Dan Sugalski <[EMAIL PROTECTED]> wrote: > While the actual compiler code is evil, doing all sorts of things it > ought not do with absolute addresses (and therefore non-JITtable), > the forth compiler does work, and will compile things. It ought, > actually, to be able to be integrated in as a compiler.
> Your task, if you're interested, is to thump it such that forth.pasm > *can* be loaded in as a compiler module such that you can do: > compile P5, ForthCompiler, "1 2 + ." > and have 3 print out. Should be easy, relatively speaking. Three Notes WRT that: 1) examples/assembly/nano_forth2.pasm uses PCC to implement such a compiler. No address hacks, but likely a bit slower. 2) I've checked in the "compreg Sx, Py" opcode, where Py is a Parrot Sub. 3) I'll check in an example soon, how to load and run a compiler written in C and residing in a shared lib. I'm just waiting for some ok WRT the platform patch, which I've pending too. leo
