feat: Make public inputs the start of the UH and MH proof

[lucasxia01]

Closes AztecProtocol/barretenberg#1089.

Removes the circuitSize, numPublicInputs, and pubInputsOffset metadata from the proof. The result is that the public inputs are at the beginning of the proof.

Followup work includes:

• Medium: making these metadata actual witnesses in the recursive VK (Should circuit_size, etc.. be witnesses in Recursive VerificationKey? barretenberg#983).
• Small: fixing verify_honk_proof acir test inputs (Fix verify_honk_proof input generation barretenberg#1279).
• Small: cleaning up or removing the proof surgeon.
• Medium: Further structuring the proof/public inputs output as Cody suggests in Move public inputs to the front of proof barretenberg#1089 to split out the accumulators in the pub inputs.
• Small Adding pubInputsOffset to the Honk Solidity VK (Add pubInputsOffset to Solidity Honk VK barretenberg#1281).

[Maddiaa0]

Question, if there is no offset anymore can it just be removed? What was it for ?

[lucasxia01]

Question, if there is no offset anymore can it just be removed? What was it for ?

No its still there in the vk, just not redundantly in the proof. The pub_inputs_offset basically represents the number of ecc_ops. I'm not sure if it needs to exist anymore if we don't use MegaHonk with an unstructured trace.

[ledwards2225]

Question, if there is no offset anymore can it just be removed? What was it for ?

TBC there are two "public input offset" variables floating around. One describes the offset of the pub inputs stored in the proof (previously 3, now 0). That one can probably go away. The other one describes the index at which the public inputs are stored in the wires. For UH it is always 1 (since we have a "zero-row" to accommodate shifts). For MH its something else. Its possible this one can also be removed but that's not effected by this PR

[ledwards2225]

Delightful, thanks for doing this. I'm assuming this will effect some noir devs who are manually splitting pub inputs from the proof using this secret magic 3? @saleel Is there any reason to hold off on this until docs are updated or something?

[ledwards2225]

thank god for GPT for things like this..

[saleel]

Delightful, thanks for doing this. I'm assuming this will effect some noir devs who are manually splitting pub inputs from the proof using this secret magic 3? @saleel Is there any reason to hold off on this until docs are updated or something?

🙏
Looks like this will this also change the output from the exposed wasm methods? If so, I think we would need to update bb.js classes

[lucasxia01]

extra call is not ideal since its more work just to fetch the number of public inputs. The vk or maybe just the number of public inputs should be returned by proof generation perhaps.

[lucasxia01]

the bb.js interface just needs to be reworked similarly to the bb interface so that prove can output both a proof and a vk...

[codygunton]

Will #11024 eventually be handled by your series of works?

[lucasxia01]

Will #11024 eventually be handled by your series of works?

@codygunton yep, Saleel brought that up as well and that's on my radar as a followup in the immediate future.