cc at informatik.uni-stuttgart.de
Thu Aug 7 03:30:58 CDT 2008
On Wed, 6 Aug 2008, Ethan Dicks wrote:
> On Wed, Aug 06, 2008 at 01:25:40PM -0400, Paul Koning wrote:
>> By the way, I meant to mention another compiler that did something
>> like P-code: RT-11 Fortran. The actual scheme was called "threaded
>> code" because it was basically a stream of function pointers. (Come
>> to think of it, that's a common Forth encoding as well.) That's about
>> as fast as straight code if most of what you need is calls to support
>> functions, and more compact because you're not including the opcode
> That scheme sort of reminds me of FORTRAN IV for the PDP-8 - IIRC, the
> compiler produces, mostly, a stream of FPP instructions that either get
> executed by a real FPP-12 (or FPP-8?) _or_, lacking math hardware, get
> emulated by wads of PDP-8 instructions.
> The difference is rather than a stream of function pointers, it's a
> stream of math co-processor instructions that are either executed by
> real hardware or by an emulator library. In either environment, though,
> the object code is the same.
And this all is similar to what compilers did in the 50s. The ACT-V
compiler (ACT-I and ACT-III, too) simply generates a stream of runtime
system calls (the runtime system is called 'basic subroutines'). You can
even compile your program by hand since these subroutine calls are
documented in the manual.
More information about the cctech