Enabling Accelerated AI and Data Workflows on CPUs, GPUs and FPGAs through oneAPI
E145 | Sun 16 Jul 4:30 p.m.–5:30 p.m.
Presented by
-
Sriram Ramkrishna
@sramkrishna
Sriram Ramkrishna is an over 20 year veteran in open source. Specializing specifically in open source communities - Sri has a solid track record in building rapport in existing communities, building new communities and meta communities. Sri has held roles from IT engineering roles to marketing and everything in between. Currently, Sri works at Intel as the community manager for oneAPI, an open standard, multi-architecture programming model that provides an open, industry driven alternative for writing AI and high performance computing on accelerators.
In community work, Sri is the brain child behind Linux App Summit, a conference that brings developers from the desktop community, distributions, and third party developers to drive application development on the Linux platform. Sri also participates in the open compute project as a liaison between the open system firmware team and the sustainability team focusing on how to build sustainable platforms. Finally, Sri is really good at connecting people - if you need help with your project or need visibility, he loves helping out.
Sriram Ramkrishna
@sramkrishna
Abstract
AI workloads and data pipelines are rapidly growing in their demand for computing power. Accelerated computing can help maximize our ability to not just run AI models, but also maximize how quickly we can capture and extract useful information from data sources. Until recently, accelerated programming has been focused on proprietary programming models, but this is changing.
oneAPI is an open, cross-industry, standards-based, unified, multiarchitecture, multi-vendor programming model that allows you to write code and target all accelerators - GPUs, CPUs and FPGAs. oneAPI is already used to accelerate AI frameworks. With oneAPI it is also possible to write accelerated data processing code to rapidly capture and transform your data.
This talk will provide examples of how oneAPI enables performant data processing, how the project is organized and the open governance model. We will also discuss the oneAPI community. You should expect to come away with a sense of what oneAPI is and fresh ideas on what new capabilities it enables.
AI workloads and data pipelines are rapidly growing in their demand for computing power. Accelerated computing can help maximize our ability to not just run AI models, but also maximize how quickly we can capture and extract useful information from data sources. Until recently, accelerated programming has been focused on proprietary programming models, but this is changing. oneAPI is an open, cross-industry, standards-based, unified, multiarchitecture, multi-vendor programming model that allows you to write code and target all accelerators - GPUs, CPUs and FPGAs. oneAPI is already used to accelerate AI frameworks. With oneAPI it is also possible to write accelerated data processing code to rapidly capture and transform your data. This talk will provide examples of how oneAPI enables performant data processing, how the project is organized and the open governance model. We will also discuss the oneAPI community. You should expect to come away with a sense of what oneAPI is and fresh ideas on what new capabilities it enables.