mimblewimble/grin

block.hash() unnecessarily expensive to call

Open

#3372 opened on Jun 29, 2020

View on GitHub
 (0 comments) (0 reactions) (0 assignees)Rust (4,876 stars) (991 forks)batch import
enhancementhelp wanted

Description

We call write() on Writeable when serializing things to bytes. Not just for serializing full data but also when we serialize during a call to hash().

Block hash -> BlockHeader hash -> ProofOfWork hash -> Proof hash

Every call to block.hash() or header.hash() delegates to the following code to produce the data to be hashed -

https://github.com/mimblewimble/grin/blob/098d25e5696d1d97e4fbfabb80f27063caccd0f3/core/src/pow/types.rs#L455-L472

We build the bitvec "on the fly" every time we call hash(). Ideally hash() on any type is as cheap as possible.

I suspect, but have no benchmarks or data to confirm, that this is relatively expensive over time. We make a lot of calls to block.hash() and header.hash().

Possible improvement -

  1. Consider initializing the bitvec when we initially read the Proof instance.
  2. Use this "cached" value when we write it in Hash serialization mode.
  3. Ignore the "cached" value when writing full data.
  4. Make sure nonces or edge_bits cannot be written to or modified in unexpected ways, leaving the cache stale.

Related - https://github.com/mimblewimble/grin/issues/2584

Maybe we don't need to cache "hashes" in a general way (yet), maybe simply making block.hash() cheap will give us most of the benefit for the least effort.

Contributor guide

Tech stack
rust
Domain
performance
Issue type
performance
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
3-5 days
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
stale
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
mostly clear
Prerequisites
Rustbitvecserialization
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
50
Research direction
The issue is in the Grin repository. Focus on `core/src/pow/types.rs` lines 455-472 where the bitvec is built on the fly during `hash()`. The proposed solution is to cache the bitvec when the Proof instance is read and reuse it during hash serialization, ensuring the cache is invalidated if `nonces` or `edge bits` change. Study the existing `read` and `write` implementations to understand serialization modes and implement the cache carefully to avoid stale data.