Date: Thu, 30 Mar 1995 14:21:17 +0100 From: Richard Tobin <richard@cogsci.ed.ac.uk> To: Bakul Shah <bakul@netcom.com>, Dave Cornejo <dave@dogwood.com> Cc: jkh@freefall.cdrom.com (Jordan K. Hubbard), freebsd-hackers@FreeBSD.org Subject: Re: HotJava Message-ID: <199503301321.OAA05112@deacon.cogsci.ed.ac.uk> In-Reply-To: Bakul Shah's message of Wed, 29 Mar 95 19:19:50 -0800
next in thread | raw e-mail | index | archive | help
> > 2) The job is non-trivial. He was rather vague, but mumbled something
> > about possibly needing kernel support.
>
> I suspect this may have to do with on-the-fly code
> generation from the byte-code.
The usual problems with on-the-fly code generation are
(a) If you jump to the new code just after writing it, your pipeline may
get filled before the code is written. This can be solved by
executing a few NOPs after generating the code.
(b) If you have separate I/D cache, you may have to flush (part of) the
I cache in order to see the modified code. This doesn't arise on
386/486, does it? What about 586?
> I must say I am not sold on Java yet. Scheme remains my
> favorite prototyping language.
I'm hoping that the GNU scripting language (which will be a Scheme)
will come along and displace perl and tk/tcl, but I'm not holding my
breath. At least Java exists, though Sun's experience with NeWS
(which also encouraged downloading small programs [in postscript])
isn't especially encouraging.
-- Richard
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199503301321.OAA05112>