Compute-Graph Execution Throughput in Data Processors

A non-transitory computer-readable storage medium storing computer program instructions that, when executed on a processor, perform a method for improving throughput in the processor, the method comprising:

• detecting, within a high-level program executed on a processor, an outer data-processing loop that includes a first data-processing operation;
• detecting, within the high-level program, an inner data-processing loop nested in the outer data-processing loop, wherein the inner data-processing loop includes a second data-processing operation that is interdependent with the first data-processing operation;
• determining if the first data-processing operation and the second data-processing operation performs a common operation;
• and
• in response to determining that the first data-processing operation and the second data-processing operation performs the common operation, fusing the first data-processing operation and the second data-processing operation into a single operation within the inner meta-pipeline loop, thereby increasing a throughput in the processor when executing the high-level program.

A system for improving throughput in data processors, the system comprising:

• one or more processors;
• and
• a memory device coupled to the one or more processors, wherein the memory device stores computer program instructions that when executed by the one or more processors, cause the system to:
• detect, within a high-level program executed on a data processor, an outer data-processing loop that includes a first data-processing operation;
• detect, within the high-level program, an inner data-processing loop nested in the outer data-processing loop, wherein the inner data-processing loop includes a second data-processing operation that is interdependent with the first data-processing operation;
• determine if the first data-processing operation and the second data-processing operation performs a common operation;
• and
• in response to determining that the first data-processing operation and the second data-processing operation performs the common operation, fuse the first data-processing operation and the second data-processing operation into a single operation within the inner meta-pipeline loop, wherein the fusing operation increases a throughput in the data processor when executing the high-level program by eliminating redundant operations.