HEaaN.MLIR: An Optimizing Compiler for Fast Ring-Based Homomorphic Encryption
summary
A compiler that lowers ring HE programs into CPU machine code and two custom IRs Poly and ModArith IRs perform cross‑polynomial optimizations: optimized Barrett reduction with exactness propagation, loop fusion across polynomial loops, autovectorization guided by profiling, and multithreaded codegen. Builds a pipeline that parses Poly/ModArith, applies fusion and modulus analyses, and emits fast integer tensor loops, yielding measured speedups of up to 4.55x (single‑thread up to 3.06x, average 20–30%, some cases 200% and 60% on a 3‑layer network).
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:48Homomorphic encryption uses
- 4:13Optimization challenges
- 6:19Polynomial acceleration limits
- 7:23Compiler design and optimizations
- 8:30Unit ops in Poly intermediate language
- 11:44Modular reduction and loop fusion
- 13:30Autovectorization of loops
- 14:34Performance evaluation
- 15:17Q&A
related talks
59mDoubly Efficient Private Information Retrieval and Fully Homomorphic RAM Computation from Ring LWE
58mFast Cryptographic Code via Partial Evaluation
20mVectorizing Sparse Coiteration for Two-finger Loop Structure (Extended Abstract)
22mAlgorithm-Aware Hardware Optimization using E-Graph Rewriting: how should we marry(…)
23mGenerating Programs for Polynomial Multiplication with Correctness Assurance
16mA Compiler for Fused Relational Operations on Multisets
22mCircuit Optimization using Arithmetic Table Lookups
30mThe Solidity Optimizooooor