CUDA Verkle
github.com/dyslex7c/cuda-verkle- Category
- Crypto
- Rank
- No. 4133Tools index
Previous survey · No. 3913 ·
- Pricing
- Open Source
- Type
- TOOL
- GitHub
- 1 stars
- Date
About
CUDA Verkle is a C++/CUDA research implementation of GPU-accelerated Pedersen vector commitments over the Banderwagon group for Ethereum's proposed Verkle tree state format (EIP-6800). It covers the full commitment pipeline, from Montgomery field arithmetic and twisted Edwards curve operations through Pippenger multi-scalar multiplication and 256-element IPA opening proofs to a sparse key/value state tree, and is explicitly experimental, not production-ready cryptographic software.
At a glance
CUDA Verkle is an experimental C++ implementation of Verkle tree commitments with a CUDA path for the expensive group arithmetic. It is worth studying if you want to compare GPU and CPU results against known reference vectors.
Good fit
Researchers working on Banderwagon commitments, opening proofs or GPU acceleration.
Consider something else when
Use established cryptographic implementations for production or workloads involving secret inputs.
Start here
Start with the documented CPU tests and reference vectors. For GPU evaluation, compare commitment outputs with the CPU path and record the device and toolchain before interpreting timings.
Cost and requirements
MIT-licensed. The host tests use a C++17 compiler. GPU validation requires an NVIDIA device and the CUDA Toolkit.
Limits to know
The project marks the cryptography experimental and warns that strict decoding is variable-time. GPU results apply to the tested device and software configuration.
Sources checked 2026-11-05. Source-based assessment.
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.