Multilib fun and games. I'll need to take another look at that tomorrow before I write the howto. For now, edit the script and let's do this in two passes, the x86_64 stuff and then the i586 stuff. So basically, copy the script so you have backup. Remove the lines that have 'i586' in the package names and run it. It should look like this. Code: #!/bin/bash sudo yum --nogpgcheck localinstall \ buildsys-build-rpmfusion-11-0.17.x86_64.rpm \ buildsys-build-rpmfusion-kerneldevpkgs-current-11-0.17.x86_64.rpm \ akmod-rt2870-2.1.2.0-1.fc11.9.x86_64.rpm \ kmod-rt2870-2.1.2.0-1.fc11.9.x86_64.rpm \ kmod-rt2870-2.6.29.6-217.2.8.fc11.x86_64-2.1.2.0-1.fc11.9.x86_64.rpm \ akmod-nvidia-180.60-2.fc11.x86_64.rpm \ kmod-nvidia-2.6.29.6-217.2.8.fc11.x86_64-180.60-2.fc11.x86_64.rpm \ kmod-nvidia-180.60-2.fc11.x86_64.rpm \ nvidia-cuda-toolkit-2.1-8.fc11.x86_64.rpm \ nvidia-cuda-toolkit-devel-2.1-8.fc11.x86_64.rpm \ nvidia-cuda-toolkit-libs-2.1-8.fc11.x86_64.rpm \ xorg-x11-drv-nvidia-180.60-1.fc11.x86_64.rpm \ xorg-x11-drv-nvidia-devel-180.60-1.fc11.x86_64.rpm \ xorg-x11-drv-nvidia-libs-180.60-1.fc11.x86_64.rpm Now run this from the cmd line in the fedora11 directory. (Highlight all the lines with your mouse and then paste onto cmd line with middle mouse button and hit enter.) Code: sudo rpm -ihv \ nvidia-cuda-toolkit-devel-2.1-8.fc11.i586.rpm \ nvidia-cuda-toolkit-libs-2.1-8.fc11.i586.rpm \ wine-cuda-runtime-2.1-2.fc11.i586.rpm \ xorg-x11-drv-nvidia-devel-180.60-1.fc11.i586.rpm \ xorg-x11-drv-nvidia-libs-180.60-1.fc11.i586.rpm
Assuming CUDA is installed OK and './deviceQuery' shows a CUDA device after you reboot. GPU1.tar.gz. Download, extract with 'tar zxvf GPU1.tar.gz', 'cd GPU1'. Contains the folding exe for the GPU2 client and a start.sh script. First time you run, './start.sh -configonly'. Then './start.sh'. Contents of start.sh ..... Code: #!/bin/bash # export WINEDEBUG=warn+cuda,+seh # export FAH_GPU_IDLE=0 nice -n 15 wine Folding@home-Win32-GPU.exe -verbosity 9 -gpu 0 -forcegpu nvidia_g80 $*
I done what you said 2 up by doing the install in 2 halves and it worked without erroring, I rebooted but when I type ,/deviceQuery I get this still: [rob@FEDORA64 fedora11]$ ./deviceQuery ./deviceQuery: error while loading shared libraries: libcudart.so.2: cannot open shared object file: No such file or directory
Code: clivem@c7p6t:~/bin]$ ls -l /usr/lib/nvidia/libcudart.so.2 lrwxrwxrwx. 1 root root 16 2009-08-18 21:20 /usr/lib/nvidia/libcudart.so.2 -> libcudart.so.2.1 Should be a symbolic link to libcudart.so.2.1. Let's check it got installed. Code: [clivem@c7p6t:~/bin]$ rpm -q nvidia-cuda-toolkit-libs-2.1-8.fc11.i586 nvidia-cuda-toolkit-libs-2.1-8.fc11.i586 [clivem@c7p6t:~/bin]$ rpm -qf /usr/lib/nvidia/libcudart.so.2 nvidia-cuda-toolkit-libs-2.1-8.fc11.i586 If it did get installed, try .... Code: sudo ldconfig Does that fix it?
Code: sudo rpm -ihv \ nvidia-cuda-toolkit-devel-2.1-8.fc11.i586.rpm \ nvidia-cuda-toolkit-libs-2.1-8.fc11.i586.rpm \ wine-cuda-runtime-2.1-2.fc11.i586.rpm \ xorg-x11-drv-nvidia-devel-180.60-1.fc11.i586.rpm \ xorg-x11-drv-nvidia-libs-180.60-1.fc11.i586.rpm OK turns out ive been a spanner and didnt do the above bit properly. I just done it properly but got this error: [rob@FEDORA64 fedora11]$ sudo rpm -ihv \ > nvidia-cuda-toolkit-devel-2.1-8.fc11.i586.rpm \ > nvidia-cuda-toolkit-libs-2.1-8.fc11.i586.rpm \ > wine-cuda-runtime-2.1-2.fc11.i586.rpm \ > xorg-x11-drv-nvidia-devel-180.60-1.fc11.i586.rpm \ > xorg-x11-drv-nvidia-libs-180.60-1.fc11.i586.rpm error: Failed dependencies: xorg-x11-drv-nvidia = 180.60-1.fc11 is needed by xorg-x11-drv-nvidia-libs-180.60-1.fc11.i586 So no the libcudart.so.2.1 didnt get installed
Well, that doesn't make much sense. The 'yum --nogpgcheck localinstall ...' should have installed xorg-x11-drv-nvidia. Code: [clivem@c7p6t:~/bin]$ rpm -q xorg-x11-drv-nvidia xorg-x11-drv-nvidia-180.60-1.fc11.x86_64 [clivem@c7p6t:~/bin]$ rpm -qa | grep nvidia | sort -d akmod-nvidia-180.60-2.fc11.x86_64 kmod-nvidia-2.6.29.6-217.2.8.fc11.x86_64-180.60-2.fc11.x86_64 nvidia-cuda-toolkit-2.1-8.fc11.x86_64 nvidia-cuda-toolkit-devel-2.1-8.fc11.i586 nvidia-cuda-toolkit-devel-2.1-8.fc11.x86_64 nvidia-cuda-toolkit-libs-2.1-8.fc11.i586 nvidia-cuda-toolkit-libs-2.1-8.fc11.x86_64 xorg-x11-drv-nvidia-180.60-1.fc11.x86_64 xorg-x11-drv-nvidia-devel-180.60-1.fc11.i586 xorg-x11-drv-nvidia-devel-180.60-1.fc11.x86_64 xorg-x11-drv-nvidia-libs-180.60-1.fc11.i586 xorg-x11-drv-nvidia-libs-180.60-1.fc11.x86_64 Would help if you can show me which nvidia packages did get installed by running the second cmd above.
Errrrrr when I type rpm -qa | grep nvidia | sort -d nothing happens and I just get the prompt back: [rob@FEDORA64 bin]$ rpm -qa | grep nvidia | sort -d [rob@FEDORA64 bin]$ Arrrggghhhhhh sorry mate this must be really annoying
I reinstalled the SMP client again from scratch, double checked everything and im still getting the same problem as yesterday where it sits on that screen NNODES=8, MYRANK=1 etc etc, ive left it there for an hour now and nothings happened Im think about wiping it and starting again? When I first installed Fedora 64 I ran the software update program which downloaded over a gig of stuff, maybe programs that were downloaded and installed then are effecting things? Cant think why the SMP client doesnt work, I cant really do anything different to it lol
That makes even less sense. It would imply the following did not complete successfully. Try running it again. Code: sudo yum --nogpgcheck localinstall \ buildsys-build-rpmfusion-11-0.17.x86_64.rpm \ buildsys-build-rpmfusion-kerneldevpkgs-current-11-0.17.x86_64.rpm \ akmod-rt2870-2.1.2.0-1.fc11.9.x86_64.rpm \ kmod-rt2870-2.1.2.0-1.fc11.9.x86_64.rpm \ kmod-rt2870-2.6.29.6-217.2.8.fc11.x86_64-2.1.2.0-1.fc11.9.x86_64.rpm \ akmod-nvidia-180.60-2.fc11.x86_64.rpm \ kmod-nvidia-2.6.29.6-217.2.8.fc11.x86_64-180.60-2.fc11.x86_64.rpm \ kmod-nvidia-180.60-2.fc11.x86_64.rpm \ nvidia-cuda-toolkit-2.1-8.fc11.x86_64.rpm \ nvidia-cuda-toolkit-devel-2.1-8.fc11.x86_64.rpm \ nvidia-cuda-toolkit-libs-2.1-8.fc11.x86_64.rpm \ xorg-x11-drv-nvidia-180.60-1.fc11.x86_64.rpm \ xorg-x11-drv-nvidia-devel-180.60-1.fc11.x86_64.rpm \ xorg-x11-drv-nvidia-libs-180.60-1.fc11.x86_64.rpm
Done it again [rob@FEDORA64 fedora11]$ sudo yum --nogpgcheck localinstall \ > buildsys-build-rpmfusion-11-0.17.x86_64.rpm \ > buildsys-build-rpmfusion-kerneldevpkgs-current-11-0.17.x86_64.rpm \ > akmod-rt2870-2.1.2.0-1.fc11.9.x86_64.rpm \ > kmod-rt2870-2.1.2.0-1.fc11.9.x86_64.rpm \ > kmod-rt2870-2.6.29.6-217.2.8.fc11.x86_64-2.1.2.0-1.fc11.9.x86_64.rpm \ > akmod-nvidia-180.60-2.fc11.x86_64.rpm \ > kmod-nvidia-2.6.29.6-217.2.8.fc11.x86_64-180.60-2.fc11.x86_64.rpm \ > kmod-nvidia-180.60-2.fc11.x86_64.rpm \ > nvidia-cuda-toolkit-2.1-8.fc11.x86_64.rpm \ > nvidia-cuda-toolkit-devel-2.1-8.fc11.x86_64.rpm \ > nvidia-cuda-toolkit-libs-2.1-8.fc11.x86_64.rpm \ > xorg-x11-drv-nvidia-180.60-1.fc11.x86_64.rpm \ > xorg-x11-drv-nvidia-devel-180.60-1.fc11.x86_64.rpm \ > xorg-x11-drv-nvidia-libs-180.60-1.fc11.x86_64.rpm Loaded plugins: refresh-packagekit Setting up Local Package Process Examining buildsys-build-rpmfusion-11-0.17.x86_64.rpm: 10:buildsys-build-rpmfusion-11-0.17.x86_64 buildsys-build-rpmfusion-11-0.17.x86_64.rpm: does not update installed package. Examining buildsys-build-rpmfusion-kerneldevpkgs-current-11-0.17.x86_64.rpm: 10:buildsys-build-rpmfusion-kerneldevpkgs-current-11-0.17.x86_64 buildsys-build-rpmfusion-kerneldevpkgs-current-11-0.17.x86_64.rpm: does not update installed package. Examining akmod-rt2870-2.1.2.0-1.fc11.9.x86_64.rpm: akmod-rt2870-2.1.2.0-1.fc11.9.x86_64 akmod-rt2870-2.1.2.0-1.fc11.9.x86_64.rpm: does not update installed package. Examining kmod-rt2870-2.1.2.0-1.fc11.9.x86_64.rpm: kmod-rt2870-2.1.2.0-1.fc11.9.x86_64 kmod-rt2870-2.1.2.0-1.fc11.9.x86_64.rpm: does not update installed package. Examining kmod-rt2870-2.6.29.6-217.2.8.fc11.x86_64-2.1.2.0-1.fc11.9.x86_64.rpm: kmod-rt2870-2.6.29.6-217.2.8.fc11.x86_64-2.1.2.0-1.fc11.9.x86_64 kmod-rt2870-2.6.29.6-217.2.8.fc11.x86_64-2.1.2.0-1.fc11.9.x86_64.rpm: does not update installed package. Examining akmod-nvidia-180.60-2.fc11.x86_64.rpm: akmod-nvidia-180.60-2.fc11.x86_64 Examining kmod-nvidia-2.6.29.6-217.2.8.fc11.x86_64-180.60-2.fc11.x86_64.rpm: kmod-nvidia-2.6.29.6-217.2.8.fc11.x86_64-180.60-2.fc11.x86_64 Examining kmod-nvidia-180.60-2.fc11.x86_64.rpm: kmod-nvidia-180.60-2.fc11.x86_64 Examining nvidia-cuda-toolkit-2.1-8.fc11.x86_64.rpm: nvidia-cuda-toolkit-2.1-8.fc11.x86_64 Examining nvidia-cuda-toolkit-devel-2.1-8.fc11.x86_64.rpm: nvidia-cuda-toolkit-devel-2.1-8.fc11.x86_64 Examining nvidia-cuda-toolkit-libs-2.1-8.fc11.x86_64.rpm: nvidia-cuda-toolkit-libs-2.1-8.fc11.x86_64 Examining xorg-x11-drv-nvidia-180.60-1.fc11.x86_64.rpm: xorg-x11-drv-nvidia-180.60-1.fc11.x86_64 Examining xorg-x11-drv-nvidia-devel-180.60-1.fc11.x86_64.rpm: xorg-x11-drv-nvidia-devel-180.60-1.fc11.x86_64 Examining xorg-x11-drv-nvidia-libs-180.60-1.fc11.x86_64.rpm: xorg-x11-drv-nvidia-libs-180.60-1.fc11.x86_64 Nothing to do [rob@FEDORA64 fedora11]$
The nvidia packages are not being installed. Probably the global exclude. For me to get this working 2 mins. For me to explain it to you and manually list all the deps so you don't need to use yum - 1 hour. Can you port forward 22 from your router to the Linux box, PM me the your external IP address, and username/password that your using? This usually takes less than 5 mins from the reboot following DVD install to folding on both CPU and GPU.
I got the SMP client working now, done a bit of googling and answer was here, I changed host name and it worked which I find strange! http://foldingforum.org/viewtopic.php?f=44&t=10349 YGPM....
OK, we're nearly there. GPU client is working. CPU client with bigadv is working. Wireless module loads and appears to be scanning. So hopefully, that'll configure without a fight and you can decide if you want to run headless, which I think you do. In which case we'll configure for runlevel 3, create the device nodes for the graphics card from rc.local at startup (cause they won't get created by starting an X server) and decide if you want a script to start the clients, or do as I do, use scripts from screen to start them so you can either re-attach from a vnc session on a Window$ machine or ssh in from anywhere. Tomorrow is another day!
After much piddling around yesterday, a lesson is learnt. If you're gonna do it, do it properly! Use a repo, (yum localinstall sucks), don't mess with tying to exclude packages from yum, don't use epochs on packages, and make sure the dependency resolution works! It could have been so much easier........ Code: sudo rpm -Uvh http://rpm.livna.org/livna-release.rpm\ http://download1.rpmfusion.org/free/fedora/rpmfusion-free-release-stable.noarch.rpm\ http://download1.rpmfusion.org/nonfree/fedora/rpmfusion-nonfree-release-stable.noarch.rpm\ http://www.vacuumtube.org.uk/folding/fedora/11/x86_64/folding-release-1-1.noarch.rpm sudo yum install wine-cuda-runtime akmod-nvidia-180xx All dependencies for GPU folding on Fedora 11 resolved and installed. Reboot. Job done! Set up your clients and away you go. Optionally install FahMon. Code: sudo yum install FahMon Rob: can I ssh in tonight and clean up what we did yesterday and switch you over to the repo?
Nice one Clive Yeah no problem mate, machine is off at the mo but will fire it up when home so we can finish it off etc... Are you on MSN or anything? Or even facebook chat?
