EIP-5478 CREATE2COPY opcode
No call decision · EVM & execution · EIP status Stagnant · created 2022-08-17
A proposal that adds a new opcode (CREATE2COPY) to greatly reduce gas costs by referencing existing code instead of re-uploading it in full, when creating a contract with code identical to code already stored on chain.
Why it is needed
A large portion of the cost of creating a new smart contract is the data gas for writing the code to the chain byte by byte. In cases like account abstraction (AA) wallets, where each user deploys an individual contract with identical code, gas must be paid again for the full byte count every time, even if another user has already uploaded the same code. In the original example, 10,000 bytes of code costs 2,000,000 data gas, and if the code content is identical, this is an unnecessary duplicate cost. CREATE2COPY aims to lower this duplicate cost to the level of a cold/warm account access cost by allowing proof that it reuses code already stored on chain.
What changes
Today
- Every time you deploy a new contract with identical code, such as an AA wallet, you pay data gas proportional to the full byte size of the code.
- A 10,000-byte code costs 2,000,000 gas based on CREATE_DATA_GAS_PER_BYTE (200 gas per byte).
- With CREATE2 (0xf5) alone, there is no way to save cost even if the same code already exists on chain.
After EIP-5478
- With CREATE2COPY (0xf6), if you specify the address of an existing contract with the same code (existing_contract_address), you pay only an EXTCODEHASH-level access cost instead of data gas.
- If the address is accessed for the first time in the transaction (cold), it costs 2600 gas; if the address is already in the access list (warm), it costs only 100 gas.
- If the generated code differs from the code at existing_contract_address, creation fails and all gas is burned.
Who is affected
- Users: Gas costs for first-time creation of contracts with identical code, such as AA wallets, can be reduced.
- Wallet & app developers: Services that deploy large numbers of smart accounts with identical code can design deployment costs around CREATE2COPY.
- Client teams: Implementing the new opcode (0xf6), code hash comparison logic, and EIP-2929 access list integration must be added to the EVM.
In depth
Stack arguments for the new opcode CREATE2COPY
Starting from the block after FORK_BLKNUM, CREATE2COPY is added at opcode 0xf6. The arguments popped from the stack are endowment, memory_start, memory_length, salt, and existing_contract_address, five in total, with existing_contract_address added to the four arguments of the existing CREATE2 (0xf5, EIP-1014). The address calculation and creation behavior itself are identical to CREATE2.
Code hash match verification and failure handling
For CREATE2COPY to succeed, the code hash of the newly created contract must match the hash of the code stored at existing_contract_address. If the code produced by executing the initcode differs from the code at existing_contract_address, creation fails with a 'mismatched contract creation code with existing code' error, and in this case all gas allocated to the contract creation is burned.
Gas schema: replacing data gas with EXTCODEHASH cost
CREATE2 requires paying data gas equal to CONTRACT_BYTES * CREATE_DATA_GAS_PER_BYTE (200 gas per byte). CREATE2COPY replaces this data gas item with the access cost of the EXTCODEHASH opcode. If existing_contract_address is accessed for the first time in the transaction (cold), only EIP-2929's COLD_ACCOUNT_ACCESS_COST of 2600 gas is charged; if it is already in the access list (warm), only WARM_ACCOUNT_ACCESS_COST of 100 gas is charged. The rest of the gas schema is identical to CREATE2.
Prerequisites: EIP-1014, EIP-2929
CREATE2COPY reuses as is the address calculation and creation logic of CREATE2 defined by EIP-1014, and directly pulls in the cold/warm access list and cost scheme defined by EIP-2929 for its gas calculation. In other words, CREATE2COPY does not create a new gas model itself, but is designed to combine the mechanisms of two existing EIPs to cut only the cost of deploying duplicate code.
Open items and current status
In the original text, the Rationale is only noted as 'TBD' and Security Considerations as 'needs discussion,' so the design rationale and security implications have not yet been documented. The FORK_BLKNUM value is also undetermined. This EIP's status is Stagnant, having stalled since it was written on 2022-08-17 with no further discussion or fork-inclusion decision.