In the context of this research interest, there are three major projects. The first project focuses on the use of graphics processors (GPU) for general-purpose applications (GPGPU). Graphics processors, of which the development is driven by the large demands of the game applications, offer a large compute power. The interest in this area is to use these high-performance and low-cost GPUs for regular applications. The challenges are in effectively transforming the applications into the stream computing programming model supported by the GPU. The research in this project is currently on: the performance characterization of the GPU; the porting of several applications to the GPU; the use of the GPU in multiprocessor systems; and the compilation of regular programs for the GPU architecture.
The second project focuses on the efficient execution of database applications achieved through an effective use of the computer system. Examples of the research in this project include: the use of advanced memory systems, such as intelligent memory, for the execution of database applications; the use of reconfigurable processors to increase the performance of database systems; the design of efficient data filtering techniques for data stream processing; and the proposal of a less costly but still accurate database benchmark suite based on the standard TPC-H.
Finally, the third project focuses on the design of power-performance efficient adaptive microprocessors. The objective of the research is to have an architecture built out of components that are able to adapt to the needs of different applications or even different phases of the same applications. Furthermore, the adaptation is performed in order to achieve the desired high-performance but at the lowest possible power-consumption. The research in this project focuses on: the analysis of the demands of different applications on the different processor components; efficient design-space analysis techniques for power-performance optimization; design of adaptive components and complete adaptive microprocessor.