Module expr

Modules§

constraints: A number of useful constraints
prologue: You can import this module like use kimchi::circuits::expr::prologue::* to obtain a number of handy aliases and helpers

auto_clone: Auto clone macro - Helps make constraints more readable by eliminating requirement to .clone() all the time
auto_clone_array

Cache: A cache
CacheId: A key for a cached value
Constants: The collection of constants required to evaluate an Expr.
FCol 🔒: A witness-column access resolved for the fused evaluator: at row i it reads evals[(scale * i + offset) % len] – the same indexing as SubEvals.
Linearization: A “linearization”, which is linear combination with E coefficients of columns.
RowOffset
Variable: A type representing a variable which can appear in a constraint. It specifies a column and a relative position (Curr or Next)

ConstantExprInner
ConstantTerm: Define the constant terms an expression can use. It can be any constant term (Literal), a matrix (Mds - used by the permutation used by Poseidon for instance), or endomorphism coefficients (EndoCoefficient - used as an optimisation). As for challengeTerm, it has been used initially to implement the PLONK IOP, with the custom gate Poseidon. However, the terms have no built-in semantic in the expression framework. TODO: we should generalize the expression type over challenges and constants. See https://github.com/MinaProtocol/mina/issues/15287
Either 🔒: Use as a result of the expression evaluations routine. For now, the left branch is the result of an evaluation and the right branch is the ID of an element in the cache
EvalResult 🔒
ExprError
ExprInner
FOp 🔒: Flat bytecode for the fused evaluator: a stack machine walked once per domain row, with intermediates living in a register frame rather than full-domain arrays.
FeatureFlag: The feature flags that can be used to enable or disable parts of constraints.
Operations
PolishToken: For efficiency of evaluation, we compile expressions to reverse Polish notation expressions, which are vectors of the below tokens.

AlphaChallengeTerm: The Challenge term that contains an alpha. Is used to make a random linear combination of constraints
ColumnEnvironment
ColumnEvaluations
FormattedOutput
Literal

fused_eval_mode 🔒: Execution mode for the experimental fused row-wise evaluator. KIMCHI_FUSED_EVAL=1 uses it (falling back to the vectorised path for expressions it does not handle); =verify runs both and asserts they agree.
l0_1: Computes prod_{j != n} (1 - omega^j) Assure we don’t multiply by (1 - omega^n) = (1 - omega^0) = (1 - 1) = 0
mul_monomials 🔒
unnormalized_lagrange_basis
unnormalized_lagrange_evals 🔒: Compute the evaluations of the unnormalized lagrange polynomial on H_8 or H_4. Taking H_8 as an example, we show how to compute this polynomial on the expanded domain.