Computer architecture, hardware-software interaction and optimization
Research area: Computer Science
This research group addresses research topics strictly related to technological opportunities, issues and development constraints of modern and future computing systems exposing multi-/many-core parallelism both for high-performance server/HPC architectures and embedded/personal devices. Major objectives are performance and, especially, energy efficiency as the latter is the major limiting factor in the evolution of microprocessors and systems.
As a cross-cutting issue, the group deals also with their effective and, most of all productive, programmability in the complex upcoming scenario where parallelism and computing heterogeneity play a dominant role (parallel CPUs and GPUs working together) in all zones of the computing spectrum (from smartphones up to HPC systems). We are developing a single-source productivity-oriented approach for automatic and performance-portable parallel programming for multi-core CPUs and many-core GPUs (see www.phast-library.com); we are addressing multiple domains: from multimedia elaboration to finance, from physics simulation to security and neural-networks.
Furthermore, the group competences span also from architecture-aware performance analysis and characterization up to novel proposals and enhancements at the hardware, software, ISA (HW-SW interface), parallelization levels in integration with the adoption of innovative technologies (e.g. integrated photonics).
Then, the lab deals with computational challenges in various application domains: e.g. computer vision, cyber-security, block-chain and mobile applications.
ERC Keywords PE6_1 Computer architecture, parallel, distributed and pervasive computing PE6_2 Computer systems, parallel/distributed systems, sensor networks, embedded systems, cyber-physical systems PE6_8 Computer graphics, computer vision, multi media, computer games