Support for GPU scheduling?

[aifartist]

I found a huge perf difference doing SD inference just in the difference from running the APP(AUTOMATIC1111) on my P-cores vs my E-cores. In theory most of the time should be on the GPU which shows 100% where my perf is fast or slow. On fast cores I get 39 it/s and on slow cores I get 27 it/s. I use 'taskset' to control the binding of app to cores.

Someone suggested the reason the CPU perf might make a big difference is due to CPU scheduling vs GPU scheduling that is available on WIndows but surprisingly seems to have never been done for Linux drivers? If my CPU is too slow for my 4090 GPU why isn't GPU scheduling available?

The word 'scheduler' or 'scheduling' appears many times in the source but I'm not finding the term 'gpu schduling' or 'cpu scheduling'. Do the open drivers support this?