketball in KK DOK Beograd.
I also love traveling and discovering new cities and cultures. When I get tired from all these activities, I like to read a good book or play some board games with the friends.
More information
My short English curriculum vitæ from January 2017.