Re: Building a translator. Want to know if this is feasible...

"Enrico Santoemma" <enrico.santoemma@mail.com> wrote in message
> Reading this post I thought this example:
> We have an old working device which does some logic transformations
> between input and output. We also have the schematic diagram. The
> device is made by old tubes and metallic wire resistors. How to
> automatically produce a schematic which uses modern ICs?
>
> I think the only answer is "emulate".


No, it is not. Your "circuit" diagram has components that have
behavioral properties. If you idealize those properties (as the
designing engineers did), then having the circuit diagram and the
names of the parts gives you a "formal" definition of the circuit. If
you claim this is a logic device, then in fact its various
constellations of parts should be straightforwardly interpretable as
primitive logic functions, in the same way that an invertor circiut
made of several transistors can be considered simply to by logical
NOT.


Once you have the primitive logical functions, then a tool can
translate that directly to gates, thence to any of the HDLs, and get
your "modern IC". [However, if this is really a vacuum tube device,
unless it is something like the SAGE computer, I doubt if its actual
complexity economically warrants building a tool to translate its
circuit diagram; just make an engineer do it].


> A more serious answer remind me an old Esprit project in the
> european community: build software that helps the huge CEE's
> translation job.


CEE? [I don't think most of us know what this is]
  ... huge?


> The better results were translation CADs, that is software that helps
> translators to recognize and automatize only the repetitive tasks.
> For the rest, human mind was (and still is) irreplaceable.


For low-level translation tasks, you overrate yourself :-}
[We build translation tools, for which the automation is irreplacable.
People don't have the stomach, time, or reliablity to
translate "huge" 500K SLOC systems.]


> The schematic example helps me to support my opinion: we should change
> the monodimensional view of a text source code for some bidimensional
> views, which is 1 to 1 representation of the code.


I think "bidimensional" isn't the right word. I'd prefer
"graph-based" representations. (Your electrical circuit diagram is a
2D projection of the actual wiring diagram). For what it is worth,
most internal program representations aren't "one dimensional", they
are graphs. We have followed this strategy with our tools because we
believe in the value of representing/manipulating not only
"string-based languages", but "visual ones" as well as arbitrary
representations of both of these.
--
Ira D. Baxter, Ph.D., CTO 512-250-1018
Semantic Designs, Inc. www.semdesigns.com
[Note to Americans: CEE is the French or Italian acronym for the EEC,
the Common Market that preceded the European Union. I gather they're
chronically desperate for people who can translate minor languages,
e.g. Danish to Greek, but natural language translation has remarkably
little in common with computer language translation. -John]

Return to the comp.compilers page.
Search the comp.compilers archives again.