[avr-libc-dev] Math.h additions

[Colin O'Flynn]

Hello,

There are a few additions to math.h that I think would
bring it closer to ANSI specs and are very easy to
do... maybe I've missed them somewhere else or
something, but they are:

1) define NaN type (is 0xFFC0xxxx according to various
FP libs and the readme)

#define NAN 0xFFC00000

2) define positive infinity

#define INFINITY 0x7FFFFFFF

2.5) Maybe define HUGE_VAL and HUGE_VALF

#define HUGE_VAL INFINITY
#define HUGE_VALF INFINITY

3) Define function int signbit(double x); which
returns nonzero value if and only if the sign-bit is
negative.

This would be pretty easy, something along the lines
of (written in perso-code)

return (int)(R25 AND 0x80);

R25 holds the exponent + sign of the double, so you
just strip everything else.. if its positive (sign bit
= 0) it returns 0, but if the sign bit is 1
(negative), then it returns 0x80 (128 dec).


There are a few more functions as well of course, but
these seemed to be almost implimented anyway and
really just a case of adding the documentation and a
few lines of code...

  -Colin