EP2901294A1 - Protocole de socket de réseau sur puce - Google Patents

Protocole de socket de réseau sur puce

Info

Publication number
EP2901294A1
EP2901294A1 EP13842232.4A EP13842232A EP2901294A1 EP 2901294 A1 EP2901294 A1 EP 2901294A1 EP 13842232 A EP13842232 A EP 13842232A EP 2901294 A1 EP2901294 A1 EP 2901294A1
Authority
EP
European Patent Office
Prior art keywords
interface unit
network interface
chip
transaction
initiator
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
EP13842232.4A
Other languages
German (de)
English (en)
Other versions
EP2901294A4 (fr
Inventor
Philippe Boucard
Jean-Jacques Lecler
Boris BOUTILLIER
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Qualcomm Technologies Inc
Original Assignee
Qualcomm Technologies Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from US13/626,758 external-priority patent/US9471538B2/en
Priority claimed from US13/626,766 external-priority patent/US9225665B2/en
Application filed by Qualcomm Technologies Inc filed Critical Qualcomm Technologies Inc
Priority to EP22196229.3A priority Critical patent/EP4123468A1/fr
Publication of EP2901294A1 publication Critical patent/EP2901294A1/fr
Publication of EP2901294A4 publication Critical patent/EP2901294A4/fr
Ceased legal-status Critical Current

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L12/00—Data switching networks
    • H04L12/28—Data switching networks characterised by path configuration, e.g. LAN [Local Area Networks] or WAN [Wide Area Networks]
    • H04L12/40—Bus networks
    • H04L12/403—Bus networks with centralised control, e.g. polling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F15/00—Digital computers in general; Data processing equipment in general
    • G06F15/16—Combinations of two or more digital computers each having at least an arithmetic unit, a program unit and a register, e.g. for a simultaneous processing of several programs
    • G06F15/163—Interprocessor communication
    • G06F15/173—Interprocessor communication using an interconnection network, e.g. matrix, shuffle, pyramid, star, snowflake
    • G06F15/1735—Network adapters, e.g. SCI, Myrinet
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F15/00—Digital computers in general; Data processing equipment in general
    • G06F15/76—Architectures of general purpose stored program computers
    • G06F15/78—Architectures of general purpose stored program computers comprising a single central processing unit
    • G06F15/7807—System on chip, i.e. computer system on a single chip; System in package, i.e. computer system on one or more chips in a single package
    • G06F15/7825—Globally asynchronous, locally synchronous, e.g. network on chip

Definitions

  • This disclosure is related generally to the field of semiconductor technology and more specifically to network-on-chip interconnects for systems on chip. BACKGROUND
  • a network-on-chip is a packet based interconnection for transporting read and write transactions between socket interfaces.
  • a NoC comprises at least one initiator network interface unit (NIU) and at least one target NIU.
  • Initiator NIUs convert transaction requests at the initiator socket interface to request packets and converts response packets to transaction responses at the initiator socket interface.
  • Target NIUs convert request packets to transaction requests at the target socket interface and converts transaction responses at the target socket interface to response packets.
  • Initiator NIUs and target NIUs are connected through a topology of switches.
  • the packets may contain a field known as a sequence ID, which encodes some or all of: the initiator ID, target ID, transaction sequence ID, and transaction tag.
  • a system and method are provided that connect two NoCs with less logic and faster timing paths using a transaction interface protocol.
  • a transaction interface protocol is disclosed wherein the interface protocol has a transaction ID signal in each of the request and response channels.
  • the protocol is used between a target NIU master and an initiator NIU slave that are directly connected through a transaction interface.
  • the target NIU response channel uses the transaction ID signal to identify the entry in a context array associated with the corresponding request.
  • the coupling of target NIU and initiator NIU enable the formation of an on-chip interconnect comprising multiple NoCs wherein the topology of the interconnect is simpler, smaller, faster, and has lower latency.
  • FIG. 1 illustrates a full chip interconnect
  • FIG. 2 illustrates a transaction interface protocol
  • FIG. 3 illustrates an initiator NIU.
  • FIG. 4 illustrates a target NIU that stores echo fields.
  • FIG. 5 illustrates a transaction interface protocol, according to the teachings of the present invention, with a transaction ID signal.
  • FIG. 6 illustrates an initiator NIU, according to the teachings of the present invention, which stores a transaction ID.
  • FIG. 7 illustrates an initiator NIU, according to the teachings of the present invention that has no context allocation unit, splitting unit, context array, or reassociation unit.
  • FIG. 8 shows a target NIU according to the teachings of the present invention.
  • FIG. 9 shows a NoC in according to the teachings of the present invention.
  • FIG. 10 shows a flowchart for the method of operation of one embodiment of the target NIU.
  • FIG. 1 shows a full chip interconnect comprising a first NoC 120 and a second NoC 140.
  • the NoC 120 comprises an initiator NIU 122 that is coupled to an initiator 110 through a socket interface 112.
  • the first NoC 120 is coupled to the initiator 110.
  • the NoC 120 also comprises a target NIU 124 that is coupled to a target 130 through a socket interface 132.
  • a target NIU 126 in the NoC 120 is coupled to an initiator NIU 142 in the NoC 140 through a socket interface 160.
  • the NoC 140 comprises a first target NIU 144 coupled to a target 150 and a target NIU 146 coupled to a target 152.
  • a NoC uses layered communication.
  • Sockets present a transaction layer protocol such as the Advanced Microcontroller Bus Architecture (AMBA) Advanced extensible Interface (AXI) and Open Core Protocol (OCP).
  • a transaction layer protocol is illustrated in FIG. 2. It has a request channel 200 with signals for opcode, address, write data, burst length, and sequence id. The protocol also has a response channel 210 with signals for data and sequence id.
  • a transport layer protocol encapsulates the transaction information in one or more packets that are transferred through the topology of switches. Packets carry data along with a header that can have fields such as address bits, route id, opcode, and sequence id.
  • a physical layer implements flow control and a simple connection of wires that transfer packet headers and data.
  • a physical layer protocol can have signals such as ready, valid, and data.
  • NIUs can perform the functions of:
  • an initiator NIU 300 is shown in FIG. 3.
  • requests are accepted through a socket interface request channel 302.
  • Responses are presented by the initiator NIU 300 through a socket interface response channel 304.
  • Request packets are sent on a request transport interface 306 and response packets are received on a response transport interface 308.
  • Transaction addresses are decoded in a decode unit 310 to produce a route id field in packet headers.
  • Packets are associated with contexts in a context allocation unit 312. If necessary, transactions are split into multiple packets in a packet splitting unit 314. The packets are sent on the request transport interface 306.
  • the initiator NIU 300 includes the splitting unit 314.
  • the splitting unit 314 causes the creation of a number of packets to transport the transaction.
  • the initiator NIU 300 can split transactions into multiple packets in order to ensure a desired byte alignment, burst alignment, target address range boundaries, address interleaving, or protocol boundary requirements, such as the AXI requirement of bursts not to cross 4KB aligned address boundaries.
  • a context array 320 comprises entries that are allocated to pending transactions. If the context array 320 is full then the initiator NIU 300 asserts backpressure on the socket interface request channel 302. The physical layer ready signal is deasserted. The context array 320 stores, among other information, the sequence id and opcode of each pending transaction. The context array 320 has four entries and therefore supports up to four pending transactions.
  • a context reassociation unit 330 When response packets are received on the response transport interface 308 a context reassociation unit 330 reassociates the packet with the context of the pending transaction. The reassociation by the reassociation unit 330 ensures that the sequence ID of the response on the response channel 304 of the initiator socket interface matches the sequence ID of the corresponding request.
  • a target NIU 400 is shown that can correspond with the initiator NIU 300.
  • Request packets are accepted on a request transport interface 402 and response packets are sent on a response transport interface 404.
  • Transaction requests are presented through a socket interface request channel 406 and transaction responses are accepted by the NIU through a socket interface response channel 408.
  • the target NIU 400 includes a context allocation unit 412, a context array 420, and a context reassociation unit 430.
  • the packet header includes an echo field that is produced by the allocation unit 312 in the initiator NIU 300 and echoed back from the target NIU 400 unaltered.
  • the echo field stores an index into the context array 320.
  • the context array is accessed as a table lookup.
  • the echo field E is stored in the target NIU context array 420 while each transaction from the target NIU 400 is pending.
  • the echo field which is part of the transport layer protocol, is included in the header of response packets for the transaction.
  • the target NIU 400 reassociation unit 430 does not have the luxury of an index the echo field.
  • Responses on the socket interface response channel 408 must be reassociated with the header information stored from the request packet, stored in the context array 420 in order to form the response packet.
  • the target NIU reassociation unit 430 must perform a lookup of the appropriate context array entry based on the oldest transaction with the sequence ID of the response packet.
  • the context array 420 is organized as a chained list of pending transactions with a chain for each sequence ID.
  • a convention transaction interface including signals for opcode, address, write data, burst length, and sequence id, is enhanced in the request channel 500 with the additional signals Request TrID and in the response channel 510 with Response TrID.
  • the TrID signals convey a transaction ID.
  • This enhanced signal interface is used by an enhanced NoC socket protocol.
  • the master of the socket interface asserts a value on TrID when a transaction request is presented and the slave of the interface gives exactly the same value on the Response TrID signal when presenting the transaction response.
  • the slave of the interface gives a value on the Response TrID signal that is transformed from the corresponding value Request TrID value.
  • a transformation can be the changing of the order or sense of bits.
  • the key is associability to a unique master context array entry.
  • each of multiple pending transactions has a unique TrID value.
  • TrID values are reused for transactions that must be executed in order, despite downstream buffering.
  • TrID values are given for successive transactions in a non-sequential order.
  • the asserted TrID value is mapped from a field in a packet header within the upstream NoC. In one embodiment the TrID value is mapped to a field in a packet header within the downstream NoC.
  • An optimal protocol configuration varies from one NoC to another and from one chip to another.
  • the NoC socket protocol is configurable at chip design time. Configuration options include, among others, the width of the Request and Response TrID signals.
  • the register transfer level (RTL) language logic is generated by a configuration tool. The tool is used by chip designers to generate customized configurations and produce RTL language code for chip synthesis.
  • FIG. 6 An aspect of the invention is shown in FIG. 6.
  • requests are accepted through a socket interface request channel 602.
  • Responses are presented by the NIU 600 through a socket interface response channel 604.
  • Request packets are sent on a request transport interface 606 and response packets are received on a response transport interface 608.
  • Transaction addresses are decoded in a decode unit (D) 610 to produce a route id field in packet headers.
  • Packets are associated with contexts in a context allocation unit A 612. If necessary, transactions are split into multiple packets in a packet splitting unit S 614. The packets are sent on the request transport interface 606.
  • D decode unit
  • a context array 620 comprises entries that are allocated to pending transactions. If the array is full, then the initiator NIU 600 asserts backpressure on the socket interface request channel 602. In that state the physical layer ready signal is deasserted.
  • the context array 620 stores, among other information, the sequence id of each pending transaction.
  • the context array 620 has four entries and therefore supports up to four pending transactions.
  • a context reassociation unit (R) 630 reassociates the packet with the context of the pending transaction.
  • the context array 620 carries TrID information.
  • the context array stores a TrID field (T) with each array entry.
  • T TrID field
  • the value stored in the field is that of a signal on the socket interface request channel 602 at the time that a transaction request is granted.
  • a signal on the interface is driven from the TrID field in the context array entry associated with the transaction for which the response is presented.
  • TrID field of the context array is included in the header of packets generated to transport the associated transaction. It forms the Echo field in the packet header.
  • the master and slave side of the socket interface have the same number of contexts and the TrID values are unique per context.
  • the TrID from the master forms a direct index into the context array in the initiator NIU.
  • the request packet header produced by and the response packet header received by the initiator NIU have an Echo field that is a direct index into the context array. In this way, no storage is required for TrID value in the context array and no remapping is required between TrID signals and packet header Echo fields.
  • an initiator NIU 700 is connected to an upstream target NIU through a NoC to NoC socket having a request channel 702, which is connected to a decode unit (D) 710, and a response channel 704.
  • the upstream target NIU (not shown) supports the same number of contexts as the initiator NIU 700. In this case, a context array is not needed within the initiator NIU. Contexts are managed only in the target NIU.
  • the TrID value of the NoC socket protocol is used directly as the Echo field in packets sent on a NoC transport request channel 706 and received on a NoC transport response channel 708. The echo field of response packets directly drives the response TrID signal on the socket response channel 704.
  • An optimal configuration of an initiator NIU varies from one NoC to another and from one chip to another.
  • the initiator NIU is configurable at chip design time. Configuration options include, among others, the width of the request and response TrID signals, the width of the TrID field in the context array of the NIU, the number of context array entries the number of pending transactions, and the mapping of TrID signals to a packet header Echo field. The number of pending transactions should be matched between an initiator NIU and its connected IP.
  • an initiator NIU If an initiator NIU supports more pending transactions than the IP then it will never use all of its context array entries. If the initiator NIU supports fewer pending transactions than the IP then it will assert back pressure on the IP even if there is network availability in the NoC. If the number of pending transactions is not a power of two then the upstream target NIU and downstream initiator NIU must agree on the encoding of valid TrIDs. If TrIDs are unique to transactions and the number of unique TrID encodings is less than the maximum number of pending transactions then the downstream initiator NIU must correctly reassociate responses to the TrID of their corresponding requests.
  • the initiator NIU logic described in a RTL language, is generated by a configuration tool.
  • the tool is used by chip designers to generate customized configurations and produce RTL language code for chip synthesis.
  • a chip design strategy that is beneficial for synthesis and layout is to create one or more client NoCs and one memory NoC that supports interleaved access to one or more memories such as double data rate DDR dynamic random access memory DRAM.
  • the memory NoC has one or more NoC to NoC initiator NIUs to receive and service requests from the client NoCs.
  • An embodiment of initiator NIU for such a configuration comprises a reorder buffer.
  • a reorder buffer is like an extended context array with an ability to store partial transaction responses.
  • TrID may be stored in the context array within the reorder buffer.
  • an initiator NIU in the downstream NoC that is coupled to a target NIU in the upstream NoC consists of no splitting unit. Splitting is performed at one or more initiator NIUs in the upstream NoC with knowledge of the splitting requirements of the initiator NIU in the downstream NoC. In this way the embodiment of the initiator NIU in the downstream NoC is smaller, faster, and has less transaction latency.
  • Some parameters of the downstream NoC considered by the initiator NIU of the upstream NoC are the address map seen by the target NIU of the downstream NoC, the maximum burst length of targets in the downstream NoC, the data width of target socket interfaces in the downstream NoC, and downstream NoC target socket protocol restrictions on bursts crossing boundaries such as the AXI protocol restriction that bursts not cross a 4k byte aligned address. Eliminating the splitting logic of the initiator NIU of the downstream NoC minimizes logic area, timing path length, and transaction latency.
  • An aspect of the invention relates to the embodiment of the target NIU that transacts with the initiator NIU at the NoC socket interface.
  • a target NIU that uses the NoC socket protocol maps the TrID signal as a function of an Echo field in a packet header.
  • the mapping function is a direct copy.
  • the target NIU comprises a context array.
  • Array entries store information about the transaction, which can include information from the request packet that caused the generation of the transaction. For each transaction request an entry in the array is allocated.
  • the value of the Response TrID signal at the socket interface forms an array index. It points to the array entry that holds the information related to the corresponding transaction.
  • the number of entries in the context array defines a maximum number of simultaneously pending transactions that can be supported by the target NIU.
  • the number of bits of the TrID signal which forms an unsigned binary index into the array, is the base two logarithm of the number of pending transactions, rounded up to an integer. Every pending transaction has a unique TrID value.
  • An optimal configuration of a target NIU varies from one NoC to another and from one chip to another.
  • the target NIU is configurable at chip design time. Configuration options include, among others, the width of the request and response TrID signals, the number of context array entries, the number of pending transactions, and the mapping of TrID signals to a packet header Echo field. The number of pending transactions should be matched between a target NIU and its connected IP. If a target NIU supports more pending transactions than the IP then it will have more context array entries than necessary. If the target NIU supports fewer pending transactions than the IP then it will never use all of its context array entries.
  • the target NIU logic is generated by a configuration tool.
  • the tool is used by chip designers to generate customized configurations and produce RTL language code for chip synthesis.
  • the combination of an initiator NIU and target NIU that use the enhanced NoC socket protocol enables the implementation of a superior multi-NoC composition of an upstream NoC and downstream NoC within an on-chip interconnect.
  • a packet header includes an Echo field that is used by an initiator to efficiently map responses to context array entries.
  • the NoC composition uses the enhanced NoC socket protocol above for the socket interface between a target NIU of an upstream NoC and a connected initiator NIU of a downstream NoC.
  • the value of the TrID signal of the transaction interface carries the Echo field of request packet of the upstream NoC to the initiator NIU of the downstream NoC.
  • the initiator NIU of the downstream NoC in turn uses the value of the TrID signal as the Echo field within one or more packets that it creates to carry out the protocol interface transaction.
  • FIG. 9 an embodiment of a NoC composition is shown in FIG. 9 and includes an upstream NoC 910 coupled to downstream NoC 950, both together transmitting a transaction request from left to right and a response from right to left.
  • Initiator NIU 920 receives a transaction request from an initiator IP (not shown), allocates a context entry in context array 922, and sends a request packet to target NIU 930.
  • Target NIU 930 receives the request packet and allocates a context entry in context array 932 that includes the Echo value from the request packet header.
  • Target NIU 930 issues transaction request 982, including a TrID signal, to initiator NIU 960.
  • Initiator NIU 960 allocates a context entry in context array 962, in which it stores the TrID according to the invention. Initiator NIU 960 sends request packet 984 to target NIU 970. Target NIU 970 receives the request packet and allocates a context entry in context array 972 that includes the Echo value from the request packet header. Target NIU 970 issues a transaction request to a target IP (not shown). The target IP issues a transaction response to target NIU 970, wherein reassociation logic 974 retrieves the context entry corresponding to the transaction from context array 972.
  • Target NIU 970 sends response packet 986 to initiator NIU 960, wherein table lookup module 964 uses the Echo packet header field to simply retrieve the context entry corresponding to the transaction from context array 962.
  • initiator NIU 960 sends transaction response 988 to target NIU 930, wherein table lookup module 934 uses the TrID signal to simply retrieve the context entry corresponding to the transaction from context array 932.
  • Target NIU 930 sends response packet 990 to initiator NIU 920, wherein table lookup module 924 uses the Echo packet header field to simply retrieve the context entry corresponding to the transaction from context array 922.
  • Initiator NIU 920 sends the transaction response to the initiator IP.
  • the value of the TrID signal is stored in the context array of the initiator NIU of the downstream NoC.
  • the stored TrID value is presented at the transaction interface with the response.
  • TrID is not stored but implicit in the echo field.
  • the initiator NIU of the upstream NoC configured with awareness of the address map and transaction protocol support of targets in the downstream NoC, performs all required splitting of initiator IP requests into multiple packets.
  • the initiator NIU of the downstream NoC therefore requires no packet splitting logic.
  • This has the further benefit of a simple mapping of the Echo field of the upstream NoC packet header to the Echo field of the downstream NoC packet header.
  • context management within the target NIU of the upstream NoC is simplified to a table access. That is a much faster structure than the chained list lookup of a conventional NoC composition.
  • the complexity of address decoding, splitting, and context association is present in the upstream initiator NIU, near the initiator IP, and generally distributed away from congested parts of the chip.
  • the upstream initiator NIU must therefore know certain properties of the targets of the downstream NoC such as address cross boundary restrictions and maximum bust lengths, among others.
  • the initiator NIU of the upstream NoC is unaware of the address map implemented in the initiator NIU of the downstream NoC.
  • the downstream NoC has a non-trivial address decode module and splitting and association logic as necessary.
  • a cross boundary parameter for the socket interface is agreed between the upstream and downstream NoC such that the upstream initiation NIU will split transactions in a way that the downstream initiation NIU need not.
  • the complexity of address decoding, splitting, and context association is present in the downstream NoC.
  • downstream NoCs are closer to performance critical memories and necessarily more central to the chip.
  • An aspect of the disclosed invention relates to the method of managing contexts within the target NIU of a socket interface between a target NIU of an upstream NoC and an initiator NIU of a downstream NoC.
  • One embodiment comprises allocating a free context for each new incoming Request using a pooling or a stack-based algorithm.
  • the context is identified by an Echo and/or a sequence ID field within the packet header.
  • Each context array entry contains the necessary information for building the response header:
  • the target NIU issues a request for a transaction, associated with the allocated context array entry.
  • the request includes a Request TrID signal.
  • the target NIU assigns the signal according to the index of the context array entry that is allocated.
  • Eventually the target NIU receives a response.
  • the response includes a Response TrID signal.
  • the target NIU uses the value of the Response TrID signal as an index value to look up the associated entry of the context array.
  • the context is freed when either the header of a write response transport packet is sent or the last data of a read response transport packet is sent.
  • the Response TrID is a transformed copy of the Request TrID signal for the same transaction.
  • One such transformation is the reordering of bits. This enables the use of bank based storage in the context array.
  • Another transformation is the addition of information indicating which of a number of data words in a burst, separated by response interleaving, the partial response word is delivering.
  • An aspect of the invention is the elimination of the needs to perform reassociation of transactions to context array entries in the target NIU of the socket interface between NoCs. This benefit is furthermore useful for types of target IPs other than NoCs such as memory controller.
  • Memory controller like downstream NoC initiator NIUs, can also be designed to respond with simply mapped TrID signals that do not require complex reassociation logic.
  • target NIUs reassociate transaction responses to context array entries based on chained list of a sequence ID. This requires a costly chained list lookup.
  • using a TrID signal on a transaction interface allows a simple table index lookup to reassociate responses with context array entries.
  • Target NIU 800 is shown that can correspond with an initiator NIU such as initiator NIU 600 or 700.
  • Request packets are accepted on a request transport interface 802 and response packets are sent on a response transport interface 804.
  • Transaction requests are presented through a socket interface request channel 806 and transaction responses are accepted by the NIU through a socket interface response channel 808.
  • Target NIU 800 comprises a context allocation unit 812, a context array 820, and a context reassociation unit 830.
  • a unique address for the table entry chosen by the context allocation unit 812 is sent as request TrID signal 840.
  • Response TrID signal 842 is used as the selector in a mux 850 to select the echo field E of the entry chosen by the context allocation unit 812. This is a lookup mechanism that eliminates the need for the traditional complex reassociation logic as well as simplify the overall system and enhance system performance.
  • Chips are increasingly complex. They can no longer be designed by a single engineer or even a single team of engineers. Chips are necessarily designed modularly. It is therefore necessary to design modules using separate NoCs. Top level integration of within the full chip requires interfaces between NoCs. Using a NoC socket protocol allows NoC interfaces to run faster and/or run with fewer pipe stages to close timing. Furthermore, using a NoC socket protocol allows teams to design independently without the need for a time consuming process of negotiating between teams on the best interface. In one embodiment a tool is used to configure and generate the NoC RTL. It automatically generates an optimized protocol specifically for each socket interface between NoCs. Among other parameters, the configuration determines the size of the TrID signal and the number of pending transactions supported.
  • the tool also accepts the address map of targets within the address space of initiators for each NoC. Using the address map of a downstream NoC the tool determines the type of splitting required to produce the one or more packets in the downstream NoC needed to complete a transaction. The tool configures the initiator NIU RTL to implement such packet splitting. The effect of this is to create a unified address map in the address decoding of initiator NIUs. This is done without requiring separate design teams to know the address map of each other's NoC. [0062] The tool also generates a verification testbench and tests to exercise all transaction types on all routes between initiators and targets with interconnectivity. With awareness of the NoC composition, the tool generates a unified testbench and tests to exercise accesses from each initiator to all accessible targets even through NoC to NoC sockets and with the resulting hierarchies of address mappings.
  • the tool also generates a performance exploration simulation environment. With awareness of the of NoC composition, it generates a simulation environment that models the passage of transactions and their associated packets between NoCs.
  • a further benefit of easy integration of multiple NoCs is that the transport packet header format in a NoC is optimized based on its configuration. For example a NoC with many initiators and targets requires a larger routelD field or a NoC with many sideband signals will require a larger user bit field. A large header increases the size of datapaths and complexity of logic throughout the NoC. By using multiple NoCs, with a transaction interface in between, the header formats of the separate NoCs can be optimized separately and generally made smaller.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Hardware Design (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Microelectronics & Electronic Packaging (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)
  • Communication Control (AREA)
  • Computer And Data Communications (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

L'invention concerne un protocole d'interface de transaction tel que le protocole d'interface comprend un signal identificateur de transaction dans chacun des canaux de requête et de réponse. Il est utilisé entre une unité d'interface réseau cible (NIU) maîtresse et une NIU esclave initiatrice qui sont directement connectées par une interface de transaction. Le canal de réponse de la NIU cible utilise le signal d'ID de transaction pour identifier l'entrée dans un tableau de contexte associé à la requête correspondante. L'accouplement de la NIU cible et de la NIU initiatrice permet la formation d'une interconnexion sur puce comprenant plusieurs réseaux sur puce (NoC), la topologie de l'interconnexion étant plus simple, plus petite, plus rapide et ayant une latence plus faible.
EP13842232.4A 2012-09-25 2013-09-24 Protocole de socket de réseau sur puce Ceased EP2901294A4 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
EP22196229.3A EP4123468A1 (fr) 2012-09-25 2013-09-24 Protocole de socket de réseau sur puce

Applications Claiming Priority (3)

Application Number Priority Date Filing Date Title
US13/626,758 US9471538B2 (en) 2012-09-25 2012-09-25 Network on a chip socket protocol
US13/626,766 US9225665B2 (en) 2012-09-25 2012-09-25 Network on a chip socket protocol
PCT/US2013/061295 WO2014052261A1 (fr) 2012-09-25 2013-09-24 Protocole de socket de réseau sur puce

Related Child Applications (1)

Application Number Title Priority Date Filing Date
EP22196229.3A Division EP4123468A1 (fr) 2012-09-25 2013-09-24 Protocole de socket de réseau sur puce

Publications (2)

Publication Number Publication Date
EP2901294A1 true EP2901294A1 (fr) 2015-08-05
EP2901294A4 EP2901294A4 (fr) 2016-08-10

Family

ID=50388890

Family Applications (2)

Application Number Title Priority Date Filing Date
EP22196229.3A Pending EP4123468A1 (fr) 2012-09-25 2013-09-24 Protocole de socket de réseau sur puce
EP13842232.4A Ceased EP2901294A4 (fr) 2012-09-25 2013-09-24 Protocole de socket de réseau sur puce

Family Applications Before (1)

Application Number Title Priority Date Filing Date
EP22196229.3A Pending EP4123468A1 (fr) 2012-09-25 2013-09-24 Protocole de socket de réseau sur puce

Country Status (6)

Country Link
EP (2) EP4123468A1 (fr)
JP (1) JP6144348B2 (fr)
KR (1) KR101690568B1 (fr)
CN (1) CN104685480B (fr)
IN (1) IN2015MN00441A (fr)
WO (1) WO2014052261A1 (fr)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB2529217A (en) * 2014-08-14 2016-02-17 Advanced Risc Mach Ltd Transmission control checking for interconnect circuitry
US11436185B2 (en) * 2019-11-15 2022-09-06 Arteris, Inc. System and method for transaction broadcast in a network on chip
CN117389931B (zh) * 2023-12-12 2024-05-03 芯动微电子科技(武汉)有限公司 适用于总线访问gpu核内存储器的协议转换模块及方法

Family Cites Families (22)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6477646B1 (en) * 1999-07-08 2002-11-05 Broadcom Corporation Security chip architecture and implementations for cryptography acceleration
US6826191B1 (en) * 1999-10-01 2004-11-30 Stmicroelectronics Ltd. Packets containing transaction attributes
US8087064B1 (en) * 2000-08-31 2011-12-27 Verizon Communications Inc. Security extensions using at least a portion of layer 2 information or bits in the place of layer 2 information
WO2002069575A1 (fr) * 2001-02-28 2002-09-06 Gotham Networks, Inc. Procedes et systeme utiles pour dispositif de routage de reseau
US6996651B2 (en) * 2002-07-29 2006-02-07 Freescale Semiconductor, Inc. On chip network with memory device address decoding
US7277449B2 (en) * 2002-07-29 2007-10-02 Freescale Semiconductor, Inc. On chip network
US6671275B1 (en) 2002-08-02 2003-12-30 Foundry Networks, Inc. Cross-point switch with deadlock prevention
CN100342370C (zh) * 2002-10-08 2007-10-10 皇家飞利浦电子股份有限公司 用于交换数据的集成电路和方法
US7181556B2 (en) * 2003-12-23 2007-02-20 Arm Limited Transaction request servicing mechanism
EP1735712A1 (fr) * 2004-03-26 2006-12-27 Koninklijke Philips Electronics N.V. Circuit integre et procede d'abandon d'une transaction
US7716409B2 (en) * 2004-04-27 2010-05-11 Intel Corporation Globally unique transaction identifiers
WO2007033363A2 (fr) * 2005-09-13 2007-03-22 Ist International, Inc. Systeme et procede permettant d'obtenir une connectivite par paquets entre des reseaux heterogenes
CN101379841A (zh) * 2005-09-13 2009-03-04 Ist国际公司 为提供异种网络之间的小包连通性和组分和小包之系统和方法
US20070245033A1 (en) * 2006-04-14 2007-10-18 Microsoft Corporation Link layer discovery and diagnostics
CN101501651A (zh) * 2006-08-08 2009-08-05 皇家飞利浦电子股份有限公司 电子设备和控制通信的方法
US8285912B2 (en) * 2009-08-07 2012-10-09 Arm Limited Communication infrastructure for a data processing apparatus and a method of operation of such a communication infrastructure
FR2951342B1 (fr) * 2009-10-13 2017-01-27 Arteris Inc Reseau sur puce a latence nulle
EP2333830B1 (fr) * 2009-12-07 2014-09-03 STMicroelectronics (Research & Development) Limited Un ensemble comprenant une première et une seconde matrice couplées par un bus multiplexé
EP2388707B1 (fr) * 2010-05-20 2014-03-26 STMicroelectronics (Grenoble 2) SAS Procédé et dispositif d'interconnexion, par exemple pour des systèmes sur puce
WO2011148925A1 (fr) * 2010-05-24 2011-12-01 日本電気株式会社 Dispositif semi-conducteur et procédé et système d'acheminement des données
FR2961048B1 (fr) 2010-06-03 2013-04-26 Arteris Inc Reseau sur puce avec caracteristiques de qualite-de-service
EP2444903A1 (fr) * 2010-09-29 2012-04-25 STMicroelectronics (Grenoble 2) SAS Agencement de réordonnancement

Also Published As

Publication number Publication date
CN104685480B (zh) 2017-07-14
CN104685480A (zh) 2015-06-03
KR101690568B1 (ko) 2016-12-28
IN2015MN00441A (fr) 2015-09-11
KR20150063433A (ko) 2015-06-09
JP6144348B2 (ja) 2017-06-07
EP2901294A4 (fr) 2016-08-10
EP4123468A1 (fr) 2023-01-25
JP2015535991A (ja) 2015-12-17
WO2014052261A1 (fr) 2014-04-03

Similar Documents

Publication Publication Date Title
JP6522663B2 (ja) ハイブリッドメモリキューブリンクを用いる相互接続システムおよび方法
US8478982B2 (en) Media access control security management in physical layer
US7111101B1 (en) Method and system for port numbering in an interconnect device
US9537488B1 (en) Apparatus for configurable interface and associated methods
JP6201065B2 (ja) 再構成可能なメモリシステムのための仮想化された物理アドレス
TWI772279B (zh) 用於具有可重配置多端口的快捷外設互聯標準儲存系統的服務品質感測輸入輸出管理的方法、系統及設備
US9021169B2 (en) Bus system including ID converter and converting method thereof
US7277975B2 (en) Methods and apparatuses for decoupling a request from one or more solicited responses
US9225665B2 (en) Network on a chip socket protocol
CN110442534A (zh) 用于相干消息的高带宽链路层
US9471538B2 (en) Network on a chip socket protocol
US9361239B2 (en) System on chip including unified input/output memory management unit
CN110795376A (zh) 具有用于访问验证的基于查询的地址转换的系统架构
EP4123468A1 (fr) Protocole de socket de réseau sur puce
US8533377B2 (en) System and method for allocating transaction ID in a system with a plurality of processing modules
US20220121452A1 (en) On-demand programmable atomic kernel loading
CN210038775U (zh) 片上系统
US7818468B2 (en) Method and system for processing control information
CN117280331A (zh) 用于处理来自通信网络中的外围设备的请求传输的技术
REDDY et al. High Speed AMBA AXI4 Interface Protocol for System on Chip Data Transactions

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

17P Request for examination filed

Effective date: 20150304

AK Designated contracting states

Kind code of ref document: A1

Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

AX Request for extension of the european patent

Extension state: BA ME

DAX Request for extension of the european patent (deleted)
RIC1 Information provided on ipc code assigned before grant

Ipc: G06F 13/42 20060101ALI20160425BHEP

Ipc: G06F 13/00 20060101AFI20160425BHEP

Ipc: G06F 15/78 20060101ALI20160425BHEP

RA4 Supplementary search report drawn up and despatched (corrected)

Effective date: 20160712

RIC1 Information provided on ipc code assigned before grant

Ipc: G06F 13/00 20060101AFI20160704BHEP

Ipc: G06F 13/42 20060101ALI20160704BHEP

Ipc: G06F 15/78 20060101ALI20160704BHEP

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: EXAMINATION IS IN PROGRESS

17Q First examination report despatched

Effective date: 20200319

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: EXAMINATION IS IN PROGRESS

REG Reference to a national code

Ref country code: DE

Ref legal event code: R003

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION HAS BEEN REFUSED

18R Application refused

Effective date: 20220807