Skip to content

Simulates a call to Convex Finance's system shutdown method, which uses about 16M gas

Notifications You must be signed in to change notification settings

mds1/convex-shutdown-simulation

Repository files navigation

Convex System Shutdown Benchmarks

This repository benchmarks performance of various Ethereum development frameworks by simulating a call to Convex's systemShutdown method. This method uses about 16M gas and performs a number of token transfers

Benchmarks

Benchmarks were run at mainnet block 14,445,961. Blocknative only simulates against the latest block (support for simulation at historical blocks is planned), so the Blocknative simulation was run first and other benchmarks were run against the same block.

Notes on benchmarks:

  • "Remote RPC" benchmarks used Alchemy as the RPC provider, connected over HTTP.
  • "Local RPC" benchmarks used a local Erigon node.
  • "Cached" benchmarks use RPC responses that the framework caches locally instead of making HTTP requests. We assume that since Blocknative and Tenderly are infrastructure providers, they aggressively cache and load hot data in-memory to improve performance, which is why their results are also in the "Cached" column.
  • Benchmarks were performed on macOS 11.6.2 with a 2.3 GHz 8-Core Intel Core i9 and 32 GB 2667 MHz DDR4.
Framework Version
Blocknative HTTP request on 2021-03-23
Dapptools dapp 0.35.0, hevm 0.49.0
Ganache 7.5.0
Hardhat 2.12.3
Foundry forge 0.2.0 (250cc85 2022-12-02T00:18:26.200251Z)
Tenderly HTTP request on 2021-03-23
Framework Remote RPC Local RPC Cached
Blocknative N/A N/A 0m3.529s
Dapptools 52m17.447s 17m34.869s 3m25.896s
Ganache 7m24.070s 0m29.631s 0m15.665s
Hardhat 6m17.896s 0m25.642s 0m9.110s
Foundry 5m55.141s 0m11.606s 0m0.999s
Tenderly N/A N/A 0m1.9315s

Notes on gas usage:

  • Ganache, Hardhat, and Tenderly all agree on gas usage after refunds and are therefore likely the truth value.
  • Foundry usage matches those values: The debugger shows the actual call costs 22,558,945 gas. Adding 21,000 intrinsic gas and 64 gas for calldata (both are excluded from the reported gas number by default) gives the same amount reported by Hardhat and others. The remaining 2995 gas is overhead from the contract based setup and Solidity-generated call checks.
  • Blocknative's gas usage does not account for refunds as this is not yet supported by their platform.
  • Dapptools' gas usage also does not account for refunds.
Framework Gas Used
Blocknative 26,668,845
Dapptools 24,066,128
Ganache 22,580,009
Hardhat 22,580,009
Foundry 22,561,940
Tenderly 22,580,009

Usage

  1. Run cp .env.example .env, and in the resulting .env file enter a URL to an Ethereum archive node in the ETH_RPC_URL environment variable. (Alchemy provides free archive node data). Also fill out the TENDERLY_* variables to benchmark the Tenderly API

  2. Run yarn to install dependencies for Ganache and Hardhat

  3. Install Foundry's forge and Dapptools using the installation instructionshere and here respectively

  4. Run dapp update to install dependencies for Dapptools and Foundry

  5. Run any command in the Makefile to benchmark that tool. For example, use make benchmark-hardhat to run the simulation against Hardhat. Alternatively, run make benchmark-all to run all tools

Tips

  • Set export CLEAR_CACHE=1 in your .env file to clear the Ganache and Hardhat caches
  • Consider running the benchmarks via Docker. See the comment header in the Dockerfile for details.

About

Simulates a call to Convex Finance's system shutdown method, which uses about 16M gas

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published