• Non ci sono risultati.

Extending the Actor Model with Parallel Patterns: a new model for multi-/many-core platforms

N/A
N/A
Protected

Academic year: 2021

Condividi "Extending the Actor Model with Parallel Patterns: a new model for multi-/many-core platforms"

Copied!
4
0
0

Testo completo

(1)

Report on the Ph.D. Activities

Luca Rinaldi

April 17, 2021

1

Research Activities

Luca Rinaldi’s Ph.D. research is focused on building a coherent combina-tion of two different parallel programming models, namely the Actor Model and the Parallel Patterns. Although multiple parallel programming models and frameworks were proposed over the year, parallel programming is still a challenging and error-prone task. Therefore, building a general, flexible, and performance parallel abstraction is still an open issue. Several existing models suffer from being specialized for a given class of parallel problems, forcing the programmers to rethink or adapt their applications to use the selected model.

The combination of the Actor Model with the structured parallel pro-gramming based on parallel patterns reduces the individual limits of these two approaches and gets the better of the two worlds. On the one hand, the introduction of Parallel Patterns in the Actor Model enables the possibility to introduce performance optimizations and structured composition of Ac-tors. On the other hand, the Actor Model provides flexibility and dynamicity features, both necessary in complex parallel application development.

The works of thesis focus on studying the viability and the advantages of building a new parallel programming library, which extends the Actor Model semantics with the structured parallel programming methodology.

(2)

2

Training Activities

Courses:

• An introduction to Deep Learning Lecturer: Antonio Gull`ı (Google)

• Skills Boosting for New (Research) Horizons

Lecturer: Paolo Ferragina (UniPI), Michele Padrone (UniPI), Davide Morelli (BioBeats), Flavio Tosi (Business Exploration),Camilla Van den Boom (Eindhoven University), Nicola Redi (Venture Factory), Ray Garcia (Buyont Capital, NY)

• Modeling methods

Lecturer: Egon Boerger (UniPI)

• Genomic data analysis with applications

Lecturer: Ugo Borello (Biologia, UniPI), Filippo Geraci (IIT CNR) • Data Stream Processing from the Parallelism Perspective

Lecturer: Gabriele Mencagli (UniPI), Matteo Andreozzi, ARM • Design-by-Contract and Behavioral Types

Lecturers: Roberto Bruni (UNIPI), Hernan Melgratti (University of Buenos Aires)

Seminars Cycles:

• Research, Innovation and Future of ICT (2017) • Academic English (2018)

• Mauriana Pesaresi seminar series (2018)

3

Publications

• Luca Rinaldi, Massimo Torquati, Gabriele Mencagli, and Marco Dane-lutto. “High-Throughput Stream Processing with Actors”. In: ceedings of the 10th ACM SIGPLAN International Workshop on Pro-gramming Based on Actors, Agents, and Decentralized Control. AGERE

(3)

2020. event-place: Virtual, USA. New York, NY, USA: Association for Computing Machinery, 2020, pp. 1–10. isbn: 978-1-4503-8185-7. doi: 10.1145/3427760.3428338

• Luca Rinaldi, Massimo Torquati, Daniele De Sensi, Gabriele Mencagli, and Marco Danelutto. “Improving the Performance of Actors on Multi-cores with Parallel Patterns”. In: International Journal of Parallel Programming (June 4, 2020). issn: 1573-7640. doi: 10.1007/s10766-020-00663-1

• Luca Rinaldi, Massimo Torquati, and Marco Danelutto. “Enforcing Reference Capability in FastFlow with Rust”. In: Parallel Comput-ing: Technology Trends, Proceedings of the International Conference on Parallel Computing, PARCO 2019, Prague, Czech Republic, September 10-13, 2019. Ed. by Ian T. Foster, Gerhard R. Joubert, Ludek Kucera, Wolfgang E. Nagel, and Frans J. Peters. Vol. 36. Advances in Parallel Computing. IOS Press, 2019, pp. 396–405. doi: 10.3233/APC200064 • Luca Rinaldi, Massimo Torquati, Daniele De Sensi, Gabriele Mencagli,

and Marco Danelutto. “Are Actors Suited for HPC on Multi-Cores?” In: 12th International Symposium on High-Level Parallel Programming and Applications. Peer reviewed with internal proceedings. Link¨oping, Sweden, June 2019, p. 21

• Luca Rinaldi, Massimo Torquati, Gabriele Mencagli, Marco Danelutto, and Tullio Menga. “Accelerating Actor-Based Applications with Par-allel Patterns”. In: 2019 27th Euromicro International Conference on Parallel, Distributed and Network-Based Processing (PDP). 2019 27th Euromicro International Conference on Parallel, Distributed and Network-Based Processing (PDP). Pavia, Italy: IEEE, Feb. 2019, pp. 140– 147. isbn: 978-1-72811-644-0. doi: 10.1109/EMPDP.2019.8671602 • Antonio Brogi, Stefano Forti, Ahmad Ibrahim, and Luca Rinaldi.

“Bon-sai in the Fog: An active learning lab with Fog computing”. In: Third International Conference on Fog and Mobile Edge Computing, FMEC 2018, Barcelona, Spain, April 23-26, 2018. IEEE, 2018, pp. 79–86. doi: 10.1109/FMEC.2018.8364048

• Antonio Brogi, Davide Neri, Luca Rinaldi, and Jacopo Soldani. “Or-chestrating incomplete TOSCA applications with Docker”. In: Sci.

(4)

Comput. Program. 166 (2018), pp. 194–213. doi: 10.1016/j.scico. 2018.07.005

• Antonio Brogi, Luca Rinaldi, and Jacopo Soldani. “TosKer: A synergy between TOSCA and Docker for orchestrating multicomponent appli-cations”. In: Softw. Pract. Exp. 48.11 (2018), pp. 2061–2079. doi: 10.1002/spe.2625

• Antonio Brogi, Andrea Canciani, Davide Neri, Luca Rinaldi, and Ja-copo Soldani. “Towards a Reference Dataset of Microservice-Based Applications”. In: Software Engineering and Formal Methods - SEFM 2017 Collocated Workshops: DataMod, FAACS, MSE, CoSim-CPS, and FOCLASA, Trento, Italy, September 4-5, 2017, Revised Selected Papers. Ed. by Antonio Cerone and Marco Roveri. Vol. 10729. Lec-ture Notes in Computer Science. Springer, 2017, pp. 219–229. doi: 10.1007/978-3-319-74781-1_16

Riferimenti

Documenti correlati

The results of tests performed in the rural areas of Tanzania show that a standardized use of the prototype can lower the dissolved fluoride from an initial concentration of 21

1 characterization of dosimetric performances of Fricke gels irradiated with very high dose- per-pulse electron beams generated by a Novac11 under reference conditions

Example: 2D (Block, *) partitioning with 5P stencil Periodic

– If the tasks synchronize at the end of the computation, the execution time will be that of the slower task. Task 1

● Strong Scaling: increase the number of processors p keeping the total problem size fixed. – The total amount of work

● Si applica un partizionamento a grana fine (numero di task >> numero di core). ● Il master distribuisce il lavoro ad una serie

◦ a written test: 4 questions, 1.30 hours (2/5 of the final mark). ◦ a practical project (3/5 of the