Re: armelfp: new architecture name for an armel variant
+++ Konstantinos Margaritis [2010-07-08 15:38 +0300]:
> Even now, libc arch specific optimizations -like libc6-i686
> that you mentioned- are undocumented, very few packages actually provide
> support for them, and in short, software ends up totally unoptimized for no
> good reason.
That's too harsh. There is a very good reason for the current armel
arch build settings - maximum coverage. Debian has always picked the
most general settings that are still useful to a reasonable number of
users. So it's always rather unoptimised for newer hardware. In the
past that hasn't really made much difference. If your 30% really is
across the board then I agree that's a big deal. But how much of that
do we get from softfp +VFP as opposed to just soft (emulation)? And
what proportion of the archive really is affected. Everything a bit,
or largely a few packages wher it really hurts (e.g. mesa, libm).
> It doesn't have to be official right from the start, we just have to decide
> names and triplets. I can do the building here and provide a working port as I
> have the cpu power
This is very true of course, and negates all the 'it's too much work'
arguments. If you are prepared to do the work, and do it in a
compatible manner for easy adoption later then that is commendable and
no-one should be getting in your way.
> Would you suggest the changes necessary to dpkg regardless? Which files should
> be changed?
ostable,cputable,triplettable provide all the mappings. I don't think
you have to change anything else. It is quite important not to make
the mess in there any worse than we have to though...
Principal hats: Linaro, Emdebian, Wookware, Balloonboard, ARM