TorchLean API

Docs HomeInstallationGuideExamplesGraphsTools

FloatLib.Floats.Formats.DecimalInterchange.Quantize.Runtime

Rounding to a specified decimal quantum #

IEEE 754-2019 §5.3.2 specifies quantize independently of ordinary projection: the target quantum is fixed, and a coefficient that does not fit produces invalid instead of changing that quantum. Underflow and overflow flags are never raised.

source
def FloatLib.Floats.Formats.DecimalInterchange.Arithmetic.quantizeMagnitude (f : Format) (mode : RoundingMode) (s : Bool) (x : ℚ) (q : ℤ) :
Outcome

Round a nonnegative magnitude to a fixed quantum, retaining the supplied sign. An invalid quantum or an oversized rounded coefficient produces a quiet NaN.

Instances For
    source
    def FloatLib.Floats.Formats.DecimalInterchange.Arithmetic.quantize (f : Format) (mode : RoundingMode) :
    Datum → Datum → Outcome

    Quantize the first operand to the second operand's quantum. The second operand's sign and coefficient do not affect a finite result.

    Instances For