GPU Texture Pipeline Bypass for Shared Texel Convolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing graphics processing systems face inefficiencies in workload distribution between texture and shader hardware due to redundant texel fetches and the use of bilinear filter hardware, leading to imbalanced utilization and reduced overall performance during convolution operations.
Innovation Solution
A method and unit for a GPU pipeline that collectively fetches all required texels for a group of output pixels, bypasses the bilinear filter hardware, and performs convolution operations in the shader hardware, shifting the workload from texture to shader hardware to balance the load and improve efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If bilinear filter hardware is used in the texture pipeline, then filtering quality is improved, but workload imbalance between texture and shader hardware occurs
Solution Approach 1:
The patent extracts the bilinear filtering operation from the texture hardware pipeline and relocates it to the shader hardware. This allows the texture hardware to focus on texel fetching while the shader hardware performs both filtering and convolution operations, resolving the workload imbalance caused by having filtering in the texture pipeline.
2Adaptability or versatility
If independent texel fetches are performed for each output pixel, then per-pixel processing flexibility is maintained, but redundant texel fetches increase memory bandwidth usage
Solution Approach 1:
The patent merges the texel fetching operations for multiple output pixels into a single collective fetch operation. By identifying and combining redundant memory access patterns across pixels that share common texels, the system reduces memory bandwidth usage while maintaining the flexibility to perform per-pixel convolution operations in the shader hardware.
Solution Approach 2:
The patent performs preliminary collective fetching of all required texels for a group of output pixels before the shader hardware processes them. This preliminary action of gathering shared texel data once and making it available to multiple pixels eliminates redundant memory accesses while preserving per-pixel processing capabilities.
3Ease of operation
If texture hardware performs both fetching and filtering operations, then pipeline integration is improved, but shader hardware becomes underutilized during convolution operations
Solution Approach 1:
The patent introduces an intermediary mechanism where the texture hardware performs collective fetching and passes raw texel data to the shader hardware, which then acts as the intermediary for performing both filtering and convolution operations. This repositioning of the filtering operation to the shader hardware ensures full utilization of the shader pipeline while maintaining efficient memory access through collective fetching.
Data Source
AI summary
A method of operation of a texturing/shading unit in a GPU pipeline is used for efficient convolution operations. The method uses texture hardware to collectively fetch all the texels required to calculate properties for a group of output pixels without any duplication. The method then bypasses bilinear filter hardware in the texture hardware and passes the fetched and unfiltered texel data from the texture hardware unit to shader hardware in the texturing/shading unit. The shader hardware uses the fetched texel data to perform a plurality of convolution operations to calculate the properties of each of the output pixel.