Closed loop performance controller work interval instance propagation

A computer implemented method comprising:

• creating a work interval instance associated with a work interval object for a frame of content to be created, the work interval instance having an identifier;
• receiving notice of a submission of a workload for the frame of content on a co-processor, the notice including the identifier for the work interval instance;
• associating a token to identify the workload with the identifier for the work interval instance;
• propagating the identifier for the work interval instance via the association with the token that identifies the workload;
• subsequent to assignment at least one thread to a thread group, associating metrics for the workload executing on the co-processor with the frame of content with the thread group based on the identifier propagated for the work interval instance;
• adjusting performance of the co-processor or the one or more processors based on the accumulated metrics for the thread group;
• determining a control effort based on the accumulated metrics for the thread group, wherein the control effort maps to a core type for the at least one thread of the thread group;
• and
• adjusting the core type based on a control effort limit for the thread group, wherein the core type comprises performance cores (P-cores) and efficiency cores (E-cores).

A data processing system comprising:

• a processor complex including a plurality of core types, the plurality of core types to execute threads of a thread group;
• a co-processor to execute a workload offloaded by a thread of the thread group;
• and a closed loop performance controller (CLPC) to monitor and control performance of the threads of the thread group, to measure performance of the workload on the co-processor, and associate the performance of the workload with the thread group, wherein the CLPC is further to:
• receive notice of a submission of a workload for a frame of content on a co-processor, the notice including an identifier for a work interval instance;
• associate a token to identify the workload with an identifier for a work interval instance associated with the frame of content on the co-processor;
• subsequent to assignment of at least one thread to the thread group, associate metrics for the workload executing on the co-processor with the frame of content with the thread group based on an identifier propagated for the work interval instance, the identifier propagated via an association with the token that identifies the workload;
• adjust performance of the co-processor or the one or more processors based on the accumulated metrics for the thread group;
• determine a control effort based on the accumulated metrics for the thread group, wherein the control effort maps to a core type for the at least one thread of the thread group;
• and
• adjust the core type based on a control effort limit for the thread group, wherein the core type comprises performance cores (P-cores) and efficiency cores (E-cores).