AVM PIL: tracking ticket

[Maddiaa0]

Overview

As part of the full vision for Aztec, we are building a Public VM that will go inside a Snark. This will take a lot of engineering time to conceive.
In our current state of dev tooling, the VM will need to be written as a series of relations within barretenberg.

The scope of this tracking issue is to explore tooling that will make this work simpler to complete, the main idea is to create a compilation pipeline around existing tooling called PIL ( Polynomial Identity Language ) that can work with barretenberg.

My hypothesis is that writing the relations in such a way will greatly decrease the amount of work that the engineering team will need to complete. Decreasing the amount of time required to iterate.

Critical Tasks

Beta Give feedback

1. Add barretenberg as a backend for PIL. #2500
   C-avm +1
   
2. feat(avm-pil-backend): public input columns #3783
   C-avm crypto +2
   
3. public-vm: define a "public input"-like mechanism in backend #3790
   C-avm +1
   
4. AVM PIL: constant columns (precompute commitments) #7115
   C-avm +1
   
5. AVM PIL: sync to upstream powdr #7114
   +0
   
6. AVM PIL: pull Powdr into monorepo #7113
   +0
   
7. AVM PIL: spike implemenation of boundary conditions #4307
   C-avm crypto +2
   

Optimisation Tasks

Beta Give feedback

1. AVM PIL: add way to combine lookup inverse columns in pil (optimization) #3879
   C-avm T-optimisation +2
2. AVM PIL: kernel output selector counter can be infered (optimization) #6484
   C-avm +1
3. AVM PIL: enable writing of constant columns in PIL