Beginning of dialog window. Escape will cancel and close the window.
End of dialog window.
详情
字幕
Low-Latency Numerical Computing With MatX
, Principal Developer Technology Engineer, NVIDIA
, Solutions Architect, NVIDIA
, Sr. Developer Technology Engineer, NVIDIA
This lab will introduce MatX and demonstrate how to accelerate latency-sensitive workflows with its easy-to-use C++ template library for numerical computing. MatX's powerful Tensor language enables rapid prototyping and a seamless transition to production. Using this library, we'll demonstrate a powerful set of well optimized transforms and operations and walk users through optimizing their pipelines to speed of light performance. We'll then use our environment's profiling tools to measure and ensure compliance with critical performance budgets. From this solid foundation, we'll build a real-world workflow, showing how MatX's simple syntax can still provide outstanding acceleration. Finally, we'll use MatX's extensible operators to develop our own custom operator, seamlessly extending MatX for our own unique function. Prerequisite(s):
Basic understanding of C++ and the ability to work with the command line.