CPU/GPU synchronization mechanism

• US 9,892,481 B2
• Filed: 09/28/2016
• Issued: 02/13/2018
• Est. Priority Date: 07/29/2011
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Ginzburg, Boris, Natanzon, Esfirush, Osadchiy, Ilya, Zach, Yoav
• Primary Examiner(s): Hoang, Peter
• Application Number:
  US15/278,316
• Publication Number:
  US 20170018051A1
• Time in Patent Office: 503 Days
• Field of Search: None
• CPC Class Codes: G06T 1/20 Processor architectures; Processor configuration, e.g. pipelining
• CPC Class Codes: G06T 1/60 Memory management

Abstract

A thread on one processor may be used to enable another processor to lock or release a mutex. For example, a central processing unit thread may be used by a graphics processing unit to secure a mutex for a shared memory.

Claims

1. A processing apparatus comprising:

a host central processing unit (CPU) to execute a first thread;

a graphics processing unit (GPU) coupled to the host CPU, the host CPU and the GPU to share access to a shared virtual address space, the first thread to synchronize access to the shared virtual address space;

wherein the first thread is to synchronize access between a second thread to execute on the host CPU and a third thread to execute on the GPU; and

wherein the first thread is to synchronize access via an acquire operation and a release operation, the acquire operation and release operation to be performed in response to a request received from the third thread.

2. The processing apparatus as in claim 1, wherein the third thread is to access data in the shared virtual address space via the acquire operation and to wait to access the data until the second thread is to release the data.

3. The processing apparatus as in claim 2, wherein the third thread is to release the data after the access via the release operation, the release to enable the second thread to access the data.

4. The processing apparatus as in claim 1, wherein the GPU is to wake the third thread when a signal message is received from the host CPU.

5. The processing apparatus as in claim 4, wherein the first thread is to enable the signal message between the host CPU and the GPU.

6. The processing apparatus as in claim 1, wherein the first thread is additionally to synchronize access to virtual memory shared between the third thread and the first thread.

7. The processing apparatus as in claim 1, wherein the host CPU and the GPU are integrated within one integrated circuit.

8. The processing apparatus as in claim 7, wherein the host CPU includes multiple processing cores.

9. The processing apparatus as in claim 8, wherein the GPU is a multi-threaded, multi-core parallel processor.

10. A heterogeneous processing system comprising:

a plurality of heterogeneous processors including a host CPU and a GPU coupled to the host CPU;

a memory shared by the host CPU and the GPU, wherein the memory includes a shared virtual address space; and

a first thread to execute on the host CPU to synchronize memory accesses by a second thread on the host CPU and a third thread on the GPU, wherein the first thread is to synchronize access via an acquire operation and a release operation, the acquire operation and release operation to be performed in response to a request received from the third thread.

11. The heterogeneous processing system as in claim 10, the third thread on the GPU to request to acquire access to data in the shared virtual address space and wait to acquire access to the data until the data is released by the first thread.

12. The heterogeneous processing system as in claim 10, the GPU to wake the third thread when a signal message is received from the host CPU, the signal message to enable the third thread to acquire access to the data.

13. The heterogeneous processing system as in claim 10, the first thread to synchronize access between the third thread and the first thread.

14. The heterogeneous processing system as in claim 10, the plurality of heterogeneous processors integrated within one integrated circuit.

15. The heterogeneous processing system as in claim 10, additionally including a driver associated with the GPU, the driver to execute on the host CPU to transfer synchronization messages between the first thread and the third thread.

16. The heterogeneous processing system as in claim 10, wherein the host CPU includes multiple processing cores.

17. The heterogeneous processing system as in claim 16, wherein the GPU is a multi-threaded, multi-core parallel processor.

18. A data processing system comprising:

a plurality of heterogeneous processors including a host CPU and a GPU coupled to the host CPU, the plurality of heterogeneous processors integrated within one integrated circuit;

a display device to display output from the GPU;

a memory shared by the host CPU and the GPU, wherein the memory includes a shared virtual address space; and

a first thread to execute on the host CPU to synchronize memory accesses by a second thread on the host CPU and a third thread on the GPU, wherein the first thread is to synchronize access via an acquire operation and a release operation, the acquire operation and release operation to be performed in response to a request received from the third thread.

19. The data processing system as in claim 18, the third thread on the GPU to request to acquire access to data in the shared virtual address space, wait to acquire access to the data until the data is released by the first thread, and the GPU to wake the third thread when a signal message is received from the host CPU, the signal message to enable the third thread to acquire access to the data.

20. The data processing system as in claim 18, the first thread to synchronize access between the third thread and the first thread.

21. The data processing system as in claim 18, additionally including a driver associated with the GPU, the driver to execute on the host CPU to transfer synchronization messages between the first thread and the third thread.

22. The data processing system as in claim 18, wherein the host CPU includes multiple processing cores and the GPU is a multi-threaded, multi-core parallel processor.