CKKS Kernel Profiler

Profiling framework for secure ML primitives.

This project provides a profiling and analysis toolkit for CKKS-based workloads. It focuses on identifying runtime hotspots in automorphism and NTT-heavy operator pipelines.

Highlights

  • detailed operator-level tracing for encrypted inference paths
  • latency and memory breakdown across kernel stages
  • reusable experiment scripts for architecture studies

Outcome

The framework makes it easier to compare optimization strategies and supports reproducible reporting for paper artifacts.

References