by ysapir » Tue Jun 11, 2013 4:21 pm
The Epiphany architecture is essentially similar to the description in the thread you link to. This means that each eCore is a separate CPU, which has it own register set, peripherals and local memory. However, the core's memory can be easily accessed by any other core.
As opposed to an x86 processor, the Epiphany will more often than not will be a co-processor. This means that normally, code will be loaded on an idle core and then the host will generate a "SYNC" interrupt to get the core start fetching the program code (this can be done just as easily by a neighbor core).
Once started, each core runs a program totally separated from the other cores. However, inter-core communications is possible via memory movements and/or interrupts.