On June 3rd, 2022  at 1.00 pm "Democratising the acceleration of sparse computations"  a new appointment of  NECSTFridayTalk, will be held  by Luigi Fusco, Master student in Computer Science and Engineering at NECSTLab, Politecnico di Milano.

Sparse computations are a crucial component of many real-world applications. As these applications become pervasive and the amount of data to be processed increases, more and more people will need access to fast computation of these workloads. Specific applications that take advantage of sparse data structures include matrix factorization, graph neural network training, and eigenproblems. GPUs are often the go-to tool to accelerate these computations, but their utilization requires experienced engineers and many lines of code to manage communication and scheduling. GrCUDA aims to help programmers at dealing with these complexities in a transparent way, by simplifying allocation and scheduling, both in the single and multi-GPU scenario. This project aims to add support to sparse data structures to GrCUDA, integrating the state of the art cuSPARSE library and adding support to multiple formats and operations. This approach has already proved successful, with a paper presented at ISCAS 2022 about the acceleration of Multi-GPU Large-scale Top-K Sparse Eigenproblems, leveraging GrCUDA and achieving performance speedup of 67x over CPU and 1.9x over FPGA implementations.

The NECSTLab is a DEIB laboratory, with different research lines on advanced topics in computing systems: from architectural characteristics, to hardware-software codesign methodologies, to security and dependability issues of complex system architectures.

Every week, the “NECSTFridayTalk” invites researchers, professionals or entrepreneurs to share their work experiences and projects they are implementing in the “Computing Systems”.


Streaming will be available via Facebook