Small gas optimization for burn/transfers/mint

[kadenzipfel]

first commit:

• saves 9 gas per burn
• saves 3 gas per transfer
• increases deployment cost but imo worth it

second commit:

• saves 6*(quantity-1)-8 gas per mint
• imo super worth the additional upfront cost

https://www.diffchecker.com/rYLkEpYd

[Vectorized]

I think the burn and transfer parts shouldn't be in assembly. Gas savings not really worth the extra function.

Plus, it may make it harder to extend (e.g. for the storage hitchhiking PR #323 ).

Nice catch on the gas savings in the _mint loop.
I changed it to make it cheaper for single mints, but overlooked the increase for multiple mints.

Maybe we can consider using assembly for emitting the Transfer event in _mint.

uint256 end = startTokenId + quantity;

// Use assembly to save around 30+ gas per iteration.
assembly {
    // Emit the `Transfer` event.
    // The event signature is given by:
    // `keccak256(bytes("Transfer(address,address,uint256)"))`.
    log4(
        0, // Start of data. Zero, as no data.
        0, // End of data. Zero, as no data.
        0xddf252ad1be2c89b69c2b068fc378daa952ba7f163c4a11628f55a4df523b3ef, // Signature.
        0, // `address(0)`.
        to, // `to`.
        startTokenId // `tokenId`.
    )

    for {
        let tokenId := add(startTokenId, 1)
    } iszero(eq(tokenId, end)) {
        tokenId := add(tokenId, 1)
    } {
        log4(0, 0, 0xddf252ad1be2c89b69c2b068fc378daa952ba7f163c4a11628f55a4df523b3ef, 0, to, tokenId)
    }
}

_currentIndex = end;

[kadenzipfel]

I think the burn and transfer parts shouldn't be in assembly. Gas savings not really worth the extra function.

if we're thinking in terms of overall costs, offsetting the deployment cost would only require ~500 transfers, ~166 burns or a combination of the two. for most collections, 500 transfers is trivial and thus imo super worth it

if we're thinking in terms of devex, imo it's much more readable to have the extra function describing the process

Plus, it may make it harder to extend (e.g. for the storage hitchhiking PR #323 ).

it's quite easy to extend actually and in the case of the storage hitchhiking PR the gas savings are even more significant (41 gas for burn, 6 gas for transfer)
see: https://github.com/kadenzipfel/ERC721A/blob/d69b41f83614db42ec9dec97ce3a8d68cebfb843/contracts/ERC721A.sol#L569
https://www.diffchecker.com/xUhWcd6r

Maybe we can consider using assembly for emitting the Transfer event in _mint.

+1

[cygaar]

Renamed the PR to be more descriptive

[kadenzipfel]

updated gas savings: https://www.diffchecker.com/BS9LslEP

[Vectorized]

LGTM