An ARM64 Guide for Pythonistas

Ram Iyengar (~ram40)


1

Vote

Description:

ARM64 is quite the rage right now. It is prevalent in local compute on laptops as well as in the cloud with popular IaaS providers/hyperscalars. In this talk, attendees will understand what the advantages are when using ARM64, especially in the context of Python work. The talk shall highlight the benefits of ARM64 for say ― data science workloads ― emphasizing its power efficiency and potential performance gains for computationally intensive tasks like machine learning and data analysis. There will also be certain specifics like NEON (SIMD instructions), a powerful feature of ARM64 that can significantly accelerate data science libraries like NumPy and SciPy. The talk will also focus on how these NEON instructions can be leveraged for common data science work such as vectorized calculations and matrix operations. The transition to ARM64 has not been without challenges. Tools like conda and pip have become crucial for managing ARM64-compatible environments. Package repositories like conda-forge are actively working on providing pre-built ARM64 binaries for popular data science libraries like NumPy, Pandas, and Scikit-learn. Severe performance bottlenecks have been identified when working with emulators such as Rosetta on Apple silicon. Libraries relying heavily on AVX instructions not supported by ARM64 have demonstrated the most degradation in performance. Come listen to more such anecdotes and nerd speak in this talk.

Speaker Info:

Ram Iyengar is an engineer by practice and an educator at heart. He was (cf) pushed into technology evangelism along his journey as a developer and hasn’t looked back since! He enjoys helping engineering teams around the world discover new and creative ways to work. He is a proponent of product development and engineering teams that put the community first.

Section: Python on Hardware
Type: Talk
Target Audience: Intermediate
Last Updated: