[MLton] Question about Library Calls

[Matthew Fluet]

>> > So... after a long period of trying to debug the Gamma-CFA code that
>> > I've written, I've reached a point where I need to know a little bit
>> > more about MLton.  We've gotten it to run on a variety of test cases
>> > involving no application of the standard library (such as a factorial
>> > program that just computes the value of factorial, but doesn't print
>> > it).  When run on code that contains a print statement (even just
>> > hello world) it loops forever.
>> 
>> What "loops forever"?  The Gamma-CFA code or the final executable?
>
> The Gamma-CFA code.

O.k.  I'll reiterate that we would be happy to create a Gamma-CFA branch 
in the mlton.org repository; many eyes make light work. ;-)

>> > I was wondering what you could tell me about how MLton implements the
>> > library call (such as the print statement), and what operations it
>> > performs that are not present in code that doesn't use any library
>> > calls.
>> 
>> Even your factorial function is using the Basis Library, in order to
>> access the definitions of Int.- and Int.*.  Nonetheless, these kinds of
>> functions become primitives (Prim.Name.Word_sub, Print.Name.Word_mul) in
>> the compiler ILs.  Other Basis Library functions that correspond to
>> interaction with the outside world (e.g., TextIO.print) are implemented
>> via normal SML code plus some external C calls.  These external C calls
>> also become primitives in the compiler ILs, but they are of the
>> Prim.Name.FFI variant.
>
> Hmm... ok.  So there isn't anything significantly different between
> the factorial function and a similar function with that prints the
> result, except for the FFI call?

The significant differences will be the FFI call and the amount of code to 
be analyzed.  There is a dead-code pass that runs before the conversion to 
the XML IL, that, in the absence of IO in the user program, will discard 
vast quantities of the Basis Library implementation.