cancel
Turn on suggestions
Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.
Showing results for 
Search instead for 
Did you mean: 

Archives Discussions

carl_morris
Journeyman III
‎09-18-2008 06:57 PM

Radeon HD 4870 and Catalyst 8.9, is Fedora 7 with CAL possible?

Trying to get 4870 working with Fedora 7 and CAL

Is it possible to use Fedora 7 to host a 4870 card for CAL evaluation?  I've got the card installed.  I've got Catalyst 8.9 installed.  I can see the fglrx module is loaded with lsmod.  The card produces video output.  But when I try to run:

    /usr/local/amdcal/samples/tutorial/OpenCloseDevice/OpenCloseDevice

I get the following:

    Opening Device : 0

        There was an error opening device 0.

Suggestions?

    -- Carl

0 Likes
4 Replies
carl_morris
Journeyman III
‎09-18-2008 07:37 PM

Another clue...

After "aticonfig --init" the system hung when it tried to start X.  The xorg.conf that ended up working after the debug process is for a generic VESA device.  Maybe fglrx and X aren't playing nice?

    -- Carl

0 Likes
carl_morris
In response to carl_morris
Journeyman III
‎09-19-2008 08:00 PM

Here's what worked for me.  Upgrade to Fedora 8, then do "yum update" and reboot until nothing else needs to be updated.  First update was 20 or so packages, second update was almost 500 packages.  Then go through the install of Catalyst, CAL, and Brook+ and you'll be able to run examples.  You will get "locking assertion failure" messages, but they are benign.

    -- Carl

0 Likes
jfkong
In response to carl_morris
Journeyman III
‎09-30-2008 11:16 PM

hey, thanks for your info. I did exactly what you said.  It seems to work.  But that "locking assertion failure" is really annoying.   Any one knows how to fix this?

0 Likes
carl_morris
In response to jfkong
Journeyman III
‎10-01-2008 07:23 PM

From what I read, a locking mechanism in lower-level library routines is being used incorrectly by higher-level code, causing a warning to be generated.  Older versions of the library just ignored the misuse, but the library developers needed to make the problems visible so higher-level code developers could fix them.  I believe there used to be a #define in the library code or an environment variable you could set to disable the output of the warning message, but that seems to have been removed... probably to encourage fixing the logic problems!   

It's not clear to me if the higher-level code that needs to be fixed in our case belongs to AMD or something they rely on to build their examples.  Anyone know?

    -- Carl

0 Likes