Optimizing ScrollChain._popL1Messages

[barakman]

On the one hand, this function looks as if it was written with gas-optimization in mind.
On the other hand, it looks like the optimization was rather poorly applied.

First off all, instead of checking the value of isCalldata on every iteration, it makes since to split this function into two functions, one for the case of isCalldata = 1 and another for all other cases.
Alternatively, you could run that check outside the loop, and then implement two separate loops (one for each case).

Second, the logic around the calculation of _count seems like something which can be done once instead of on every iteration.
You can do this by splitting the loop into two parts:

1. A loop over all 256-length chunks
2. A final iteration for the last chunk (if larger than 0)

For example:

unchecked {
    uint256 numOfChunks = totalL1MessagesPoppedInBatch / 256;
    for (uint256 i = 0; i < numOfChunks; ++i) {
        IL1MessageQueue(messageQueue).popCrossDomainMessage(startIndex, 256, bitmap);
        startIndex += 256;
    }
    uint256 _lastChunkSize = totalL1MessagesPoppedInBatch % 256;
    if (_lastChunkSize > 0) {
        IL1MessageQueue(messageQueue).popCrossDomainMessage(startIndex, _lastChunkSize, bitmap);
    }
}