fhewsyn

Netlist repository for the paper Improved Circuit Synthesis with Multi-Value Bootstrapping for FHEW-like Schemes by Johannes Mono, Kamil Kluczniak, and Tim Güneysu [1].

Netlists

The directory data/ contains the following netlists:
- addX: Add two X-bit numbers.
- calculator: Simple 16-bit calculator supporting addition, substraction and multiplication.
- constX: Add a X-bit constant to a X-bit number.
- image: Process an 8x8 grayscale image using a Gaussian blur, sharpening, or ricker wavelet.
- relu: Computes the ReLU of a 16-bit integer.
- sqrt: Computes the square root of a 16-bit integer.
- strrev: Reverses an array of characters of up to 8 characters.
- structs: Computes the sum for a 32-bit integer field in a multi-dimensional array of structs, each dimension is upper bounded by 2.
- sum: Computes the sum of two 32-bit integers.
- sum3d: Computes the sum for a three-dimensional array of 8-bit integers with maximum dimensions 2, 3, 2.

Reproduce

You need Yosys (we used commit 7ce5011c24bc4c605578bb16576cadf79373e23f) and HAL (we used commit fdbea579c21d56028730b86740cf124d2bc38639).
We also use the python packages functools, natsort, collections, and re.
You can reproduce netlist generation as follows where X is the maximum look-up table size:

./build.sh boolean -a
./build.sh lut -aX
./build.sh lutfa -aX

For X = 2, you have to comment in line 3 in src/lut.ys and src/lutfa.ys, otherwise Yosys will still generate three-bit look-up tables.

For FHE-Deck's dependencies, see fhe-deck-core/README.md.
You can generate, build, and run the benchmarking code as follows:

./bench.sh
cd fhe-deck-core
cmake -S . -B build
cmake --build build
bench/run.sh

Cite

@article{fhewsyn,
  author       = {Johannes Mono and Kamil Kluczniak and Tim G{\"{u}}neysu},
  title        = {Improved Circuit Synthesis with Amortized Bootstrapping for {FHEW}-like Schemes},
  journal      = {IACR Cryptol. ePrint Arch.},
  pages        = {1223},
  year         = {2023},
  url          = {https://eprint.iacr.org/2023/1223},
  timestamp    = {Fri, 08 Sep 2023 15:28:09 +0200},
  biburl       = {https://dblp.org/rec/journals/iacr/MonoKG23.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}