Interleaved Rasterization for VR GPU Bottlenecks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Virtual reality systems with Head-Mounted Displays (HMDs) require generating two frame buffers for a properly immersive experience, doubling the burden on GPUs unless clever techniques are used, as existing rasterization methods do not efficiently handle the high frame rates needed for VR applications.
Innovation Solution
An interleaved architecture for rasterization and pixel shading is introduced, where pairs of primitives and their pixel data are interleaved during processing, increasing temporal locality and reducing the burden on GPUs by optimizing texture data access and processing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional rasterization methods are used for VR systems, then two separate frame buffers must be generated for left and right eyes, but this doubles the processing burden on the GPU and reduces productivity
Solution Approach 1:
The patent merges the rendering of left and right eye frame buffers into a single interleaved processing pipeline. By alternating between rendering primitives for the left eye and right eye within the same rasterization and pixel shading passes, the system reduces the number of separate processing passes required while maintaining the ability to generate both frame buffers simultaneously.
Solution Approach 2:
The patent segments the frame buffer generation process into interleaved time slices, where the GPU alternates between processing primitives for the left eye and right eye. This segmentation allows the system to divide the workload into manageable chunks that can be processed sequentially within a single rendering pass, reducing the overall processing burden.
2Speed
If high frame rates are generated for immersive VR experience, then processing speed must increase, but this increases power consumption and heat generation
Solution Approach 1:
The patent implements continuous processing by eliminating idle passes between left and right eye rendering. The interleaved architecture ensures that the GPU continuously processes primitives and pixels without requiring separate rendering passes, maintaining high frame rates while reducing the total energy consumption by keeping the processing pipeline constantly utilized rather than repeatedly starting and stopping.
3Productivity
If texture data is accessed separately for each eye's rendering, then memory access patterns are inefficient, but improving temporal locality requires complex memory management
Solution Approach 1:
The patent merges texture data access operations for both eyes into a unified memory access pattern. By processing primitives and pixels from both eyes in an interleaved sequence, the system naturally accesses texture data in a pattern that improves temporal locality without requiring separate memory management logic for each eye, as the same texture resources are accessed alternately within the same processing context.
Data Source
AI summary
One embodiment provides for a graphics processing apparatus comprising first logic to rasterize pixel regions associated with multiple interleaved primitives; second logic to shade pixel regions covered by one or more of the multiple interleaved primitives; and third logic to interleave output of the second logic for the multiple interleaved primitives to a single render target, the single render target including output associated with the multiple interleaved primitives.