Closed dantaik closed 16 hours ago
Thank you for the changes! I have raised an alternative PR (#171) with two additional commits on top of yours:
IPreconfTaskManager
LOOKAHEAD_BUFFER_SIZE
: I decided to bump it up to 128
from 64
. If the dispute period (for a lookahead slot) itself is two epochs, having just 2 epochs worth of lookahead in the buffer is not enough. Except the first slot in the first lookahead, no other slot gets to go through the full dispute period.BLOCK_ID_TO_PROPOSER_BUFFER_SIZE
: Our system has a dispute period of 2 epochs, so ideally we would not need to store more than 2 epochs worth of blocks in the buffer. Our nodes have a cap of 4 L2 blocks per slot, making it 256
blocks in 2 epochs.LOOKAHEAD_POSTER_BUFFER_SIZE
: I did not completely understand the reason behind using PreconfConstants.SECONDS_IN_SLOT * BLOCK_ID_TO_PROPOSER_BUFFER_SIZE
as the value in your commit. The rationale I used here is that since we have 4 epochs worth of lookahead at any moment, we may safely have this buffer contain the posters of just those 4 epochs.Closed and will continue on https://github.com/NethermindEth/Taiko-Preconf-AVS/pull/171
LOOKAHEAD_POSTER_BUFFER_SIZE
: I did not completely understand the reason behind using PreconfConstants.SECONDS_IN_SLOT
*BLOCK_ID_TO_PROPOSER_BUFFER_SIZE
as the value in your commit.
This is incorrect in my previous PR. No need to have SECONDS_IN_SLOT
in the calculation at all. Good catch.
Lets switch to discuss in your new PR.
See https://github.com/taikoxyz/Taiko-Preconf-AVS-copy/discussions/11
Reduce SSTORE from 20K to 5K for
lookaheadPosters
andblockIdToProposer