improve feedback when GPU already used.
For now a clear bottleneck is having nabu / GPU job failing in case of concurrency jobs.
Here are some ways ideas provided by @broche:
- get a dedicated message error from tomwer to clearly notify that there is a cuda mem-alloc error, dedicated color...
- improve error message in nabu. Mention that other user can / are using the GPU. @paleo maybe something more advanced could also be imagine. Like a semaphore or something like that (it must stay simple and robust)
- get a widget listing all the jobs using GPU running or nabu jobs
- get a 'diode' widget that is becoming red when the gpu is running. It can be integrated or not to nabu widget.