Whole-file compilation in gfortran

[FX Coudert <fxcoudert at gmail dot com>]

Hi gfortranners, hi Michael,

A short mail about the topic of whole-file compilation. This is not  
an urgent mail, I simply thought about it a bit more, and I wanted to  
share my impressions, check that I'm not too confused about the goals  
and means here, and then share links to two patches ([1] and [2]  
below) by Steven Bosscher and Michael Matz aiming at this (I didn't  
know about these patches, or forgot, so maybe the notice will be good  
for others as well).

As far as I understood, there are 3 different issues to be tackled  
that this would help us :
  (a) perform better inter-file checking of interfaces,
  (b) inline non-contained procedures, and
  (c) module loading performance

I will not talk about (a) here, as it's probably the most tricky. As  
far as I see, (c) requires us to store modules into memory in  
addition to writing them to disk, so that we can load symbols from  
them without actually reading/parsing the file back. There might be  
more to it, Paul could tell.

Now, I'm kind of puzzled about (b). Following the approach of Michael  
Matz at [2], means that every time we would create a function decl,  
we start by looking whether there already is one with this name in  
the toplevel namespace, and if there is, we use this instead of  
creating a new. Michael apparently never follow this patch up, but I  
can't think of any fundamental objection to it, of anything that  
would make it completely flawed. (The one thing I can think of is  
that, without further type-checking of function decls, we might end  
up create ICEs on invalid code that the front-end doesn't catch yet,  
like mismatching arguments; but we can add type-checking, can't we?

So, I hope my point of view here is not to naïve, but I'd like to  
hear your opinions on all this.

FX


[1] http://gcc.gnu.org/ml/gcc/2005-07/msg00863.html
[2] http://gcc.gnu.org/ml/gcc/2005-07/msg00940.html