Communication system between different modules and method

The present invention relates to the field of the communication between modules.

In the field of IT, conventional systems are centered around a specialized bus that interconnects several Agents them. Here, the term "agent", likely a factor receiving or transmitting data on a bus, for example a CPU, a memory, an input / output device, etc.

Following a specific protocol and own the bus, only one agent can simultaneously transmit a message on the bus. Running a Treatment includes having operands present in the agent that will perform the treatment and consequently demands the transfer data between agents through the system communication. Here, the term "operand" an element on which a door operation.

For example, a system comprising a coprocessor arithmetic and SDRAM memory that contains data must, for a transaction to be executed, transfer operands stored in the memory via the controller of memory, then via bus to the coprocessor. The results of the operation will be transferred on the same bus.

This results in a limitation of the number of exchanges due to the use of a bus which forms a communication resource unique and limited, the bandwidth of the communication system formed by the bus being finished.

In applications where high performance is necessary, the execution of a program requires organization specific instructions to optimize and sequence to best various exchanges between the agents that are connected to the bus communication, to optimize the use of bandwidth. The compilers can handle this type of problem, but a non-optimal way and with a portion of the resources system is used for the management of these exchanges. The system loses then efficiency and performance degrades more quickly a high number of agents is present on the bus communication.

To overcome these drawbacks, it is desirable that several requests can be transmitted simultaneously between various agents and that all agents working in parallel and sequentially without the need to have a General Sequencer to sequence the general flow of data.

The invention aims to remedy the drawbacks of the system vector mentioned above.

The invention provides a system and method of particularly efficient communication, allowing extremely fast communication between different agents while using wisely the capacity of the system communication.

The communication system between the modules, according to a aspect of the invention comprises a starter element, at least one intermediate element and a slave element. The initiator element equipped with means for generating a first request and means to send the first query to the intermediate element.

The first request includes the address of the initiator element, the address of the intermediate element, data and at least one instruction to be processed by the intermediate element. The element intermediate is equipped with means for receiving a request in from the starter element, means to develop a second request in response to the first request and means for envoyer la deuxième requête à un élément esclave.

The second request includes the address of the slave element, at least one instruction to be processed by the slave element of data, and the address of the initiator element. The slave element equipped with means for receiving a request from the intermediate element, and means for executing at least one statement contained in the second query.

It thus allows complex operations that require a small number of requests or messages. It can generally be pass messages sent by the intermediate element to the element initiator or messages sent by the slave element to the intermediate element.

In one embodiment of the invention, the slave element includes means for sending an acknowledgment to the element initiator in response to the second request. The acknowledgment will sent without passing through the element. This reduces the number of messages that go through the communication system.

In one embodiment of the invention, the slave element comprises means for sending data to the initiator element.

In one embodiment of the invention, the system comprises a plurality of intermediate elements, an element intermediate nth comprising means for sending a intermediate element of rank n + 1 query a rank n + 1 in response a request of rank n. The n + 1 ranking query includes the address of through element rank n + 1, at least one instruction before be processed by the intermediate element of rank n + 1, and the address of the initiator element.

A query can therefore move in a significant number intermediate elements, each performing there for treatment wherein there is provided and sending to the element following a new request.

Different applications may include a field address of the sending element of said request, this field being preferably filled with the element of the initiator address. So, an intermediate element sees the request as being directly issued by the initiator element and not by the intermediate member, immediately preceding row.

In one embodiment of the invention, the element intermediate comprises means for generating a third query, the third query comprising the element address initiator, the address of an intermediate element, data, and least one instruction to be processed by the intermediate element, and means for sending a third request to said element intermediate.

In one embodiment of the invention, said element intermediate includes means for developing a fourth application the fourth request including the address of an element slave, data, and at least one instruction to be processed by the slave element, and means for sending the fourth request to said slave element.

The slave element may include means for receiving the fourth request from the intermediate element, and means for executing at least one instruction contained in the fourth request. Such a system is particularly well suited in case the slave elements are memories, and elements Intermediaries are memory controllers.

Advantageously, the modules are part of an integrated circuit. It is particularly advantageous that a single integrated circuit include the entire system, including the modules.

However, it is expected that the modules are part of a plurality of integrated circuits disposed in a common carrier, and / or This removable media connected. said Removable media can be connected by a son ply, a wire coaxial, of the differential pairs, laser diodes, etc. A support may be formed by a printed circuit board provided with elements welded or detachably fixed.

In one embodiment of the invention, a request subsequently, as the second request includes the address of the initiator element as the address of the transmitter of the second request.

The invention also provides a communication method between modules in an interconnection system comprising at least three elements, wherein:<ul><li>an initiator element sends a first request to intermediate member, said first request comprising the address of the initiator element, the element address intermediate, data, and at least one instruction to be processed by the intermediate element,</li><li>in response to the first request, the intermediate element sends to the slave element comprising a second request the address of the initiator element, the element address slave, data, and at least one instruction before be processed by the slave element, and</li><li>in response to the second request, the slave element performs the statement contained in the second query.</li></ul>

It is thus possible to optimize the use of a resource communication system by limiting the number of queries or issued messages.

In one embodiment of the invention, in response to the second request, the slave element sends further acknowledgment of receipt to the initiator element. The slave element can send data to the initiator element or not to issue a response. We means slave element, an element at the end of a string agents involved in a transaction, likely to perform instruction, and to report to the initiator element of the transaction. In other words, the response sent by a slave element to the initiator element is uneducated. Thus, the transaction may include a plurality of requests each comprising at least one instruction and, where appropriate, a response consisting of a transaction ID, the address of the initiator element, an accused for receiving and / or data.

In one embodiment of the invention, in response to the first request, sends the intermediate element to another element through a request including the address of the initiator element, the address of said another intermediate element, data, and at least an instruction to be processed by said another member intermediate.

In one embodiment of the invention, in response to a ranking query n, an intermediate element of rank n sends a intermediate element of rank n + 1, a request of rank n + 1 including the address of the initiator element, the element address intermediate of rank n + 1, data and at least one instruction to be treated through element row n + 1.

Advantageously, the elements, initiators, intermediate and slave of the interconnection system, according communicate the same protocol.

In one embodiment of the invention, the first request includes the address of the slave element, and the number of data to transfer. In other words, the initiator element develops a query Copy comprising the address of the initiator element, the address of the intermediate element, an instruction to be processed by the intermediate element, the slave element address and the number of data used, and sends the request to the intermediate element. The intermediate element is developing a storage request including the address of the initiator element, the address of the slave element, a instruction to be processed by the slave element, data resulting from the execution of the instruction processed by the element intermediate storage and sends the request to the slave element. The slave unit executes the instruction by storing data.

The instruction to be processed by the slave element is an instruction said data loading. The data contained in the copy request includes the address of the intermediate element, and the address of the slave element. A copy of statement of an item to another can be performed with a small number of issued requests, thus increasing the availability of resources of the communication system.

In one embodiment of the invention, a request includes the address of the initiator element, the element address recipient of the request, an operational instruction, the address of data in a repository of the receiving element of the application, and an identifier of the application. The operational instruction can be selected from a load instruction, storage or copying. An application may contain the number of bytes of data in said request. An application may contain an element for make the connection between a request and a response. An initiator element can send a request before receiving a response to a request previously sent. This is particularly advantageous since the initiator element is not slowed by awaiting a response to a request for sending a subsequent request. This achieves increase data processing speed by the element initiator.

The invention therefore provides not only a saving on communication resources of a communication system, but also an increase in the data processing speed by different agents.

The invention will be better understood on reading the description Detail of some embodiments taken as examples in no way limiting and illustrated by the appended drawings, in which:<ul><li>Figure 1 is a block diagram of a management device communication messages of an integrated circuit chip;</li><li>Figure 2 is a block diagram of a variation of a communication management device of an integrated circuit chip;</li><li>Figure 3 is a block diagram of a management module asynchronous communication messages; and</li><li>Figures 4 to 7 are block diagrams of a sys me communication according to one aspect of the invention.</li></ul>

In Figure 1, there is shown a one chip integrated circuit with an orientation indicator 2, and 3 feet for communication with the outside of said chip 1. The components of the 1 chip, more and more, are grouped into multi-synchronous modules 4. by multi-synchronous modulus means a subset clocked circuit through one or more dedicated clocks said module.

An asynchronous communication management module 5 messages, connected to at least two multi-synchronous modules 4, here all multi-synchronous modules 4, by point-to-point connections 6, manages communication between the various modules 4 of the chip 1 integrated circuits. Modules 4 communicate with the outside of the chip 1 by connections 7 between the modules and legs 3 of the chip 1. Two multi-synchronous modules 4 can also, if design constraints require it, communicate directly with a point-to-point connections 6 identical to point-to-point 6 between the asynchronous communication management module 5 and each module multi-synchronous 4. However, it is preferable that the communications between two synchronous modules 4 pass through the management module Asynchronous communication 5.

8 a set of at least one clock, dedicated to a single synchronous multi-module determines the timing references operating frequency of said synchronous multi-modulus 4. It is Note that the module 5 is a special case of multi-synchronous module and it also has its own set of 8 to least one clock.

A single specific predetermined communication protocol is used for communication between said chip 1 modules taking place on routes 6.

Each point-to-point 6 comprises, for each end, means for transmitting and receiving means messages, each receiving means including a memory Type FIFO, commonly known as "FIFO".

On a link 6, the clock data and other data are sent simultaneously. During transport, data clock undergo a phase shift, and the finish clock data do not have the same phase as the outgoing clock data. The Other data then undergo a temporal dispersion. If the time dispersion is too large, a re-synchronization is performed.

2 shows a variant of the device of Figure 1. The device comprises at least two management modules Asynchronous communication 5, the next two are represented. He can be necessary, in view of the complexity of a circuit and its partition a large number of multi-synchronous modules, have several asynchronous communication management modules 5. Moreover, the circuit may include one or more modules 9, only one of which is shown here, not connected to the communication management modules Asynchronous 5. This device can be seen from the outside as a single asynchronous communication management module for recover a device similar to that of Figure 1.

However there is a difference, because the fact of putting several asynchronous communication management modules requires additional steps for communication between two modules synchronous, including 6 additional connection for the transfer data exchange, and several successive treatments corresponding to the number of communication management modules asynchronous.

Figure 3 is a block diagram of a management module asynchronous communication 5. This includes inputs 5a receiving data messages from other modules and 5b outputs transmission data messages to other modules. The input messages are stored temporarily by example in a first type of first-out memory, commonly known as "FIFO" receiving means messages in a point-to-point link 6. This temporary storage enables to manage the incoming messages, even if they manage to a rate greater than the message processing capacity Module 5. The asynchronous communication management module 5 further comprises means 5c for routing messages output to their destination module, as well as means for 5d manage referrals output messages. An input message thus stored in the FIFO of the message receiving means a point-to-point 6, and then taken into account by the module 5, which will manage its referral to the synchronous module 4 by destination 5c and 5d means. This message then leaves the module 5 by output 5b and passes through the means for transmitting a connection message point-to-point 6 to synchronous module 4 destination of the message.

As illustrated in Figure 4, the communication system comprises an element 11, called "initiator component" capable of generating a transaction resulting in a plurality of requests, an element Intermediate 12, able to handle a request and send a request, an intermediate element 13, adapted to receive a request and transmitting a query, and a slave element 14, adapted to send the item initiator 11 data and / or acknowledgment. The element initiator initiates the transaction request including by developing a the address of the initiator element, the address of the intermediate element 12, an operational instruction to be executed by the element Intermediate 12 and, in general, data. The initiator element 11 also includes said request issuing means destination of intermediate element 12.

The intermediate element 12 receives the request sent by 15 the initiator element 11 and performs processing on the received data according to the operational instruction contained in the Application 15 received. The intermediate element 12 comprises means to develop a second query 16 for the intermediate element 13 of the the Offeror element 11, and a means for sending said 16. The request 16 request sent via member 12 the intermediate element 13 includes the element address Intermediate 13, the address of the starter element 11 as an address of the sender of said request, an instruction operational to be executed via 13 agent, and data.

The intermediate element 13, which also includes means for generating and sending request means, reception the request 16 sent through element 12, performs the treatment defined operational instruction entered in said Query 16 and develops a request 17 containing, as the Previous, the destination address of said request 17 address the initiator element 11 as the address of sender of said request, data and operational instructions. The request generated 17 by the intermediate element 13 can be sent directly to the slave member 14 or to another intermediate element, not shown, the transaction can go through an unlimited number of intermediate elements between the intermediate member 13 and the slave element 14.

The slave element 14, upon receiving a request from 17, or of the intermediate member 13 or either of the last element through the chain of intermediate elements, performs treatment defined operational instruction contained in the received request and in response to said request, may transmit to the initiator element 11 data and / or acknowledgment. The slave element 14 can also transmit anything to the element initiator 11.

It is understood that any request received by an element intermediate or by the slave element comprises, as an address transmitter, the address of the starter element 11. Thus, each element intermediate and slave element receives an identical request to the he would have received if it was sent directly to the element initiator 11 to the element. It is thus possible to avoid operating protocol changes.

For example, one can provide various types operational instructions may be contained in the requests: a store instruction without acknowledgment, an instruction storage with acknowledgment of receipt, an instruction procurement and storage in memory, an instruction supply a copy of statement without acknowledgment receipt and a copy instruction with return receipt.

For example, a storage request may include a field transmitter addresses, filled with the element address initiator 11, an address field of the receiving element of the request, message size field, an operational code defining operational training and a data field. A supply request has the same structure, with the exception of data field. A copy of the request has the same structure, this except that the data field is filled with the element address slave destination.

It is particularly advantageous that each request further includes a relatively short field that allows to the relationship between a request and a response, to the extent that the element Offeror does not expect the response to a first request for issue another query.

5 illustrates the case of an application of the invention in telecommunications. In this area, some equipment act as data disintegrating. A cutterhead data input receives line broadband communications, carrying a very large number of flows, for example thousand, which are multiplexed. It sends out a thousand independent lines. Such disintegration data requires the management of a context dependent on each line. here means context, the definition of the requirements of each line. memories of Type FIFO are generally used to storing incoming packets associated with a line. Then, Periodically, these memories, FIFO are emptied to rows interfaces. If the number of memories is high, thousand example, the hardware implementation with a stream through memory particularly expensive.

The invention provides a first memory manager entered first out, reducing the number of memories with Dynamic allocation of memory to a line resulting from the execution of a management software with a hardware module.

In the example illustrated in Figure 5, there is provided an element Intermediate 12 formed by the memory manager and an element Slave 14 formed by a memory. The initiator sends a member 11 15 to request the intermediate element 12.

In the case of a copy transaction, the query 15 can include the address of the initiator element 11, the element address Intermediate 12, the number of bytes of data, operational code the desired instruction, eg storage with return reception, the address in the intermediate element 12, the address of the slave member 14, the address in the slave member 14, and a transaction identifier. The intermediate element 12, manager memories, transforming the memory address selected in a physical address present in the memory formed by the element slave 14.

The intermediate element 12 makes a request 16 including, as sender address, the address of the initiator element 11, as recipient address, the memory containing the data of the address selected, the number of transmitted data, the operation code storage with acknowledgment of receipt, the real address of the representative memory selected, the data and the transaction identifier.

Upon receipt of the request 16 of the intermediate element 12, the slave element 14, here a memory containing data line selected, updates the data of the selected memory then sends an acknowledgment to the initiator element 11.

Figure 6 illustrates the case of an unstacking instruction of data present in a memory and sending said data to the starter element 19, which here is another initiator element. The initiator element 19, initiator of the transaction, initiates a request 20 loading data from a corresponding memory virtually to the treated telecommunications line. The query the initiator 19 is addressed to the memory manager, alias the intermediate element 12, and includes the element address initiator 19, the address of the intermediate element 12, the number data bytes, the operation code of the store instruction with return receipt, the virtual address of the selected memory and the identifier of the transaction.

The intermediate element 12 performs the transformation of the virtual address of selected memory to a physical address present in the memory means the slave member 14 and containing the line of data processed by the initiator element 19. The element Intermediate 12 issues a request to supply 21 the slave element 14. The request includes as sender address, the address of the initiator element 19, the memory address containing the necessary data for the processed line, the number of bytes, operational instruction storage with return receipt, the real address representing the selected memory the transaction identifier.

Upon receipt of such request, the slave member 14 which contains updates the desired data from memory data selected then sends the data to the initiator element 19 in the as a response 22 containing the address of the initiator element 19, data and the identifier of the transaction.

7 illustrates the case of an intermediate element 12 receiving a request 16 from an intermediate element 12 who himself received a request from a 15 element initiator 11. The intermediate element 12 is developing a response and sends it to a slave element 23. The element 23 is developing a slave response 24 and sends it to the starter element 11. It is thus possible to transit a response by a number of elements intermediate according to the system architecture.