Data Pointer Table Pre-Fetcher for Memory Latency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methodologies for identifying data pointers in memory are time-consuming, leading to power and performance losses in systems due to inefficient data retrieval and storage processes.

Innovation Solution

A system and method utilizing a data pointer table pre-fetcher that assembles a data pointer table within main memory, pre-fetches it, samples data from it, and stores it in a cache and content-addressable memory to facilitate fast lookups and efficient data transfers.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of time

If data pointers are identified using current methodologies, then data retrieval and storage operations can be performed, but the process takes relatively long time resulting in power and performance losses

Engineering Contradiction:
Improvedata pointer identification timeVSAvoidsystem performance
Core Design Contradiction:
Loss of timeVSProductivity

Solution Approach 1:

The patent pre-fetches the data pointer table from main memory into a dedicated cache structure before actual data access operations are needed. This preliminary action stores frequently accessed data pointer information in a faster access medium, eliminating the time penalty of repeated main memory accesses and improving overall system productivity

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent introduces a data pointer table cache as an intermediary structure between main memory and the data access operations. This intermediary holds pre-fetched data pointers, allowing rapid lookup without directly accessing main memory, thus resolving the contradiction between fast access and memory hierarchy constraints

Inventive Principle:
Principle #24Intermediary (Mediator)

2Loss of energy

If data pointers are identified using current methodologies, then data retrieval and storage operations can be performed, but power losses occur due to the time-consuming process

Engineering Contradiction:
Improvepower loss during data pointer identificationVSAvoiddata pointer identification time
Core Design Contradiction:
Loss of energyVSLoss of time

Solution Approach 1:

The data pointer table is pre-fetched and cached in advance, so that subsequent data access operations can retrieve pointers from the cache rather than performing time-consuming main memory accesses. This reduces both the time and power consumption of repeated operations

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system serves itself by maintaining a local cache of data pointers that can be quickly replenished from main memory when needed. The cache structure enables the system to handle common data access patterns without external intervention or repeated expensive memory operations

Inventive Principle:
Principle #25Self-service

Data Source

PatentUS9223708B2System, method, and computer program product for utilizing a data pointer table pre-fetcher
Publication Date: 2015.12.29 NVIDIA CORP
  • US9223708B2 patent drawing
  • US9223708B2 patent drawing
  • US9223708B2 patent drawing

AI summary

A system, method, and computer program product are provided for utilizing a data pointer table pre-fetcher. In use, an assembly of a data pointer table within a main memory is identified. Additionally, the data pointer table is pre-fetched from the main memory. Further, data is sampled from the pre-fetched data pointer table. Further still, the sampled data is stored within a data pointer table cache.