Bump Plonky3 - #77
Merged
Merged
Conversation
Adapt the CUDA PCS to Plonky3's pruned multiproof protocol while preserving the recursive verifier's compact per-query circuit. Add a verified advice re-encoder that expands native pruned authentication frontiers only at the recursion boundary, leaving native proof serialization and verification unchanged. Bring the CUDA PCS in line with upstream domain semantics by enforcing the minimum committed LDE height, restoring row bit reversal for domain evaluations, and covering Goldilocks log-height 32. Transfer the initial reduced FRI opening into its resident LDE without an extra allocation or device copy. Replace per-column cudaMemcpy2D submissions during Merkle construction with one coalesced gather kernel and cap staging at 32 MiB. Serialize concurrent quotient scheduling to prevent transient recursive-prover OOMs while retaining parallel GPU kernel execution. Update the compatibility digest and keep the CPU-only build independent of CUDA. The CUDA smoke suite passes 50 tests and produces byte-identical 17,213-byte CPU/GPU proofs. On the RTX PRO 6000 q50 Vector.extract_append workload, inner plus outer STARK proving is 65.87s on CPU and 8.81s on CUDA (7.48x), with identical proof sizes.
Update the fixed-output hash used by Fenix to match the Rust 1.98 channel manifest selected by rust-toolchain.toml. This restores both the packaged Nix build and the development shell without changing the pinned toolchain or flake inputs.
arthurpaulino
force-pushed
the
ap/bump-p3
branch
from
September 1, 2026 12:50
bc69d37 to
fac47fa
Compare
samuelburnham
previously approved these changes
Sep 2, 2026
| use crate::cuda::mmcs::CudaCommitMmcs; | ||
| let ldes: Option<Vec<_>> = inputs | ||
| .par_iter() | ||
| .iter() |
Member
Author
There was a problem hiding this comment.
Yes. It only affects the cuda accelerated path. It tries to prevent OOMs.
`advice.rs` re-encoded a pruned FRI multiproof into one authentication path per query, for a recursive verifier that consumed the legacy per-query transport. That verifier now consumes the native pruned multiproof directly, so the expansion has no caller: nothing in this crate references `advice`, and the only downstream user (aiur) imported just its `AdviceError` type, which it can define locally. Drop the module and its `pub mod advice;` declaration. The library is self-consistent without it (lib + tests build clean; 35 tests pass). The frontier expansion remains recoverable from history if a fallback to the per-query transport is ever wanted.
arthurpaulino
force-pushed
the
ap/bump-p3
branch
from
September 2, 2026 17:14
8f3c50d to
2a1d3a5
Compare
arthurpaulino
enabled auto-merge (squash)
September 2, 2026 17:17
gabriel-barrett
approved these changes
Sep 2, 2026
| assert_eq!( | ||
| proof_bytes.len(), | ||
| 77_637, | ||
| 22_017, |
Member
There was a problem hiding this comment.
Interesting, it seems like proof_bytes is much smaller now. What is the reasoning?
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Bump the Plonky3 dependency, fix breaking changes and patch the CUDA pipeline accordingly.