[mercury-users] mmc --make and --trans-intermod-opt

[Julien Fischer]

On Tue, 2 May 2006, Michael Day wrote:

> > It works but what remains to be done is convert a lot of the existing
> > optimisations to use it.  In particular inlining, higher-order
> > specialization and type specialization don't currently work with it, so
> > it's unlikely you'll see much benefit from it at the moment (of course
> > don't let that stop you testing it for us - and reporting any bugs ;-) )
>
> It seems to work, at least compiling Prince with it creates a bunch of
> analysis files and it doesn't crash or produce any unusual errors, so
> that's a vote of confidence I guess :)
>
> Now... will --smart-recompilation work with --intermodule-analysis?
> Because at the moment, touch one line and the whole damn thing rebuilds :)

The will _run_ together (well, at any rate, they did the last time I checked),
they probably won't _work_ together, in the sense of minimising the
amount of recompilation done, since smart recompilation doesn't know
anything about .analysis files (yet).

In general, --intermodule-analysis, will increase the compilation time
by quite a bit (for the standard library I think we end up running four
or five analysis passes on most of the modules).  The upper limit on the
number of passes is controlled by the `--analysis-repeat' option.  (Each
analysis pass is roughly equivalent to compiling a file with
--make-optimization-interface).

Julien.
--------------------------------------------------------------------------
mercury-users mailing list
post:  mercury-users at cs.mu.oz.au
administrative address: owner-mercury-users at cs.mu.oz.au
unsubscribe: Address: mercury-users-request at cs.mu.oz.au Message: unsubscribe
subscribe:   Address: mercury-users-request at cs.mu.oz.au Message: subscribe
--------------------------------------------------------------------------