Previous | Next --- Slide 17 of 35
Back to Lecture Thumbnails
malsup

To use a sort of classroom demo to illustrate snooping cache coherence, let's say in 418 lecture everyone were a processor working on some calculation. Whenever anyone finished and wrote to the "main memory", they would yell out that that address had been updated. Everyone else would be listening for the yell, and if they had that address in their personal "memory" they would stop using it. With all the shouting you can imagine that it would get pretty loud and difficult to hear. The job of the interconnect is to handle all the yelling back and forth. For instance, if we had a webservice that handled all the shouting instead, that would be the interconnect in our classroom analogy.

fangyihua

The individual cache must now handle 2 things: 1. load and store from its own processor. 2. handle the broadcast from other processors.

For this example, the interconnect will serialize all the broadcast by all the processor such that there will be only 1 broadcast request handled by the interconnect.