Chainfeeds Guide:
ETH parallel execution is expected to go online in 2026, potentially increasing TPS by 3-5 times.
Article Source:
https://x.com/NPC_Leo/status/1930474563152081251
Article Author:
Leo
Perspective:
Leo: ETH parallel execution is expected to go online in 2026, potentially increasing TPS by 3-5 times. Vitalik and the entire foundation have refocused their short to medium-term (2-3 years) efforts on L1 upgrades, bringing parallel execution back to the agenda. Vitalik has already included the related EIP proposal in the 2026 plan during his speech in HK. Currently, there is an EIP draft, and the implementation is expected to be relatively straightforward, with minimal modifications to ETH itself and low technical complexity. It is currently in the draft stage, undergoing discussion and early testing. This EIP proposal adds a block-level access list (BAL). By including a complete list of all addresses and storage keys accessed during block execution, as well as the values written after execution, parallel disk reading and parallel transaction verification can be achieved. This will improve execution efficiency, accelerate block verification, and potentially support future gas limit increases. Previously, EIP-2930's optional access list entered the mainnet but was not mandatory, so it was not widely supported by clients, nodes, RPC, and wallets. If it becomes a mandatory ETH standard, improvements can be quickly realized. In simple terms: Before execution, the client already knows which states the entire block will access, prepares in advance, and prepares the required materials in the block header. This allows parallel execution during the process, as reading and calculating states is the most time-consuming step. This can leverage parallel capabilities and significantly reduce execution and verification time. This is one of the two main parallel technologies. Solana requires developers to specify the required states when building Dapps, enabling parallel processing. The other parallel technology is optimistic parallelism, which assumes all transactions do not conflict, processes them in parallel, and switches to sequential processing if conflicts are detected. Mainstream examples include Sui and Monad. The first approach allows precise parallelism but requires someone to calculate the required data and state locations/changes in advance, relying on developers or entities with strong computational capabilities. For optimistic execution, the advantage is that developers do not need to know the accessed states in advance, but performance can slow down during severe state conflicts (such as everyone competing for the same Non-Fungible Token). Currently, transactions without clear access lists cannot be efficiently parallelized because the execution engine cannot determine which addresses and storage slots will be accessed in advance. Although EIP-2930 provided transaction-level access lists, they were not mandatory, making pipeline optimization difficult. This EIP further upgrades the technology and makes it mandatory, enabling parallel capabilities and TPS improvements. It also considers having large Builders calculate the entire list before giving the block to block producers, forcing this standard into the block header, allowing nodes to execute in parallel when receiving blocks. Reth is developing an ETH parallel client led by Paragdim, with a parallel technology route similar to Sui/Monad's optimistic execution, first analyzing state conflicts and executing non-conflicting transactions in parallel. Rise chain L2 is already using this client. Arbitrum has two virtual machines, EVM and Stylus, which can collaborate and process some transactions in parallel. However, this requires developers to use this mechanism, moving slow EVM tasks to another virtual machine to achieve parallel effects.
Content Source
