Workshop: Thirteenth Workshop on Accelerator Programming and Directives (WACCPD 2026)
Authors: William Godoy (Oak Ridge National Laboratory), Antonio Peña (Barcelona Supercomputing Center)
Abstract: Hardware heterogeneity is not anecdotal anymore, but rather the norm. Furthermore, it is not any longer limited to the now-customary CPU+GPU system architectures. These days we haveSmartNICs, processing-in-memory (and its variants), quantum accelerators, FPGAs, vector processors, and a plethora of domain-specific accelerators, to name only a few, admittedly indifferent degrees of maturity and adoption. Application developers always struggle to figure out the right programming model(s), language(s), strategies (…) to make their codes future-proof.While AI-programming has quickly become the norm, we still want to understand AI-generated code (at least for the time being). In this talk, I will go over the current landscape of hardwareheterogeneity and its programming, my expectations for the upcoming future, and some solutions that my group has been working on for a variety of accelerator technologies.
Back to Thirteenth Workshop on Accelerator Programming and Directives (WACCPD 2026) Archive Listing
Back to Full Workshop Archive Listing