Floating point support in Jato

Arthur Huillet has been working on floating point support for Jato. Today, I merged some initial FPU support patches which brings our bytecode coverage to 70%! There's still some work to do (we don't support doubles, for example) but I'm pretty confident we'll have complete FPU support in Jato soon.

We're using SSE1 instructions for the FPU support because the stack-based x87 model is just really weird and inefficient. We wanted to use SSE2 but as it turns out, 32-bit AMD CPUs only support SSE1. But I guess that's a good thing as we now support old Intel CPUs all the way back to Pentium III.