[Engine] Batched L1 consolidation

[frisitano]

Overview

Instead of consolidating a single L1 derived block at a time, we can sync them to the EN in batches.

rollup-node/crates/engine/src/future/mod.rs

Lines 211 to 242 in 6e18a23

	async fn handle_payload_attributes<EC, P>(
	client: Arc<EC>,
	execution_payload_provider: P,
	safe_block_info: BlockInfo,
	mut fcs: AlloyForkchoiceState,
	payload_attributes: ScrollPayloadAttributesWithBatchInfo,
	) -> Result<(L2BlockInfoWithL1Messages, IsReorg, BatchInfo), EngineDriverError>
	where
	EC: ScrollEngineApi + Unpin + Send + Sync + 'static,
	P: ExecutionPayloadProvider + Unpin + Send + Sync + 'static,
	{
	tracing::trace!(target: "scroll::engine::future", ?fcs, ?payload_attributes, "handling payload attributes");
	let ScrollPayloadAttributesWithBatchInfo { mut payload_attributes, batch_info } =
	payload_attributes;
	let maybe_execution_payload = execution_payload_provider
	.execution_payload_for_block((safe_block_info.number + 1).into())
	.await
	.map_err(|_| EngineDriverError::ExecutionPayloadProviderUnavailable)?
	.filter(|ep| matching_payloads(&payload_attributes, ep, safe_block_info.hash));
	if let Some(execution_payload) = maybe_execution_payload {
	// if the payload attributes match the execution payload at block safe + 1,
	// this payload has already been passed to the EN in the form of a P2P gossiped
	// execution payload. We can advance the safe head by one by issuing a
	// forkchoiceUpdated.
	let safe_block_info: L2BlockInfoWithL1Messages = (&execution_payload).into();
	fcs.safe_block_hash = safe_block_info.block_info.hash;
	forkchoice_updated(client, fcs, None).await?;
	Ok((safe_block_info, false, batch_info))
	} else {

In the code above we would change the function input argument to payload_attributes: Vec<ScrollPayloadAttributesWithBatchInfo>. We would issue a single fcu following the successful consolidation of the batch. This would result in a reduction in fcu invocations by a factor of 1/(average blocks per batch), hopefully leading to higher consolidation speed. We need to think through the case of failing mid-batch.