128 bits Floats

NUTS – Numerical Tools – is a library of fixed size numbers written for the 64-bit Free Pascal compiler (version 3.0.0 or higher).

Currently the library contains only 128-bit floating-point numbers with the basic operations QAdd, QSub, QMul, QDiv, QRem, QSqrt, and QFMA.
There are no transcendental functions yet.

Does this library follow some ‘754’ standard? Clearly, no. There are no so-called ‘signaling NaNs’ (all NaNs are regarded as quiet NaNs).
Any other difference (except the number of implemented functions) with the ‘754-2008’ standard would be a bug since it would not be intentional.

The current version is an alpha version. Many things can still be modified. There is no documentation yet but the code is commented and there is a Lazarus demo project: nuts.Demo.Float128s.

• NUTS was tested with x86_64 processors only.
• NUTS cannot run on big endian systems (adapting the Pascal code should not be too difficult but I have no big endian machine in order to check the code).
• Though it would not be intentional, it is possible that NUTS does not work with an other OS than Linux (I have no other OS).

French

Réels 128 bits