The CrossWorks C library uses IEEE floating point format as specified by the ISO 60559 standard with restrictions.
This library favors code size and execution speed above absolute precision. It is suitable for applications that need to run quickly and not consume precious resources in limited environments. The library does not implement features rarely used by simple applications: floating point exceptions, rounding modes, and subnormals.
NaNs and infinities are supported and correctly generated. The only rounding mode supported is round-to-nearest. Subnormals are always flushed to a correctly-signed zero. The mathematical functions use stable approximations and do their best to cater ill-conditioned inputs.