Pixel Serialization for Conservative Depth Estimation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional z-testing in graphics pipelines leads to unnecessary processing and memory bandwidth consumption due to inadequate early-stage occlusion detection, resulting in many pixels being processed that are ultimately discarded as not visible, with a high incidence of 'indeterminate' outcomes.

Innovation Solution

Implementing z-culling operations within the graphics processing pipeline by determining z-ranges for coarse raster tiles, comparing these ranges to corresponding stored ranges, and culled or further processing pixels based on occlusion status, thereby reducing the number of 'indeterminate' outcomes and unnecessary processing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If z-testing is performed at the end of the graphics pipeline, then the depth comparison is accurate, but unnecessary processing and memory bandwidth consumption occur due to processing occluded pixels

Engineering Contradiction:

Improvedepth comparison accuracyVSAvoidprocessing efficiency

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent performs z-testing earlier in the graphics pipeline before shading and texturing operations. By determining the z-value during rasterization and comparing it with the stored z-value in the frame buffer at an intermediate stage, the system identifies occluded pixels before they undergo expensive processing operations, thereby eliminating unnecessary work while maintaining accurate depth comparison

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent divides the graphics pipeline into distinct stages with z-testing occurring at an intermediate point between rasterization and the final pixel processing stage. This segmentation allows the system to perform depth comparison at the most beneficial point in the pipeline, separating the culling function from both early rasterization and late pixel processing

Inventive Principle:

Principle #1Segmentation

2Reliability

If z-testing is performed at the end of the graphics pipeline, then the depth comparison is complete, but memory bandwidth is consumed unnecessarily by data transfers for every pixel

Engineering Contradiction:

Improvecompleteness of depth comparisonVSAvoidmemory bandwidth consumption

Core Design Contradiction:

ReliabilityVSLoss of energy

Solution Approach 1:

The patent performs the z-value comparison with the frame buffer earlier in the pipeline, before all pixel processing is complete. By doing so, it eliminates the need to transfer data for occluded pixels to and from the frame buffer, reducing memory bandwidth consumption while still ensuring complete depth comparison for all potentially visible pixels

Inventive Principle:

Principle #10Preliminary action

3Measurement precision

If z-testing is performed at the end of the graphics pipeline, then the depth comparison is finalized, but power is consumed unnecessarily by processing and memory operations for occluded pixels

Engineering Contradiction:

Improvefinality of depth comparisonVSAvoidpower consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by stationary object

Solution Approach 1:

The patent performs depth comparison and identifies occluded pixels at an intermediate stage in the graphics pipeline, before the expensive shading, texturing, and final pixel processing operations. This preliminary z-testing eliminates power-wasting operations on occluded pixels while maintaining final and accurate depth comparison results for visible pixels

Inventive Principle:

Principle #10Preliminary action

4Productivity

If z-testing is performed earlier in the graphics pipeline, then processing efficiency is improved by eliminating occluded pixels, but the granularity is insufficient to accurately discern visible versus occluded pixels

Engineering Contradiction:

Improveprocessing efficiencyVSAvoidpixel visibility discrimination accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent uses the z-value determined during rasterization as feedback to perform an intermediate z-test before pixel processing. This feedback mechanism allows the system to use the already-computed depth information to cull occluded pixels early, improving efficiency while maintaining accurate visibility discrimination through the use of precise z-value comparisons

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS9684998B2Pixel serialization to improve conservative depth estimation

Publication Date: 2017.06.20 NVIDIA CORP

AI summary

One embodiment includes determining a first z-range for a first portion of a coarse raster tile, where the first portion includes a plurality of pixels having a first set of pixel locations, retrieving from a memory a corresponding z-range related to a second set of pixel locations associated with the coarse raster tile, where the first set of pixel locations comprises a subset of the second set of pixel locations, and comparing the first z-range to the corresponding z-range to determine whether the plurality of pixels is occluded. If the plurality of pixels determined to be occluded, then the plurality of pixels is culled. If the plurality of pixels is determined to not be occluded, then the plurality of pixels is transmitted to a fine raster unit for further processing. The coarse raster tile comprises a plurality of portions, including the first portion, and those portions are processed serially.