# US 2005/0238052 A1 · Method and apparatus for providing an interconnection network function

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2005/0238052 A1
- Document ID: US20050238052A1
- Status: Published application
- Kind: A1
- Application number: 11/108,971
- Filed: 2005-04-19
- Published: 2005-10-27
- Earliest claimed: 2004-04-22
- Assignees: Sony Computer Entertainment Inc.
- Inventors: Takeshi Yamazaki
- Web page: https://patentgrep.com/patent/US20050238052A1

## CPC

- H04Q 3/521: {using semiconductors in the switching stages}
- G06F 15/17393: {having multistage networks, e.g. broadcasting scattering, gathering, hot spot contention, combining/decombining}
- H04L 49/101: using crossbar or matrix
- H04L 49/358: {Infiniband Switches}
- H04L 69/08: Protocols for interworking; Protocol conversion
- H04L 49/25: Routing or path finding in a switch fabric
- H04L 69/18: Multiprotocol handlers, e.g. single devices capable of handling multiple protocols
- H04L 69/323: in the physical layer [OSI layer 1]
- H04L 69/324: in the data link layer [OSI layer 2], e.g. HDLC
- H04L 69/325: in the network layer [OSI layer 3], e.g. X.25 (H04L69/16 takes precedence)
- H04L 69/326: in the transport layer [OSI layer 4] (H04L69/16 takes precedence)
- H04Q 2213/1302: Relay switches
- H04Q 2213/1304: Coordinate switches, crossbar, 4/2 with relays, coupling field
- H04Q 2213/13076: Distributing frame, MDF, cross-connect switch
- H04Q 2213/13215: Code checking, CRC
- H04Q 2213/13292: Time division multiplexing, TDM
- H04Q: SELECTING (switches, relays, selectors H01H; wireless communication networks H04W)
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)

## Abstract

Method and apparatus are provided for improved connection of devices and lower latency of communications between devices of a massively parallel network. In particular, method and apparatus are provided for cross-bar switches, a multiple protocol interface device, a low latency upper communication protocol layer, addressing and remote direct memory access over a massively parallel network.

## Claims

### Claim 1

A cross-bar switch for permitting communication between a multiplicity of devices, comprising:
an input interface operable to receive data from a first device selected from a group of devices;
an output interface operable to output the data to a second device selected from said group of devices;
an m×n array of switch elements, m and n each being greater than or equal to two, coupled to transfer the data between said input interface and said output interface, each said switch element having a plurality of input ports operable to receive at least one bit per cycle, and a plurality of output ports operable to output at least one bit per cycle, such that said array is operable to transfer the data between said input interface and said output interface in units of said at least one bit per cycle.

### Claim 2 (depends on 1)

A cross-bar switch as claimed in claim 1, wherein each said switch element has a plurality X of said input ports, a plurality X of said output ports, and at least a plurality X/2 of data paths connectable between ones of said input ports and said output ports, such that each said switch element is operable to simultaneously conduct at least X/2 transfers between said input ports and said output ports in units of at least one bit.

### Claim 3 (depends on 1)

The cross-bar switch of claim 1 wherein said array includes an input row of said switch elements coupled to said input interface, an output row of said switch elements coupled to said output interface, and at least one interior row coupled between said input row and said output row, wherein each of said switch elements of said input row and said interior row are coupled to provide output to two switch elements of said array, and each of said switch elements of said interior row and said output row are coupled to receive input from two switch elements of said array.

### Claim 4 (depends on 1)

The cross-bar switch of claim 1, wherein said input interface is operable to receive the data from the first device as parallel data and said output interface is operable to output the data to the second device as parallel data, and said array is operable to transfer the data between said input interface and said output interface as a plurality of bits per cycle through respective ones of said switch elements.

### Claim 5 (depends on 1)

The cross-bar switch of claim 1, wherein said input interface is operable to receive a single bit per cycle of the data from the first device and said output interface is operable to output a single bit per cycle of the data to the second device, and said array is operable to transfer a single bit per cycle of the data between said input interface and said output interface through respective ones of said switch elements.

### Claim 6 (depends on 5)

An interconnection apparatus including a plurality of cross-bar switches as claimed in claim 5, wherein said input interfaces of said cross-bar switches are each operable to receive a different single bit per cycle of a plurality of bits output per cycle as parallel data by the first device.

### Claim 7 (depends on 6)

An interconnection apparatus including a plurality of cross-bar switches as claimed in claim 6, wherein said output interfaces of said cross-bar switches are each operable to output a different single bit per cycle of the parallel data to the second device.

### Claim 8 (depends on 7)

An interconnection apparatus as claimed in claim 7, wherein said input interfaces are further operable to transfer the parallel data to said arrays on a time-division multiplexed basis and said output interfaces are further operable to receive the parallel data from said arrays on a time-division multiplexed basis.

### Claim 9 (depends on 1)

An interconnection apparatus including a plurality of cross-bar switches as claimed in claim 1, wherein said input interfaces of said cross-bar switches are each operable to transfer a different portion per cycle of a plurality of bits output per cycle as parallel data by the first device to said input ports of said switch elements of respective ones of said cross-bar switches, and said output interfaces of said cross-bar switches are each operable to output a respective one of said different portions per cycle of the parallel data to the second device.

### Claim 10 (depends on 9)

An interconnection apparatus as claimed in claim 9, wherein said input interfaces are further operable to transfer the parallel data to said arrays on a time-division multiplexed basis and said output interfaces are further operable to receive the parallel data from said arrays on a time-division multiplexed basis.

### Claim 11 (depends on 1)

A cross-bar switch as claimed in claim 1, wherein said input interface is further operable to transfer the data to said array on a time-division multiplexed basis and said output interface is further operable to receive the data from said array on a time-division multiplexed basis.

### Claim 12

A method of communicating between a multiplicity of devices, comprising:
inputting data to an input interface from a first device selected from a group of devices;
transferring at least one bit per cycle of the data between said input interface and an output interface through an m×n array of switch elements, m and n each being greater than or equal to two, each said switch element having a plurality of input ports operable to receive at least one bit per cycle, and a plurality of output ports operable to output at least one bit per cycle; and
outputting the data from said output interface to a second device selected from said group of devices;

### Claim 13 (depends on 12)

A method as claimed in claim 12, wherein each said switch element has a plurality X of said input ports, a plurality X of said output ports, and at least a plurality X/2 of data paths connectable between ones of said input ports and said output ports, such that each said switch element simultaneously conducts at least X/2 transfers between said input ports and said output ports in units of at least one bit.

### Claim 14 (depends on 12)

A method as claimed in claim 12 wherein said array includes an input row of said switch elements coupled to said input interface, an output row of said switch elements coupled to said output interface, and at least one interior row coupled between said input row and said output row, wherein each of said switch elements of said input row and said interior row is coupled to provide output to two switch elements of said array, and each of said switch elements of said interior row and said output row are coupled to receive input from two switch elements of said array.

### Claim 15 (depends on 12)

A method as claimed in claim 12, wherein the data is input from the first device to said input interface as parallel data, the data is output to the second device from said output interface as parallel data, and the data is transferred between said input interface and said output interface as a plurality of bits per cycle through respective ones of said input ports and said output ports of said switch elements.

### Claim 16 (depends on 12)

A method as claimed in claim 12, wherein the data is input a single bit per cycle to said input interface, the data is output a single bit per cycle from said output interface, and the data is transferred a single bit per cycle through said array.

### Claim 17 (depends on 16)

A method as claimed in claim 16, wherein each said input port of said switch elements is operable to input a single bit per cycle and each said output port of said switch elements is operable to output a single bit per cycle.

### Claim 18 (depends on 16)

A method as claimed in claim 16, further comprising inputting respective single bits of parallel data from the first device to respective ones of a plurality of said input interfaces, and for each said input interface, transferring the respective single bit per cycle between said input interface and one said output interface by an array of a plurality of said arrays, and outputting respective single bits of the parallel data per cycle to the second device through respective ones of a plurality of said output interfaces.

### Claim 19 (depends on 18)

A method as claimed in claim 18, further comprising outputting a different single bit of the parallel data from said output interface of each respective one of said cross-bar switches.

### Claim 20 (depends on 19)

A method as claimed in claim 19, wherein the parallel data is transferred to said arrays by said input interfaces on a time-division multiplexed basis and the parallel data is transferred from said arrays to said output interfaces on a time-division multiplexed basis.

### Claim 21 (depends on 12)

A method as claimed in claim 12, further comprising inputting respective portions of parallel data from the first device to respective ones of a plurality of said input interfaces, and for each said input interface, transferring the respective portion per cycle between said input interface and one said output interface by an array of a plurality of said arrays, and outputting a respective portion of the parallel data per cycle to the second device through a respective one of a plurality of said output interfaces.

### Claim 22 (depends on 21)

A method as claimed in claim 21, wherein the data is transferred through said arrays on a time-division multiplexed basis.

### Claim 23 (depends on 1)

A method as claimed in claim 1, wherein the data is transferred through said array on a time-division multiplexed basis.

### Claim 24

A multiple protocol stack interface device, comprising:
a first lower layer adapter operable to communicate according to a first lower communication protocol stack layer;
a first upper layer adapter operable to communicate according to a first upper communication protocol stack layer operating above said first lower communication protocol stack layer; and
a second upper layer adapter operable to communicate according to a second upper communication protocol stack layer operating above said first lower communication protocol stack layer,
wherein said interface device is operable to communicate according to said first lower communication protocol stack layer and said first upper communication protocol stack layer and to communicate according to said first lower communication protocol stack layer and said second upper communication protocol stack layer.

### Claim 25 (depends on 24)

An interface device as claimed in claim 24, wherein said first lower communication protocol stack layer and said first upper communication protocol stack layer are parts of a single communication protocol stack.

### Claim 26 (depends on 25)

An interface device as claimed in claim 25, further comprising a second lower layer adapter operable to communicate according to a second lower communication protocol stack layer, wherein said interface device is operable to convert a first message received according to said first lower communication protocol stack layer into a second message according to said second lower communication protocol stack layer.

### Claim 27 (depends on 26)

An interface device as claimed in claim 26, wherein said interface device is further operable to convert a third message received according to said second lower communication protocol stack layer into a fourth message according to said first lower communication protocol stack layer.

### Claim 28

A method of communication between devices which communicate according to at least different lower communication protocol stack layers, comprising:
receiving a first message according to a first lower communication protocol stack layer and a second upper communication protocol stack layer operating above said first lower communication protocol stack layer; and
converting the first message to a second message according to at least a second lower communication protocol stack layer; and
transmitting the second message.

### Claim 29 (depends on 28)

A method as claimed in claim 28, wherein said second message is converted according to said second upper communication protocol stack layer.

### Claim 30 (depends on 29)

A method as claimed in claim 29, wherein said first lower communication protocol stack layer and said first upper communication protocol stack layer are parts of a single communication protocol stack.

### Claim 31 (depends on 28)

A method as claimed in claim 28, further comprising converting a third message received according to said second lower communication protocol stack layer into a fourth message according to said first lower communication protocol stack layer.

### Claim 32

A method of communicating across a network of nodes, comprising:
assigning an address to each node of a first network, said address having a first number of address bits;
converting communications inbound to said first network from first devices from having a second number of address bits to having said first number of address bits, said first number being greater than said second number; and
converting communications outbound from said first network to said first devices from having said first number of address bits to having said second number of address bits.

### Claim 33

A method of providing remote direct memory access across a network, comprising:
generating, by a local processor, at least one packet for carrying out a direct memory access (DMA) command to a remote device;
storing said generated packet in storage local to said local processor;
transmitting said packet to said remote device without supervision by said local processor; and
receiving said packet at said remote device without supervision by said local processor.

### Claim 34 (depends on 33)

A method as claimed in claim 33, wherein said remote device includes a processor and said packet is received without supervision by said processor of said remote device.
