Sort by
Refine Your Search
-
Listed
-
Category
-
Field
-
, energy consumption, and accuracy.; ; Training deep learning models, especially in LLMs, faces critical challenges that compromise the optimal use of GPUs. These bottlenecks result in poor computational
-
main activities: ; 1. Exploration of the applicability of eBPF and its ecosystem: review and exploration of the use of eBPF in different domains (e.g., GPU), of the various libraries available for its
-
including heterogeneous accelerator devices such as GPUs, DSPs or FPGAs, requiring software to cope with concurrent and parallel synchronous and asynchronous computation. The emergence of connected autonomous
-
techniques that enable the application of eBPF to areas that currently lack direct support (e.g., GPUs, HPC systems, etc.); 2. Development of new eBPF functionalities: exploration and development of new
-
including heterogeneous accelerator devices such as GPUs, DSPs or FPGAs, requiring software to cope with concurrent and parallel synchronous and asynchronous computation.; The emergence of connected