主权项 |
1. A method, comprising:
providing a first node in a cluster-on-chip computing environment, wherein the cluster-on-chip computing environment comprises a plurality of processor cores, wherein the plurality of processor cores are in communication by way of a mesh network, wherein each processor core has access to multiple memories with varying latencies, wherein the cluster-on-chip computing environment comprises main memory that is accessible to the plurality of processor cores by way of the mesh network, wherein the cluster-on-chip system fails to support hardware cache coherency, and wherein the first node comprises a first processor core, a first operating system kernel executing thereon, and a first memory including a message buffer, wherein the first operating system kernel is coupled to the first memory and the main memory; and providing a second node in the cluster-on-chip computing environment, wherein the second node comprises a second processor core and a second operating system kernel executing thereon, the first operating system kernel directing at least a portion of a message received by the first node from the second node to either the message buffer or to the main memory based at least in part on an expected frequency with which the first processor core may access the message and a latency of the message buffer. |