HCSCA
Heterogeneous Computing Systems with Customized Accelerators
The main goal of this proposal is to investigate into future heterogeneous computing system architecture main building blocks (namely, customizable CPUs, vector engines, and specialized accelerator tiles), and to integrate the resulting architectural components into high-level system design. While these building blocks have been explored in depth in the past, we believe that this is the perfect moment to impact the hardware and software stack of high-performance computing systems for the future. This is because of the following potentially disruptive research opportunities: First, the RISC- V open hardware initiative, spearheaded by ETHZ, is growing and gaining popularity and we argue that it is at the start of an exponential growth curve much like its open software counterpart in the early 1990s. Second, increased emphasis over new killer-applications such as deep learning, crypto (blockchain, cryptocurrencies) and security, high-definition video- content delivery are requiring exponentially increasing computation power within limited energy budget in both HPC and embedded domain.
Publications:
external pageA RISC-V in-network accelerator for flexible high-performance low-power packet processingcall_made
Di Girolamo Salvatore, Kurth Andreas, Calotoiu Alexandru, Benz Thomas, Schneider Timo, Beranek Jakub, Benini Luca, Hoefler Torsten (2021), A RISC-V in-network accelerator for flexible high-performance low-power packet processing, in 2021 ACM/IEEE 48th Annual International Symposium on Computer Architecture (ISCA), Valencia, SpainIEEE, New Jersey, USA.
external pageHighly parallel GPU accelerator for HEVC transform and quantizationcall_made
Cobrnic Mate, Duspara Alen, Dragic Leon, Piljic Igor, Kovac Mario (2020), Highly parallel GPU accelerator for HEVC transform and quantization, in Third International Conference on Image, Video Processing and Artificial Intelligence, Shanghai, ChinaSPIE, Washington, USA.
external pageHow Europe Is Preparing Its Core Solution for Exascale Machines and a Global, Sovereign, Advanced Computing Platformcall_made
Kovač Mario, Notton Philippe, Hofman Daniel, Knezović Josip (2020), How Europe Is Preparing Its Core Solution for Exascale Machines and a Global, Sovereign, Advanced Computing Platform, in Mathematical and Computational Applications, 25(3), 46-46.
external pageParallel Processing and Applied Mathematics13th International Conference, PPAM 2019, Bialystok, Poland, September 8–11, 2019, Revised Selected Papers, Part Icall_made
Cobrnic Mate, Duspara Alen, Dragic Leon, Piljic Igor, Mlinaric Hrvoje, Kovac Mario (2020), Parallel Processing and Applied Mathematics13th International Conference, PPAM 2019, Bialystok, Poland, September 8–11, 2019, Revised Selected Papers, Part I, Springer International Publishing, Cham.