lava
lava copied to clipboard
Output Axons should be mappable
User Story
Core id and chip id of output axons should be mappable to correctly configure core or chip due to bad core mapping or multi chip placement
Conditions of satisfaction
- Multichip output axon works
- Bad Chip/Core output axon works
Acceptance tests
- Multichip output axon test works
- Bad Chip/Core output axon test works
I don't understand this story yet, but it is worth noting that NxCore handles bad cores automatically, and allows accessing chips by a linear index (under the hood it'll figure out the chip address).
In the final stage just before executing on the system, it replaces placeholders with the correct chip and core addresses for the specific system it is running on.
Lava/Magma would only need to worry about this if Lava is explicitly setting chip or core addresses instead of relying on NxCore.