Bug 487981

Summary: CUDA support missing
Product: [Fedora] Fedora Reporter: Edmond <ymedhui>
Component: boinc-clientAssignee: Milos Jakubicek <xjakub>
Status: CLOSED WONTFIX QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: medium Docs Contact:
Priority: low    
Version: 9CC: cheekyboinc, mmahut, xjakub
Target Milestone: ---   
Target Release: ---   
Hardware: x86_64   
OS: Linux   
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2009-03-06 15:38:59 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:

Description Edmond 2009-03-02 00:34:58 UTC
Description of problem:
In the lastest version, CUDA support is missing (require libcudart.so).

Version-Release number of selected component (if applicable):

How reproducible:

Steps to Reproduce:
1. start boinc /etc/init.d/boinc-client start
2. connect to core using boincmgr
3. observed "No CUDA device found" in message
Actual results:
message "No CUDA device found"

Expected results:
messgae "CUDA device found" and GPUGrid app is schedule and run

Additional info:
I was able to fix it with these steps:
1) copy libcudart.so from GPUGrid version of boinc-client to /var/log/boinc/
2) change BOINCUSER=root in /etc/init.d/boinc-client

Comment 1 Edmond 2009-03-02 09:23:12 UTC
Also one more step to make CUDA work:

3) put a kludge in /etc/init.d/boinc-client, just before starting boinc-client

        mknod -m 0666 /dev/nvidiactl c 195 255 &>/dev/null
        mknod -m 0666 /dev/nvidia0 c 195 0 &>/dev/null

Comment 2 Milos Jakubicek 2009-03-06 15:38:59 UTC
I'm sorry, but Nvidia's CUDA (including the libcudart.so) is not open source and, as such, will never be part of Fedora -- I've just clarified this with Fedora Legal team members. However, I'll consider packaging CUDA in rpmfusion -- but that's another story.