-
Notifications
You must be signed in to change notification settings - Fork 245
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Program hangs running on AMD Ryzen CPUs #536
Comments
Can you do 'thread apply all bt' to display where the other thread(s) are at? It sounds weird if this is CPU specific, but if it's race then it can be just good luck in the other platform that it passes. |
Also, is this is a regression, does it reproduce with 0.14? |
This is where the other threads are at:
|
Well, from a quick look, one (driver) thread is in the kernel ( |
Also, what are the global WG sizes ( global_work_size parameter to clEnqueueNDRangeKernel) that you usually use for this kernel ? |
Can you share the kernel? |
Sure. The kernel is belongs to this project I am currently working: Kernel source files are under "device" folder. Compilation: Execution: |
Tried on a AMD Ryzen, seems to work:
@L30nardoSV this seems like a hardware issue.. |
...or just hard to reproduce indeterministic (race) issue. |
I am testing pocl 0.15-pre (compiled with LLVM 4.0.0) on AMD Ryzen 5 1600X CPUs.
The problem is that my application hangs at different points in execution. Backtracing this:
Eventually the program is able to continue executing up to completion. A second problem is, however, that computing results are not correct. I am using only single precision FP, but I am not sure if this issue is related to the first one.
Also, the temporal hanging described first (taking up to ~10 minutes), is unexpected as I tried with pocl 15-pre on i5 CPU - LLVM 3.9.1 having a smooth program execution.
Any ideas on this? thank you!
The text was updated successfully, but these errors were encountered: