Re: [wwwdocs] Mention cgraph stuff in news/changes?

[Jan Hubicka <jh at suse dot cz>]

> On 13 Jul 2003, Steven Bosscher wrote:
> > Any plans to implement it or shall I correct this in the Changes page?
> 
> I don't know of any plans, though I'd love to see something done for 3.4.
> Inlining in gcj has never really worked that well in part because it could
> not inline out-of-order methods (and there is a long-standing PR for
> this).
> 
> Gcj has been compiling class-at-a-time for a some time now, so the call
> graph would be useful with or without unit-at-a-time.  (In fact many
> Java sources contain only one class, in which case unit-at-a-time would
> make no practical difference.)  I haven't yet looked at the cgraph
> interface to see how it might be done.
> 
> I'm copying Andrew Haley since he is the one most familiar with Java
> inlining.
In the case you or Andrew is interested in working on this, just fell
free to ask about the interface.
It is basically pretty simple - when you are finished with function
body, you call cgraph_finalize_function, when you are finished with
compilation unit, you call cgraph_finalize_compilation_unit and finally
you call cgraph_optimize to do the rest of work.
You need to define the hook to expand function.

Honza
> 
> Jeff
>