Abstract: Various embodiments provide for a circuit package including an electronic integrated circuit comprising a plurality of processing elements, and a plurality of bidirectional photonic channels, e.g., implemented in a photonic integrated circuit underneath the electronic integrated circuit, that connect the processing elements into an electro-photonic network. The processing elements include message routers with photonic-channel interfaces. Each bidirectional photonic channel interfaces at one end with a photonic-channel interface of the message router of a first one of the processing elements and at the other end with a photonic-channel interface of the message router of a second one of the processing elements and is configured to optically transfer messages (e.g., packets) between the message routers of the first and second processing elements.

Abstract: A package includes a bridging element (an OMIB), and first and second photonic paths, forming a bidirectional photonic path. The OMIB has first and second interconnect regions to connect with one or more dies. Third and fourth unidirectional photonic paths may couple between the first interconnect region and an optical interface (OI). A photonic transceiver has a first portion in the OMIB and a second portion in one of the dies. The first and the second portions may be coupled via an electrical interconnect less than 2 mm in length. The die includes compute elements around a central region, proximate to the second portion. The OMIB may include an electro-absorption modulator fabricated with germanium, silicon, an alloy of germanium, an alloy of silicon, a III-V material based on indium phosphide (InP), or a III-V material based on gallium arsenide (GaAs). The OMIB may include a temperature compensation for the modulator.

Abstract: A package includes a bridging element (an OMIB), and first and second photonic paths, forming a bidirectional photonic path. The OMIB has first and second interconnect regions to connect with one or more dies. Third and fourth unidirectional photonic paths may couple between the first interconnect region and an optical interface (OI). A photonic transceiver has a first portion in the OMIB and a second portion in one of the dies. The first and the second portions may be coupled via an electrical interconnect less than 2 mm in length. The die includes compute elements around a central region, proximate to the second portion. The OMIB may include an electro-absorption modulator fabricated with germanium, silicon, an alloy of germanium, an alloy of silicon, a III-V material based on indium phosphide (InP), or a III-V material based on gallium arsenide (GaAs). The OMIB may include a temperature compensation for the modulator.

Abstract: A package includes a bridging element (an OMIB), and first and second photonic paths, forming a bidirectional photonic path. The OMIB has first and second interconnect regions to connect with one or more dies. Third and fourth unidirectional photonic paths may couple between the first interconnect region and an optical interface (OI). A photonic transceiver has a first portion in the OMIB and a second portion in one of the dies. The first and the second portions may be coupled via an electrical interconnect less than 2 mm in length. The die includes compute elements around a central region, proximate to the second portion. The OMIB may include an electro-absorption modulator fabricated with germanium, silicon, an alloy of germanium, an alloy of silicon, a III-V material based on indium phosphide (InP), or a III-V material based on gallium arsenide (GaAs). The OMIB may include a temperature compensation for the modulator.

Abstract: One embodiment is a method that includes generating a request for a data item in a memory, obtaining the data item from the memory with a photonic interface, sending the data item to a fabric using a transmit unit of the photonic interface, and routing the data item through a portion of the fabric coupled to the memory, the portion of the fabric including one or more additional transmit and receive units between the photonic interface and a destination receive unit.

Abstract: A package includes a bridging element (an OMIB), and first and second photonic paths, forming a bidirectional photonic path. The OMIB has first and second interconnect regions to connect with one or more dies. Third and fourth unidirectional photonic paths may couple between the first interconnect region and an optical interface (OI). A photonic transceiver has a first portion in the OMIB and a second portion in one of the dies. The first and the second portions may be coupled via an electrical interconnect less than 2 mm in length. The die includes compute elements around a central region, proximate to the second portion. The OMIB may include an electro-absorption modulator fabricated with germanium, silicon, an alloy of germanium, an alloy of silicon, a III-V material based on indium phosphide (InP), or a III-V material based on gallium arsenide (GaAs). The OMIB may include a temperature compensation for the modulator.

Abstract: The present disclosure relates to thermal control systems, photonic memory fabrics, and electro-absorption modulators (EAMs). For example, the thermal control systems efficiently move data in a memory fabric based on utilizing and controlling thermally controlling optical components. As another example, the EAMs are instances of optical modulators used to efficiently move data within digital circuits while maintaining thermally-stable optical modulation across a wide temperature range.

Abstract: Various embodiments provide for a circuit package including an electronic integrated circuit comprising a plurality of processing elements, and a plurality of bidirectional photonic channels, e.g., implemented in a photonic integrated circuit underneath the electronic integrated circuit, that connect the processing elements into an electro-photonic network. The processing elements include message routers with photonic-channel interfaces. Each bidirectional photonic channel interfaces at one end with a photonic-channel interface of the message router of a first one of the processing elements and at the other end with a photonic-channel interface of the message router of a second one of the processing elements and is configured to optically transfer messages (e.g., packets) between the message routers of the first and second processing elements.

Abstract: Various embodiments provide for clock signal distribution within a processor, such as a machine learning (ML) processor, using a photonic fabric.

Abstract: Various embodiments provide for computational systems including multiple circuit packages, each circuit package comprising an electronic integrated circuit having multiple processing elements and intra-chip bidirectional photonic channels connecting the processing elements into an electro-photonic network, with inter-chip bidirectional photonic channels connecting the processing elements across the electro-photonic networks of the multiple circuit packages into a larger electro-photonic network.

Abstract: Various embodiments provide for digital neural network (DNN) that form part of a machine learning (ML) processor and that perform a compute intensive function, such as convolutions or matrix multiplies, which can facilitate operation of a ML model. According to some embodiments, the DNN comprises a combinatorial tree that uses multiple-accumulate (MAC) units, and a sequencer that reads values from memory devices and feeds the combinatorial tree.

Abstract: Various embodiments provide for electro-photonic networks, including a plurality of processing elements connected by bidirectional photonic channels, suited for implementing neural-network models. Weights of the model may be preloaded into memory of the processing elements based on assignments of neural nodes to processing elements implementing them, and routers of the processing elements can be configured to stream activations between the processing elements based on a predetermined flow of activations in the model.

Abstract: Vector and matrix multiplications can be accomplished in photonic circuitry by coherently combining light that has been optically modulated, in amplitude and/or phase, in accordance with the vector and matrix components. Disclosed are various beneficial photonic circuit layouts characterized by loss- and delay-balanced optical paths. In various embodiments, loss balancing across paths is achieved with suitable optical coupling ratios and balanced numbers of waveguide crossings (using dummy crossings where needed) across the paths. Delays are balanced in some embodiments with geometrically delay-matched optical paths.

Abstract: Vector and matrix multiplications can be accomplished in photonic circuitry by coherently combining light that has been optically modulated, in amplitude and/or phase, in accordance with the vector and matrix components. Disclosed are various beneficial photonic circuit layouts characterized by loss- and delay-balanced optical paths. In various embodiments, loss balancing across paths is achieved with suitable optical coupling ratios and balanced numbers of waveguide crossings (using dummy crossings where needed) across the paths. Delays are balanced in some embodiments with geometrically delay-matched optical paths.

Abstract: A novel method of operating a Distributed Cable Modem Termination System is provided. Each branch CMTS node in a distributed CMTS supports a complete CMTS system and with full-spectrum ports. One or more MAC domains are defined at each branch CMTS node. A MAC domain defined at a branch CMTS node includes only service flows of the CMs that are connected to the branch CMTS node. Identifiers of service flows coming from a branch CMTS node are always unique, i.e., two different service flows of the branch CMTS would always have different SIDs, even if they belong to different MAC domains. On the other hand, service flows belonging to different branch CMTS nodes are free to reuse the same identifiers.

Abstract: A novel method for provisioning network services for a cable system is provided. The method provides a configuration command interpreter/compiler that receives configuration commands of the cable system and generates configuration commands understood by the actual physical devices implementing the cable system. The interpreter transforms the configuration commands of the cable system into the configuration commands of the actual physical devices based on a set of normalized data models describing the cable system. The normalized data models are applicable to the cable system regardless of the actual devices implementing the cable system. The normalized data models are specified using normalized parameters that are generally applicable to different types devices that can be used to implement the cable system.

Abstract: Some embodiments provide a method for dynamically creating a service flow for an Ethernet node (EN) in a distributed cable management system that includes a cable headend and several in-the-field ENs for connecting several service nodes to the headend. For a particular device of a particular service node, the method receives a request to create a set of parameters for a service flow that is to be dynamically created. In some embodiments, the received request is in response to a request for a phone call that is to have a quality of service (QoS) guarantee and the service flow is for a PacketCable (PC) connection session. For the service-flow parameter request, the method identifies the EN that connects to the particular service node from a group of several EN that the method manages.

Abstract: Some embodiments provide a method for dynamically creating a service flow for an Ethernet node (EN) in a distributed cable management system that includes a cable headend and several in-the-field ENs for connecting several service nodes to the headend. For a particular device of a particular service node, the method receives a request for a Packet Cable Multi-Media (PCMM) connection session that is to have a quality of service (QoS) guarantee. For the request, the method identifies the EN that connects to the particular service node from a group of several EN that the method manages. The method then forwards a set of authorized dynamic QoS (DQoS) parameters to the identified EN, so that the identified EN can use the forwarded DQoS parameter set to create a service flow for the particular device. In some embodiments, the identified EN creates two service flows for the two different directions of traffic associated with the particular device at the particular service node.

Abstract: A novel method of handling network traffic for cable service flows in a distributed cable system is presented. Such a cable systems use remote distribution nodes in the fields to handle RF communications with cable modems in a distributed fashion. A packet engine is configured to assign a logical interface to each cable service flow in the cable system. Each logical interface in the packet engine is uniquely identifiable by a compound identifier that includes the identifier of the corresponding service flow and the identifier of the remote distribution node. The packet engine is configurable to selectively provide L3 level routing or L2 level switching/bridging between different logical interfaces. In some embodiments, the controller selects between configuring the packet engine to perform L3 routing or configuring the packet engine to perform L2 bridging based on whether the packet engine support unnumbered interfaces and integrated routing and bridging (IRB).

Abstract: A novel method of handling network traffic for cable service flows in a distributed cable system is presented. Such a cable systems use remote distribution nodes in the fields to handle RF communications with cable modems in a distributed fashion. A packet engine is configured to assign a logical interface to each cable service flow in the cable system. Each logical interface in the packet engine is uniquely identifiable by a compound identifier that includes the identifier of the corresponding service flow and the identifier of the remote distribution node. Upon receiving upstream data packet from a particular cable service flow, the remote distribution node applies a set of tags or labels to the data packet identifying the data packet as being from the particular cable service flow. The remote distribution node then forwards the tagged packet toward the packet engine, where the tags/labels are used to direct the packet toward the corresponding logical interface of the particular cable service flow.