Maximizing Kernel Development Productivity Under Performance Constraints, Philippe Tillet, OpenAI

Опубликовано 18 декабря 2024, 17:32

Machine Learning research workflows are often bottlenecked by the development of specialized compute kernels for novel algorithms. This process can be daunting and often requires a careful trade-off between productivity and performance: the most valuable code may not be that which is faster today, but that which will be easier to modify tomorrow.

In the Advancing AI 2024 Luminary Developer Keynote, Dr. Philippe Tillet discusses how Triton---a mid-level programming language for kernel development--approaches this multi-objective optimization challenge, and the design decisions made. In particular, he highlights how Triton enabled OpenAI to build a fast, yet portable kernel library understood by researchers.

Speaker: Philippe Tillet, OpenAI

Learn how to install Triton for ROCm rocm.docs.amd.com/projects/rad...

The information contained in this video represents the view of AMD or the third-party presenter as of the date presented. AMD and/or the third-party presenters have no obligation to update any forward-looking content in the above presentations. AMD is not responsible for the content of any third-party presentations and does not necessarily endorse the comments made therein. GD-84.

© 2024 Advanced Micro Devices, Inc. All rights reserved. AMD, the AMD Arrow logo, EPYC, ROCm, and Instinct and combinations thereof are trademarks of Advanced Micro Devices, Inc.