Region testing of geometry while rendering for efficient multi-GPU rendering

A computer system comprising:

• a processor;
• memory coupled to the processor and having stored therein instructions that, if executed by the computer system, cause the computer system to execute a method, comprising:
• dividing responsibility for rendering geometry of graphics between a plurality of graphics processing units (GPUs) based on a plurality of screen regions corresponding to a display, each of the plurality of GPUs having a corresponding division of the responsibility for rendering which is known to the plurality of GPUs;
• rendering a plurality of pieces of geometry of an image frame using the plurality of GPUs executing a plurality of commands loaded to a rendering command buffer;
• begin rendering a first piece of geometry by a first GPU using the plurality of commands;
• determining at the first GPU that the first piece of geometry overlaps a first screen region of which the first GPU has a first division of the responsibility for rendering;
• executing to completion the rendering of the first piece of geometry at the first GPU based on the determining that the first piece of geometry overlaps the first screen region;
• determining that the first piece of geometry does not overlap any screen region of which a second GPU has a second division of the responsibility for rendering;
• generating a hint at the first GPU indicating that the second GPU can skip rendering a second piece of geometry;
• delivering the hint to the second GPU from the first GPU;
• and
• skipping the rendering of the second piece of geometry at the second GPU based on the hint.

A method, comprising:

• dividing responsibility for rendering geometry of graphics between a plurality of graphics processing units (GPUs) based on a plurality of screen regions corresponding to a display, each of the plurality of GPUs having a corresponding division of the responsibility for rendering which is known to the plurality of GPUs;
• rendering a plurality of pieces of geometry of an image frame using the plurality of GPUs executing a plurality of commands loaded to a rendering command buffer;
• begin rendering a first piece of geometry at a first GPU using the plurality of commands;
• determining at the first GPU that the first piece of geometry overlaps a first screen region of which the first GPU has a first division of the responsibility for rendering;
• executing to completion the rendering of the first piece of geometry at the first GPU based on the determining that the first piece of geometry overlaps the first screen region;
• receiving at the first GPU a hint from a second GPU that a second piece of geometry does not overlap any screen region of which the first GPU has the first division of the responsibility for rendering;
• and
• skipping rendering of the second piece of geometry at the first GPU.

A method, comprising:

• dividing responsibility for rendering geometry of graphics between a plurality of graphics processing units (GPUs) based on a plurality of screen regions corresponding to a display, each of the plurality of GPUs having a corresponding division of the responsibility for rendering which is known to the plurality of GPUs;
• rendering a plurality of pieces of geometry of an image frame using the plurality of GPUs executing a plurality of commands loaded to a rendering command buffer;
• begin rendering a first piece of geometry at a first GPU using the plurality of commands;
• determining at the first GPU that the first piece of geometry overlaps a first screen region of which the first GPU has a first division of the responsibility for rendering;
• executing to completion the rendering of the first piece of geometry at the first GPU based on the determining that the first piece of geometry overlaps the first screen region;
• and
• wherein the determining at the first GPU that the first piece of geometry overlaps the first screen region includes testing the first piece of geometry for overlap of the first screen region during a geometry processing stage of a graphics pipeline.

A method, comprising:

• dividing responsibility for rendering geometry of graphics between a plurality of graphics processing units (GPUs) based on a plurality of screen regions corresponding to a display, each of the plurality of GPUs having a corresponding division of the responsibility for rendering which is known to the plurality of GPUs;
• rendering a plurality of pieces of geometry of an image frame using the plurality of GPUs executing a plurality of commands loaded to a rendering command buffer;
• begin rendering a first piece of geometry at a first GPU using the plurality of commands;
• determining at the first GPU that the first piece of geometry overlaps a first screen region of which the first GPU has a first division of the responsibility for rendering;
• executing to completion the rendering of the first piece of geometry at the first GPU based on the determining that the first piece of geometry overlaps the first screen region;
• begin rendering a second piece of geometry at the first GPU using the plurality of commands;
• determining at the first GPU that the second piece of geometry overlaps a second screen region of which a second GPU has a second division of the responsibility for rendering;
• generating a hint at the first GPU indicating that the second GPU can render the second piece of geometry;
• delivering the hint to the second GPU from the first GPU;
• and
• rendering of the second piece of geometry at the second GPU based on the hint.