Re: [avr-libc-dev] Math.h additions

[Joerg Wunsch]

[That'll be my last mail before finally leaving for vacation.]

As Colin O'Flynn wrote:

> 1) define NaN type (is 0xFFC0xxxx according to various
> FP libs and the readme)
> 
> #define NAN 0xFFC00000

Well, that won't work.  Remember, these numbers are supposed to be FP
numbers, not integer numbers, yet they gotta compare equal to a certain
bit pattern of an FP number...

I once thought about that but don't have a solution.  GCC makes some
assumptions about the internal bit patterns which it can compare (IIRC),
but our NaN is different from GCC's assumptions.

> 2.5) Maybe define HUGE_VAL and HUGE_VALF
> 
> #define HUGE_VAL INFINITY
> #define HUGE_VALF INFINITY

Same problem here.

I think a major polish of the FP library is due after 1.0 left the door.

-- 
J"org Wunsch                                           Unix support engineer
address@hidden        http://www.interface-systems.de/~j/