Skip to content

Conversation

@arthurpaulino
Copy link
Member

Implement the whole pipeline for prove and verify in Aiur2, from Lean to Rust through C and back to Lean.

gabriel-barrett and others added 25 commits July 15, 2025 13:38
* Add the output size on the call Op
* Compute the numbers of auxiliars and lookups
* Simplify the inclusion of memory widths with an RBTree
Use a subtype whose inner value is an `UInt64`, constrained to be
smaller than the size of the Goldilocks field.

This definition of `G` entails a much better memory layout for
interfacing with Rust, as it is not conditioned on the size of the
number being wrapped like `Nat` is.
"Claim" is now just a vector/array of Goldilocks values. This
simplifies the FFI glue a bit.

Extra: remove the execution test workaround and define the IO test
by checking the whole claim returned by `AiurSystem.prove`.
@arthurpaulino arthurpaulino enabled auto-merge (squash) July 24, 2025 12:57
@arthurpaulino arthurpaulino merged commit e0efff7 into main Jul 24, 2025
12 checks passed
@arthurpaulino arthurpaulino deleted the aiur2-prover branch July 24, 2025 15:07
johnchandlerburnham pushed a commit that referenced this pull request Aug 11, 2025
* prove and verify for Aiur2
* env var to opt out of parallelism
* bump the Rust toolchain

---------

Co-authored-by: Gabriel Barreto <gabriel.aquino.barreto@gmail.com>
Co-authored-by: Samuel Burnham <45365069+samuelburnham@users.noreply.github.com>
johnchandlerburnham added a commit that referenced this pull request Oct 27, 2025
* WIP: ix hash command

* make Proof serialize as an Ixon Const

* broken: connect canonicalizer to Ix.Meta

* fix: Nix devShell & light refactor (#55)

* ix store cli command

* WIP: store command and new IxM structure

* megacommit implementing Ix commitment API

* fixup merge

* fixup merge 2

* ix prove check command

* implement ix prove CLI and simplify metaprogramming

- CompileM is now persistent with the filemapped store
- constructing evaluation claims is done through MetaM rather than CompileM
- CLI now features the Prove command

* provecmd examples

* cleanup repo

* simplify ZKVoting example app

* remove IO from the compiler core

* add level params to Ix IR to support decompilation

* refactor Ix IR to support decompilation

* implement decompilation mutdefs, compilation of mixed mutdefs, and tests for new mutdef format

* wip

* decompilation roundtrip of Ix get_env working

* WIP: new mutual definition changes

* roundtrip compilation decompilation tests now passing

* add count to roundtrip

* fixup merge

* fixup merge

* fixup merge test

* remove style guide

* merge

* Aiur2 prove/verify (#183)

* prove and verify for Aiur2
* env var to opt out of parallelism
* bump the Rust toolchain

---------

Co-authored-by: Gabriel Barreto <gabriel.aquino.barreto@gmail.com>
Co-authored-by: Samuel Burnham <45365069+samuelburnham@users.noreply.github.com>

* Remove binius (#185)

* chore: drop Binius

* Drop the dependency on Binius
* Remove Aiur
* Rename Aiur2 to Aiur

* chore: deactivate Iroh

* update the hash for the Rust toolchain

* satisfy ci

* Bench (#186)

* add a `clock` executable for performance analysis

* clean up for a benchmarking implementation

* initial ixon serialization framework

* rustfmt

* ixon univ/expr roundtrip

* new ixon format passing proptest roundtrip

* remove old files

* implement Ixon v2 in Lean

* Ixon reconstruction in Rust from Lean pointers

* use standard formatting

* fix Ixon Lean serialization

* WIP: Lean -> Rust round-trip serialization test

* fix FFI

* WIP: synchronized Lean and Rust implementations, failing on metadata FFI?

* fix FFI for ReducibilityHints::Regular

* WIP: improved performance of mutual def sorting, but compilation still OOM on large terms

* very WIP compiler rewrite

* compiler refactored

* proptest for `Ixon` serde

* try reverting caches to RBMap

* bitblast now compiling

* full compilation of Lean->Ixon, but slow

* Tarjan's SCC algorithm for mutuals, so we can compile Recursors

* simplify mutual compilation, remove IR

* compile unsafe Lean?

* decompilation wip

* decompilation progress, but failing on _cstage2

* preprocessing GroundM step, but CondenseM breaking

* fix CondenseM, GroundM breaking

* wip

* GroundM and CondenseM working

* CompileM/DecompileM roundtrip provisionally working up to 5% of get_env!

* wip

* fixup for comments and clippy

* disable aiur benchmarks due to Ixon format change

* xclippy warnings

* disable Cli test suite

---------

Co-authored-by: Samuel Burnham <45365069+samuelburnham@users.noreply.github.com>
Co-authored-by: Arthur Paulino <arthurleonardo.ap@gmail.com>
Co-authored-by: Gabriel Barreto <gabriel.aquino.barreto@gmail.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants