Gas Cost Estimator

The project aims to provide a tool for measuring, analysing and comparing gas costs of EVM operations across different implementations. The reproducibility is the key to the research. We have provided a complete setup guide and tooling to make the execution of the benchmarks as easy as possible.

The result of the analysis is a new gas cost schedule.

Research progress

The project is divided into stages. The progress of each stage is described in the corresponding document:

• Stage I - Initial research and methodology proposal
• Stage II - Benchmarking and data collection
• Stage III - Data analysis and report generation
• Stage IV - Comprehensive analysis of the gas cost and reproducibility (work in progress)
• Stage V - Improved tooling and report generation (planned)

Introduction and project scope

EVM Implementations

We have included the following EVM implementations in the research:

• EvmOne
• Go Ethereum
• Erigon
• EthereumJS
• Nethermind
• Revm
• Besu

More implementations can be added in the future, depending on the community feedback, implementation maturity and availability of the resources.

Measured OPCODEs and precompiles

We measure all OPCODEs together with the precompiles as of hard fork Cancun.

Tooling and automation

The release contains precompiled binaries for easy execution. The binaries are available for Linux x64, MacOS x64 and Windows.

Additionally, we provide a complete setup guide to compile the EVM implementations and run the benchmarks.

Quick start

If you plan to compile and run benchmarks on your own, you need the following tools installed:

• Python 3.8+
• Go 1.22.8+
• Rust 1.80.0+
• Node.js 18.0.0+
• .NET 7.0+
• Java 21.0.0+

You can use setup_tools.sh script to install the required tools on Linux.

To download and compile the EVM implementations, run the following commands:

./src/instrumentation_measurement/setup_clients.sh

If your configuration is different, follow the steps in the script. The end results should be the same - you should have all the EVM implementations compiled and copied to the ../gas-cost-estimator-clients/build directory.

To run the benchmarks, use the provided Python script:

python3 ./src/instrumentation_measurement/measurements.py measure --input_file ./src/stage4/pg_marginal_full5_c50_step5_shuffle.csv --evm evm_name --sample_size 10

Where evm_name is the name of the EVM implementation you want to measure.

Docker - reports

In order to build locally the docker image execute in the repository root

docker build ./src/analysis -f Dockerfile.reports -t imapp-pl/gas-cost-estimator/reports:4.0

Note that the context is ./src/analysis in order to decrease the data size. The image includes the report notebooks -- files. But the bytecode programs and measurement resutls need to be provided. For now, use /data volume to pass input files and retrieve an output report.

To render measure_marginal report provide your params and an output file and execute the command:

docker run -it -v /your/path/to/data:/data --rm imapp-pl/gas-cost-estimator/reports:4.0 Rscript -e "rmarkdown::render('/reports/measure_marginal_single.Rmd', params = list(env = 'erigon', programs='pg_marginal_full5_c50_step1_shuffle.csv', results='erigon_pg_marginal_full5_c50_step1_shuffle_size_10.csv', output_estimated_cost='erigon_marginal_estimated_cost.csv'), output_file = '/data/erigon_measure_marginal_single.html')"

To render final_estimate report provide your params and an output file and execute the command:

docker run -it -v /your/path/to/data:/data --rm imapp-pl/gas-cost-estimator/reports:4.0 Rscript -e "rmarkdown::render('/reports/final_estimation.Rmd', params = list(estimate_files='besu_marginal_estimated_cost.csv, erigon_marginal_estimated_cost.csv, ethereumjs_marginal_estimated_cost.csv, geth_marginal_estimated_cost.csv, nethermind_marginal_estimated_cost.csv, revm_marginal_estimated_cost.csv', current_gas_cost='current_gas_cost.csv'), output_file = '/data/final_estimation.html')"