Is there a known bug with glMemoryBarrier() when using the latest AMD Catalyst driver (14.7)?

Started by
15 comments, last by davychen 9 years, 5 months ago

Hmm, its really strange. Driver version 14.9.2 beta says that cl_khr_gl_event is supportet, but when I try to enable it in the kernel with the following line of code the compilation of the kernel fails saying that its an unrecognized extension. Any other supported extension works fine. Even if I try to enable all supported extensions the compilation fails.

#pragma OPENCL EXTENSION cl_khr_gl_event : enable

Advertisement

Same happens to me here on 280x sad.png

It compiles when using Intel CPU driver, but also does not when using AMD CPU driver.

Now you have a nice collection of driver bugs to submit.

I hope AMD soon releases a new driver with generic 2.0 support and fixing this.

Also when both AMD and Intel have 2.0 then, maybe NV is willing to update their OpenCL too.

Edit: Same for cl_khr_spir, cl_khr_image2d_from_buffer, cl_khr_image2d_from_buffer, cl_khr_dx9_media_sharing, ... ?

What a mess

Edit2:

See http://devgurus.amd.com/thread/155539

Very similar problem, so maybe you can just remove the pragma line and use the extension?

cl_khr_gl_event changes the behaviour of clEnqueueAcquireGLObjects() and clEnqueueReleaseGLObjects() so I cant really USE this extension. Either these functions synchronize between OpenGL and OpenCL or not. For me they dont do it. sad.png

Because of this removing the pragma line is no option.

I just reported these issues to AMD. Lets wait and hope that it does not take too long because I really need this for my master thesis.

In the mean time I will tripple buffer my data to hopefully bypass this problem.

hi thegabman, i have great interest for this issue, i tried to reproduce it on my R9 280X, but it is hard to reproduce. can you send me your project (source and shader)? my email: cndv3996@163.com thank you. regards davy

hi thegabman, i have great interest for this issue, i tried to reproduce it on my R9 280X, but it is hard to reproduce. can you send me your project (source and shader)? my email: cndv3996@163.com thank you. regards davy

Hi, right now I don't have the original project where the issue occurs, but you can look here for a minimal version of the code.

http://stackoverflow.com/questions/26763106/is-there-a-known-bug-with-glmemorybarrier-when-using-the-latest-amd-catalyst-d

thank you gabman, i compiled your code, still can't reproduce. my card: R9 280X my system: win8-64 my driver: latest catalyst i see your SSBO size is 100000(uint), i tried even with size of 100M(uint), can't reproduce. and gabman, i see on AMD website, latest catalyst is 4.9, beta is 4.11. can you upgrade your driver and give another try?
hi dear gabman, i tried with Hawaii Pro (R9 290) today, with your code, and i raised the size of SSBO to 10M (uint) at a later case. and i can't reproduce with both cases. sorry, i think the only difference between our environment would be the driver version. thank you for your help. regards davy

This topic is closed to new replies.

Advertisement