Real-time hardware-assisted gpu tuning using machine learning

A method, comprising:

• executing, by a multiprocessing unit, a multithreaded application according to one or more operating parameters;
• measuring, by a performance monitor circuits within the multiprocessing unit, performance monitor values during execution of the multithreaded application;
• reporting, by the performance monitor circuits within the multiprocessing unit, the performance monitor values during execution of the multithreaded application;
• and
• generating, by a control unit, one or more operating parameters based on the performance monitor values,
• wherein the control unit includes a machine learning model configured to receive the performance monitor values as inputs and update the one or more operating parameters as outputs during execution of the multithreaded application.

A non-transitory, computer-readable storage medium storing instructions that, when executed by a processor, cause the processor to:

• execute, by a multiprocessing unit, a multithreaded application according to one or more operating parameters;
• measure, by a performance monitor circuits within the multiprocessing unit, performance monitor values during execution of the multithreaded application;
• report, by the performance monitor circuits within the multiprocessing unit, the performance monitor values during execution of the multithreaded application;
• and
• generate, by a control unit, one or more operating parameters based on the performance monitor values,
• wherein the control unit includes a machine learning model configured to receive the performance monitor values as inputs and update the one or more operating parameters as outputs during execution of the multithreaded application.