# Project proposal: CraneZisk

Index document: [CraneZisk index](/1NqZLstgS2KpTASWRgOulg)

## :beginner: Project Info

:small_blue_diamond:Project Name: CraneZisk

:small_blue_diamond:Team: Andrew Borg in collaboration with Polygon

:small_blue_diamond:Expected timeline: October 2024 - February 2025

## :star: Summary

The proposal to build an efficient compiler from Rust to Zisk zkVM based on the [Cranelift toolchain](https://cranelift.dev).

## :triangular_flag_on_post: Background

During collaboration between Polygon Hermez and NEAR we've built a [backend for Cranelift compiler](https://github.com/near/wasmtime/blob/main/docs/zkasm/architecture.md) that can compile Rust and WASM programs to [zkEVM and zkASM](https://docs.polygon.technology/zkEVM/spec/zkasm/).

This approach showed a lot of promise, being able to generate zkASM for [Keccak hash function](https://github.com/near/wasmtime/blob/ca2f1d0d48045ea3991aa11349e126a8c7aa3480/cranelift/zkasm_data/benchmarks/keccak/src/lib.rs) benchmark with performance within 10x (cycles) of the hand-written state machine.

The next leap in performance can come from switching to zkVM that is [better tailored to compilation from Cranelift IR](https://github.com/near/wasmtime/issues/158) (efficient 32-bit operations, efficient access to registers and memory).

Zisk is the new zkVM memory-based architecture that would be a great target for Cranelift compilation.

## :spiral_note_pad: Stages and objectives

**Problem Formulation: Enable efficient compilation of Rust and WASM programs to Zisk**

### 1. PoC Cranelift-based compiler from Rust to Zisk

**Goal:** Build a proof-of-concept Cranelift Zisk backend that can compile multiple representative Rust benchmarks and surpass the performance of RiscV based backend:
- Iterative SHA256
- Iterative Keccak
- ECDSA verification
- Fibonacci numbers

Note, that we don't aim to support all possible Rust programs at this point and instead focus on benchmarks only.

**Time-estimate:** 2 months of work by Andrew with basic support from Zisk team

### 2. Feature-complete compiler from Rust to Zisk

**Goals:**
- Extend the PoC Zisk backend to support all relevant Rust programs (e.g. feature parity with SP1)
- Onboard engineer(s) from Polygon to co-own of the project

**Time-estimate:** 2 months of work by Andrew + 1 collaborator from Polygon

### 3. Production-ready compiler from Rust to Zisk

**Goals:**
- Compiler easily available to the end-users
- High test coverage
- Performance regression tests

**Time-estimate:** 2 months of work by Andrew + 2 collaborators from Polygon

## :timer_clock: Timeline

| **Item** | **Date** | **Note** |
|:----------------------:|:--------:|:--------:|
| Submit Proposal |13 August | |
| Review Proposal | | |
| Start implementation | | |
| Present results | | |

## :book:Reference

1. [ZKWASM project index](http://go/zkwasm-index)
2. [Risc0 benchmarks](https://benchmarks.risczero.com/main/applications-benchmarks)
3. [Risc0 examples](https://github.com/risc0/risc0/tree/main/examples)