[HiPEAC-announce] PhD at INRIA: Infrastructures and Compilation Strategies for Heterogeneous Manycore Systems

[Erven Rohou]

The ALF research group at INRIA Rennes (http://www.irisa.fr/caps) has an opening 
for a PhD position.


Context
Multicore processors have now become mainstream for both general-purpose and 
embedded computing. In the near future, every hardware platform will feature 
thread level parallelism. Within a decade, it will become technologically 
feasible to implement 1000's of cores on a single chip. It is also expected that 
various types of cores will be implemented on the same die. At the same time, 
one can expect that legacy sequential codes will be progressively replaced by 
parallelized versions or newly developed parallel applications.
In order to execute efficiently these parallel applications on a large spectrum 
of parallel hardware platforms, the computer science community has to invent a 
new form of application portability that will replace the traditional binary 
compatibility. Processor virtualization combined with split-compilation can be 
used to address this portability issue. Applications are no longer directly 
compiled in the final native code, but in a target independent bytecode format. 
The final native code generation is delayed till the executing target is known. 
It can even occur at run-time through just-in-time compilation (JIT). JIT can 
leverage all dynamic information on the application execution context: actual 
target processor, current machine load, input values, etc.

Thesis
While virtualization is a good alternative to binary compatibility for allowing 
application portability, performance portability is a major issue. The objective 
of this thesis is to explore the benefits of virtualization and 
split-compilation in the context of heterogeneous manycore systems, for 
achieving portable performance. The objective is to enrich processor 
virtualization to allow both functional portability and high performance using 
JIT at runtime, or bytecode-to-native code offline compiler. Split compilation 
can be used to annotate the bytecode with relevant information that can be 
helpful to the JIT at runtime or to the bytecode-to-native code offline 
compiler. Because the first compilation pass occurs offline, aggressive analyses 
can be run and their outcomes encoded in the binary. For example, such 
information include vectorizability, memory references (in)dependencies, 
suggestions derived from iterative compilation, polyhedral analysis, or integer 
linear programming. Virtualization allows to postpone some optimizations to run 
time, either because they increase the code size and would increase the cost of 
an embedded system, or because the actual hardware platform characteristics are 
unknown.

References
[1] K. Asanovic, R. Bodik, B. Catanzaro, J. Gebis, P. Husbands, K. Keutzer, D. 
Patterson, W. Plishker, J. Shalf, S. Williams, and K. Yelik. The Landscape of 
Parallel Computing Research: A View from Berkeley. Technical Report 
UCB/EECS-2006-183, EECS Department, University of California at Berkeley, 
December 2006.
[2] Koen De Bosschere, Wayne Luk, Xavier Martorell, Nacho Navarro, Mike O'Boyle, 
Dionisios Pnevmatikatos, Alex Ramirez, Pascal Sainrat, André Seznec, Per 
Stenström, and Olivier Temam. High-Performance Embedded Architecture and 
Compilation Roadmap, volume 4050/2007 of LNCS, pages 5-29. 2007.
[3] M. Cornero, R. Costa, R. Fernández Pascual, A. Ornstein, and E. Rohou. An 
Experimental Environment Validating the Suitability of CLI as an Effective 
Deployment Format for Embedded Systems. In International Conference on HiPEAC, 
volume 4917 of LNCS, pages 130-144, Göteborg, Sweden, January 2008.
[4] Roberto Costa and Erven Rohou. Comparing the size of .NET applications with 
native code. In CODES+ISSS, pages 99-104, 2005.
[5] Chandra Krintz and Brad Calder. Using annotations to reduce dynamic 
optimization time. In Proc. of PLDI, pages 156-167, 2001.
[6] Piotr Lesnicki, Albert Cohen, Grigori Fursin, Marco Cornero, Andrea 
Ornstein, and Erven Rohou. Split compilation: an application to just-in-time 
vectorization. In International Workshop on GCC for Research in Embedded and 
Parallel Systems, Brasov, Romania, September 2007.
[7] Computing Systems Consultation Meeting. Research Challenges for Computing 
Systems - ICT Workprogramme 2009-2010. European Commission - Information Society 
and Media, Braga, Portugal, November 2007.
[8] Patrice Pominville, Feng Qian, Raja Vallée-Rai, Laurie Hendren, and Clark 
Verbrugge. A framework for optimizing Java using attributes. In Proc. of CC, 
volume 2027 of LNCS, pages 334-354, 2001.


Contacts: erven.rohou at irisa.fr or andre.seznec at irisa.fr