Closed Christian-B closed 6 months ago
I don't forsee this being needed, but happy to leave it just in case!
We need to decide if we will or will not support more than one none user core.
There is also a higher order question. On the Spin 2 boards will the none user core always be core 0?
I don't see a need for more than one non-user core so far.
In terms of virtual core support, for SpiNNaker 2 this will likely have to be done host-side, as writing to core memory and starting a core is likely to be direct rather than via SCAMP. This means that we can still have virtual core 0 for SCAMP, but mapped on the host side. Note also that the current plan is to use the special peripheral processor (an additional process on SpiNNaker 2 that is currently guaranteed to exist) for SCAMP, so the real address of the SCAMP processor will always be the same.
https://github.com/SpiNNakerManchester/SpiNNMachine/pull/237 does a better job so if that goes in close this one.
other pr merged so this not needed.
Either we allow Spin2 to define more than 1 n_non_user_cores Then we need this PR
Or we dont and then should remove that call from version