Peripheral Completion Report Aggregation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

High-performance computing systems face bottlenecks due to the large number of bus transactions required for transmitting completion reports from peripheral devices to host memory, consuming significant bandwidth and causing performance issues.

Innovation Solution

The implementation of a computer peripheral device with processing circuitry that generates completion reports of varying sizes, eliminates redundant data fields, and transmits multiple reports in a single bus transaction or merges reports for successive work items, reducing data volume and transaction overhead.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If completion reports are transmitted individually for each work item, then completion information is delivered promptly, but the number of bus transactions increases significantly, consuming bandwidth and reducing system performance

Engineering Contradiction:
Improvecompletion information deliveryVSAvoidsystem performance
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent combines multiple completion reports into a single aggregated report that is transmitted in one bus transaction. The completion report includes a list of completed work items, allowing the peripheral device to report multiple completions simultaneously rather than individually, thereby reducing bus transaction overhead and improving system performance.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The peripheral device prepares and aggregates completion reports in advance before transmission. By collecting completion information for multiple work items and packaging them into a single report structure with appropriate formatting and length fields, the device performs preliminary organization of data, reducing the need for multiple separate transactions.

Inventive Principle:
Principle #10Preliminary action

2Loss of information

If complete completion reports with all data fields are transmitted for each work item, then all completion information is available, but the data volume increases, consuming more bandwidth and transaction resources

Engineering Contradiction:
Improvecompletion information completenessVSAvoiddata volume
Core Design Contradiction:
Loss of informationVSQuantity of substance

Solution Approach 1:

The patent extracts and transmits only the essential completion information needed by the host processor. The completion report structure includes selectively included data fields such as work item identifiers, completion status, and relevant result data, while omitting redundant or unnecessary information, thereby reducing data volume while maintaining information completeness.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent dynamically adjusts the content and size of completion reports based on the specific completion event. The report structure allows for variable-length fields and selective inclusion of data elements, changing the parameter composition of each report to match the actual completion information required, thus optimizing data volume.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentUS8924605B2Efficient delivery of completion notifications
Publication Date: 2014.12.30 MELLANOX TECHNOLOGIES LTD(IL)
  • US8924605B2 patent drawing
  • US8924605B2 patent drawing
  • US8924605B2 patent drawing

AI summary

A computer peripheral device includes a host interface, which is configured to communicate over a bus with a host processor and with a system memory of the host processor. Processing circuitry in the peripheral device is configured to receive and execute work items submitted to the peripheral device by client processes running on the host processor, and responsively to completing execution of the work items, to generate completion reports and to write a plurality of the completion reports to the system memory via the bus together in a single bus transaction.