Re: mesa: Changes to 'debian-experimental'
On Son, 2011-06-19 at 18:28 +0000, Cyril Brulebois wrote:
>
> commit e3e4cde22d4c6eedaf183103e22fca9b63fefa6c
> Author: Cyril Brulebois <kibi@debian.org>
> Date: Sun Jun 19 16:06:28 2011 +0200
>
> Restrict r300g to amd64 and i386.
>
> diff --git a/debian/changelog b/debian/changelog
> index e2c956b..53955c3 100644
> --- a/debian/changelog
> +++ b/debian/changelog
> @@ -34,7 +34,10 @@ mesa (7.11~0-1) UNRELEASED; urgency=low
> * Replace radeon with r300 in GALLIUM_DRIVERS to match upstream renaming
> (5722286be2).
> * Build r300 classic (through DRI_DRIVERS) everywhere, since r300g comes
> - with a few additional requirements.
> + with a few additional requirements: LLVM is needed for this driver,
> + and apparently only works fine on x86 platform. As a consequence, only
> + build r300g on amd and i386, and add llvm-2.9-dev build-dep on those
> + platforms accordingly. Disable it explicitly on other platforms.
> * Stop building i915g at all, it's apparently never going to be a
> suitable replacement for i915c.
This is wrong. r300g works fine without LLVM. The reason for the
dependency is that without LLVM, Gallium's software vertex processing is
very slow. But this only affects integrated GPUs without vertex shaders,
which are only available for x86 CPUs.
> commit 7d332507bbffa1fdb59723808ef39b9e50983e16
> Author: Cyril Brulebois <kibi@debian.org>
> Date: Sun Jun 19 15:37:18 2011 +0200
>
> Stop building i915g at all.
>
> It's apparently never going to be a suitable replacement for i915c.
According to whom? My impression from the sideline has been that i915g
has been improving quite rapidly recently and already performs better
than i915c in some cases.
Or is it because i915g doesn't support some of the chipsets supported by
i915c (those without pixel shaders)?
--
Earthling Michel Dänzer | http://www.vmware.com
Libre software enthusiast | Debian, X and DRI developer
Reply to: