Graphical processing unit (GPU) implementing a plurality of virtual GPUs

A non-transitory computer readable storage medium storing program instructions that, when executed by a graphics processing unit (GPU), cause the GPU to carry out operations comprising:

• implementing a first virtual GPU having a first priority level that is variable;
• implementing a second virtual GPU having a second priority level that is variable;
• modifying the first priority level, the second priority level, or both based on an arbitration scheme;
• executing a first thread corresponding to a first GPU command, wherein the first GPU command has the first priority level based on the first virtual GPU having the first priority level;
• and
• subsequent to executing the first thread, executing, by the GPU, a second thread corresponding to a second GPU command, wherein the second GPU command has the second priority level based on the second virtual GPU having the second priority and wherein the first priority level is higher than the second priority level.

A method, comprising:

• running a plurality of virtual graphics processing units (GPUs) on a GPU;
• receiving a plurality of virtual GPU indicators that indicate, for a plurality of GPU commands, corresponding ones of the plurality of virtual GPUs;
• executing, by the GPU, a first thread corresponding to a first GPU command of the plurality of GPU commands, wherein the first thread is stored in an instruction buffer, and wherein the first GPU command has a first priority level based on a corresponding first virtual GPU having the first priority level;
• and
• subsequent to executing the first thread, executing, by the GPU, a second thread corresponding to a second GPU command of the plurality of GPU commands, wherein the second GPU command has a second priority level based on a corresponding second virtual GPU having the second priority level, and wherein the first priority level is higher than the second priority level.