Justin Thaler

Results 10 issues of Justin Thaler

Many zkVMs today use continuations, which means they break the execution of a computer program into chunks and prove each chunk independently before recursively aggregating the proofs into one. One...

enhancement
optimization

This is a major issue involving the following sub-issues. *Switch the finite field to GF[2^128]. **This will force changing all the "subtables and combining function" in the Instruction Lookups (Lasso)....

enhancement
optimization

Zeromorph paper: https://eprint.iacr.org/2023/917 This will make the Jolt proofs much shorter than when using the Hyrax commitment, and will also slightly speed up the prover (Hyrax commitments are big enough...

optimization

Jolt currently uses the grand product argument from [Thaler13](https://eprint.iacr.org/2013/351.pdf), which has roughly log^2(n)-sized proofs for products of n values. Section 6 of the [Quarks paper](https://eprint.iacr.org/2020/1275) reduces this proof size to...

optimization

More instructions: The Jolt codebase currently implements the RISC-V 32-bit Base Integer instruction set (RV32I), but the Jolt construction is highly flexible. Add support for the RISC-V “M” extension for...

enhancement

The on-chain verifier will be much cheaper after switching the polynomial commitment scheme to one with logarithmic verifier costs (see [this other issue](https://github.com/a16z/Lasso/issues/208))

optimization

Incorporating precompiles into Jolt is essential to achieve developer adoption, since performance of all zkVMs, Jolt included, is simply too slow today without them. The most natural approach is to...

enhancement
optimization

The current Rust verifier is not optimized at all (no parallelization, etc.)

optimization

Zeromorph is described in this paper: https://eprint.iacr.org/2023/917. But it's very focused on hiding/zk, which is not an initial priority for Jolt. HyperKZG is described and implemented here: https://github.com/microsoft/Nova/blob/main/src/provider/hyperkzg.rs Switching to...

enhancement
optimization

About 93% of the factors arising the Lasso grand product arguments are 1. We are currently not leveraging this fact when committing to the partial products in the Hybrid or...