Altera Corporation

Intel Corporation

Lsi Logic Corporation

Divya Vijayaraghavan

Chong H Lee

Curt Wortman

Vinson Chan

Keith Duwel

Ning Xue

Nageen Himayat

Min Suet Lim

Kooi Chi Ooi

Florence R Pon

Denica N Larsen

Michael Menghui Zheng

Lady Nataly Pinilla Pico

Fatema Adenwala

Ankitha Chandran

Martin Langhammer

Peng T Li

Huy X Ngo

Haiyun Yang

Gopi Krishnamurthy

Lana May Chan

David Wolk Mendel

Gregg William Baeckler

Tam Minh Nguyen

Curtis Ridgeway

Apparatuses, methods and storage medium associated with computer-assisted or autonomous driving (CA/AD) vehicles are disclosed herein. In embodiments, CA/AD vehicles are members of a CA/AD vehicle social network (CASN) in which various CA/AD vehicles may form connections or relationships with one another. CA/AD vehicles that have an existing relationship or connection may share CASN information with one another. The CASN information may include authenticated and/or proprietary information. Other embodiments are also described and claimed.

Computer-assisted or autonomous driving vehicles social network

Methods, apparatus, systems, and articles of manufacture are disclosed to improve data training of a machine learning model using a field-programmable gate array (FPGA). An example system includes one or more computation modules, each of the one or more computation modules associated with a corresponding user, the one or more computation modules training first neural networks using data associated with the corresponding users, and FPGA to obtain a first set of parameters from each of the one or more computation modules, the first set of parameters associated with the first neural networks, configure a second neural network based on the first set of parameters, execute the second neural network to generate a second set of parameters, and transmit the second set of parameters to the first neural networks to update the first neural networks.

Methods and apparatus to improve data training of a machine learning model using a field programmable gate array

Methods, apparatus, systems, and articles of manufacture for allocating a workload to an accelerator using machine learning are disclosed. An example apparatus includes a workload attribute determiner to identify a first attribute of a first workload and a second attribute of a second workload. An accelerator selection processor causes at least a portion of the first workload to be executed by at least two accelerators, accesses respective performance metrics corresponding to execution of the first workload by the at least two accelerators, and selects a first accelerator of the at least two accelerators based on the performance metrics. A neural network trainer trains a machine learning model based on an association between the first accelerator and the first attribute of the first workload. A neural network processor processes, using the machine learning model, the second attribute to select one of the at least two accelerators to execute the second workload.

Methods and apparatus for allocating a workload to an accelerator using machine learning

Systems and methods are disclosed for buffering data using a multi-function, multi-protocol first-in-first-out (FIFO) circuit. For example, a data buffering apparatus is provided that includes a mode selection input and a FIFO circuit that is operative to buffer a data signal between a FIFO circuit input and a FIFO circuit output, wherein the FIFO circuit is configured in an operating mode responsive to the mode selection signal.

Multi-function, multi-protocol FIFO for high-speed communication

One embodiment relates a method of receiving data from a multi-lane data link. The data is encoded with an FEC code having a block length. The data is FEC encoded at a bus width which is specified within particular constraints. One constraint is that the FEC encoder bus width in bits is an exact multiple of a number of bits per symbol in the data. Another constraint may be that the FEC code block length is an exact multiple of the FEC encoder bus width. Another constraint may be that the FEC encoder bus width is an exact multiple of a number of serial lanes of the multi-lane interface. Other embodiments and features are also disclosed.

Circuit structure and method for high-speed forward error correction

A configurable multi-protocol transceiver implemented in an integrated circuit ('IC') includes configurable deskew circuitry. The transceiver has various configurable deskew settings to facilitate effectively adapting transmit and/or receive communications corresponding to a selected one of a plurality of high-speed communication protocols and/or adapt to different implementations in which a deskew block addresses either just static skew or both static and dynamic skew. Configurable circuitry is adapted to control an allowed data depth of a plurality of buffers. Configurable circuitry is adapted to control a deskew character transmit insertion frequency. A programmable state machine is adapted to control read and write pointers in accordance with selectable conditions for achieving an alignment lock condition. Configurable circuitry is adaptable to select between logic and routing resources in the transceiver and logic and routing resources in a core of the IC in which the transceiver is implemented for controlling at least certain deskew operations.

Multi-protocol configurable transceiver including configurable deskew in an integrated circuit

Embodiments include a configurable multi-protocol transceiver including configurable deskew circuitry. In one embodiment, configurable circuitry is adapted to control an allowed data depth of a plurality of buffers. In another embodiment, configurable circuitry is adapted to control a deskew character transmit insertion frequency. In another embodiment, a programmable state machine is adapted to control read and write pointers in accordance with selectable conditions for achieving an alignment lock condition. In another embodiment, configurable circuitry is adaptable to select between logic and routing resources in the transceiver and logic and routing resources in a core of the IC in which the transceiver is implemented for controlling at least certain deskew operations. In another embodiment, configurable selection circuitry allows deskew processing to occur in a data path either before or after clock compensation processing depending on a communication protocol for which the transceiver is to be configured.

One embodiment relates a method of receiving data from a multi-lane data link. The data is encoded with an FEC code having a block length. The data is FEC decoded at a bus width which is specified within particular constraints. One constraint is that the FEC decoder bus width in bits is an exact multiple of a number of bits per symbol in the data. Another constraint may be that the FEC code block length is an exact multiple of the FEC decoder bus width. Another constraint may be that the FEC decoder bus width is an exact multiple of a number of serial lanes of the multi-lane interface. Other embodiments and features are also disclosed.

One embodiment relates to an integrated circuit configured to perform dynamic transmit equalization of a bi-directional lane. The integrated circuit including an interface between the physical coding and media access control circuitry, and an equalization control circuit which is external to the physical coding circuitry and which is configured to perform the dynamic transmit equalization using said interface. Another embodiment relates to a transceiver circuit which includes physical coding circuitry and media access control circuitry. The transceiver circuit further includes an interface between the physical coding circuitry and the media access control circuitry and an equalization controller which is external to the physical coding circuitry and which is configured to perform dynamic transmit equalization using said interface. The interface is configured to provide transmit coefficient data in a time-multiplexed signal format from the media access control circuitry to the physical coding circuitry. Other embodiments, aspects, and features are also disclosed.

Apparatus and methods of dynamic transmit equalization

A configurable interface includes a transmitter module and a receiver module, each configured to operate according to at least three different interface standards. The configurable interface further includes an interface module configured to determine a physical medium attachment (PMA) standard of a PMA coupled to the configurable interface and activate at least one component of the configurable interface based on the PMA standard. In an arrangement, the device interface supports a CAUI-4 standard.

Configurable multi-standard device interface

The present disclosure provides physical coding sublayer architectures that enable high-speed serial interfaces capable of operating at data rates ranging from 400 gigabits per second (Gbps) to 1 terabit per second (Tbps). A first embodiment relates to an architecture that provides an aggregated physical coding sublayer (PCS) that provides multiple virtual lanes. A second embodiment relates to an architecture that has a channel-based PCS and provides an aggregation layer above the PCS channels. A third embodiment relates to an architecture that, like the second embodiment, has a channel-based PCS and provides an aggregation layer above the PCS channels. However, each channel-based PCS in the third embodiment provides multiple virtual lanes. Other embodiments, aspects and features are also disclosed.

Growing community of inventors

Los Altos, CA, United States of America

Divya Vijayaraghavan