Abstract: A technique for executing commands for an accelerated processing device is provided. The technique includes obtaining an iteration number and predication data from metadata for an iterative indirect command buffer; for each iteration indicated by the iteration number, performing commands of the iterative indirect command buffer as specified by the predication data; and ending processing of the iterative indirect command buffer in response to processing a number of iterations equal to the iteration number.

Abstract: A technique for executing commands for an accelerated processing device is provided. The technique includes obtaining an iteration number and predication data from metadata for an iterative indirect command buffer; for each iteration indicated by the iteration number, performing commands of the iterative indirect command buffer as specified by the predication data; and ending processing of the iterative indirect command buffer in response to processing a number of iterations equal to the iteration number.

Abstract: This disclosure describes techniques for context switching. In one example, a graphics processing unit may be configured to generate one or more signatures for context information stored in on-chip memory of the graphics processing unit, determine whether the one or more signatures match any previously generated signatures for context information stored in one or more memories accessible by the graphics processing unit, store, to at least one of the one or more memories, any signature of the one or more signatures that is determined not to match any previously generated signature stored in at least one of the one or more memories, and store, to at least one of the one or more memories, the context information respectively corresponding to the one or more signatures determined not to match any previously generated signature stored in at least one of the one or more memories.

Abstract: A graphics processing unit (GPU) may dispatch a first set of commands for execution on one or more processing units of the GPU. The GPU may receive notification from a host device indicating that a second set of commands are ready to execute on the GPU. In response, the GPU may issue a first preemption command at a first preemption granularity to the one or more processing units. In response to the GPU failing to preempt execution of the first set of commands within an elapsed time period after issuing the first preemption command, the GPU may issue a second preemption command at a second preemption granularity to the one or more processing units, where the second preemption granularity is finer-grained than the first preemption granularity.

Abstract: A method of data processing, the method comprising receiving, at a graphics processing unit (GPU), a command stream, the command stream including one or more commands to be performed by the GPU and at least one command stream marker, the at least one command stream marker indicating a workload type of the command stream, determining, by the GPU, an operation algorithm for the GPU based on the at least one command stream marker prior to executing the command stream, and executing, by the GPU, the command stream based on the operation algorithm.

Abstract: Techniques are described with respect to preemption in which a graphics processing unit (GPU) may execute a first set of commands in response to receiving a draw call, the draw call defining a plurality of primitives that are to be rendered by the first set of commands, receive a preemption notification during execution of the first set of commands, and preempt the execution of the first set of commands, prior to completing the execution of the first set of commands to render the plurality of primitives of the draw call, for executing a second set of commands.

Abstract: A graphics processing unit (GPU) may dispatch a first set of commands for execution on one or more processing units of the GPU. The GPU may receive notification from a host device indicating that a second set of commands are ready to execute on the GPU. In response, the GPU may issue a first preemption command at a first preemption granularity to the one or more processing units. In response to the GPU failing to preempt execution of the first set of commands within an elapsed time period after issuing the first preemption command, the GPU may issue a second preemption command at a second preemption granularity to the one or more processing units, where the second preemption granularity is finer-grained than the first preemption granularity.

Abstract: A method of data processing, the method comprising receiving, at a graphics processing unit (GPU), a command stream, the command stream including one or more commands to be performed by the GPU and at least one command stream marker, the at least one command stream marker indicating a workload type of the command stream, determining, by the GPU, an operation algorithm for the GPU based on the at least one command stream marker prior to executing the command stream, and executing, by the GPU, the command stream based on the operation algorithm.

Abstract: A command processor may process a command stream for execution by at least one processor, including storing data associated with a first set of one or more operations in the command stream in a trace buffer, wherein the first set of one or more operations accesses one or more memory locations in memory, and wherein the data include an indication of contents of the one or more memory locations associated with the first set of one or more operations. The command processor may interrupt the processing of the command stream. The command processor may, in response to resuming processing of the command stream subsequent to the interrupting of the processing of the command stream, replay at least a portion of the command stream, including processing a second set of one or more operations of the command stream based at least in part on the data stored in the trace buffer.

Abstract: Techniques are described with respect to preemption in which a graphics processing unit (GPU) may execute a first set of commands in response to receiving a draw call, the draw call defining a plurality of primitives that are to be rendered by the first set of commands, receive a preemption notification during execution of the first set of commands, and preempt the execution of the first set of commands, prior to completing the execution of the first set of commands to render the plurality of primitives of the draw call, for executing a second set of commands.

Abstract: This disclosure describes techniques for context switching. In one example, a graphics processing unit may be configured to generate one or more signatures for context information stored in on-chip memory of the graphics processing unit, determine whether the one or more signatures match any previously generated signatures for context information stored in one or more memories accessible by the graphics processing unit, store, to at least one of the one or more memories, any signature of the one or more signatures that is determined not to match any previously generated signature stored in at least one of the one or more memories, and store, to at least one of the one or more memories, the context information respectively corresponding to the one or more signatures determined not to match any previously generated signature stored in at least one of the one or more memories.

Abstract: A command processor may process a command stream for execution by at least one processor, including storing data associated with a first set of one or more operations in the command stream in a trace buffer, wherein the first set of one or more operations accesses one or more memory locations in memory, and wherein the data include an indication of contents of the one or more memory locations associated with the first set of one or more operations. The command processor may interrupt the processing of the command stream. The command processor may, in response to resuming processing of the command stream subsequent to the interrupting of the processing of the command stream, replay at least a portion of the command stream, including processing a second set of one or more operations of the command stream based at least in part on the data stored in the trace buffer.