GPU Memory Copy Engine for Parallel Data Offloading

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current graphics processing units (GPUs) face inefficiencies in accessing and managing memory, as they often need to perform memory-related tasks concurrently with data processing, leading to idle times and reduced performance due to the lack of separation between memory management and data processing tasks.

Innovation Solution

The integration of a memory copy engine within the GPU allows for independent execution of memory management tasks, such as data retrieval and storage, parallel to data processing, enabling the GPU to handle memory operations without interrupting data processing, and vice versa, through a system bus interface.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If the GPU performs memory management tasks concurrently with data processing, then the GPU can handle both memory access and computation, but the GPU experiences idle times and reduced performance due to lack of task separation

Engineering Contradiction:
Improveprocessing efficiencyVSAvoididle times
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The patent segments the GPU into distinct functional units: a memory copy engine dedicated to memory management tasks and a shader processor dedicated to data processing tasks. This segmentation allows each unit to specialize in its respective function, eliminating the need for the GPU to switch between memory management and computation, thereby eliminating idle times and improving overall processing efficiency.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent merges the memory copy engine and shader processor into a single integrated GPU system, allowing them to work in close coordination while maintaining functional independence. This merging enables efficient data transfer between system memory and local memory while the shader processor simultaneously performs computations, maximizing resource utilization and eliminating idle times.

Inventive Principle:
Principle #5Merging (Combining)

2Productivity

If the GPU integrates memory management and data processing in a single unit, then the structure is simplified, but the GPU cannot perform multiple tasks in parallel, reducing overall performance

Engineering Contradiction:
Improveparallel processing capabilityVSAvoidGPU structure
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent divides the GPU into specialized segments: a memory copy engine for memory management and a shader processor for data processing. This segmentation enables parallel execution of memory operations and computational tasks, significantly improving productivity while the integrated architecture keeps the overall structure manageable.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The memory copy engine is designed as a universal unit that can handle various memory management operations including data retrieval, storage, and transfer between system memory and local memory. This multi-functional design allows the GPU to perform diverse tasks in parallel without requiring separate specialized hardware for each operation, balancing complexity with capability.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP2754055B1Memory copy engine for graphics processing
Publication Date: 2018.06.27 QUALCOMM INC
  • EP2754055B1 patent drawingFigure 1
  • EP2754055B1 patent drawingFigure 2
  • EP2754055B1 patent drawingFigure 3

AI summary

The example techniques described in this disclosure may be directed to interaction between a graphics processing unit (GPU) and a system memory. For example, the GPU may include a memory copy engine that handles tasks related to accessing data that is stored or is to be stored in the system memory. In addition, in some examples, the memory copy engine may perform additional tasks such as modification tasks to increase the performance of the GPU.