Distributed compute work parser circuitry using communications fabric

A method, comprising:

• sending, by a master workload parser circuit, kernel state information for a compute kernel to each of a plurality of distributed workload parser circuits over a communications fabric;
• for multiple iterations, the master workload parser circuit:
• determining a next position in multiple dimensions for a next batch of one or more workgroups from the kernel, wherein the kernel includes a plurality of workgroups in each of multiple dimensions;
• and
• sending batch information to the distributed workload parser circuits via the communications fabric, wherein the batch information indicates one of the distributed workload parser circuits assigned to process the batch of workgroups;
• maintaining, by the distributed workload parser circuits, coordinate information for the kernel and updating the coordinate information in response to the master workload parser circuit sending batch information, including updating the coordinate information of a distributed workload parser circuit that is not assigned to execute the batch;
• and
• dispatching, by the distributed workload parser circuits, workgroups from assigned batches to ones of a respective set of a plurality of shader processors.

A non-transitory computer readable storage medium having stored thereon design information that specifies a design of at least a portion of a hardware integrated circuit in a format recognized by a semiconductor fabrication system that is configured to use the design information to produce the circuit according to the design, including:

• a plurality of shader processors configured to perform operations for compute workgroups included in compute kernels processed by the circuit, wherein one or more of the kernels include a plurality of workgroups in each of multiple dimensions;
• and
• a plurality of distributed workload parser circuits;
• a communications fabric connected to the plurality of distributed workload parser circuits and a master workload parser circuit;
• and
• the master workload parser circuit configured to:
• send kernel state information for a compute kernel to each of the distributed workload parser circuits over the communications fabric;
• for multiple iterations:
• determine a next position in the multiple dimensions for a next batch of one or more workgroups from the kernel;
• and
• send batch information to the distributed workload parser circuits via the communications fabric, wherein the batch information indicates one of the distributed workload parser circuits assigned to process the batch of workgroups;
• wherein the distributed workload parser circuits are configured to:
• maintain coordinate information for the kernel and update the coordinate information in response to the master workload parser circuit sending batch information, including when the distributed workload parser circuit is not assigned to execute the batch;
• and
• dispatch workgroups from assigned batches to ones of a respective set of the plurality of shader processors.