> Neil's FP Tester and the JCK :-) You need to make sureyou get the more
> awkward stuff like NaN and infinity handling just right. This was the
> reason initially I ported this library rather than use the ARM one---the
> ARM one was a bit broken in the IEEE respsect at the time.

Well, there still appear to be "issues" in the EGCS library; the response I
got at work was that the EGCS-provided ARM library uses a middling-performance
machine independent floating point implementation, which does not pass
paranoia.

> Rather than beefing up that code, it'd be better to at least upgrade
> what's there to the latest version first. This has been on my to-do list
> for ages, it's just never got done.

I don't suppose this code is available for download anywhere?

Todd Whitesel
toddpw @ best.com