Section: Computer Science and Engineering
Tutor: LANZI PIER LUCA Major Research topic
:Computer Aided Design framework for High Performance Reconfigurable Systems
Advisor: SANTAMBROGIO MARCO DOMENICOAbstract:
In order to meet the stringent performance requirements of future exascale HPC applications, HPC systems will require highly power efficient heterogeneous compute nodes. It is foreseeable that different types of nodes will be available in such systems, where high performance CPU nodes will be present side by side to application specific nodes in order to maximize the exploitation of low-level parallelism.
Within this context, Field-Programmable Gate Arrays (FGPAs) are gaining considerable attention from the community thanks to their flexibility in being reconfigured as needed by the application designer and their relative low power consumption with respect to provided computational capacity. Nevertheless, the main limiting factor towards a wide adoption of such devices for HPC systems, relies on the application design process, which requires highly experienced users and a considerable amount of time due to hardware synthesis. Furthermore, the heterogeneity offered by the system highly increases the size of the design space in terms of application tasks mapping and scheduling solutions, thus increasing the probability of the adoption of sub-optimal implementations.
The main objective of this research is the development of a framework able to guide the application developer in the implementation of efficient hardware-software solutions for high performance reconfigurable systems. The framework aims at assisting the designer starting from the high level analysis of the code, towards the definition of the functionalities to be accelerated on the reconfigurable nodes, the selection of a suitable architectural template for the FPGA devices and the final implementation of the system together with its runtime support.