Hi all,
I am using SDK 2.4 and Visual Studio 2010. My kernel runs fine in debug mode but I am getting access violations in write when I run the code in release mode.Any idea why this would happen only in release mode?
I have put printf statements in the kernel and kernel fails for some thread id > 0. Is it possible to put assert statements in the kernel and break the kernel execution when assertion is violated?
How can I debug it? The problem is whenever violation occours, 90% of the time my system just freezes down and I have to reboot. So debugging has become a pain in this case.
Any help would be appreciated.
Regards,
Richeek
Try using GDB to debug it in CPU mode.
Does the samples also fail in release mode?
Also mention your system details:CPU,GPU,SDK,CAtalyst,OS.
Originally posted by: himanshu.gautam Try using GDB to debug it in CPU mode.
Does the samples also fail in release mode?
Also mention your system details:CPU,GPU,SDK,CAtalyst,OS.
No, samples are passing. My kernel code fails but it is not repetitive. Sometime it fails but when I re-run the with exactly the same settings it passes. It shows access violation in writing. I check the call stack and aticaldd.dll seems to be crashing. My scratch register usage is > 0, could it be the cause of crash? I found out a old thread having similar issues but it seems that this bug has been taken care of now... so I am not sure why I have intermittant failure problem.
http://forums.amd.com/devforum/messageview.cfm?catid=390&threadid=135108&highlight_key=y
My system details are as follows:
AMD Athlon II Quad Core
ATI Radeon 5450 Graphics Card
Windows 7 64 bit
SDK v2.4
Catalyst Version: 11.5
I am running my simulation in 32 bit mode.
Regards,
Richeek
aticaldd.dll fails on this instuction (if this helps)
6914C858 mov dword ptr [eax],edx
Video drive crashes too when I get access violation...
richeek,
I suggest you to post some test case here or file a ticket and attch you code there.