* Parrot Raiser <1parr...@gmail.com> [2014-12-07 22:40]:
> The practical distinction, surely, is that the output of a compiler
> is usually kept around, to be run one or more times, whereas the an
> interpreter always works with the original human-readable source.

Yes, surely that’s it. We all consider Python a compiler, after all.


Go on, tweak your definition to pin it down. :-)

* Gerard ONeill <oobl...@usa.net> [2014-12-08 15:10]:
> How about an interpreter interprets input directly into action (even
> if there is some optimization going on), while a compiler converts
> instructions from one set to another set to be interpreted later.

That’s just an unnecessarily concrete rephrasing of the definitions
I mentioned.

> Which would make perl both at the perl source level,

Perl never interprets raw perl code without first parsing it into an

> and an interpreter at the bytecode level.

Well yeah, bytecode always implies an interpreter.

> Thinking of execution as interpretation, this allows for the transmeta
> concept, where the CPU was just an interpreter / just in time compiler
> that interpreted x86 instructions.
> Although modern CISC CPU's have a step where the input to the chip was
> still converted to microcode which was actually what was run.  So
> a compilation step, and an interpreting step.

Execution takes a program as input and produces the program’s output as
output. So it’s interpretation. By definition.

Sometimes there are dedicated hard-wired circuits that do it, and
sometimes there are other layers of abstraction around the hard-wired
circuitry. The layers can be in hardware, and even then at different
degrees of abstraction (FPGA vs microcode, say), or in software, and
really what is software and what is hardware depends merely on your
perspective. There are plenty of coprocessors that internally run code
which is opaque from the outside; is that software or hardware?

That’s what I meant by fuzzy ideas. You don’t get anywhere trying to
nail this pudding to the wall. You only get somewhere if you accept that
which is which is relative to your point of view and that the difference
is defined in terms of the output: compilers transform programs to other
programs and interpreters transform a program into its output. That’s it.

E.g. if you have something perl running a Perl program then you have the
CPU interpreting a program (perl) that itself interprets another program
(the optree), which in turn was compiled from the user Perl program
earlier on.

Once you stop trying to artificially force everything into a single
absolute distinction, the entire debate about which is which vanishes.

Aristotle Pagaltzis // <http://plasmasturm.org/>

Reply via email to