CUDA-on-andLinux install problems

It would make me really happy to get CUDA working under andLinux . But so far I’m stuck on the first step: installing the CUDA driver.

I’m running a near-virgin andLinux install (which approximates Ubuntu 7.10), except that I did the not-root-by-default fix suggested in the andLinux wiki. It’s running on top of WinXP 32bit. Hardware includes an 8800GTS G80 640MB.

When I switch to root and run NVIDIA-Linux-x86-177.67-pkg1.run , I get

[root@andLinux andLinux]# ./NVIDIA-Linux-x86-177.67-pkg1.run
Verifying archive integrity… OK
Uncompressing NVIDIA Accelerated Graphics Driver for Linux-x86 177.67……
Received signal SIGSEGV; aborting.

Between the last dot and the SIGSEGV, some other screen flashes briefly, perhaps with a grey strip near the bottom. I’m not sure.

Maybe in a few days I’ll try booting into my non-andLinux partition, which is very similar, and see how that goes. But I’d really like to get it working under andLinux, to lower my barrier to post-work coding.

Any advice?

I doubt that this configuration will work. Most hardware drivers need direct access to the hardware. Virtual machines (such as what you’re running) abstract away the real hardware.

If you want to run Ubuntu-7.10, then you’ll need to install it as a standalone OS, not inside of a virtual machine.

andLinux (and its core, colinux) aren’t really running in a virtual machine. I admit I don’t understand the details as clearly as I’d like to, but that’s what they tell me! =]

“Virtual machine” or not, you may be right about it being infeasible. But it seems like I should be able to at a minimum get the simulator running under andLinux, even if I have to reboot into standalone Ubuntu to run it against my physical GPU hardware.

Couldn’t a Xen-domU with a patched NVIDIA-driver work, too? AFAIK Xen has a feature to pass devices to a client OS. Of course you cannot use Win XP as a dom0.