Subject: Re: gcc4.0
To: None <tech-kern@netbsd.org>
From: Christos Zoulas <christos@astron.com>
List: tech-kern
Date: 11/02/2005 23:31:17
In article <20051102201333.GA6446@blackmilk.localzone.fr>,
fab  <fab@gnux.info> wrote:
>> >My personal feeling is that we should wait for 4.1 before switching, but 
>> >that's mainly down to my paranoia about .0 releases.
>> 
>> Several months ago, I tested a pre-release against building the kernel. 
>> At first, it did not work. Then it did. There was a bug, though, that I 
>> reported and that was corrected and several syntax modifications.
>> 
>> But, more important, I wasn't able to really see a difference between a 
>> kernel built by GCC 3.3 and by GCC 4.0. There is some fuss about the new 
>> optimizing feature, but I remain a bit sceptical, until someone 
>> definitely proves it really goes faster (on a i386 machine). I did not 
>> do extensive testing though, so my advice is not worth a penny.
>
>AFAIK 4.0 was the first with a support for the new "core". The next (4.1
>probably) will be focused on optimisation, 4.0 is just for testing the
>new core and permit developpers to port the code without waiting a
>version with real advantages.
>I've seen some kind of benchmark that show C++ is faster with 4.0, maybe
>this is why people think 4.0 is faster.

The biggest obstacle is that the gcc-4.0 build uses gmake features and it
is difficult to make it build with another make variant. We'll either have
to add gmake features to our make, or change the build process.

christos