Mappings, workloads and accelerators are now defined as yaml files. The files are validated in corresponding the Validator, and converted into the internal representation in Factory.
Notes:
IMC architectures are not yet supported, and will need to be added later on.
Layer Operands can only be called I or W (and not X or Y or ...). This just makes it easier for the validator. Do we need the freedom to rename Layer Operands or are there no use cases for this?
I could use some feedback on the new user format in particular. Can it be simplified further in any way?
Mappings, workloads and accelerators are now defined as yaml files. The files are validated in corresponding the
Validator, and converted into the internal representation inFactory.Notes:
IorW(and notXorYor ...). This just makes it easier for the validator. Do we need the freedom to rename Layer Operands or are there no use cases for this?Dependency: Pull request https://github.com/KULeuven-MICAS/zigzag/pull/49 (Reworked internals)