# EPF Cohort 6 Updates – Week 9

This week I focused on completing the remaining work for benchmarking Ream's Beacon State, with major developments taking place in this area.

## Benchmarking Progress

The existing [consenzero-bench](https://github.com/ReamLabs/consenzero-bench) repository only benchmarked `process_operations` and was using an outdated RISC-Zero version. I successfully added functionality to benchmark `process_epoch` and, after discussing with Unnawut, submitted a pull request:

**PR Link:** https://github.com/ReamLabs/consenzero-bench/pull/25

## Pico zkVM Integration

Pico released their new version, eliminating the need to use a previous release of the `ream-consensus-beacon` crate. I took this opportunity to implement end-to-end benchmarks for both `process_operations` and `process_epoch` for Pico zkVM.

**Repository:** https://github.com/x-senpai-x/consenpico-bench

This implementation successfully tracks the guest's RISC-V execution cycles!

## Contributing to Pico

I discovered that Pico's new version still generates starter kits using an outdated repository. After discussing with Brevis support, I submitted a PR to fix this issue:

**PR Link:** https://github.com/brevis-network/pico-zkapp-template/pull/2

## ethProofs Call #4 Notes

I attended the fourth ethProofs call and learned about the current proposal to replace EVM with RISC-V and the crucial role ZKVMs will play.

### Why RISC-V for ZKVMs?

**Advantages:**
- Better performance with custom ISA
- Broader compiler support
- Simple instruction set (just 40 instructions)
- Majority of ZKVMs already use RISC-V
- Easy to formally verify

**Comparison with other architectures:**
- **x86/ARM:** Thousands of instructions, making formal verification difficult
- **MIPS:** Lesser support, not actively maintained, MIPS itself is moving to RISC-V


**Adoption challenges:**
- Compilers are hard to develop
- Getting adoption of new systems is difficult
- Not everything uses LLVM

### Vitalik's Perspective

The current bottleneck is VM execution and hashing:
- Hashing issues will hopefully be resolved with ongoing Poseidon work
- For VM execution: Instead of RISC-V serving as a wrapper over EVM, it needs to directly replace (enshrine) EVM to avoid overhead

## 3SF Updates

Since the goal is to build a prototype to identify key bottlenecks rather than a full-fledged consensus protocol using 3-slot finality, I'll start directly with Ream's 3SF-mini and figure out how to generate mock consensus spec test data.

## Next Week's Goals

- [ ] Figure out how to generate mock consensus spec test data for 3SF mini and set up RISC-Zero codebase
- [ ] Add timing, flamegraphs, and other benchmarking data for `consenPico-bench` and `consenZero-bench`

## Long Term Goal
I'll be working on building a detailed writeup on my experience across all the zkVms out there, something like this
https://x.com/Kautukkundan/status/1867589273446232368