Investigate incorporating PEGASUS for CKKS / FHEW scheme switching
- Dominant language
- MLIR
- Stars
- 906
- Forks
- 171
- Avg merge
- 4d 12h
- Merged PRs (30d)
- 32
Description
https://eprint.iacr.org/2020/1606.pdf
See this comment:
Thanks for the information. We are current working on designing a FHE accelerator that could simutaneously accelerator both arithmetic (e.g., CKKS) and boolean (e.g., TFHE) schemes and technologies to convert between them (e.g., PEGASUS https://eprint.iacr.org/2020/1606.pdf). To achieve the goal, we need to analyze the common abstractions of these schemes, which is similar to what you mentioned. Moreover, we have some RTL code for the NTT/iNTT and transpose (not open sourced yet). I would like to see if these code is useful to the project.
For the hardware benchmarking system, we have several Xilinx U250 accelerators with VU13P FPGA and could use them for benchmarking. By the way, many people (at least in the academic field) are now working on ASIC-based accelerator (details are in our survey), how to benchmark these accelerators brings a large question (maybe a simulator is needed).
_Originally posted by @snowzjx in https://github.com/google/heir/discussions/39#discussioncomment-7998502_
Contributor guide
Assessment
This issue has not been assessed yet.