'A deep dive into optimizing Multi-Scalar Multiplication' (Niall Emmart, Yrrid)

Abstract:
Multi-Scalar Multiplication is a computationally expensive operation that is heart of many ZK proof systems. In this talk I will do a deep dive into the techniques and optimizations that were used to win the GPU MSM and WASM MSM ZPrize competitions. The talk has two parts - the first will cover the high level algorithms used to minimize the number of expensive EC point and finite field operations. In the second part, we will take an in depth look into different multiple precision (MP) integer representations and their impact on performance.

Bio:
Niall Emmart recently won first place in two ZPrize competition categories for MSM. He completed his PhD in CS from the University of Massachusetts. His research focused on MP arithmetic on GPUs. He then joined NVIDIA, where he worked on MP libraries and accelerating DL GEMMs and convolutions.