Multiresolution Priority Queue for High-Performance Computing

Solution Overview

Problem

Existing priority queue technologies face inefficiencies in high-performance computing applications due to limitations in handling a wide range of priorities, leading to suboptimal performance and increased computational complexity.

Innovation Solution

The multiresolution priority queue system introduces a flexible data structure that trades off resolution error in the priority space to achieve improved performance by discretizing the priority space into slots, allowing partial ordering and concurrent operations, thereby reducing computational complexity to O(1) and enhancing queue performance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional priority queue implementations (e.g., binary heap, Fibonacci heap) are used to handle a wide range of priorities, then the queue can maintain correct priority ordering, but the computational complexity increases (O(log n) or O(1) amortized) and performance deteriorates

Engineering Contradiction:

Improvepriority ordering correctnessVSAvoidqueue operation performance

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The patent segments the priority space into discrete resolution levels (coarse and fine granularity). The data structure maintains multiple queues at different resolution levels, allowing elements to be processed at coarser resolutions for high-performance operations and only refined to finer resolutions when necessary, thus resolving the contradiction between maintaining correct ordering and achieving high performance

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent implements dynamic resolution adjustment where the priority queue can operate at different resolution levels depending on the operational context. The system dynamically switches between coarse and fine resolution modes, allowing O(1) performance at coarse resolution while maintaining correctness through fine resolution refinement when needed

Inventive Principle:

Principle #15Dynamics

2Productivity

If bucket queues or calendar queues are used to achieve O(1) insert and remove operations, then performance is improved, but the resolution of priority set cannot be tuned and they do not support changing the resolution of the priority set

Engineering Contradiction:

Improvequeue operation speedVSAvoidpriority resolution flexibility

Core Design Contradiction:

ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent implements dynamic resolution adjustment where the priority queue can operate at different resolution levels depending on the operational context. The system dynamically switches between coarse and fine resolution modes, allowing O(1) performance at coarse resolution while maintaining correctness through fine resolution refinement when needed

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent changes the resolution parameter dynamically based on operational requirements. The system can adjust the priority resolution from coarse to fine based on the current state of the queue and operational needs, providing both high performance and adaptability

Inventive Principle:

Principle #35Parameter changes

3Productivity

If soft heaps are used to achieve constant amortized cost by introducing error in the key space, then performance is improved, but all keys could be corrupted depending on the sequence of insertions and extractions

Engineering Contradiction:

Improvecomputational efficiencyVSAvoidpriority accuracy

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent segments the priority space into discrete resolution levels (coarse and fine granularity). The data structure maintains multiple queues at different resolution levels, allowing elements to be processed at coarser resolutions for high-performance operations and only refined to finer resolutions when necessary, thus resolving the contradiction between maintaining correct ordering and achieving high performance

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies partial refinement by operating at coarse resolution for most operations and only applying fine resolution when necessary. This partial action approach maintains correctness without the excessive computational cost of always operating at full precision, avoiding the key corruption issue of soft heaps

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS11704153B2Systems and methods for multiresolution priority queues

Publication Date: 2023.07.18 QUALCOMM INC

AI summary

A system for storing and extracting elements according to their priority takes into account not only the priorities of the elements but also three additional parameters, namely, a priority resolution pΔ and two priority limits pmin and pmax. By allowing an ordering error if the difference in the priorities of elements are within the priority resolution, an improvement in performance is achieved.