libfastm(3M)

NAME
     libfastm - fast transcendental library

SYNOPSIS
     #include <math.h>
     libfastm.a contains various hand-coded routines
     tuned to the specific architecture of the system.
     Currently, on MIPS 1 systems libfastm.a contains
     only routines sqrt and fsqrt.  On MIPS 2 systems
     libfastm.a contains routines sqrt and fsqrt,
     which are identical to those in libm.a .
     (Hardware square root instructions exist on MIPS 2
     and later architectures.)

     On MIPS 4 systems, libfastm.a also contains the
     following routines: sin, cos, tan, atan2, exp, log,
     pow, sinf, cosf, tanf, expf, logf, and powf.


LIMITATIONS:
     On MIPS 4 systems, routines in libfastm.a must be run in round to nearest
     rounding mode to give correct results.  This is the default rounding
     mode.

     Routines in this library do not set global errno, and may not give
     correct results for NaNs and Infinities.

     Routines sin, cos, and tan give accurate results for arguments between
     +-2**22 radians.

     Routines sinf, cosf, and tanf give accurate results for arguments between
     +/-2**15 radians.  Precision may be lost near multiples of pi/2 for these
     three routines.

     Routines sin, cos, tan, and atan2 should be run with underflow and
     overflow floating point traps disabled (the default mode) to avoid
     trapping in these routines.