FP overhaul

[The Beez]

Hi 4tH-ers!

You must have been wondering what I've been doing lately, since there are MANY changes in SVN.

Well, I've been testing the accuracy of FP 4tH on a 64 bit platform and found there was LOTS of room for improvement.

First, working under 64bit means that ANS float now handles full 128 bit long floats. That's 38 digits of precision. Zen float now uses 63 bit long floats (excl. sign bit), so we had to update several constants. Where ever we could of course.

We've also looked into FEXP and FLN functions, where all of a sudden flnflogb.4th began to shine where precision is concerned. We also tuned the "epsilon" parameters, both 32 bit and 64 bit.

We replaced F>D D>F calls with FTRUNC where possible - which helps when your integer part doesn't fit a double number. That brought along the need for a FAST double word SHIFT routine - which brought along a quirk in GCC, and we fixed that one.

So a lot of work has been going on into that, which you won't notice on first sight, but may appear in higher accuracy, some speedups and the occasional (potential) bugfix. Note that NONE of this should effect 32 bit compatibility.

I hope you appreciate the effort - and you'll hear from me again.

Hans Bezemer