Multi-dimensional routing architecture
AssigneeArm Limited
InventorsAlejandro Rico Carro, Saurabh Pijuskumar Sinha, Douglas James Joseph, Tiago Rogerio Muck
Life of this patent
Estimating the term…
Abstract
Various implementations described herein refer to a device having a multi-layered logic structure with multiple layers including a first layer and a second layer arranged vertically in a stacked configuration. The device may have a first network that links nodes together in the first layer. The device may have a second network that links the nodes in the first layer together by way of the second layer so as to reduce latency related to data transfer between the nodes.
Claims20
A device comprising:
- a multi-layered logic structure with multiple layers and multiple networks including a first layer having a first network and a second layer having a second network arranged vertically in a stacked configuration, wherein the first layer and the second layer include different logic structures;
- and
- first repeater logic disposed between nodes in the first layer, wherein:
- the first network links the nodes together in the first layer by way of a first data path that passes through the first layer,
- the first repeater logic is used for data transfer between the nodes in the first layer,
- the second network links the nodes in the first layer together by way of a second data path that bypasses the first data path in the first layer and passes through the second layer so as to reduce latency related to data transfer between the nodes, and
- the second data path links the nodes with reduced latency such that data transfer between the nodes with the second data path is faster than data transfer between the nodes with the first data path, and
- reducing latency related to data transfer between the nodes is associated with load balancing between the first network and the second network.
A device comprising:
- a multi-layered logic structure with multiple layers including a first layer and a second layer arranged vertically in a stacked configuration;
- a first network that links nodes together in the first layer;
- and
- a second network that links the nodes in the first layer together by way of the second layer so as to reduce latency related to data transfer between the nodes,
- wherein the nodes are configured to insert data in the first network or the second network,
- wherein reducing latency related to data transfer between the nodes is associated with load balancing between the first network and the second network, and
- wherein each node that inserts a message packet into the first network or the second network decides whether to inject the message packet through a link in the first network or the second network in a randomized or reactive manner based on a signal that informs the node that congestion is present in the first network or the second network.
A device comprising:
- a multi-layered logic structure with multiple layers including a first layer and a second layer arranged vertically in a stacked configuration;
- a first network that links nodes together in the first layer;
- and
- a second network that links the nodes in the first layer together by way of the second layer so as to reduce latency related to data transfer between the nodes,
- wherein the nodes include a first node and a second node that is separate and distinct from the first node,
- wherein the first node has first router logic disposed in the first layer,
- wherein the second node has second router logic disposed in the first layer,
- wherein the nodes include a third node that is separate and distinct from the first node and the second node,
- wherein the third node has third router logic disposed in the first layer between the first router logic in the first node and the second router logic in the second node,
- wherein the first network has a first data path formed in the first layer that passes from the first node to the second node by way of the third node, and
- wherein the second network has a second data path formed in the second layer that passes from the first node to the second node with reduced latency such that data transfer between the first router logic and the second router logic with the second data path is faster than data transfer with the first data path.
A device comprising:
- an on-chip network structure having multiple levels including a first level with a first data network and a second level with a second data network that are arranged vertically in a stacked configuration, wherein the first level and the second level include different network structures;
- and
- first intra-layer repeater logic disposed between nodes in the first level, wherein:
- the first data network has a first data path formed in the first level that links nodes together by way of the first level including a first node and a second node of the nodes,
- the first intra-layer repeater logic is used for data transfer between the nodes in the first level,
- the second data network has a second data path formed in the second level that links the nodes together by way of the second level so as to reduce latency related to data transfer between the nodes, and
- wherein the second data path links the nodes with reduced latency such that data transfer between the nodes with the second data path bypasses the first data path in the first level and is faster than data transfer between the nodes with the first data path.
A device comprising:
- an on-chip network structure having multiple levels including a first level and a second level arranged vertically in a stacked configuration;
- a first data path formed in the first level that links nodes together including a first node and a second node that is separate from the first node;
- and
- a second data path formed in the second level that links the nodes together by way of the second level so as to reduce latency related to data transfer between the nodes,
- wherein the first node has first router logic disposed in the first level,
- wherein the second node has second router logic disposed in the first level,
- wherein the nodes include a third node that is separate and distinct from the first node and the second node,
- wherein the third node has third router logic disposed in the first level between the first router logic in the first node and the second router logic in the second node,
- wherein the first data path passes from the first node to the second node by way of the third node, and
- wherein the second data path passes from the first node to the second node with reduced latency such that data transfer between the first router logic and the second router logic with the second data path is faster than data transfer with the first data path.
A method comprising:
- providing a multi-layered logic structure by joining multiple layers together having multiple networks including a first layer having a first network and a second layer having a second network that are arranged vertically in a stacked configuration, wherein the first layer and the second layer include different logic structures;
- providing first repeater logic disposed between nodes in the first layer, wherein the first repeater logic is used for data transfer between the nodes in the first layer;
- linking the nodes together with the first network disposed in the first layer by way of a first data path;
- and
- linking the nodes together with the second network disposed in the second layer by way of a second data path that bypasses the first data path in the first layer and passes through the second layer, which is different than the first data path, so as to reduce latency related to data transfer between the nodes,
- wherein the second data path links the nodes with reduced latency such that data transfer between the nodes with the second data path is faster than data transfer between the nodes with the first data path.
A method comprising:
- providing a multi-layered logic structure by joining multiple layers together including a first layer and a second layer arranged vertically in a stacked configuration;
- linking nodes together with a first network disposed in the first layer;
- and
- linking the nodes together with a second network disposed in the second layer so as to reduce latency related to data transfer between the nodes,
- wherein the second layer is separate and distinct from the first layer, wherein the second layer is disposed above the first layer, and wherein the second layer is joined to the first layer with a face-to-face bond (F2FB),
- wherein each layer of the multiple layers has a frontside face and a backside face, and
- wherein a backside face of the first layer is joined to a backside face of the second layer so as to form the face-to-face bond (F2FB) between the first layer and the second layer.
Dependencies are inferred from the claim text.