Systems and methods for GPU virtualization
View Patent ↗A method is provided for sharing access to graphics processing unit (GPU) hardware between multiple client virtual machines, wherein each of the client virtual machines has a high-level application programming interface (API) associated therewith for communicating with the GPU hardware. The method includes virtualizing the GPU by intercepting GPU-specific commands from the plurality of client virtual machines, wherein the commands specific to the GPU are at a lower level than that of the high-level API, and providing the intercepted commands to the GPU hardware.
1. A method of sharing access to graphics processing unit (GPU) hardware between a plurality of client virtual machines, wherein each of the client virtual machines has a high-level application programming interface (API) associated therewith for communicating with the GPU hardware, the method comprising:
virtualizing the GPU by intercepting GPU-specific commands from the plurality of client virtual machines, wherein the GPU-specific commands are at a lower level than that of the high-level API and wherein translation of the GPU-specific commands occurs separately from a GPU driver that directly communicates with the GPU hardware; and
providing the intercepted commands to the GPU hardware.
2. The method of claim 1 , wherein the GPU-specific commands are at a user-kernel application programming interface (API) level.
3. The method of claim 1 , wherein the high-level API commands are translated to the GPU-specific commands separate from a GPU driver communicating with the GPU hardware.
4. The method of claim 1 , wherein the intercepted commands are communicated to a server virtual machine configured to support concurrent connections with the plurality of client virtual machines.
5. A graphics processing unit (GPU) virtualization system comprising:
graphics processing unit (GPU) hardware;
a first client virtual machine and a second client virtual machine, wherein each of the client virtual machines has a high-level application programming interface (API) associated therewith for communicating with the GPU hardware; and
a virtualization module for virtualizing the GPU by intercepting GPU-specific commands from the first and second client virtual machines, wherein the GPU-specific commands are at a lower level than that of the high-level API and wherein translation of the GPU-specific commands occurs separately from a GPU driver that directly communicates with the GPU hardware.
6. The system of claim 5 , wherein the GPU-specific commands are at a user-kernel application programming interface (API) level.
7. The system of claim 5 , wherein the high-level API commands are translated to the GPU-specific commands separate from a GPU driver communicating with the GPU hardware.
8. The system of claim 5 , further including a server virtual machine configured to support concurrent connections with the first and second client virtual machines and receive the intercepted commands therefrom.
9. The system of claim 8 , further including a shared memory region for GPU-related allocations communicatively coupled between the server virtual machine and the first client virtual machine.
10. The system of claim 9 , further including an argument-shared-buffer residing within the shared memory region.
11. The system of claim 9 , further including a graphics heap residing within the shared memory region.
12. The system of claim 5 , further including a swap chain linked to a virtual framebuffer associated with the first client virtual machine.
13. Non-transitory computer-readable media bearing software instructions configured to instruct a processor to share access to graphics processing unit (GPU) hardware between a plurality of client virtual machines, wherein each of the client virtual machines has a high-level application programming interface (API) associated therewith for communicating with the GPU hardware by performing the steps of:
virtualizing the GPU by intercepting GPU-specific commands from the plurality of client virtual machines, wherein the GPU-specific commands are at a lower level than that of the high-level API and wherein translation of the GPU-specific commands occurs separately from a GPU driver that directly communicates with the GPU hardware; and
providing the intercepted commands to the GPU hardware.
14. The non-transitory computer-readable media of claim 13 , wherein the GPU-specific commands are at a user-kernel application programming interface (API) level.
15. The non-transitory computer-readable media of claim 13 , wherein the high-level API commands are translated to the GPU-specific commands separate from a GPU driver communicating with the GPU hardware.
16. The non-transitory computer-readable media of claim 13 , wherein the intercepted commands are communicated to a server virtual machine configured to support concurrent connections with the plurality of client virtual machines.
17. The non-transitory computer-readable media of claim 13 , further including linking a swap chain to a plurality of virtual framebuffers respectively associated with the plurality of client virtual machines.