I resume blogging after a long time and after moving the blog everywhere..I am back to blogger.
It has been a week without classes and among other things I have been working with, I tried my hand at figuring out how to work with Nvidia CUDA. The supported ubuntu 8.04 was a major blow. Installation on ARCH linux was a breeze. Ended up setting up a pseudo cluster with 2 nodes and 2 kick a** GPUs.
1) Install Archlinux
2) Install nvidia drivers in the repo ( Which are new unlike ubuntu)
3) Install CUDA with yaourt
4) Install Open MPI with yaourt
5) NFS mount the dir on the other machine
6) Enable key based login on ssh and do a ssh-keygen in your home ( Note the NFS mounting before)
7) cat .ssh/id_rsa.pub > .ssh/authorized_keys
8) Done :)
Hey Praneeth, how is it going buddy. So you're using CUDA too..great. What gpu and application?
I tried to install CUDA via AUR. Unfortunately nividia-beta is an dependence and nvidia-beta/nividia is in conflict with libgl. How did you solve this problem.
By the way: I just want use cuda in emulation mode because I don't have an nvidia adapter in my laptop.
Is it possible in virtual box.
After installing the drivers on Linux,
I was trying to find an easy way to install drivers on Windows, after all, my father's computer is
still on Windows, this can't be avoided. So after searching the forums, I found NVidia driver http://driverscenter.com/manufacturers/nvidia I've got an update and shouldn't need any patch anymore and use the Nvidia driver instead. And Praneeth, thank you,
this information helped me a lot!
Post a Comment