gcc news #8

[Mathieu Lacage <Mathieu dot Lacage at sophia dot inria dot fr>]

I would like to apologize for the current irregular releases which are
now more or less bi-monthly. I will do my best to stick to this
schedule.

As usual:
http://gccnews.chatta.us/

meta
----
Due to various personal events, I have been unable to keep the original
strictly weekly schedule for this newsletter. I do apologize for this
and will do my best to stay on track. If you enjoy reading this, you
might consider helping me a bit: a great way to do so would be to send
me links to the start of specific mailing list discussions you found
interesting. Although not currently covered (ie: I don't really read
them), the fortran, libstdc++, java and gcc-patches mailing lists might
be interesting candidates to start such an experiment.


gcc release
-----------
Mark Mitchell sent a status update on gcc 3.4.2 and 3.5. It looks like
3.42 is scheduled for september 1st while 3.5 targets early 2005. Stage
3 for 3.5 is planned for early september.

gcc development
---------------
What kind of optimizations are turned on or off at various optimization
levels is a recuring topic. This week, it came back when a user
complained (wrongly) about invalid code generated by gcc. Although
everyone agrees it would be nice for gcc to generate better code at -O0,
-O1 does not seem to be reasonable candidate to become -O0. It remains
to be seen if someone will do the work needed to add a few simple
optimization passes to -O0.

Although I do not always understand the details of the optimization
passes, Jeffrey A. Law sent a nice summary about the implementation
challenges associated with a new algorithm presented at the gcc summit.
His second and the third followup by Kenneth Zadeck provide some
interesting insight into how Jump threading is performed within the
compiler. Diego Novillo sent a nice summary about some background issues
related to this threading rework.

regards,
Mathieu
-- 
Mathieu Lacage <mathieu.lacage@sophia.inria.fr>