/pevm

Blazingly fast Parallel EVM in Rust

Primary LanguageRustMIT LicenseMIT

RISE Parallel EVM

CI

Blazingly fast Parallel EVM in Rust.

⚠️ This repository is a work in progress and is not production ready 🚧

Banner

RISE PEVM is a parallel execution engine for EVM transactions heavily inspired by Block-STM. Since Blockchain transactions are inherently sequential, a parallel execution engine must detect dependencies and conflicts to guarantee the same deterministic outcome with sequential execution. Block-STM optimistically executes transactions and re-executes when conflicts arise using a collaborative scheduler and a shared multi-version data structure. Since it does not require prior knowledge or constraints on the input transactions, replacing an existing sequential executor with Block-STM is easy for substantial performance boosts.

Block-STM was initially designed for the Aptos blockchain that runs MoveVM. We must consider several modifications to make it work well with EVM. For instance, all EVM transactions in the same block read and write to the beneficiary account for gas payment, making all transactions interdependent by default. We must carefully monitor reads to this beneficiary account to lazily evaluate it at the end of the block or when an explicit read arises.

While Polygon has adapted a version of Block-STM for EVM in Go, it is still slower than sequential execution in Rust & C++. On the other hand, our redesign & Rust implementation has achieved the fastest execution speed of any public EVM executor.

Finally, while Aptos and Polygon embed their PEVM implementation directly into their nodes, this dedicated repository provides robust versions and a playground for further advancements. For instance, we can introduce static-analysed metadata from an optimised mempool, support multiple underlying executors, track read checkpoints to re-execute from upon conflicts and hyper-optimise the implementation at low system levels.

Goals

  • Become the fastest EVM (block) execution engine for rapid block building and syncing — 10 Gigagas/s and beyond.
  • Provide deep tests and audits to guarantee safety and support new developments.
  • Provide deep benchmarks to showcase improvements and support new developments.
  • Complete a robust version for syncing and building blocks for Ethereum, RISE, Optimism, and more EVM chains.
  • Get integrated into Ethereum clients and ZK provers like Reth, Helios, and Zeth to help make the Ethereum ecosystem blazingly fast.

Development

  • Install cmake for building snmalloc (highly performant memory allocator).
$ cargo build

Alpha Done

  • Build a Block-STM foundation to improve on.
  • Preprocess dependencies among transactions with the same sender or explicitly interacting with the beneficiary account.
  • Atomically update gas payments to the beneficiary account as implicit reads & writes.
  • Atomically update raw transfer recipients' balance as implicit reads & writes.
  • Improve scheduler design & aggressively find tasks to save scheduling cycles.
  • Complete the first test & benchmark suites.

Alpha TODO

  • Complete a vital test suite.
  • More granular memory locations (like breaking AccountInfo down into balance, nonce, and code) to avoid false positive dependencies.
  • More low-hanging fruit optimizations.
  • Robust error handling.
  • Better types and API for integration.
  • Benchmark a Reth integration for syncing and building Ethereum and RISE blocks.

Beta TODO

  • Optimize concurrent data structures to maximize CPU cache and stack memory.
  • Optimize the scheduler, worker threads, and synchronization based on common block scenarios.
  • Add pre-provided metadata from a statically analysed mempool or upstream nodes.
  • Better memory management:
    • Custom memory allocators for the whole execution phase and the multi-version data structure. Early experiments with jemalloc, mimalloc, and snmalloc show potential up to 50% improvements.
    • Dedicated thread (pool) for cleaning up memory between blocks' execution.
  • Track read checkpoints to re-execute from instead of re-executing the whole transaction upon conflicts.
  • Support multiple EVM executors (REVM, JIT & AOT compilers, etc.).
  • Hyper-optimise at low system levels.
  • Propose an EIP to “tax” late dependencies in blocks for validators to put them up front to maximize parallelism.

Testing

We have three test groups:

$ git submodule update --init
# Running our heavy tests simultaneously would congest resources.
# Each parallel test still executes parallelly anyway.
$ cargo test --release -- --test-threads=1

Benchmarks

See the dedicated doc here.