Parallella Community

Hypothetically speaking, is there anything stopping an eCore reading or writing to/from the local memory of an eCore in a different workgroup?

I know it's bad practice and I'm aware that the eSDK is not set up in this way (in that, the majority of functions deal with eCores within the caller core's workgroup) but if you could calculate the local memory address space of an eCore in one particular workgroup, could you read or write to that eCore from an eCore in another, separate, workgroup?

My hunch is that you can but I just wanted to double check.

No. The memory is not protected. Hack away.

Nice! Any pointers on how you would calculate the address you would need to write to a core in another workgroup?

For example, I have a 1 x 3 workgroup (let's call it Group A) beginning at core (0,0) and a 1 x 1 workgroup (Group B) at core (0,3). What address would the cores in Group A need in order to write to the local address 0x4000 of the core in group B? How could I calculate that?

Any help would be much appreciated - this is on a Parallella-16 Microserver by the way.

EDIT: If I hardcode the address - 0x80B00000 as the base address for writes to Group B it works, but I'm interested to know if this address can be calculated on the fly from the another workgroup.

Thanks jar!