cancel
Showing results for 
Search instead for 
Did you mean: 

Archives Discussions

pisoir
Journeyman III

strange behaviour in for_loop sample in Brook+ 1.3.0

In tutorial sample "for_loop" in Brook+ 1.3.0, if I change the width and height of the 2D buffer to bigger values (height,width = 2048,2048), the monitor goes black for quite long time. I have ATI Radeon 4850, 512 MB. Maybe there is only some memory overflow, but the behaviour is kind of scary.
0 Likes
6 Replies
jean-claude
Journeyman III

Have a look at your system timer watch-dog parameter.

This may help!

0 Likes

I forgot to say that not only everything goes black, the problem is also, that the graphic card restarts itself (I always get the "report crash" pop-up window), and also I get no results after computation (just zeros). Do you think, that this has something to do with the timer? Sorry for a stupid question, but how can I change the watch-dog timer? Thanks.

0 Likes

0 Likes

Hi Pisoir

 
I went to similar issues a couple of months ago. My recommendation:


(1)   Just make sure than having increased the size of your computation, doesn’t lead to a watchdog situation where the OS is worrying about the GPU not responding and so merely decides to reset it.

In this case, just expand the timer duration appropriately.

If you’re under Vista, have a look at http://forums.amd.com/devforum/messageview.cfm?catid=328&threadid=100142&highlight_key=y


(2) If (1) is not fixing the issue, then I’m afraid you’ve entered into a situation I met where a multi-loop in Brook generates crashes and blue-screen (see http://forums.amd.com/devforum/messageview.cfm?catid=328&threadid=104937&highlight_key=y )


At this point I’m still expecting some feedback from AMD developers about this bug.

 
Being stuck on (2) I moved from Brook+ to CAL and as of today I’m barely seeing this phenomenon happening (even if from time to time it shows up and crashes the PC).


Good luck and let us know how it worked out for you


Jean-Claude

0 Likes

Hi Pisoir

 
I went to similar issues a couple of months ago. My recommendation:


(1)   Just make sure than having increased the size of your computation, doesn’t lead to a watchdog situation where the OS is worrying about the GPU not responding and so merely decides to reset it.

In this case, just expand the timer duration appropriately.

If you’re under Vista, have a look at http://forums.amd.com/devforum/messageview.cfm?catid=328&threadid=100142&highlight_key=y


(2) If (1) is not fixing the issue, then I’m afraid you’ve entered into a situation I met where a multi-loop in Brook generates crashes and blue-screen (see http://forums.amd.com/devforum/messageview.cfm?catid=328&threadid=104937&highlight_key=y )


At this point I’m still expecting some feedback from AMD developers about this bug.

 
Being stuck on (2) I moved from Brook+ to CAL and as of today I’m barely seeing this phenomenon happening (even if from time to time it shows up and crashes the PC).


Good luck and let us know how it worked out for you


Jean-Claude

0 Likes

Thank you for your advice. I have Win XP, so I'm not sure if the proposed solution for vista also works for win XP. I tried it anyway (add TdrLevel in win registry), but with no success. I couldn't find any explicit solution for win xp, so if you know how to set this timer in win xp, please let me know. Thank you once more.

0 Likes