The command is a shorthand for set-threading-mode multicore.
If cell is given, then "multicore" threading mode is set for that particular cell, otherwise it will be used for all cells in the configuration.
The protocol argument chooses between different memory protection schemes and can be one of: msi, ww, or wwp. The msi protocol means that there can be multiple simultaneous readers for each memory region (currently the MMU page size of the target architecture) but only one writer. The ww protocol allows multiple simultaneous writers to the same memory region. This requires that the host has an equal or stricter memory consistency model then the simulated target architecture. In particular, target write operations must be as atomic on the host as on the target. Both these protocol requests exclusive read/write permissions for atomic operations. The wwp protocol uses host atomic operations to implement target atomic operations if possible. Currently this is not 100% accurate but will only fail for very obscure cases not normally found in parallel workloads.
The max-time-span argument specifies the maximum virtual time span (in seconds) that is allowed between clocks in an multicore accelerator enabled cell. This corresponds to the min latency that can be set between cells in sync domains.
NOTE: The multicore accelerator mode is not deterministic. This means that running the same workload multiple times will most likely behave differently due to different interleaving of memory accesses.