Port to whir PR #215 + PR #217 with SHA-256 by Bisht13 · Pull Request #288 · worldfnd/provekit

Bisht13 · 2026-02-13T14:19:48Z

Ports provekit to whir's PR #215 transcript API and PR #217 Weights trait. Uses SHA-256 for Merkle tree hash, transcript sponge, and proof of work.

Changes

Transcript: IOPattern → DomainSeparator/ProverState/VerifierState. Spongefish v1.0.0-rc1 (byte-oriented).
Weights: Weights enum → Covector<F> + dyn Weights<F> trait objects.
Hash: SHA-256 for Merkle tree (whir::hash::SHA2), transcript sponge (spongefish::instantiations::SHA256), and PoW. Skyscraper code retained for future use.
Build: Added release-fast profile, jemalloc, reduced tracing noise.
Cleanup: Removed SumcheckIOPattern, WitnessIOPattern, witness_io_pattern.rs.

Tested

Prepare → Prove → Verify on complete_age_check.

Replace IOPattern-based transcript with whir's DomainSeparator/ProverState/ VerifierState API (rev 246dae28). Update spongefish to v1.0.0-rc1 with byte-oriented Permutation<64> interface. Proof now carries narg_string + hints instead of a single transcript blob. Key changes: - Sponge: Permutation<64> with U=u8, WIDTH=64, RATE=32 - PoW: PoWSolution return type, solution() method - Prover: config.commit()/config.prove() replacing CommitmentWriter/Prover - Verifier: config.receive_commitment()/config.verify() replacing readers - Public weights use Weights::geometric (non-deferred) on both sides - Cross-evaluation hints for dual-commitment path - NTT registration via global type-map for ark_bn254::Fr - Clippy clean on prover/verifier crates

Propagate whir's debug_assertions transcript pattern through WhirR1CSProof so in-process prove→verify (bench tests) passes in debug builds. Add register_ntt() call in bench tests. Point whir to 22efa1b2 (px/reduce-tracing-noise branch).

Replace ark-crypto-primitives-based SkyscraperCRH/SkyscraperMerkleConfig with a whir HashEngine implementation that registers Skyscraper for leaf hashing, internal nodes, and proof-of-work. Uses block4 NEON SIMD on aarch64 for batched compression, with a stack-allocated pair buffer to avoid heap allocation in the hot path.

…ization

ashpect · 2026-02-16T01:05:30Z

tooling/provekit-gnark/src/gnark_config.rs

 #[derive(Debug, Serialize, Deserialize)]
-
 pub struct WHIRConfigGnark {
-    /// number of rounds


There does not seem any logical reason to remove comments. Maybe its better to keep it for more context when fixing gnark.

ashpect · 2026-02-16T01:27:02Z

tooling/cli/src/main.rs

 fn main() -> Result<()> {
    let args = argh::from_env::<cmd::Args>();
-    let subscriber = Registry::default().with(SpanStats);
+    let subscriber = Registry::default().with(SpanStats.with_filter(LevelFilter::INFO));


I think its better to define level instead of defaulting to INFO.

for eg :
// Debug builds: track ALL spans for detailed profiling
// Release builds: only INFO+ to reduce overhead
#[cfg(debug_assertions)]
let level = LevelFilter::TRACE;
#[cfg(not(debug_assertions))]
let level = LevelFilter::INFO;

ashpect · 2026-02-16T01:28:27Z

tooling/provekit-ffi/src/ffi.rs

 pub extern "C" fn pk_init() -> c_int {
-    // Initialize tracing/logging if needed
-    // For now, we'll keep it simple and just return success
+    provekit_common::register_ntt();


Keep the tracing/logging as a TODO comment

ashpect · 2026-02-16T01:38:04Z

provekit/common/src/skyscraper/pow.rs

-    challenge: [u64; 4],
-    bits:      f64,
+    challenge:     [u8; 32],
+    challenge_u64: [u64; 4],


Storing u64 challenge seems redundant. transmuting when required might be a better approach.

For eg : see verify code below

ashpect · 2026-02-16T01:38:12Z

provekit/common/src/skyscraper/pow.rs

        Self {
-            challenge: transmute!(challenge),
+            challenge,
+            challenge_u64: transmute!(challenge),


ashpect · 2026-02-16T03:27:45Z

provekit/prover/src/whir_r1cs.rs

+                .map(|w| covector_dot(w, &c1r_evals))
+                .collect();
+
+            merlin.prover_hint_ark(&(f_sums_1.clone(), g_sums_1.clone()));


unnecessary cloning, these 4 aren't used again hence can be directly passed

ashpect · 2026-02-16T03:29:03Z

provekit/prover/src/whir_r1cs.rs

+            // weights_2 on c1's polynomials. Whir's prove() expects evaluations
+            // for ALL (weight, polynomial) pairs in row-major order.
+            let c1m_evals = coeffs_to_evals(&c1.masked_polynomial_coeff);
+            let c1r_evals = coeffs_to_evals(&c1.random_polynomial_coeff);


redundant call , its same as c1m_evals

ashpect · 2026-02-16T03:30:54Z

provekit/prover/src/whir_r1cs.rs

+                &c2.random_polynomial_coeff,
+            ];
+
+            let poly_evals: Vec<Vec<FieldElement>> =


c1m and c1r evals are already calculated before.

use let poly_evals = vec![c1m_evals, c1r_evals, c2m_evals, c2r_evals];

ashpect · 2026-02-16T03:38:48Z

provekit/verifier/src/whir_r1cs.rs

-                (sums_2.0.try_into().unwrap(), sums_2.1.try_into().unwrap());
-
-            let mut statement_1 = prepare_statement_for_witness_verifier::<3>(
+            _public_weights_challenge,


This variable starts with _ , but it is used and not unused, fix naming

ashpect · 2026-02-16T05:08:53Z

provekit/verifier/src/whir_r1cs.rs

+        // Check deferred linear constraints.
+        // The public weight is Geometric (non-deferred), so it's not in the deferred
+        // list.
+        let offset = 0;


Irrelevent comment and remove offset.

- Remove redundant challenge_u64 field in SkyscraperPoW, inline transmute - Add SAFETY comment for unsafe block in SkyscraperHashEngine - Add PoW unit test for Skyscraper - Move coeffs_to_evals and covector_dot to common/utils/zk_utils - Derive Debug on WhirR1CSScheme (WhirConfig now supports it) - Call register_ntt() in create_domain_separator for robustness - Remove unnecessary clones on prover hint submissions - Reuse pre-computed evaluations instead of redundant recomputation - Remove unused _public_weights_challenge from verifier destructuring - Remove dead offset variable in deferred eval checks - Use cfg-based LevelFilter (TRACE debug, INFO release) in CLI - Add TODO for FFI tracing initialization - Restore doc comments on GnarkConfig and WHIRConfigGnark fields - Restore comment about arbitrary no-public-inputs hint

… sponge

Bisht13 added 3 commits February 13, 2026 19:44

Add release-fast profile, jemalloc allocator, and reduce tracing noise

19a371a

Bisht13 changed the title ~~Port to whir PR #215 transcript API~~ Port to whir refactor transcript API Feb 14, 2026

Bisht13 added 2 commits February 14, 2026 19:37

Port to whir Weights dyn trait API (PR #217) with pre-transform optim…

9b91018

…ization

Bisht13 mentioned this pull request Feb 14, 2026

Update gnark recursive verifier for new whir API #289

Open

Bisht13 changed the title ~~Port to whir refactor transcript API~~ Port provekit to whir PR #215 transcript API + PR #217 Weights trait + Skyscraper HashEngine Feb 14, 2026

Bisht13 changed the title ~~Port provekit to whir PR #215 transcript API + PR #217 Weights trait + Skyscraper HashEngine~~ Port to whir PR #215 + PR #217 with Skyscraper HashEngine Feb 14, 2026

Bisht13 requested a review from ashpect February 15, 2026 17:11

ashpect requested changes Feb 16, 2026

View reviewed changes

Bisht13 force-pushed the px/whir-pr215-compat branch from e17c47c to ac651c1 Compare February 16, 2026 06:11

Switch from Skyscraper to SHA-256 for Merkle tree hash and transcript…

e79f619

… sponge

Bisht13 changed the title ~~Port to whir PR #215 + PR #217 with Skyscraper HashEngine~~ Port to whir PR #215 + PR #217 with SHA-256 Feb 16, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Port to whir PR #215 + PR #217 with SHA-256#288

Port to whir PR #215 + PR #217 with SHA-256#288
Bisht13 wants to merge 7 commits intomainfrom
px/whir-pr215-compat

Bisht13 commented Feb 13, 2026 •

edited

Loading

Uh oh!

ashpect Feb 16, 2026

Uh oh!

ashpect Feb 16, 2026

Uh oh!

ashpect Feb 16, 2026

Uh oh!

ashpect Feb 16, 2026

Uh oh!

ashpect Feb 16, 2026

Uh oh!

ashpect Feb 16, 2026

Uh oh!

ashpect Feb 16, 2026

Uh oh!

ashpect Feb 16, 2026

Uh oh!

ashpect Feb 16, 2026

Uh oh!

ashpect Feb 16, 2026

Uh oh!

ashpect Feb 16, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

Bisht13 commented Feb 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changes

Tested

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Bisht13 commented Feb 13, 2026 •

edited

Loading