Okay, It's been a looooooong time since I added anything, but I'd like to make it up by finally publishing something that I have been working on, quite literally, for years. I have just added a file to the project that fully describes all of the ins and outs of a workable balanced ternary floating point data type. It is presented in the form of a standards document. If you've ever read an IEEE standards document, you've got the right idea.
It doesn't tell you how to build a floating point unit or library, it tells you what a floating point unit or library would need to do. That might not seem like much, but it encompasses a level of detail that required years of work and many revisions to perfect (fingers crossed). The document says version 3.1, (two major re-designs before publishing it) but I was taking notes, doing research and generally sorting out the basics long before version 1 of the document.
Anyways, it's finally done and I'm rather proud of it. I've already got some basic proof of concept work done on an FPU (very preliminary) and I'm sure I'll publish that stuff somewhere along the line.
I'm just about finished with floating point for a while. Instead I think I'll develop some work I started a long while back on a balanced ternary algebra. I'd kind of like to have a fully developed algebra that could be used to work out minimization of truth tables. In other words, using math alone to determine the most efficient hardware design to achieve a specified truth table. I already have a good chunk of it, but I think I'm still missing some basic axiom. We'll see how it goes.