Re: RFA: Java --indirect-dispatch

[Per Bothner <per at bothner dot com>]

Bryce McKinlay wrote:

> I would like to get approval for the compiler part of this patch:
> 
> http://gcc.gnu.org/ml/java/2001-12/msg00209.html
> 
> I should point out that I have bootstrapped libjava on x86 and PowerPC 
> linux, and run libjava's "make check" both with and without the 
> --indirect-dispatch option without regressions.
> 
> OK to commit?

Yes.  I do have a couple of comments.

First, I have not studied the code in detail, though I think this
is a very nice feature to have. I think it is significant enough
that it merits mention on the GCJ News page.  Thanks for doing this!

I assume you will update the documentation.

Some possible hybids to consider:  Assume the standard java.*
and javax.* are "direct dispatch".  Use direct dispatch for calls
to classes in the same compilation unit.

There have been past attempts at ABIs for C++ that are more robust
in the face of the kind of changes we're talking about.  It would
not surprise me if someone has implemented it for G++.  Does COM
support something similar (I have no idea)?  If there is prior art,
it would be good if we were compatible with it, or at least are
aware of it, so we can improve compatibility.  Could you try to
find out - at least ask on the gcc mailing list?
-- 
	--Per Bothner
per@bothner.com   http://www.bothner.com/per/