Subject: Re: DRAFT toolchain upgrade plan
To: Perry E. Metzger <perry@wasabisystems.com>
From: None <cgd@broadcom.com>
List: tech-toolchain
Date: 05/28/2002 14:33:03
At 28 May 2002 17:05:59 -0400, Perry E. Metzger wrote:
> Or do you think we really need a newer toolchain soon?

Well, depends on what users want, really.

3.1 has a bunch of features that are "desirable," over and above what
are provided by 2.95.x...

My assumption is that not going to 3.0.x is a no-brainer.  But I'm a
fan of going to 3.1.

(A few things that pop into mind: 3.0 has a bunch of stuff which would
be useful for kernel code, e.g. the __builtin_expect functionality to
helpe w/ branch prediction.  3.0 also included the new c++ library,
which is a fairly huge change and has a new ABI.  3.1 brings
"stability for the previous," plus I'm a fan of the prefetch builtin
support in 3.1, too...)


If the plan is for the next release to roll out a bunch of new system
features, I'd say 3.1 would be a good thing to include in it.



cgd