Systolic array including fused multiply accumulate with efficient prenormalization and extended dynamic range

    公开(公告)号:US12067375B2

    公开(公告)日:2024-08-20

    申请号:US17932537

    申请日:2022-09-15

    发明人: Thomas Elmer

    IPC分类号: G06F7/544 G06F15/80

    CPC分类号: G06F7/5443 G06F15/8046

    摘要: Systems and methods are provided to perform multiply-accumulate operations of at least one normalized number in a systolic array. The systolic array can obtain a first input and detect that the first input is denormal. Based on determining the first input is denormal, the systolic array can generate a first normalized number by normalizing the first input. Processing elements of the systolic array can include a multiplier and an adder. The multiplier can multiply the first normalized number by a second normal or normalized number to generate a multiplier product and the adder can add an input partial sum to the multiplier product to generate an addition result.

    Highly available modular hardware acceleration device

    公开(公告)号:US12066964B1

    公开(公告)日:2024-08-20

    申请号:US17548386

    申请日:2021-12-10

    摘要: A system includes a rack with multiple hardware acceleration devices and multiple modular controllers coupled together into a single system implementing one or more servers. Each modular hardware acceleration device includes multiple hardware accelerators, such as graphical processing units, field programmable gate arrays or other specialized processing circuits. In each modular hardware acceleration device, hardware accelerators are communicatively coupled to a multi-port connection device, such as a switch, and also communicatively coupled to at least two external ports. A modular controller of a particular server coordinates operation of hardware accelerators of multiple hardware acceleration devices included in the particular server to provide advanced processing capabilities. Hardware accelerators may be dynamically assigned to particular processing servers to adjust processing capabilities of those servers. A particular server may be assigned one or more standby controller to enhance availability of the server.

    Managing failover region availability for implementing a failover service

    公开(公告)号:US12066906B2

    公开(公告)日:2024-08-20

    申请号:US17814761

    申请日:2022-07-25

    IPC分类号: G06F11/20

    CPC分类号: G06F11/2025

    摘要: The present disclosure generally relates to managing a failover service. The failover service can receive a list of regions and a list of rules that must be satisfied for a region to be considered available for failover. The failover service can then determine the regions that satisfy each rule of the list of rules and are available for failover. The failover service can then deliver this information to a client. The failover service can determine the regions that do not satisfy one or more of the rules from the list of rules and deliver this information to a client. The failover service can perform automatic remediation to the unavailable failover regions and client remediation to the unavailable failover regions.

    CRYPTOGRAPHIC ASSERTIONS FOR CERTIFICATE ISSUANCE

    公开(公告)号:US20240275615A1

    公开(公告)日:2024-08-15

    申请号:US18642700

    申请日:2024-04-22

    IPC分类号: H04L9/32 H04L9/14

    CPC分类号: H04L9/3268 H04L9/14

    摘要: Components of a public certificate authority (CA) generate respective cryptographic assertions during performance of respective tasks of a certificate issuance workflow and a workflow approval component approves/rejects certificate issuance, based upon verification of the cryptographic assertions. For example, a workflow manager may assign tasks of a certificate workflow process to a number of components that process the tasks. The components generate responses and sign the respective responses with keys particular to each component. The workflow manager gathers the cryptographic assertions and sends them to a workflow approval component that validates the assertions, verifies the assertions indicate successful completion of the workflow and approves or rejects certificate issuance.

    Resource management for services
    75.
    发明授权

    公开(公告)号:US12063166B1

    公开(公告)日:2024-08-13

    申请号:US17709265

    申请日:2022-03-30

    IPC分类号: H04L47/80 H04L9/40 H04L47/78

    摘要: Systems and methods for resource management are disclosed. A search request may be received at a resource management service of a provider network. The search request may be received from a client device that does not have permission to access resources in a protected region of a provider network. The search request may specify a query associated with at least one operational health indicator in the protected region. It may be determined, using a secure query service, that the at least one operational health indicator does not exist in the protected region. The secure query service enables the client device to obtain information about the resources in the protected region without gaining access to the resources in the protected region. Sending of a notification indicating that the at least one operational health indicator does not exist in the protected region to the client device may be caused.

    Device detection and device operation sensor

    公开(公告)号:US12062911B1

    公开(公告)日:2024-08-13

    申请号:US17037202

    申请日:2020-09-29

    IPC分类号: H02J13/00 G01R21/00

    摘要: Described are systems, methods, and apparatus that monitor potential voltage and/or radiated electromagnetic energy at a premises to determine device usage at the premises. The monitored potential voltage and/or radiated electromagnetic energy may be processed to generate voltage signature data and radiated electromagnetic energy signature data, a signature data pair, of an operational device and that signature data pair may be used to identify at least one of a device type, a device, a device element of the device, and/or an operational state of a device.

    De-duplicating redundant data for data aggregation

    公开(公告)号:US12061944B1

    公开(公告)日:2024-08-13

    申请号:US17845089

    申请日:2022-06-21

    摘要: De-duplicating redundant data for data aggregation is described herein. In an example, a computing device receives first data identifying first items. The computing device determines that a first data portion and a second data portion indicate that a same item is detected to be within a location. The computing device generates second data by removing the second data portion from the first data. The computing device determines, based on a query to a data store storing previously stored data identifying second items, that (i) a no match exists between the first data portion and the previously stored data, (ii) a match exists between a third data portion of the second data and the previously stored data. The computing device generates a count of items detected within the location by incrementing a previous count based on the no match and by foregoing a count increment based on the match.

    Cloud-based job scheduling and management service

    公开(公告)号:US12061928B1

    公开(公告)日:2024-08-13

    申请号:US17031469

    申请日:2020-09-24

    发明人: Akash Anjanappa

    摘要: Devices, systems, and methods are provided for cloud-based job scheduling and management. A method may include receiving, by at least one serverless processor of a cloud-based network, data from a first interface. The method may include determining, based on the data, a first activity and a second activity, the first activity associated with a first service of the cloud-based network, and the second activity associated with a second service of the cloud-based network. The method may include determining, based on the data, a precondition associated with causing execution of the first activity and the second activity. The method may include determining that an event associated with the cloud-based network has occurred, and generating an instance associated with causing performance of the first activity and the second activity. The method may include discarding the instance after performance of the first activity and the second activity.