Need a better way of doing remote OpenCL computation
Hi,
I have a ATI HD 5970 on CentOS 5.5 (x64_64 and up-to-date) on which I installed ATI SDK v2.3 and Catalyst 11.2. OpenCL works from the console. I need it remotely, without actually logging in at the server. I have tried remote display, verified opening that it was working with obtaining xterm from the server on to my local system. But FindNumDevices shows zero devices! This remains so after following ATI's instruction as given in http://developer.amd.com/gpu_assets/App_Note-Running_ATI_Stream_Apps_Remotely.pdf.
After spending hours I was able to find a trick through which I could see the devices. That was to login to the server from my local client and then login from the serer to server itself! This of course is not the best way to use it. I was able to actually get the OpenCL working earlier before I did a reinstall directly. I tried to fidn the same source again to help but could not. All I remember is that the site mentioned changing fglrx in such a way that remote access will be assumed not to have ATI card or something to that effect.
Could anyone suggest a better way to get remote access working for OpenCL please?
TIA,
-S
PS: Logging from server to server itself works from Windows as well! (I have tried with putty.)
[user@localhost ~]$ ssh ati Last login: Sun Mar 20 10:25:46 2011 from ati.iiita.ac.in [user@ati ~]$ FindNumDevices Supported CAL Runtime Version: 1.3.185 Found CAL Runtime Version: 1.4.1016 Use -? for help CAL initialized. Finding out number of devices :- Device Count = 0 CAL shutdown successful. Press enter to exit... [user@ati ~]$ ssh ati user@ati's password: Last login: Sun Mar 20 10:27:58 2011 from office.iiita.ac.in [user@ati ~]$ FindNumDevices Supported CAL Runtime Version: 1.3.185 Found CAL Runtime Version: 1.4.1016 Use -? for help CAL initialized. Finding out number of devices :- Device Count = 2 CAL shutdown successful. Press enter to exit... [user@ati ~]$ cd /usr/local/ati-stream-sdk-v2.3-lnx64/samples/opencl/bin/x86_64/ [user@ati x86_64]$ ./MatrixMulDouble --device gpu inputA 0.468961 3.29865 7.22117 1.76034 8.3174 5.53392 1.41339 7.70584 1.7714 9.50188 8.73543 2.54885 8.07439 8.41388 7.41171 9.2991 2.34384 4.29576 3.20228 10.5797 7.99043 5.40295 1.04537 9.11623 0.764542 6.49287 5.24888 9.08263 10.6554 9.28057 6.60218 0.124327 1.57922 2.82335 1.88467 9.89663 8.35727 3.29806 6.60247 10.1287 1.79994 4.3379 1.67753 9.87433 1.75179 9.08924 8.17342 4.09563 2.385 0.375705 3.67536 10.3754 5.77865 4.72073 8.49166 6.54319 0.213596 2.74054 4.62583 10.869 1.02111 0.228005 10.9933 2.60034 inputB 0.468961 3.29865 7.22117 1.76034 8.3174 5.53392 1.41339 7.70584 1.7714 9.50188 8.73543 2.54885 8.07439 8.41388 7.41171 9.2991 2.34384 4.29576 3.20228 10.5797 7.99043 5.40295 1.04537 9.11623 0.764542 6.49287 5.24888 9.08263 10.6554 9.28057 6.60218 0.124327 1.57922 2.82335 1.88467 9.89663 8.35727 3.29806 6.60247 10.1287 1.79994 4.3379 1.67753 9.87433 1.75179 9.08924 8.17342 4.09563 2.385 0.375705 3.67536 10.3754 5.77865 4.72073 8.49166 6.54319 0.213596 2.74054 4.62583 10.869 1.02111 0.228005 10.9933 2.60034 Platform :Advanced Micro Devices, Inc. Device 0 : Cypress Device 1 : Cypress Executing kernel for 1 iterations ------------------------------------------- Output 1896.97 1582.6 1829.29 1820.91 2045.8 1516.58 2039.15 1825.62 1784.83 1749.03 1793.08 1716.81 1998.3 1948.26 2050.13 2042.94 1822.02 2103.93 2158.2 1819.27 1997.26 1755.34 1828.73 1692.75 1810.57 1875.35 1816.79 1806.83 2154.7 1996.99 2025.09 1720.31 1470.83 1991.42 1729.9 1993.12 1849.63 1987.74 1971.02 1716.79 1813.25 1899.83 1780.76 2051.36 1918.26 1750.86 1856.36 2011.21 1892.25 1699.33 1908.12 1904.67 1831.24 1907.13 2025.19 1620.93 1856.91 1654.13 1735.45 2122.66 1839.02 2017.7 2090.03 1809.83 [user@ati x86_64]$
