Struct SoundnessCalculator 

pub struct SoundnessCalculator {
    pub logup_bits: f64,
    pub gkr_sumcheck_bits: f64,
    pub gkr_batching_bits: f64,
    pub zerocheck_sumcheck_bits: f64,
    pub constraint_batching_bits: f64,
    pub stacked_reduction_bits: f64,
    pub whir_bits: f64,
    pub whir_details: WhirSoundnessCalculator,
    pub total_bits: f64,
}

Fields

logup_bits: f64

Security bits from LogUp α/β sampling and PoW.

gkr_sumcheck_bits: f64

Security bits from ordinary GKR sumcheck rounds.

gkr_batching_bits: f64

Security bits from GKR μ/λ batching per layer.

zerocheck_sumcheck_bits: f64

Security bits from ZeroCheck sumcheck after the fused input-layer boundary.

constraint_batching_bits: f64

Security bits from the fused GKR/ZeroCheck input boundary and constraint batching.

stacked_reduction_bits: f64

Security bits from stacked reduction sumcheck.

whir_bits: f64

Security bits from WHIR (minimum across all rounds and error sources).

whir_details: WhirSoundnessCalculator

Detailed WHIR soundness breakdown.

total_bits: f64

Total security bits (minimum of all components).

Implementations

impl SoundnessCalculator

pub fn calculate( params: &SystemParams, base_field_order: f64, challenge_field_bits: f64, max_num_constraints_per_air: usize, num_airs: usize, max_constraint_degree: usize, max_log_trace_height: usize, num_trace_columns: usize, num_stacked_columns: usize, n_logup: usize, ) -> Self

Calculates soundness for the given system parameters.

Arguments

• params - System parameters including WHIR config and LogUp parameters.
• base_field_order - Order p of the base field that sample_bits reduces. For BabyBear: 2^31 - 2^27 + 1. Used to charge the sample_bits sampling bias (query bad-set argument and proof-of-work).
• challenge_field_bits - Bits in the challenge field. For BabyBear4: ~124 bits.
• max_num_constraints_per_air - Maximum constraints in any single AIR.
• num_airs - Number of AIRs being batched.
• max_constraint_degree - Maximum degree of any constraint polynomial.
• max_log_trace_height - Maximum log₂(trace height) across all AIRs.
• num_trace_columns - Total columns batched in stacked reduction.
• num_stacked_columns - Total columns across all commitments (for WHIR μ batching).
• n_logup - GKR depth (log₂ of total interactions), or 0 if no interactions.
• proximity_regime - Unique decoding or other regimes (for WHIR-related calculations).

pub fn logup_soundness( max_interaction_count: u32, log_max_message_length: u32, challenge_field_bits: f64, log2_list_size: f64, ) -> f64

LogUp soundness from α/β sampling (before proof-of-work grinding).

• α: Random evaluation point to test whether Σ p(y)/q(y) = 0. If interactions are unbalanced, the sum is a non-zero rational function with a bounded number of roots. By Schwartz-Zippel, a random α detects this with high probability.
• β: Random challenge for compressing interaction messages into field elements. Degeneracy would allow distinct message tuples to collide.

Security = |F_ext| - log₂(2 × max_interaction_count) - log_max_message_length - log₂(L_PCS)

Proof-of-work grinding is an orthogonal amplification that callers add on top via Self::effective_pow_bits. This is the single source of truth for the formula; SDK parameter selection calibrates against it.

Reference: Section 4 of docs/Soundness_of_Interactions_via_LogUp.pdf

pub fn whir_proximity_gap_security( proximity_regime: ProximityRegime, challenge_field_bits: f64, log_degree: usize, log_inv_rate: usize, batch_size: usize, ) -> ProximityGapSecurity

Computes WHIR proximity gap security bits.

Per WHIR paper: err*(C, ℓ, δ) = (ℓ - 1) · 2^m / (ρ · |F|)

• UniqueDecoding: Security bits = |F_ext| - log₂(ℓ - 1) - log₂(degree) - log₂(1/rate)
• ListDecoding { m }: Uses BCHKS25/TR25-169 Theorem 1.5 (contrapositive) to bound the “bad z set” size by a, with Haböck’s global extension introducing a linear factor (ℓ - 1).

impl SoundnessCalculator

pub fn calculate_from_vk<SC: StarkProtocolConfig>( vk: &MultiStarkVerifyingKey<SC>, ) -> Self

Computes a conservative soundness estimate for the verifier defined by the given vk.

The verifying key does not fix a single proof shape: optional AIRs may be absent and trace heights can vary within verifier-enforced bounds. This function therefore uses verifier-side upper bounds derived from the key.

Trait Implementations

impl Clone for SoundnessCalculator

fn clone(&self) -> SoundnessCalculator

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for SoundnessCalculator

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.