[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: [Fwd: Re: libbsf-java]



Robert Lougher <rob.lougher <at> gmail.com> writes:

> 
> Hi Dalibor,
> 
> Dalibor Topic <robilad <at> kaffe.org> writes:
> 
> > Kaffe uses its own sysdepCallMethod code but can also use libffi as an
> > additional option. See config/$arch/sysdepCallMethod.h for details. See
> > config/sysdepCallMethod-ffi.h for the wrapper for ffi. Feel free to merge it
> > into JamVM, if you think it's useful.
> > 
> 
> I'll definately have a look at it.  Thanks for the pointer.  Have you done any 
> benchmarks to see if there's any difference in method invocation speed?  It's 
> certainly a good way to get a port up and running fast.

Not me, others may have. We should take that to the Kaffe mailing list, though.

> > My personal plan is to merge in libffi into Kaffe as well, and use it as a
> > default on Linux at least, and then to gradually switch to it for other 
> > platforms.
> > 
> 
> To clarify, you mean to put libffi itself into kaffe, i.e. not use it as a 
> separate library?  I'm probably completely wrong, but I had heard something 
> about libffi being part of gcc, and hard to find/install as a separate entity, 
> i.e. on an embedded system.  Is this why you'd merge it into kaffe?

Yeah. Kaffe is a bit of a everything in a box, VM, class libs, tools, and all
that, but it's also pretty highly customizable. So for people looking for an
'out of the box' solution for the common Java-related problems, merging some of
the core dependencies, and resyncing with them reduces the burden on the
user/packager to do some sophisticated packaging work. Kaffe runs on a lot of
platforms that don't have such excellent packaging support that Debian enjoys. :)

The other, more important side is of course encouraging people to hack on
libffi, as with 1.7 M loc in my hands in Kaffe, I'm a firm believer in using
other people's good code rather than inventing my own from scratch, and
promoting those efforts :)

cheers,
dalibor topic



Reply to: