Vulcan

Model sizes must fit in the GPU memory. If the model is too large like most cutting age models, then they split parts of the model and submit the work...

here is the output: ``` pocld -a localhost -p 7777 --log_filter=err,warn,general [2024-01-12 17:40:11.121665366]POCL: in fn launch at line 440: | GENERAL | SERVER: PID 14900 Listening on command=localhost (127.0.0.1):7777, stream=localhost...

Thanks, I did that, and the errors went away. But example1 hasn't finished since minutes. I notice 4 pocld processes taking up 100% cpu.

Stuck with this: ``` OCL_ICD_VENDORS=/etc/OpenCL/vendors/mesa.icd pocld -a localhost -p 7777 --log_filter=err,warn,general [2024-01-12 18:12:19.033840677]POCL: in fn launch at line 440: | GENERAL | SERVER: PID 15729 Listening on command=localhost (127.0.0.1):7777, stream=localhost...

okay did that couple of times, looks like its hanging , no gpu load with radeontop, not really computing, so I fired up wireshark and saw lot of keepalive packets...

will wait for pr to be merged. thanks a ton. great work you folks have done! its like magic.

I built this branch: https://github.com/jansol/pocl/tree/pocld-structure and tested it. It hangs as before. What could be the reason? Is pocl remote working for others? Maybe I should build this on a...

Thank you for the responses. I would love to test on AMD APU/GPU with clover. I had similar issues with another vintage remote opencl project before, https://gitlab.com/illwieckz/clara/-/issues/23 https://gitlab.com/illwieckz/clara which didn't...

@pjaaskel @jansol > Just in case, I recommend checking if it works when running pocld on top of PoCL-CPU. With pocld with cpu works! Output of example 1: ``` ./example1...

I'll send you that in a while.