Hardware-Based Scheduling of GPU Work

• US 20110050713A1
• Filed: 09/03/2009
• Published: 03/03/2011
• Est. Priority Date: 09/03/2009
• Status: Active Grant

• Current Assignee: Advanced Micro Devices, Inc., ATI Technologies ULC (Advanced Micro Devices, Inc.)
• Original Assignee: Advanced Micro Devices, Inc.
• Inventors: Cheng, Gongxian Jefferey, McCrary, Rex, Liljeros, Frank
• Granted Patent: US 8,310,492 B2
• Time in Patent Office: Days
• US Class Current: 345/522
• CPC Class Codes: G06F 9/3851 from multiple instruction streams, e.g. multistreaming
• CPC Class Codes: G06F 9/3877 using a slave processor, e.g. coprocessor peripheral processor G06F13/12; vector processor G06F15/8053
• CPC Class Codes: G06F 9/3888 controlled by a single instruction for multiple threads [SIMT] in parallel
• CPC Class Codes: G06F 9/4881 Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues

Abstract

An apparatus and methods for scheduling and executing commands issued by a first processor, such as a CPU, on a second processor, such as a GPU, are disclosed. In one embodiment, a method of executing processes on a graphics processing unit (GPU) includes monitoring one or more buffers in a memory, selecting a first subset from the one or more buffers for execution on the GPU based on a workload profile of the GPU, and executing the first subset on the GPU. The GPU may also receive a priority ordering of the one or more buffers, where the selecting is further based on the received priority ordering. By performing prioritization and scheduling of commands in the GPU, system performance is enhanced.

Claims

1. A method of processing work items on a graphics processing unit (GPU), comprising:

selecting, based on a workload profile of the GPU, a first subset of buffers from one or more buffers in memory for execution of work items from the first subset on the GPU; and

executing said work items from the first subset on the GPU.

2. The method of claim 1, further comprising:

receiving a priority ordering of the one or more buffers, wherein the selecting is further based on the received priority ordering.

3. The method of claim 1, wherein the executing comprises:

executing on the GPU, work items from a first buffer from the first subset;

determining that work items from a second buffer is to be executed on the GPU;

pre-empting work items from the first buffer from execution; and

initiating execution of work items from the second buffer on the GPU.

4. The method of claim 3, wherein the pre-empting comprises:

saving a context of the first ring buffer to a context save area in a local memory.

5. The method of claim 3, wherein the determining comprises:

comparing a time elapsed in execution of work items from the first buffer to a predetermined time slice value.

6. The method of claim 3, wherein the determining comprises:

monitoring at least one of said buffers; and

detecting a work item in said at least one of said buffers having a priority level higher than the first buffer.

7. The method of claim 1, wherein the monitoring comprises:

reading at least one memory location written by a second processor; and

detecting a command write event based on a value read from the at least one memory location.

8. The method of claim 1, wherein the selecting comprises:

analyzing work items in each of the buffers;

determining a priority of said each of the buffers; and

selecting the first subset based on the determined priority.

9. The method of claim 1, wherein the memory is a system memory coupled to a central processing unit (CPU) and the GPU.

10. The method of claim 1, wherein at least one of the one or more buffers is a ring buffer.

11. The method of claim 1, wherein each of the said first subset includes one or more command buffers written by a central processing unit (CPU).

12. A system for executing work items on a graphics processing unit (GPU), the GPU configured for:

selecting, based on a workload profile of the GPU, a first subset of buffers from one or more buffers in memory for execution of work items from the first subset on the GPU; and

executing said work items from the first subset on the GPU.

13. The system of claim 12, wherein the GPU is further configured for:

receiving a priority ordering of the one or more buffers, wherein the selecting is further based on the received priority ordering.

14. The system of claim 12, wherein the GPU is further configured for:

executing a first buffer from the first subset;

determining that a second buffer is to be executed on the GPU;

pre-empting the first buffer from execution; and

initiating execution of the second buffer on the GPU.

15. The system of claim 14, wherein the pre-empting comprises:

saving a context of the first ring buffer to a context save area in a local memory.

16. The system of claim 14, wherein the determining comprises:

monitoring at least one of said buffers; and

detecting a work item in said at least one of said buffers having a priority level higher than the first buffer.

17. The system of claim 12, wherein at least one of the one or more buffers is a ring buffer.

18. The system of claim 12, further comprising:

a central processor unit (CPU); and

the memory, coupled to the CPU and the GPU.

19. The system of claim 12, wherein the GPU comprises:

a local memory configured with one or more context save areas.

20. A computer readable media storing instructions wherein said instructions when executed are adapted to process work items on a graphics processing unit (GPU) using a method comprising:

selecting, based on a workload profile of the GPU, a first subset of buffers from the one or more buffers in memory for execution of work items from the first subset on the GPU; and

executing said work items from the first subset on the GPU.

21. The computer readable media of claim 20 wherein said instructions comprise hardware description language instructions.

22. The computer readable media of claim 20 wherein said instructions are adapted to configure a manufacturing process through the generation of maskworks/photomasks to generate a device for processing said work items on a GPU.