Control Flow in Circuits

Circuits support a subset of Achronyme’s control flow. Every construct must resolve at compile time — the compiler needs to know the exact structure of the constraint system before any inputs are provided.

for Loops

for loops are statically unrolled — the compiler expands every iteration into separate constraints. Two forms are supported.

Range form — iterate over start..end:

circuit range_form(out: Public, a: Witness, b: Witness) {
    let product = a * b
    assert_eq(product, out)

    let sum = a + b
    assert_eq(sum, a + b)
}

Array form — iterate over a declared array:

circuit array_form(expected_sum: Public, vals: Witness[3]) {
    let acc = vals[0]
    let acc = acc + vals[1]
    let acc = acc + vals[2]
    assert_eq(acc, expected_sum)
}

The constraint cost of a loop equals the number of iterations multiplied by the cost of the body.

Nested Loops

Nested for loops are supported. The total iteration count is the product of all loop bounds:

circuit nested_loops(matrix: Witness[9]) {
    for i in 0..3 {
        for j in 0..3 {
            range_check(matrix[i * 3 + j], 8)
        }
    }
}

This unrolls to 9 iterations, each with a range_check (9 constraints in R1CS, 9 lookups in Plonkish).

if/else as mux

In circuits, if/else compiles to a mux (multiplexer) — both branches are always evaluated, and the condition selects which result to use. This costs 2 constraints (one boolean enforcement for the condition, one selection constraint).

circuit conditional(cond: Witness, a: Witness, b: Witness) {
    let result = if cond { a } else { b }
}

This is equivalent to mux(cond, a, b). There is no short-circuit evaluation — both a and b are computed regardless of cond.

An if without else is not supported in circuits because both branches must produce a value for the mux.

let Bindings

let bindings create aliases — they do not emit constraints. A let just gives a name to an existing linear combination.

circuit let_demo(a: Witness, b: Witness) {
    let sum = a + b        // 0 constraints — just an alias
    let doubled = sum + sum // 0 constraints — still a linear combination
}

Bindings only cost constraints when they involve operations that require multiplication gates (e.g., let product = a * b costs 1 constraint for the multiplication, not for the let).

Rejected Constructs

These constructs are rejected at compile time with clear error messages:

Construct	Error	Reason
while	WhileInCircuit	Iteration count unknown at compile time
forever	ForeverInCircuit	Same — unbounded loops can’t be unrolled
break	BreakInCircuit	Dynamic exit not representable in constraints
continue	ContinueInCircuit	Dynamic skip not representable in constraints
print()	—	Side effect with no circuit equivalent

Circuits require a fixed, deterministic constraint structure. Any construct that depends on runtime values to determine control flow is incompatible with this model.