Redundant Thread Gating in Graphics Processors for Power Efficiency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Graphics processing units (GPUs) face challenges in power efficiency as they integrate more compute resources, leading to increased power consumption, which is unsustainable under given constraints, necessitating improved power management to enhance performance.

Innovation Solution

Implementing a method to gate redundant threads in GPUs by determining data similarity between threads and using output data from one thread as output for similar threads, thereby reducing dynamic power consumption by disabling unnecessary arithmetic logic unit (ALU) execution units.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If more compute resources are integrated and operated at higher frequencies to increase performance, then GPU performance is improved, but power consumption increases

Engineering Contradiction:
ImproveGPU performanceVSAvoidpower consumption
Core Design Contradiction:
ProductivityVSUse of energy by moving object

Solution Approach 1:

The patent applies partial action by selectively executing only necessary threads while gating redundant ones. The system determines data similarity between threads and gates execution of threads with similar data, performing only the minimal required computational actions rather than executing all scheduled threads, thus reducing power consumption while maintaining performance for unique computations

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The patent changes the operational state parameter of ALU execution units from active to gated based on data similarity detection. By dynamically adjusting the execution state parameter according to whether thread data falls within a predetermined similarity threshold, the system optimizes power consumption while maintaining performance for dissimilar threads

Inventive Principle:
Principle #35Parameter changes

2Loss of energy

If redundant thread processing is disabled to reduce power consumption, then power efficiency is improved, but computational redundancy increases

Engineering Contradiction:
Improvepower efficiencyVSAvoidcomputational output
Core Design Contradiction:
Loss of energyVSProductivity

Solution Approach 1:

The patent uses copying by replicating output data from executed threads to replace results that would have been produced by gated redundant threads. When threads are gated due to data similarity, their output is copied from the executed thread's result, maintaining computational output without performing redundant energy-consuming computations

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent discards redundant thread executions that would produce duplicate results and recovers the necessary computational output by copying results from executed threads. This approach eliminates wasted energy on redundant computations while recovering the required computational productivity through result replication

Inventive Principle:
Principle #34Discarding and recovering

Data Source

PatentUS10360177B2Method and processing apparatus for gating redundant threads
Publication Date: 2019.07.23 ATI TECHNOLOGIES ULC
  • US10360177B2 patent drawing
  • US10360177B2 patent drawing
  • US10360177B2 patent drawing

AI summary

Described is a method and processing apparatus to improve power efficiency by gating redundant threads processing. In particular, the method for gating redundant threads in a graphics processor includes determining if data for a thread and data for at least another thread are within a predetermined similarity threshold, gating execution of the at least another thread if the data for the thread and the data for the at least another thread are within the predetermined similarity threshold, and using an output data from the thread as an output data for the at least another thread.