add rlp decode circuit

[smtmfft]

RLP decoding circuit for txlist. It supports both correct rlp bytes and incorrect rlp bytes.
Design doc: https://www.notion.so/taikoxyz/RLP-circuit-design-doc-e59ddfda4b5a40e6ac0f170f23154ed8

So far the majority decoding logic seems ok, rest jobs are:
TODOs:

• lookup to keccak table to insure txlist input, which may also involve hi-lo change.
• design output params (maybe a lookup table??) for other circuits.
• logic completeness check, for valid bytes at least.
• refactory using circuit-tool.
• more test coverage, especially for invalid cases.
• 1559 tx type support, with or without legacy depends on future design.

[ggkitsas]

first non-extensive and shallow pass - minor comments

[ggkitsas]

Question for my understanding :)

I see these 3 tables belong to different regions. Is it necessary to distinguish them with tags?

I genuinely don't know!

[smtmfft]

I think lookup uses columns which contains all regions it covers. like this:

So, based on my understanding, I can setup region A/B/C separately, and lookup will take them all into account.
correct me if I am wrong @Brechtpd .

[ggkitsas]

Suggested change

	// output txlist hash check
	// meta.lookup_any("comsumed all bytes correctly", |meta| {
	// let is_enabled = meta.query_fixed(q_first, Rotation::next());
	// let input_rlc = meta.query_advice(acc_rlc_value, Rotation::cur());
	// let input_len = meta.query_advice(rlp_remain_length, Rotation::cur());
	// let hash_rlc = meta.query_advice(value, Rotation::cur());
	// let table = &aux_tables.keccak_table;
	// let table_is_enabled = meta.query_advice(table.is_enabled, Rotation::cur());
	// let table_input_rlc = meta.query_advice(table.input_rlc, Rotation::cur());
	// let table_input_len = meta.query_advice(table.input_len, Rotation::cur());
	// let table_hash_rlc = meta.query_advice(table.output_rlc, Rotation::cur());
	// vec![
	// (is_enabled.expr(), table_is_enabled.expr()),
	// (is_enabled.expr() * input_rlc.expr(), table_input_rlc.expr()),
	// (is_enabled.expr() * input_len.expr(), table_input_len.expr()),
	// (is_enabled.expr() * hash_rlc.expr(), table_hash_rlc.expr()),
	// ]
	// });

[ggkitsas]

What is the intention behind this value? I see it being used as a default value but there are no constraints using it.

Is it mean to be seen in the witness at all? If not, would it make sense to prohibit this value from appearing using constraints?

[smtmfft]

Fixed, changed to a unified way of telling length/size of a enum.

[smtmfft]

Created a new PR for 1559 tx support, all fixes are already there.
smtmfft#15
This one is closed!