Future Wei Technologies, Inc.

Huawei Technologies Co., Limited

National University of Singapore

Huawei International Pte. Ltd.

Naxin Zhang

Chenghong He

Iulin Lih

Hongbo Shi

Zhiguo Ge

Yuan Yao

Tulika Mitra

Mihai Pricopi

This invention is related to an Express Traversal (EXTRA) Network on Chip (NoC) comprising a number of EXTRA routers. The EXTRA NoC comprises a Buffer Write and Route Computation (BW/RC) pipeline, a Switch Allocation-Local (SA-L) pipeline, a Setup Request (SR) pipeline, a Switch Allocation-Global (SA-G) pipeline, and a Switch Traversal and Link Traversal (ST/LT) pipeline. The BW/RC pipeline is configured to write an incoming flit to an input buffer(s) of a start EXTRA router and compute the route for the incoming head flit by selecting an output port to depart from the start EXTRA router. The SA-L pipeline is configured to arbitrate the start EXTRA router to choose an input port and an output port for a winning flit. The SR pipeline is configured to handle the transmission of a number of SR signals from the start EXTRA router to downstream EXTRA routers.

Dedicated SSR pipeline stage of router for express traversal (EXTRA) NoC

A data processing method includes allocating a tag entry in a tag array for a data block; allocating a data entry in a data array for the data block when the data block is actively shared; and de-allocating the data entry when the data block is temporarily private or gets evicted in the data array.

Cache coherence directory architecture with decoupled tag array and data array

A reconfigurable cache architecture is provided. In processor design, as the density of on-chip components increases, a quantity and complexity of processing cores will increase as well. In order to take advantage of increased processing capabilities, many applications will take advantage of instruction level parallelism. The reconfigurable cache architecture provides a cache memory that in capable of being configured in a private mode and a fused mode for an associated multi-core processor. In the fused mode, individual cores of the multi-core processor can write and read data from certain cache banks of the cache memory with greater control over address routing. The cache architecture further includes control and configurability of the memory size and associativity of the cache memory itself.

Fusible and reconfigurable cache architecture

A network node comprises a receiver configured to receive a first packet, a processor coupled to the receiver and configured to process the first packet, and prioritize the first packet according to a scheme, wherein the scheme assigns priority to packets based on phase, and a transmitter coupled to the processor and configured to transmit the first packet. An apparatus comprises a processor coupled to the memory and configured to generate instructions for a packet prioritization scheme, wherein the scheme assigns priority to packet transactions based on closeness to completion, and a memory coupled to the processor and configured to store the instructions. A method comprises receiving a first packet, processing the first packet, prioritizing the first packet according to a scheme, wherein the scheme assigns priority to packets based on phase, and transmitting the first packet.

Phase-based packet prioritization

A computing system for handling barrier commands includes a memory, an interface, and a processor. The memory is configured to store a pre-barrier spreading range that identifies a target computing system associated with a barrier command. The interface is coupled to the memory and is configured to send a pre-barrier computing probe to the target computing system identified in the pre-barrier spreading range and receive a barrier completion notification messages from the target computing system. The pre-barrier computing probe is configured to instruct the target computing system to monitor a status of a transaction that needs to be executed for the barrier command to be completed. The processor is coupled to the interface and is configured to determine a status of the barrier command based on the received barrier completion notification messages.

System and method for barrier command monitoring in computing systems

Disclosed herein is a processing network element (NE) comprising at least one receiver configured to receive a plurality of memory request messages from a plurality of memory nodes, wherein each memory request designates a source node, a destination node, and a memory location, and a plurality of response messages to the memory requests from the plurality of memory nodes, wherein each memory request designates a source node, a destination node, and a memory location, at least one transmitter configured to transmit the memory requests and memory responses to the plurality of memory nodes, and a controller coupled to the receiver and the transmitter and configured to enforce ordering such that memory requests and memory responses designating the same memory location and the same source node/destination node pair are transmitted by the transmitter in the same order received by the receiver.

Cache coherent handshake protocol for in-order and out-of-order networks

A processing network comprising a cache configured to store copies of memory data as a plurality of cache lines, a cache controller configured to receive data requests from a plurality of cache agents, and designate at least one of the cache agents as an owner of a first of the cache lines, and a directory configured to store cache ownership designations of the first cache line, and wherein the directory is encoded to support substantially simultaneous ownership of the first cache line by a plurality but less than all of the cache agents. Also disclosed is a method comprising receiving coherent transactions from a plurality of cache agents, and storing ownership designations of a plurality of cache lines by the cache agents in a directory, wherein the directory is configured to support storage of substantially simultaneous ownership designations for a plurality but less than all of the cache agents.

Reduced scalable cache directory

A method comprising detecting at least one Quality of Service (QoS) requirement is met that indicates a very important packet (VIP) is outstanding from a source node in a multi-hop network comprising multiple nodes, sending an initiation message to an adjacent node in response to the detection that may activate a protocol in which a reserved channel is activated, and receiving the VIP via the reserved channel. Also, a method comprising receiving an initiation message from an adjacent node in a multi-hop network that comprises information identifying the VIP comprising a source node, a destination node, a packet type, wherein the initiation message activates a protocol in which a reserved channel is activated, searching for the VIP identified by the initiation message, and forwarding the VIP promptly if present via the reserved channel or forwarding an initiation message to adjacent nodes closer to the source node.

Forward progress assurance and quality of service enhancement in a packet transferring system

A home node for selecting a source node using a cache coherency protocol, comprising a logic unit cluster coupled to a directory, wherein the logic unit cluster is configured to receive a request for data from a requesting cache node, determine a plurality of nodes that hold a copy of the requested data using the directory, select one of the nodes using one or more selection parameters as the source node, and transmit a message to the source node to determine whether the source node stores a copy of the requested data, wherein the source node forwards the requested data to the requesting cache node when the requested data is found within the source node, and wherein some of the nodes are marked as a Shared state corresponding to the cache coherency protocol.

Method for peer to peer cache forwarding

A method comprising advertising to a second node a total allocation of storage space of a buffer, wherein the total allocation is less than the capacity of the buffer, wherein the total allocation is partitioned into a plurality of allocations, wherein each of the plurality of allocations is advertised as being dedicated to a different packet type, and wherein a credit status for each packet type is used to manage the plurality of allocations, receiving a packet of a first packet type from the second node, and storing the packet to the buffer, wherein the space in the buffer occupied by the first packet type exceeds the advertised space for the first packet type due to the packet.

Growing community of inventors

Singapore, Singapore

Naxin Zhang