Multi-Tile GPU Work Partitioning for Compute Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current graphics processing systems are limited in utilizing multiple tiles for workload assignment, leading to suboptimal performance in graphics processing and machine-learning operations.
Innovation Solution
A multi-tile partitioned walker is implemented to optimize compute tasks by efficiently distributing workloads across multiple graphics processing units (GPUs), allowing for improved parallel processing and resource allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple tiles are implemented in the graphics processor, then processing capacity and parallelism are improved, but the ability to utilize these tiles for workload assignment is insufficient
Solution Approach 1:
The graphics processor is divided into multiple independent tiles, each capable of executing workloads independently. The patent implements a tile-based architecture where each tile contains its own processing units and can be assigned specific workloads, enabling parallel processing while maintaining individual tile autonomy for flexible workload distribution
Solution Approach 2:
The patent introduces a new dimension of workload management by implementing a multi-tile assignment mechanism that operates at the tile level rather than the traditional processor level. This dimensional change in workload assignment enables finer-grained control and better utilization of multiple tiles for both graphics and machine-learning operations
2Productivity
If fixed function computational units are used, then specific graphics operations are processed efficiently, but the processor cannot support a wider variety of operations
Solution Approach 1:
The patent implements universal computational units within each tile that can be dynamically configured to perform different types of operations. These units support both traditional graphics processing and machine-learning workloads, replacing fixed-function units with reconfigurable processors that maintain high performance across diverse operation types
Data Source
AI summary
Embodiments are generally directed to compute optimization in graphics processing. An embodiment of an apparatus includes one or more processors including a multi-tile graphics processing unit (GPU) to process data, the multi-tile GPU including multiple processor tiles; and a memory for storage of data for processing, wherein the apparatus is to receive compute work for processing by the GPU, partition the compute work into multiple work units, assign each of multiple work units to one of the processor tiles, and process the compute work using the processor tiles assigned to the work units.