3.9 and the Full Closure Compiler
Marcus Denker
denker at iam.unibe.ch
Thu Sep 1 09:21:10 UTC 2005
Am 01.09.2005 um 08:32 schrieb Brent Pinkney:
> Hi,
>
> Does anyone else think the Full Closure Compiler is a priority for
> 3.9 ?
>
For me, yes. I'm slowly working on it, looks quite promising. The
plan is to
have the new compiler optionally available quite soon.
Then this needs a quite hard testing phase, as it's a real critical
part of the infrastructure.
This compiler can create both old-style blocks and closures, and it's
easy to turn
it off completely and just use the old one.
I'm now working on getting all the infrastructure into 3.9 (merge the
overrrides, add the code
needed for the closures-runtime code...). When that's done, the
newcompiler will
be a module: drop-in, easy to load and unload, no patch needed.
It should be made clear that the closure runtime would only be used
when compiling
code with the new compiler in closure-mode, so this change will have
no impact at
all for performance or stability.
> Traits and what-not are nice but a solidly engineered Comnpiler is
> IMHO more important.
The new compiler is quite nice... nothing is perfect, of course, but
it's a promising starting point.
One thing that it does quite well is to enable experiments: The
backend is reusable for other
compilers and makes it very easy to muck around with bytecode (see
ByteSurgeon), the RBAST
is a good startingpoint for making lots of stuff easier and cleaner,
and the SmaCC based Parser
is easy to change (if you get to now how to use SmaCC), and so on.
Marcus
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.squeakfoundation.org/pipermail/squeak-dev/attachments/20050901/14c78019/attachment.htm
More information about the Squeak-dev
mailing list
|