Optimization of broadcast operations in multiple core processing system with a distributed cache

Abstract

A multiple core processing system as described herein for optimizing broadcast operations includes multiple first-level (L1) caches, a distributed second-level (L2) cache system including multiple L2 cache partitions each having a unique identifier, and a routing network that enable communications between the L1 and L2 caches. Each L1 cache sends a request message with a memory address to a targeted L2 cache partition. When rejected, the L1 cache resends the request only in response to a retry message that incorporates an identifier that matches a locally stored identifier. Each L2 cache partition sends a reject message to a requesting Li cache when busy and later broadcasts a retry message that incorporates its identifier when not busy. The L1 cache predetermines the identifier of the target L2 cache partition when sending the request, or each L2 cache partition incorporates its identifier in the reject message.