Question about the roadmap of parallel trie

[yutianwu]

Hi, we researched trie and parallel-trie of reth recently, and wanna help with it.

It seems parallel trie has been used in some places in the codebase like:

reth/crates/blockchain-tree/src/chain.rs

Line 221 in 490fbb0

ParallelStateRoot::new(consistent_view, hashed_state)

But it has not been applied to the Merkle Stage so far.

Can you share the concerns about the parallel trie and the road map about it?

[yutianwu]

@rkrasiuk hi, can you share some insights, thanks

[rkrasiuk]

hey @yutianwu, thanks for expressing the interest! we are not planning to integrate it into the merkle stage for now, but if we were, we would need to find a way to infer how many storage roots we should pre-calculate. merkle stage computes state root in iterations with incremental commits of intermediate nodes to the database. terminating condition for single iteration depends on how many entries we already walked during current iteration

reth/crates/trie/src/trie.rs

Lines 292 to 314 in a590ed7

	// Decide if we need to return intermediate progress.
	let total_updates_len = trie_updates.len() +
	account_node_iter.walker.updates_len() +
	hash_builder.updates_len();
	if retain_updates && total_updates_len as u64 >= self.threshold {
	let (walker_stack, walker_updates) = account_node_iter.walker.split();
	let (hash_builder, hash_builder_updates) = hash_builder.split();
	let state = IntermediateStateRootState {
	hash_builder,
	walker_stack,
	last_account_key: hashed_address,
	};
	trie_updates.extend(walker_updates);
	trie_updates.extend_with_account_updates(hash_builder_updates);
	return Ok(StateRootProgress::Progress(
	Box::new(state),
	hashed_entries_walked,
	trie_updates,
	))
	}