Transparent multi-buffering in multi-GPU graphics subsystem
AssigneeVIA Technologies, Inc.
InventorsGuofeng Zhang, Xuan Zhao
Life of this patent
Estimating the term…
Abstract
This invention discloses a method and system for implementing transparent multi-buffering in multi-GPU graphics subsystems. The purpose of multi-buffering is to reduce GPU idle time. In one example, after rendering a first image by a first GPU in a back buffer, the first image is displayed by flipping to the back buffer. After that, the front buffer and back buffer are exchanged, and then shifting the back buffer and internal buffers in a predetermined sequence. A second image is rendered to current back buffer by a second GPU. The second image is displayed by flipping to a current back buffer. After that, the front buffer and back buffer are exchanged again, and shifting the back buffer and internal buffers again.
Claims12
A method for multi-buffering in a computer system with multiple graphics processing units (GPUs) for rendering images and displaying the same by flipping between two flip buffers, the method comprising:
- allocating one or more internal buffers for replacing contents of one of the two flip buffers with contents of one of the internal buffers at the end of a flip execution, wherein the internal buffers are additional buffers implemented in the GPUs in addition to the two flip buffers to provide more renderable buffers;
- rendering a first image by a first GPU to a back buffer of the two flip buffers;
- displaying the first image by flipping to the back buffer, after which a front buffer of the two flip buffers becomes a back buffer and the back buffer becomes a front buffer;
- shifting current back buffer and the internal buffers in a predetermined sequence;
- rendering a second image by a second GPU to the current back buffer, which has been replaced by one of the internal buffers;
- displaying the second image by flipping to the current back buffer, after which the front buffer and back buffer are exchanged;
- and
- shifting the current back buffer and the internal buffers again in the predetermined sequence, wherein the two flip buffers are shared by the first and second GPUs for displaying the rendered first and second images.
A computer system comprising:
- two or more graphics processing units (GPUs) for rendering images including a first image and a second image;
- two or more flip buffers providing data for displaying the rendered first and second images by flipping between two flip buffers including a front buffer and a back buffer alternatively;
- one or more internal buffers, wherein the internal buffers are additional buffers implemented in the GPUs in addition to the two or more flip buffers for replacing contents of the back buffer with contents of one of the internal buffers at the end of a flip execution to get more renderable buffers;
- and
- a graphics driver for shifting one of the two flip buffers and the internal buffers,
- wherein the GPUs include a first GPU for displaying the first image by flipping to one of the two flip buffers and a second GPU for displaying the second image by flipping to another of the two flip buffers, and the two flip buffers are shared by the two GPUs for displaying the rendered first and second images.
A system for supporting transparent multi-buffering in a computer graphics subsystem with multiple graphics processing units (GPUs) and at least two flip buffers including a front buffer and a back buffer, the system comprising:
- at least one internal buffer allocated for replacing contents of one of the two flip buffers at the end of a flip execution with contents of one of the at least one internal buffer and for providing more renderable buffers, wherein the internal buffer includes one or more additional buffers implemented in the GPUs in addition to the two flip buffers;
- a graphics driver for shifting one of the two flip buffers and the internal buffer;
- and
- at least one extra data member in each buffer structure to store original buffer structure content, and two data members allocated by the driver to store an initial information of a graphics subsystem for later being destroyed at an end of a drawing program, with one member to store pointers to the internal buffer and one to save original primary surface address,
- wherein the GPUs include a first GPU for displaying a first image by flipping to one of the two flip buffers and a second GPU for displaying a second image by flipping to another of the two flip buffers, and the two flip buffers are shared by the first and second GPUs for displaying the rendered first and second images.
Dependencies are inferred from the claim text.