Postdoc: Data Mining On Chip for Low Energy Systems

Duration: 12 monthes Expected start date: Between June and September 2010

Postdoc position for the DAMOCLES project: DAta Mining On Chip for Low Energy Systems

Co-located between TIMA lab and LIG lab

Context:
The first goal of this postdoc is to adapt existing frequent pattern mining algorithms in order to drastically reduce their computational needs, and use them to find patterns of interest in the traffic and memories access in MultiProcessor Systems-on-Chip (MPSoC). These algorithms will be used to analyze trace data of simulated MPSoC executions.

The second goal is to discover problematic memory access patterns in these traces, such as contention on a memory block or long pathes between computation core and the memory blocks holding computation's data. These patterns will be used to optimize task and data placement in order to reduce energy consumption. This novel and promising application of data mining raises important research challenges.

The candidate will work with Frédéric Pétrot from TIMA and Alexandre Termier from LIG.

Prerequisites:
The candidate is expected to have a strong background in computer science (graphs, parallel programming, ...) and be knowledgeable in multiprocessor architecture and hardware modeling. Practical use of C++ is mandatory, knowledge of SystemC is a plus.

_Contact_:
How to apply: Please mail your CV to Alexandre.Termier@imag.fr .