Method, device, and product for GPU cluster

An electronic device, comprising:

• at least one processing unit;
• and
• memory coupled to the at least one processing unit and having instructions stored therein, wherein the instructions, when executed by the at least one processing unit, cause the electronic device to perform actions comprising:
• obtaining a graph of interconnections between GPUs in a GPU cluster;
• forming a parallel hierarchical architecture of the GPU cluster based on the graph of interconnections between GPUs;
• and
• mapping parallel tasks to the parallel hierarchical architecture to execute the parallel tasks;
• wherein obtaining the graph of interconnections between GPUs in the GPU cluster comprises generating scores for respective connections between the GPUs as a function of (i) connection speeds of the respective connections between the GPUs, and (ii) determinations as to whether each such connection is an intra-domain connection or a cross-domain connection;
• and
• wherein forming the parallel hierarchical architecture of the GPU cluster based on the graph of interconnections comprises comparing the generated scores for the respective connections to one or more thresholds in order to separate the GPUs in the GPU cluster into one or more sub-groups for use in the mapping of the parallel tasks.

A computer program product comprising a non-transitory computer-readable medium having computer-executable instructions stored therein, wherein the computer-executable instructions, when executed by a computer, cause the computer to perform actions comprising:

• obtaining a graph of interconnections between GPUs in a GPU cluster;
• forming a parallel hierarchical architecture of the GPU cluster based on the graph of interconnections between GPUs;
• and
• mapping parallel tasks to the parallel hierarchical architecture to execute the parallel tasks;
• wherein obtaining the graph of interconnections between GPUs in the GPU cluster comprises generating scores for respective connections between the GPUs as a function of (i) connection speeds of the respective connections between the GPUs, and (ii) determinations as to whether each such connection is an intra-domain connection or a cross-domain connection;
• and
• wherein forming the parallel hierarchical architecture of the GPU cluster based on the graph of interconnections comprises comparing the generated scores for the respective connections to one or more thresholds in order to separate the GPUs in the GPU cluster into one or more sub-groups for use in the mapping of the parallel tasks.