aszepieniec
aszepieniec
This writeup introduces a *contiguity argument*, which is a Polynomial IOP / RAP hybrid for establishing that the set of values is distributed contiguously across an evaluation domain. Phrased crudely,...
Add TIP-0002: multi-table subset argument
Yuncong's (@yczhangsjtu's) attack may apply whenever the correct sorting of memory tables is not enforced. It reads an inconsistent value from memory even though the memory table's AIR is satisfied...
Except for the Hash Table, all constraints are represented as multivariate polynomials using the `MPolynomial` struct. This representation is both more reflective of the underlying maths that makes STARKs work...
Instruction Buckets are cumbersome. A tiny change in the ISA throws the instruction buckets into disarray. Getting the constraints right is challenging. I propose to get rid of instruction buckets...
This PR fixes the proof system with respect to memory consistency. More specifically, it updates table population and extension, and defines constraints for memory consistency in accordance with TIPs [0001](https://github.com/TritonVM/triton-vm/blob/master/tips/tip-0001/tip-0001.md)...
Implement Transition Constraints for Correct Updating of Extension Columns for Cross-Table Arguments
Cross-table arguments like Evaluation Arguments and Permutation Arguments need an extension column on both tables to compute a running evaluation or running product. This correct computation is not verified, to...
Currently, the number of challenges is hardcoded in `challenges.rs`, near the top: ``` impl AllChallenges { pub const TOTAL_CHALLENGES: usize = 130; .... } ``` All challenges have a specific...
For the purpose of U32 operations, this note proposes an alternative to relying on a separate table with corresponding table relations. It describes how the `split` instruction is powerful enough...
Bug-hunt 2aed4267c0a21d4e6b6212b263c64433abc0c0d4 identified two types of errors that can occur when running the VM. 1. Rust produces a "`failed to fill whole buffer`" message. This is probably related to stdin/secret_in...