Re: Languages don't matter. A mathematical refutation

[Bob Duff <bobduff@theworld.com>]

"J-P. Rosen" <rosen@adalog.fr> writes:

> Le 03/04/2015 15:37, Bob Duff a écrit :
>> Assembly language: Semantics are defined in terms of the generated
>> machine code.
> OK
>
>> High-level language: Semantics are defined in terms of what the program
>> does.
> I don't understand this. Do you mean "specification: observed behaviour
> of a program" ? ;-)

Yes.  In an assembly language, if you write "ADD", the semantics are
that an ADD instruction is generated in the machine code.  But in a
high-level langauge, if you write "+", the language definition doesn't
say anything about what machine instruction(s) are emitted.  And in
practice, you might get a shift, or you might get no instructions
at all.  C is clearly in the second category.

>> Those are how I understand the terms, and I think how they are commonly
>> understood.  I can't think of any other useful definitions.
>> C is clearly a high-level language, by those definitions.
> Not clearly. A number of things (handling of integer overflows,
> semantics of %, dereferencing the null pointer...) are defined in the
> language as "like the machine does" (this is spelled: implementation
> defined). Unless it has changed in recent versions of C ?

No, no C standard has ever said "like the machine does".  Those things
are defined as "undefined behavior" in C, and that does not mean "like
the machine does", it means "anything can happen" (as far as the C
standard is concerned).  "Like the machine does" is not even meaningful
in a high-level language.  What the machine does depends on what code
the compiler generates.  For example, on x86, a C compiler might
generate an ADD (so you get wraparound semantics).  Or it might generate
ADD followed by TRAPV (so the program crashes on overflow).  Or it might
turn this:

    if (X + 1 <= 0)
        do_something();

into:

    do_something();

if it knows X >= 0 before the 'if'.  gcc actually does things
of that nature.  I've seen people get really annoyed at gcc for
that sort of thing, but I think their beef is with the language
definition -- they should use a language that defines the behavior
of overflow.

If it can prove X = the max integer, it can remove the entire if
statement.

- Bob