Re: BC ABI (was "Re: [fedora-java] Re: Java OpenGL on FC4")

[Bryce McKinlay <mckinlay at redhat dot com>]

John M. Gabriele wrote:

> Thanks for the reply Bryce. I used much of it to update the
> beginning of:
> http://gcc.gnu.org/wiki/How%20to%20BC%20compile%20with%20GCJ
>  

Great, thanks for doing this.

> It sounds like, to be able to mix interpreted and natively
> compiled code at runtime, you need to have compiled your .jars
> with the -findirect-dispatch option, right?
>  

If you want to be able to call interpreted code from native code, yes. 
The other way around works fine regardless of the ABI.

> If so, why isn't -findirect-dispatch simply the default
> compilation mode?
>  

Mostly because its not finished yet. The major blockers are CNI, because 
the C++ compiler does not yet understand the BC ABI, and the bug(s) 
affecting -findirect-dispatch when used for source code compilation. 
Certainly, the goal is to make BC the default ABI as soon as the issues 
are resolved.

There is also a performance hit from the BC-ABI - the tests I have done 
show this to be relatively small (<= 10%) for most applications, but it 
would be good to have more data.



> > [snip]
> >
> > On the other 
> > hand, if you compile an application using the old style "C++ ABI", it 
> > will break as soon as any changes to the public APIs of dependent class 
> > libraries are made.
> >    
>
> Right -- since the Java runtime should be smart enough to use
> the lib as long as the signatures of the methods _it wants to
> use_ match up, correct?
>  

Right. The old ABI will break code in many situations, such as the order 
of fields or methods in a class changing. This is, obviously, 
unacceptable for Java.

> What does it mean to compile code with the "old style 'C++ ABI'"?
> (Is that referred to as the "standard ABI"?) Does that just simply
> mean to *not* use the -findirect-dispatch option?

Correct.

Bryce