Module fixed 

Deterministic integer analog of this crate’s floating-point error-bit arithmetic.

Every quantity elsewhere in this crate is -log2(error) as an f64 via crate::error::ErrorBits. A protocol that grades proof soundness inside a recursive verifier — a circuit or a VM with no floating-point unit — cannot reproduce that arithmetic bit-for-bit, and needs every rounding step biased toward understating security rather than overstating it. This module provides that: -log2(error) as a 16-fractional-bit fixed-point integer, computed with const fn integer operations only, rounding down on added logarithms and up on subtracted ones so the result is always a conservative lower bound on the true value.

Not wired into the rest of the crate’s f64 composition machinery — a consumer needing bit-exact recursive verification uses this module’s primitives directly rather than crate::error::ErrorBits.

Constants

FRACTIONAL_BITS

Number of fractional bits in the fixed-point representation.

LOG2_E

log2(e) in fixed point, rounded up.

ONE

The value 1.0 in fixed point.

Functions

bits_per_query

Returns the conjectured security contributed by a single FRI query, in fixed point.

ceil_log2

Returns log2(x) in fixed point, rounded up.

floor_log2

Returns log2(x) in fixed point, rounded down.

from_bits

Converts a whole number of bits into fixed point.

to_bits

Truncates a fixed-point quantity to whole bits, rounding down.