[m-dev.] option-finetuning (generating trans-opts)

[Nancy Mazur]

> On 12 Sep, Fergus Henderson wrote:
> 
> > Well, basically the only way to avoid this would be to redesign the
> > way trans-opt files work.  That might well be a good idea!
> 
> There was a rumor floating around recently that there was some interest
> in developing a new approach to the build process.  I'm hoping this
> would entail merging mmake and mmc into a single program that builds
> the application bottom-up one file dependency SCC (or even call graph
> SCC) at a time.  If this is done, can't the generation of .opt,
> .transopt and .c files be done at once?  Could this make it cheap
> enough to just always generate a .transopt file?  I realize there are
> potential memory footprint problems with compiling an SCC at a time,
> but this can always be handled by flushing information not immediately
> necessary from memory to disk, and reloading it when it is needed
> again.
> 
> So what are the plans and when are they likely to happen (if ever)? 
> Could Nancy take advantage of such a rewrite?

I'd love the SCC-aproach! Our analysis has the problem of dealing
with mutually recursive predicates defined over several modules. 
The SCC-compilation approach would solve this. 
But I don't think that this scheme would be useful for everyday 
compilation, just when the user requires some kind of heavy optimizations 
to be done... 

Nancy
--------------------------------------------------------------------------
mercury-developers mailing list
Post messages to:       mercury-developers at cs.mu.oz.au
Administrative Queries: owner-mercury-developers at cs.mu.oz.au
Subscriptions:          mercury-developers-request at cs.mu.oz.au
--------------------------------------------------------------------------