Cache-Efficient Indirect Illumination via Interleaved Sub-Image Blur

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Real-time global illumination in computer graphics remains challenging for large and dynamic scenes due to latency and memory bandwidth issues, particularly in rendering indirect illumination, where conventional algorithms rely on approximations and limited sampling methods.

Innovation Solution

The processor and method involve restructuring reflective shadow maps and camera-view buffers into sub-buffers and sub-images, employing interleaved sampling, and applying cross-bilateral blurring to generate and combine indirect illumination sub-images into a higher-resolution image, improving texture cache efficiency and reducing latency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If conventional algorithms are used for rendering indirect illumination, then the rendering process can be implemented, but latency and memory bandwidth issues occur reducing real-time performance

Engineering Contradiction:
Improvereal-time rendering performanceVSAvoidlatency
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The reflective shadow map is divided into multiple sub-RSMs, and the rendering process is segmented into multiple passes. Each pass processes a subset of pixels using interleaved sampling, which improves cache utilization by keeping related data in memory together. This segmentation reduces memory bandwidth requirements and latency, enabling real-time rendering performance.

Inventive Principle:
Principle #1Segmentation

2Reliability

If comprehensive illumination algorithms are used, then physically plausible indirect illumination is achieved, but the computational complexity and processing time increase

Engineering Contradiction:
Improvephysical plausibility of illuminationVSAvoidalgorithm complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The algorithm uses partial sampling of the reflective shadow map through interleaved sampling patterns. Instead of processing all pixels comprehensively, it strategically samples subsets of pixels across multiple passes, achieving physically plausible illumination results with reduced computational complexity. The multiple passes with different sampling patterns ensure adequate coverage while maintaining efficiency.

Inventive Principle:
Principle #16Partial or excessive action

3Measurement precision

If high-resolution indirect illumination is rendered, then image quality is improved, but memory bandwidth consumption and processing time increase

Engineering Contradiction:
Improveindirect illumination image qualityVSAvoidmemory bandwidth consumption
Core Design Contradiction:
Measurement precisionVSLoss of energy

Solution Approach 1:

The high-resolution rendering is achieved through segmentation into multiple passes with interleaved sampling. Each pass processes a portion of the image at reduced resolution requirements, allowing the final high-quality result to be constructed from multiple lower-bandwidth operations. This approach maintains image quality while reducing peak memory bandwidth consumption compared to processing the entire high-resolution image in a single pass.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentUS9129443B2Cache-efficient processor and method of rendering indirect illumination using interleaving and sub-image blur
Publication Date: 2015.09.08 NVIDIA CORP
  • US9129443B2 patent drawing
  • US9129443B2 patent drawing
  • US9129443B2 patent drawing

AI summary

A cache-efficient processor and method for rendering indirect illumination using interleaving and sub-image blur. One embodiment of the processor is configured to render an indirect illumination image and includes: (1) a buffer restructurer configured to organize a reflective shadow map (RSM), rendered with respect to a reference point, into a plurality of unique sub-RSMs, each having sub-RSM pixels, (2) an indirect illumination computer configured to employ interleaved sampling on the plurality of unique sub-RSMs to generate a plurality of indirect illumination sub-images, and (3) a filter operable to smooth accumulated light values of the indirect illumination sub-images for subsequent interleaving into the indirect illumination image.