Module ieee754 

IEEE 754 strict floating-point utilities. IEEE 754 float rules enforced by the parity engine.

Until vyre IR gains full float variants, this module acts as a strict guard: any code path that would require float semantics returns a deterministic error rather than falling back to undefined or driver-dependent behavior. When float support lands, this module will become the source of truth for rounding mode, NaN propagation, and subnormal handling that the conform gate checks.

Constants

BACKEND_ELEMENTARY_F32_ULP_BUDGET

Maximum accepted backend-vs-reference error for f32 programs without transcendental operations under the default parity policy.

BACKEND_TRANSCENDENTAL_ULP_BUDGET

Maximum accepted backend-vs-reference error for programs containing f32 transcendental operations.

REFERENCE_TRANSCENDENTAL_ULP_BUDGET

Maximum accepted reference-oracle error against correctly rounded f32 transcendental results.

Functions

canonical_acos

Deterministic f32 arc cosine used by the CPU parity oracle.

canonical_asin

Deterministic f32 arc sine used by the CPU parity oracle.

canonical_atan

Deterministic f32 arc tangent used by the CPU parity oracle.

canonical_cos

Deterministic f32 cosine used by the CPU parity oracle.

canonical_cosh

Deterministic f32 hyperbolic cosine used by the CPU parity oracle.

canonical_exp

Deterministic f32 exponential used by the CPU parity oracle.

canonical_exp2

Deterministic f32 base-2 exponential used by the CPU parity oracle.

canonical_f32

Flush subnormal f32 values to signed zero, preserving the sign bit.

canonical_inverse_sqrt

Deterministic f32 inverse square root used by the CPU parity oracle.

canonical_log

Deterministic f32 natural logarithm used by the CPU parity oracle.

canonical_log2

Deterministic f32 base-2 logarithm used by the CPU parity oracle.

canonical_reciprocal

Deterministic f32 reciprocal used by the CPU parity oracle.

canonical_sin

Deterministic f32 sine used by the CPU parity oracle.

canonical_sinh

Deterministic f32 hyperbolic sine used by the CPU parity oracle.

canonical_sqrt

Deterministic f32 square root used by the CPU parity oracle.

canonical_tan

Deterministic f32 tangent used by the CPU parity oracle.

canonical_tanh

Deterministic f32 hyperbolic tangent used by the CPU parity oracle.

canonical_ulp_distance

Compute the deterministic ULP distance after vyre f32 canonicalization.

pending_float_types

Return the canonical float-pending error.