080 Accelerating NTT with RISC-V Vector Extension for FHE w/ Alexandre Rodrigues скачать в хорошем качестве

080 Accelerating NTT with RISC-V Vector Extension for FHE w/ Alexandre Rodrigues

Abstract We target the optimization of FHE by leveraging vectorized implementations in RISC-V processors, using the RISC-V Vector (RVV) Extension. In particular, we accelerate the Number Theoretic Transform (NTT) and the Inverse-NTT (INTT), integrating our implementations in OpenFHE. The specific contributions of our paper are: 1) A systematic analysis of the performance and memory access patterns of four widely used NTT/INTT algorithms, namely Cooley-Tukey/Gentelman-Sande and Korn-Lambiotte/Pease; 2) RVV-optimized implementations of these algorithms, including modular arithmetic routines and tailored butterfly structures and root-of-unity accesses; 3) The integration of our vectorized implementations into the OpenFHE library; 4) A comprehensive performance evaluation using two distinct computational platforms - a Vector Processing Unit (VPU) running on an FPGA and an out-of-order core simulated in gem5. Our results show that the Korn-Lambiotte/Pease implementations outperform Cooley-Tukey/Gentelman-Sande, which are commonly used by state-of-the-art libraries (e.g. Intel HEXL). Our approach achieves significant speedups in both NTT benchmarks and complete FHE applications. About the speaker Alexandre Rodrigues is currently pursuing a PhD in Computing from Instituto Superior Técnico’s doctoral programme in Electrical and Computer Engineering. His research interests include performance modelling, microarchitectural design and RISC-V. Never miss an update The newsletter where we post community announcements: https://fheorg.substack.com/ The discord server where you can discuss FHE related topics with the community: https://discord.fhe.org Make sure to join either (or both) of these to stay informed about future events! View all resources related to this meetup on https://fhe.org/meetups/080