Innovations in Performance, Efficiency, and Productivity for FPGAs in Data Centers

Speaker: Gabriele Montanaro
DEIB PHD Student
DEIB - Beta Room (Bld. 24)
February 13th, 2025 | 2.00 pm
DEIB PHD Student
DEIB - Beta Room (Bld. 24)
February 13th, 2025 | 2.00 pm
Sommario
On February 13th, 2025 at 2.00 pm Gabriele Montanaro, PHD Student in Information Technology, will hold a seminar on "Innovations in Performance, Efficiency, and Productivity for FPGAs in Data Centers" at DEIB Beta Room (Building 24).
The adoption of Field-Programmable Gate Arrays (FPGAs) is becoming increasingly relevant within the high-performance computing (HPC) industry, driven by their potential to deliver substantial enhancements in computational performance and energy efficiency. However, traditional programming and prototyping methodologies for these large and complex devices are inadequate to meet the escalating demands for performance and energy efficiency. These limitations primarily derive from the inherent slowness of the design process and the challenges in fully optimizing device performance and resource utilization.
This seminar will explore emerging strategies for the efficient deployment of FPGAs in data centers, addressing the aforementioned challenges. Specifically, we will examine the use of System-on-Chip (SoC) prototyping platforms to accelerate the development of novel architectures, multi-tenancy frameworks to optimize the resource utilization of cloud-based FPGAs by partitioning their logic among multiple users, advanced programming techniques aimed at enhancing designer productivity, and optimization methodologies to improve quality of results. The seminar will conclude with a discussion of a practical use case that integrates several of these techniques, providing insights into their effectiveness and real-world applicability.
The adoption of Field-Programmable Gate Arrays (FPGAs) is becoming increasingly relevant within the high-performance computing (HPC) industry, driven by their potential to deliver substantial enhancements in computational performance and energy efficiency. However, traditional programming and prototyping methodologies for these large and complex devices are inadequate to meet the escalating demands for performance and energy efficiency. These limitations primarily derive from the inherent slowness of the design process and the challenges in fully optimizing device performance and resource utilization.
This seminar will explore emerging strategies for the efficient deployment of FPGAs in data centers, addressing the aforementioned challenges. Specifically, we will examine the use of System-on-Chip (SoC) prototyping platforms to accelerate the development of novel architectures, multi-tenancy frameworks to optimize the resource utilization of cloud-based FPGAs by partitioning their logic among multiple users, advanced programming techniques aimed at enhancing designer productivity, and optimization methodologies to improve quality of results. The seminar will conclude with a discussion of a practical use case that integrates several of these techniques, providing insights into their effectiveness and real-world applicability.