About Me

I am a Senior Researcher in the School of Informatics at the University of Edinburgh and
a Research Fellow of the Royal Academy of Engineering. I am part of the CArD group and the ICSA Institute.

For the previous four and a half years, I was a Research Associate working with Hugh Leather. I earned my PhD at the University of Patras, Greece in July 2011 under Stefanos Kaxiras. Before that, I earned a Diploma (BSc+MSc) in Electrical and Computer Engineering from the University of Patras.

Research

My research focuses on compilers, runtime systems, and development tools that help programmers write fast, energy efficient programs with as little effort as possible.

Electronic systems already consume 10% of global electricity, reaching 20% by 2030. This is unsustainable. More energy efficient hardware can help us tackle this problem but such hardware is too complex for the average programmer. My aim is to deliver intelligent tools that bridge the gap between the programmer and processor.

Projects

RAEng Logo

Deep Learning For Easier Compiler Analysis and Optimisation
Royal Academy of Engineering Research Fellowship.
Extracting performance from complex processing systems is difficult. Existing tools cannot make intelligent decisions. The process still requires trial and error and expert knowledge. Even then we might miss many optimisation opportunities. With highly heterogeneous processors and distributed computing becoming the norm, this gap between possible and typically achievable performance will only widen. With this fellowship, my plan is to use deep learning to lift the barriers which keep optimised applications expensive. My work will produce analysis and optimisation methodologies which are fast, easy to use, and fully automated with no supervision or expert guidance. This work has the potential to massively increase performance for every application(2018-2023)

EPSRC Logo

SUMMER - Scheduling on Heterogeneous Mobile Multicores based on Quality of Experience
EPSRC-funded project on heterogeneous processor scheduling policies that minimise energy consumption while avoiding any degradation of the user's quality of experience (2017-2018)

ALEA Logo

ALEA - Abstraction Level Energy Accounting for Many-core Programming Languages
EPSRC-funded project on fine-grained energy profiling and energy-aware compiler optimisations (2014-2016)

Compucast is the podcast from computer scientists for computer scientists. It was started a few years ago by Hugh, went into a hiatus, and it was reborn in 2016. Now it's a collaboration between the University of Edinburgh, the University of Lancaster, and the University of St Andrews, with me as the chief editor.

Each episode features an interview with a researcher in the field of Computer Science, news from academia and industry, a quiz, and the INFAMOUS JOKE! If you don't know where to start, my favourite recent episode is the Robot Wars one, where I interviewed Sethu Vijayakumar on Robotics and his experience as a Judge on the revival of Robot Wars. We are on a short break now, but we will come back in the autumn with a new team. Stay tuned!

COSMIC

Every two years, I organise the International Workshop on Code Optimisation for Multi and Many-Cores. Its acronym is COSMIC but don't ask me how this is even possible. The past three workshops took place in conjunction with CGO. If everything works out well, COSMIC'19 will be part of CGO'19 in Washington DC.

Publications

Compiler Fuzzing through Deep Learning

Christopher Cummins and Pavlos Petoumenos and Alastair Murray and Hugh Leather
Proceedings of the 27th ACM SIGSOFT International Symposium on Software Testing and Analysis (ISSTA 2018)
June 2018 Amsterdam, Netherlands
Distinguished Paper Award
bibtex     abstract     paper     presentation    

End-to-end Deep Learning of Optimization Heuristics

Christopher Cummins and Pavlos Petoumenos and Zheng Wang and Hugh Leather
Proceedings of the International Conference on Parallel Architectures and Compilation Techniques (PACT 2017)
September 2017 Portland, Oregon, US
Best Paper Award
bibtex     abstract     paper     presentation    

Synthesizing Benchmarks for Predictive Modeling

Christopher Cummins and Pavlos Petoumenos and Zheng Wang and Hugh Leather
Proceedings of the International Symposium on Code Generation and Optimization (CGO 2017)
February 2017 Austin, Texas, US
Best Paper Award
bibtex     abstract     paper     presentation     clgen    

Minimizing the Cost of Iterative Compilation with Active Learning

William Ogilvie and Pavlos Petoumenos and Zheng Wang and Hugh Leather
Proceedings of the International Symposium on Code Generation and Optimization (CGO 2017)
February 2017 Austin, Texas, US
bibtex     abstract     paper     presentation    

ALEA: A Fine-grained Energy Profiling Tool

Lev Mukhanov and Pavlos Petoumenos and Zheng Wang and Nikos Parasyris and Dimitrios Nikolopoulos and Bronis de Supinski and Hugh Leather
ACM Transactions on Architecture and Code Optimization (TACO)
2017
bibtex     abstract     paper    

Towards Collaborative Performance Tuning of Algorithmic Skeletons

Christoper Cummins and Pavlos Petoumenos and Michel Stewer and Hugh Leather
Proceedings of the International Workshop on High-Level Programming for Heterogeneous and Hierarchical Parallel Systems (HLPGPU 2016)
January 2016 Prague, Czech Republic
bibtex     abstract     paper    

Iterative Compilation on Mobile Devices

Paschalis Mpeis and Pavlos Petoumenos and Hugh Leather
Proceedings of the 6th International Workshop on Adaptive Self-tuning Computing Systems (ADAPT 2016)
January 2016 Prague, Czech Republic
bibtex     abstract     paper    

Autotuning OpenCL Workgroup Size for Stencil Patterns

Christoper Cummins and Pavlos Petoumenos and Michel Stewer and Hugh Leather
Proceedings of the 6th International Workshop on Adaptive Self-tuning Computing Systems (ADAPT 2016)
January 2016 Prague, Czech Republic
bibtex     abstract     paper    

Power Capping: What Works, What Does Not

Pavlos Petoumenos and Lev Mukhanov and Zheng Wang and Hugh Leather and Dimitrios Nikolopoulos
Proceedings of the 21st IEEE International Conference on Parallel and Distributed Systems (ICPADS), 2015
December 2015 Melbourne, Australia
bibtex     abstract     paper     presentation    

Intelligent Heuristic Construction with Active Learning

William F Ogilvie and Pavlos Petoumenos and Zheng Wang and Hugh Leather
CPC '15: Proceedings of Compilers for Parallel Computing
January 2015 London, England
bibtex     abstract    

Measuring QoE of Interactive Workloads and Characterising Frequency Governors on Mobile Devices

Volker Seeker and Pavlos Petoumenos and Hugh Leather and Bjorn Franke
IISWC '14: Proceedings of the 2014 IEEE International Symposium on Workload Characterization
October 2014 Raleigh, North Carolina, USA
Best Paper Award
bibtex     abstract     paper     presentation    

Fast Automatic Heuristic Construction Using Active Learning

William F. Ogilvie and Pavlos Petoumenos and Zheng Wang and Hugh Leather
LCPC '14: Proceedings of the Workshop on Languages and Compilers for Parallel Computing
15 September 2014 Hillsboro, Oregon, USA
bibtex     abstract     paper    

Instruction-based Reuse Distance Prediction Replacement Policy

Pavlos Petoumenos and Georgios Keramidas and Stefanos Kaxiras
JWAC-1: 1st JILP Workshop on Computer Architecture Competitions (7th Place)
June 2010 St. Malo, France
bibtex     abstract     paper     slides     Competition Code    

Where replacement algorithms fail: a thorough analysis

Georgios Keramidas and Pavlos Petoumenos and Stefanos Kaxiras
CF 2010: Proceedings of the ACM International Conference on Computing Frontiers
May 2010 Bertinoro, Italy
bibtex     abstract     paper    

MLP-aware Instruction Queue Resizing: The Key to Power-Efficient Performance

Pavlos Petoumenos and Georgia Psychou and Stefanos Kaxiras and Juan M. Cebrian Gonzalez and Juan Luis Aragon
ARCS 2010: Proceedings of the International Conference on Architecture of Computing Systems
February 2010 Hanover, Germany
bibtex     abstract     paper     slides    

Instruction-based reuse-distance prediction for effective cache management

Pavlos Petoumenos and Georgios Keramidas and Stefanos Kaxiras
SAMOS IX: Proceedings of the International Conference on Systems, architectures, modeling and simulation
July 2009 Samos, Greece
bibtex     abstract     paper     slides    

Efficient microarchitecture policies for accurately adapting to power constraints

Juan M. Cebrian Gonzalez and Juan Luis Aragon and Jose M. Garcia and Pavlos Petoumenos and Stefanos Kaxiras
IPDPS 2009: Proceedings of the IEEE International Symposium on Parallel and Distributed Processing
May 2009 Rome, Italy
bibtex     abstract     paper    

Cache Replacement Based on Reuse-Distance Prediction

Georgios Keramidas and Pavlos Petoumenos and Stefanos Kaxiras
ICCD 2007: Proceedings of the IEEE International Conference on Computer Design
October 2007 Lake Tahoe, CA, USA
bibtex     abstract     paper     slides    

Using Value Locality to Reduce Memory Encryption Overhead in Embedded Processors

Georgios Keramidas and Pavlos Petoumenos and Stefanos Kaxiras and Alexandros Antonopoulos and Dimitrios Serpanos
ETFA 2007: Proceedings of the IEEE/ACM Conference on Emerging Technologies and Factory Automation
September 2007 Patras, Greece
bibtex     abstract     paper    

Modelling Cache Sharing on Chip Multiprocessor Architectures

Pavlos Petoumenos and Georgios Keramidas and Hakan Zeffer and Stefanos Kaxiras and Erik Hagersten
IISWC 2006: Proceedings of the IEEE International Symposium on Workload Characterization
October 2006 San Jose, CA, USA
bibtex     abstract     paper     slides    

Preventing Denial-of-Service Attacks in Shared CMP Caches

Georgios Keramidas and Pavlos Petoumenos and Stefanos Kaxiras and Alexandros Antonopoulos and Dimitrios Serpanos
SAMOS VI: Proceedings of the International Workshop on Embedded Computer Systems: Architectures, Modeling, and Simulation
July 2006 Samos, Greece
bibtex     abstract     paper    

Statshare: A statistical model for managing cache sharing via decay

Pavlos Petoumenos and Georgios Keramidas and Hakan Zeffer and Stefanos Kaxiras and Erik Hagersten
MoBS 2006: Proceedings of the Annual Workshop on Modeling, Benchmarking and Simulation
June 2006 Boston, MA, USA
bibtex     abstract     paper    

Presentations and Invited Talks

How fast? How furious? Real optimizations for real people    

GPG Seminars
4 May 2016 Glasgow, Scotland

How fast? How furious? Real optimizations for real people    

AMAS-DO 2016 (CGO Workshop)
13 March 2016 Barcelona, Spain

Power Capping: What works, what does not    

ICPADS 2015
16 December 2015 Melbourne, Australia

Iterative Compilation 1.01    

PPar Lunch Lectures
10 December 2014 Edinburgh, Scotland

Instruction-based Reuse Distance Prediction Replacement Policy    

1st JILP Workshop on Computer Architecture Competitions
20 June 2010 St. Malo, France

MLP-aware Instruction Queue Resizing: The Key to Power-Efficient Performance    

ARCS 2010
23 Febryary 2010 Hanover, Germany

Instruction-based reuse-distance prediction for effective cache management    

SAMOS-IX
20 July 2009 Samos, Greece

Cache Replacement Based on Reuse-Distance Prediction    

ICCD 2007
9 October 2007 Lake Tahoe, CA, USA

Modelling Cache Sharing on Chip Multiprocessor Architectures    

IISWC 2006
27 October 2006 San Jose, CA, USA

Theses

PhD Thesis - Managing Shared Resources in Multi-core Systems

Supervisor - Stefanos Kaxiras
Thesis (pdf - Greek Only)

Contact

Pavlos Petoumenos Informatics Forum 1.02
School of Informatics
10 Crichton Street
Edinburgh
EH8 9AB
Scotland

: +44 (0) 131 650 5152
: ppetoume at inf dot ed dot ac dot uk