Archives Discussions

[douglas125]

clAcquireGLObjects seems to take more time than it should on ATI cards

Hi;

I'm using Win7 and Radeon 5770 to do some drawings using the CLGL interop. I have a question, however. The kernel execution time is very low but the clAcquire/ReleaseGLObjects takes a long time and even freezes the application.

When I run the same code using a Quadro 5800 card and WinXP 64 bits, there's no such thing, the Acquire/ReleaseGLObjects runs in no time (practically).

So I'm given to assume that this has to do with either my GPU or the current OpenCL implementation, correct? How should I solve the problem?

If you want the source code for a case study, here is the link:

http://www.cmsoft.com.br/tempFiles/CLGLSierpinskiGasket.zip

Thanks

[nou]

check if you create all GL object that you want share AFTER creation of CL context. otherwise they are not shared.

[douglas125]

No, it's not the case that the objects are not shared. The code works just fine. It's just that acquiring/releasing the objects takes too long in AMD. Strangely enough it's really fast using Quadro.

[nou]

try move OpenGL objects AFTER creation of OpenCL context.

this is important on AMD

EDIT:i timed SimpleGL example. and whole OpenCL part take few ms to execute. using SDK 2.3 and Catalyst 10.12