This is the mail archive of the
kawa@sources.redhat.com
mailing list for the Kawa project.
Re: Question on compiling to bytecode
- From: Per Bothner <per at bothner dot com>
- To: Erol Akarsu <eakarsu at cmguc dot com>
- Cc: kawa at sources dot redhat dot com
- Date: Tue, 10 Dec 2002 15:30:48 -0800
- Subject: Re: Question on compiling to bytecode
- References: <3DF674C3.2EF095C5@cmguc.com>
Erol Akarsu wrote:
I have one question. In the KAWA-related paper, it is mentioned that it
is
better to translate Scheme to java bytecode directly because of easiness
of
writing interactive programs. I agree with that. But I need GCJ to make
full optimization. Currently, Gcj has to work on bytecode generated by
Kawa to optimize, but this may not produce optimal code. What about Kawa
should have also an option to generate full Java source code translation
instead of bytecode. Therefore, Gcj will have opportunity to apply full
optimization power on it. Can we make it?
Currently Kawa is tied to producing bytecode only. One could
conceivably extend it to other "backends", such as Java source,
or C source or .net bytecodes or whatever. However, that is a
really big project, and difficult to justify. I'd only do it
somebody paid me (well) to do it.
It seems better to enhance GCJ to do better optimizations on
bytecodes.
--
--Per Bothner
per@bothner.com http://www.bothner.com/per/