Interested in implementing the CMT FBT large LUT evaluation method

Library Questions
1

Hi,

Came across issue #1228 and after reading the paper ( Evaluating Larger Lookup Tables using CKKS ) for the last week, I am interested in implementing the CMT FBT method for LUT evaluation.

I have also been going through the current CKKS functional bootstrapping code. My plan is to start with the functional digit decomposition using the existing multi value bootstrapping support, then implement the Kronecker product and collapsed multiplexer tree, and finally put the full evaluation together with tests and benchmarks against the current FBT implementation.

Wanted to check with the library maintainers first if this is roughly the way you had in mind for approaching it or if there is anything else you think I should account for. Also is #1228 open for someone to take on?

Thanks,
Ishan

2

Hi! Thank you for your interest!

We plan to add large LUT evaluation to OpenFHE v1.7.0. In the meantime, you can check the poc implementation. It is based on OpenFHE v1.6.0, and is already several times faster than in the paper thanks to improvement in the OpenFHE library, and our implementation of CMT-FBT.

3

Thanks! I didn’t realise this was already planned for v1.7.0. I’ll take a look at the PoC. Is someone currently working on the integration? I’ve spent some time going through the paper and existing FBT code, so I’d be happy to help if there’s a part I could take on.