This disclosure relates to integrated circuit devices that may include a network processor in a data processing die and an on-package memory in a base die. The data processing die may implement one or more network functionalities that may exchange data with low-latency memory, high capacity in the base die. The data processing die may be programmable fabric, which may be dynamically reconfigured during operation.

A method of commissioning a wired communication network, wherein said communication network is being configured to comprise a plurality of interconnected Data Forwarding Devices, DFDs in accordance with a network topology plan, wherein said network topology plan identifies how said plurality of DFDs are interconnected, and wherein each DFD has a plurality of ports for connecting to one or more further DFDs, wherein said method comprises the steps of generating link combination codes used to identify cables for interconnections in said network topology plan, wherein each link combination code is based on respective ports to which a respective cable is to be connected, generating unique port combination codes used to identify DFDs in said network topology plan, wherein each port combination code is based on respective ports with which a respective DFD is connected to further DFDs, and wherein said port combination codes are generated such that each DFD in said network topology plan utilizes different sets of ports for said interconnecting and applying said unique port combination codes to said plurality of DFDs.

A compute device to manage workflow to disaggregated computing resources is provided. The compute device comprises a compute engine receive a workload processing request, the workload processing request defined by at least one request parameter, determine at least one accelerator device capable of processing a workload in accordance with the at least one request parameter, transmit a workload to the at least one accelerator device, receive a work product produced by the at least one accelerator device from the workload, and provide the work product to an application.

The invention provides a converged network interface card, a message coding method and a message transmission method thereof. The converged network interface card comprises a PCIE host interface processing module, a high speed network card core logic, a crossbar switch XBAR, an Ethernet network card core logic, an Ethernet message dicing/slicing module, a physical layer, a high speed network/Ethernet message conversion module EoH, and a high speed network/Ethernet configurable network port. The invention supports customized high speed interconnection interface and a standard Ethernet interface on a set of network hardware, and supports three working modes on a set of physical hardware (high speed network mode, Ethernet mode and EoH mode transmitting Ethernet messages over the high speed network), implements seamless compatibility between the high speed network/Ethernet, and flexibly supports multimode applications such as scientific computing and cloud computing.

Methods and systems related to speculative resource allocation for routing on an interconnect fabric are disclosed herein. One disclosed method includes speculatively allocating a collection of resources to support a set of paths through an interconnect fabric. The method also includes aggregating a set of responses from the set of paths at a branch node on the set of paths. If a resource contention is detected, the set of responses will include an indicator of a resource contention. The method will then further include transmitting, from the branch node and in response to the indicator of the resource contention, a deallocate message downstream and the indicator of the resource contention upstream, and reallocating resources for the multicast after a hold period.

In one embodiment, a computer network system, includes a plurality of mesh networks, each mesh network including at least three interconnected respective internal switches with each respective internal switch being connected to each other one of the respective internal switches via a respective internal network connection, and Clos topology network connections connecting the mesh networks in a Clos topology arrangement.

Systems and methods for providing multicast group (MCG) membership relative to partition membership in a high performance computing environment. In accordance with an embodiment, by allowing a subnet manager of a local subnet to be instructed that all ports that are members of the relevant partition should be set up as members for a specific multicast group, the SM can perform a more efficient multicast-routing process. It is also possible to limit the IB client interaction with subnet administration conventionally required to handle join and leave operations. Additionally, subnet manager overhead can be reduced by creating a spanning tree for the routing of multicast packets that includes each of the partition members added to the multicast group, instead of creating a spanning tree after each multicast group join request is received, as conventionally required.

This invention discloses a hierarchical switching fabric and deadlock avoidance method for ultra high radix network routers. The hierarchical switching fabric comprises a network-on-chip and K multi-port components. The multi-port component comprises a port module configured to receive packets by a high speed serializer/deserializer, code and format the packets, send the packets to a corresponding hyper packet module after coding and format conversion, and send the packets sent by the hyper packet module to the network; and the hyper packet module configured to perform protocol processing for the received data link level packets, discard illegal packets, forward legitimate packets to the network-on-chip, perform data error correcting, format conversion and channel mapping for the packets received from the network-on-chip, and send the packets to the corresponding port module.

In some examples, a switching system includes a plurality of fabric endpoints and a multi-stage switching fabric having a plurality of fabric planes each having a plurality of stages to switch data units between any of the plurality of fabric endpoints. A fabric endpoint of the fabric endpoints is configured to send, to a switch of a first one of the stages and within a first fabric plane of the plurality of fabric planes, a self-ping message destined for the fabric endpoint. The fabric endpoint is configured to send, in response to determining the fabric endpoint has not received the self-ping message after a predetermined time, an indication of a connectivity fault for the first fabric plane.

Systems and methods for supporting inter subnet control plane protocol for consistent multicast membership and connectivity across multiple subnets in a high performance computing environment. In accordance with an embodiment, by associating a multicast group with an inter-subnet partition, and enforcing a dedicated router port for the multicast group, multicast loop avoidance can be provided for between connected subnets. Because only a single router port is selected as being capable of handling the MC packet, no other router port in the subnet can then pass a multicast packet back to the originating subnet.