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

The multiple core processing system 1, further comprising a plurality of core clusters, each comprising:

• a subset of the plurality of first-level caches;
• a corresponding one of the plurality of second-level cache partitions;
• and
• wherein the routing network comprises:
• a local internal router coupled to each of the subset of the plurality of first-level caches and the corresponding one of the plurality of second-level cache partitions;
• and
• a local external router coupled to each of the subset of the plurality of first-level caches and the corresponding one of the plurality of second-level cache partitions.

The multiple core processing system 8, further comprising at least one meta-tile, each comprising:

• a subset of the plurality of core clusters;
• and
• wherein the routing network comprises:
• a meta-tile internal router coupled to a local internal router of each of the subset of the plurality of core clusters;
• and
• at least one meta-tile external router coupled to a local external router of at least one of the subset of the plurality of core clusters.

A method, comprising:

• in a multiple core processing system, comprising:
• a plurality of first-level caches;
• a second-level cache system comprising a plurality of second-level cache partitions, each associated with a corresponding one of a plurality of identifiers;
• and
• a routing network that is configured to enable communication between the plurality of first-level caches and the second-level cache system;
• sending, by a first-level cache, a request message to a targeted one of the second-level cache partitions;
• receiving, by the first-level cache, a reject message from the targeted one of the second-level cache partitions;
• storing, by the first-level cache, a corresponding identifier associated with the targeted one of the second-level cache partitions;
• receiving, by the first-level cache, a retry message incorporating an identifier associated with a second-level cache partition that broadcasted the retry message;
• comparing, by the first-level cache, the stored identifier with the identifier incorporated within the retry message;
• and
• resending, by the first-level cache, the request message only if the stored identifier matches the identifier incorporated within the retry message.