This talk will explore how RISC-V architecture can accelerate custom workloads, focusing on AI/ML applications. We’ll start by examining the RISC-V ecosystem and its increasing relevance in the software development landscape. By looking at OpenBLAS, a highly optimized linear algebra library, we'll demonstrate how it's been optimized for RISC-V hardware, leading to significant performance gains in AI/ML tasks.
We’ll also highlight the ongoing work at RISC-V International on the matrix multiply extension, which promises to further enhance performance for these critical operations. Through practical examples, the talk will highlight why RISC-V is a strong candidate for optimizing workloads requiring high computational efficiency.
This session will offer valuable insights into how RISC-V can meet demanding performance needs while providing flexibility for custom workload optimization.
Speaker
![](https://qconsf.com/sites/qcon_sf/files/styles/medium/public/pictures/2024-08/Ludovic_profile_small.jpg?itok=nJwSDtyq)
Ludovic Henry
Member of Technical Staff @Rivos, Performance-Minded Engineer, Hardware & Software, Previously @Xamarin, @Microsoft, @Datadog
Ludovic is currently working at Rivos, a RISC-V hardware company, focused on accelerating AI/ML and Data Analytics workloads. His focus is on ensuring software stacks like Java, Python, Go, and various System Librariers are optimal in their usage of the hardware, as well as making sure the right hardware gets developed to accelerate these workloads.
His passion for hardware has developed over the past decade where he had the chance to work on multiple runtime technologies: OpenJDK, Go, Mono, .NET, Ruby, and more. These experiences have taught him how important hardware is to the performance of software, and how to make best use of it.