L2 multicore request arbitration

[SergiVilardell]

Hello,

I am studying contention on the L2 for the OpenPiton+Ariane architecture. I'm interested in knowing which is the arbitration or queue system for the requests that are send from multiple cores to the L2. Is it a global queue for all cores? Does each core has its own queue?
Or maybe there is a queue for each slice of the L2? Thanks.

Sergi.

[Jbalkind]

There isn't any arbitration insofar as each L2 has its 3 NoC connections and messages arrive in order according to network routing. Between the NoCs there is a fixed priority where noc3 has the highest priority and noc1 the lowest, but they are generally connecting to different pipelines of the L2 anyway.