Two GPUs, two purposes - libgl and nvidia-utils conflict - lib32-libgl

Question

Hi all!
After realizing there is no way (by now) to perform a vgs switch (between integrated Intel and discrete Nvidia GPUs) using nvidia propertary driver (as I read, the only one supporting gpgpu with cuda/openCL), I wondered if there is a way of letting them co-exist: graphics on the integrated one and applications on the nvidia. After all, I might not need to be using the discrete card for the video if I want to use it for other purposes.
Obeviously, the tho drivers conficlt (nvidia-utils conflicts with libgl), but in this post a workaround is suggested.
Thing is, also lib32-nvidia-utils and lib32-libgl (Skype depends on it) conflict, so I thought I could just extend the suggestion, but I couldn't find lib32-libgl's pkgbuild. Where can I find it? If it doesn't exist at all, is there a way I can workaround?
By the way, why aren't libgl and lib32-libgl on abs? In what way are they different from the others?
There is libgl-git in AUR, but I couldn't find the 32-bit version.
Do you think I have a chanche it'll work?
Thanks a lot
Davide
running Arch linux 64 bit
on a Dell Vostro 3500
$ lspci|grep VGA
00:02.0 VGA compatible controller: Intel Corporation Core Processor Integrated Graphics Controller (rev 18)
01:00.0 VGA compatible controller: nVidia Corporation GT218 [GeForce 310M] (rev a2)

Answer 1

Xehoz wrote:
davide wrote:
ngoonee wrote:Using both at the same time? No, of course not.
So you think it'll never work :sad:
Actually, it is working now. It's a recent development posted in the mailing lists, but the solution came up yesterday and tests have been positive so far. It's confirmed to work in at least two Optimus machines.
Progress is being posted in the mailing lists, but the author, Martin Juhl, is also blogging it. I'll just sum up what the solution: It resorts to turbojpeg  VirtualGL and two xorgs. The intel card does everything, and apps meant to use the nvidia are specifically run with vglrun.
Unfortunately, the solution removes hardware acceleration in the intel card (immediately noticeable by the absence of desktop effects). In arch that translates the libgl / nvidia-utils conflict.
intel-dri demands libgl.
xf86-video-intel demands intel-dri.
In order to install nvidia + nvidia-utils in Arch, without removing intel-dri and xf86-video-intel, you have to:
# pacman -Sd nvidia nvidia-utils
This will still remove libgl, though.
Now THAT is just cool... would make optimus laptops so much more useful if/when it works reasonably well. Sounds like quite a bit of hackery involved though.