Date: 5 August 2021
Author: Ying Tong ([email protected])
An EVM word (256 bits) is represented as a linear combination of 32 bytes:
encode(word) = b_0 + r * b_1 + ... + r^{31} * b_{31},
where word is a 256-bit word, b_i's are bytes, and r is a random factor.
This helper returns an expression of encode(word).
In the zkevm circuit, this encode(word) expression will not be directly
looked up. Instead, it will be folded into the bus mapping lookup.
Merged PR: privacy-scaling-explorations/zkevm-circuits#21
Previously, we had planned to use lookup tables to implement byte-wise addition of two EVM words. However, we realised that this can be handled more directly by a custom gate.
WIP PR: privacy-scaling-explorations/zkevm-circuits#15
The "bus mapping" data structure must be consistent across the EVM circuit and
the state circuit. It is derived from the execution trace and consists of the
actual operations executed in the block. Further, it is consumed by subcircuits
such as ADD, and thus is a blocker.
The contents of the bus mapping will possibly change as we build out the circuits; I think one way forward is to use an incomplete bus mapping for now to unblock other circuits, and adapt the struct as needed later on.