zkEVM Benchmarking Workload

This repository benchmarks Ethereum stateless-validator guests across multiple zkVMs. The normal workflow has two phases:

1. Obtain canonical EEST blockchain_tests fixtures containing statelessInputBytes and statelessOutputBytes.
2. Pass a fixture file, fixture directory, or EEST fixture checkout to ere-hosts and write execution metrics, cost estimates, proofs, or verification results.

Workspace At a Glance

• crates/ere-hosts: benchmark CLI for execution, estimation, proving, and verification jobs.
• crates/benchmark-runner: shared orchestration for canonical fixture loading, guest resolution, execution, proof flow, and verification.
• crates/metrics: serializable result types such as BenchmarkRun.
• crates/witness-generator-spec-cli: separate CLI and library for producing and publishing benchmark-ready EEST stateless fixtures from CL/EL RPC endpoints.

Reth v0.1.0-rc.3 and Ethrex v26.0.0 support OpenVM, SP1, and ZisK. Zesu tests-glamsterdam-devnet@v8.1.4 supports ZisK only. The workspace pins ere-guests v0.17.0 and Ere v0.17.0. Default guest downloads use release assets. GitHub authentication is optional. Compatible custom artifacts can use --bin-path or --guest-artifact-base-url.

Prerequisites

• Rust via rustup
• Docker
• Canonical EEST blockchain_tests fixtures
• Python 3.10 or later for the comparison reports

Quickstart

Inspect both maintained CLIs:

cargo run -p ere-hosts -- --help
cargo run -p witness-generator-spec-cli -- --help

The witness generator produces benchmark-ready EEST fixtures from live CL/EL networks. Use generate for one block or collect for continuous per-block collection. Exported live batches contain a blockchain_tests/ tree and can be passed to ere-hosts immediately after extraction.

Obtain the tests-zkevm@v0.8.4 fixtures_zkevm.tar.gz bundle, whose blockchain_tests cases contain canonical stateless bytes. Then benchmark either the extracted fixture root, a directory of EEST JSON files, or one EEST JSON file:

cargo run -p ere-hosts --release -- --zkvms sp1 \
    stateless-validator --execution-client reth \
    --input-folder /path/to/execution-specs/fixtures

Execution, cost estimation, and proving require --input-folder. Verification reads saved proofs and does not require fixtures.

Estimate proving costs without generating a proof:

cargo run -p ere-hosts --release -- --zkvms sp1 --action estimate-cost \
    stateless-validator --execution-client reth \
    --input-folder /path/to/execution-specs/fixtures

Actions merge their results into each fixture JSON. Execution records duration. Estimation records component costs and optional heap usage. Cost units differ by zkVM. Compare compatible baseline and candidate runs with python3 scripts/compare_costs.py BASELINE CANDIDATE.

Guides

• Documentation map
• Benchmark execution, proofs, and verification guide
• Benchmark input reference
• Benchmark output reference
• Stateless input publication guide

The root README is intentionally short. Detailed workflow documentation lives under docs/.

License

Licensed under either of

• MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)
• Apache License, Version 2.0 (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)

at your option.