CN106980595B - Multiprocessor communication system and communication method for sharing physical memory - Google Patents

Abstract

A multiprocessor communication system and a communication method thereof for sharing physical memory are provided, the multiprocessor communication system includes: a plurality of processors, wherein the plurality of processors send and receive data with each other; the physical memory is divided into be a plurality of physical memory blocks, so that each processor has a dedicated physical memory block; wherein, the sending processor of the plurality of processors that will send data sends data to the one of the plurality of processors that will receive data The receiving processor's dedicated physical memory block, and the receiving processor reads the data from its dedicated physical memory block. According to the multi-processor communication system and the communication method thereof, when multiple processors transmit data to each other, the occupation of memory space can be reduced, the use efficiency of memory can be improved, and the transmission speed of data can be increased.

Description

Multiprocessor communication system and communication method for sharing physical memory

This application is a divisional application of the application submitted to the China Intellectual Property Office with the application date of December 5, 2014, the application number 201410734202.9, and the invention title "Multiprocessor Communication System and Communication Method for Shared Physical Memory".

technical field

The present invention relates generally to the design of multiprocessor communication systems. More specifically, it relates to a communication system and a communication method for realizing communication between multiprocessors by sharing physical memory.

Background technique

In the existing multiprocessor communication system, data communication is carried out between multiprocessors through serial communication interfaces (such as UART, USB, SPI, etc.), and the peripheral circuits are complex and the transmission speed is slow when using serial communication interfaces for communication. It is difficult to realize communication with a large amount of data or communication with high speed requirements between multiprocessors.

However, in the existing multi-processor communication system sharing physical memory, no peripheral circuit is required for data communication between multi-processors, and the access speed is fast and the real-time performance is strong. Specifically, in systems where multiprocessors communicate through shared physical memory, the physical memory is divided into dedicated blocks of physical memory so that each processor has a dedicated block of physical memory exclusively for running its own In addition, for inter-processor data communication, a physical memory block will be reserved in the physical memory, which is a global memory shared area, including a data shared cache area for storing multi-processor communication data Shared buffer with the state used to save and update the state of the data buffer.

Usually, in order to communicate in the existing multiprocessor communication system that shares physical memory, it is generally necessary to reserve a relatively large global memory shared area, and this part of memory space needs to be constructed according to a specific hardware system (for example, product design needs, the actual number of processors, the application scenarios of each processor, etc.) to formulate corresponding usage plans. There are several obvious problems with this approach:

First, the global memory shared area is reserved for a long time, which is a waste of memory (once the global memory shared area is statically planned, the determined data shared buffer area and state shared buffer area cannot be dynamically released or recycled for other purposes);

Second, the usage plan of the global memory shared area is inflexible (in the case of changing the memory layout for some reason, it is necessary to calculate and modify the parameters in the software system on each processor, which is troublesome and error-prone);

Third, the data communication process between multiprocessors requires two data copy operations (data needs to be copied from the sending processor to the global memory shared area, and then copied from the global shared memory area to the receiving processor), which is not efficient.

Therefore, when transferring data in an existing system that communicates among multiple processors by sharing physical memory, the usage efficiency of physical memory is low, and the speed of data transmission is relatively slow.

Contents of the invention

Exemplary embodiments of the present invention provide a multi-processor communication system and communication method for sharing physical memory, so as to solve the problems of low memory usage efficiency and slow data transmission speed in the existing multi-processor communication system.

According to an aspect of an exemplary embodiment of the present invention, there is provided a multiprocessor communication system sharing physical memory, including: a plurality of processors, wherein the plurality of processors send and receive data from each other; physical memory, divided into be a plurality of physical memory blocks, so that each processor has a dedicated physical memory block; wherein, the sending processor of the plurality of processors that will send data sends data to the one of the plurality of processors that will receive data The receiving processor's dedicated physical memory block, and the receiving processor reads the data from its dedicated physical memory block.

In the system, the sending processor may request the receiving processor to allocate a temporary data buffer for caching data in its dedicated physical memory block, and the receiving processor may allocate temporary data in response to the sending processor's request cache area.

In the system, the data may be burst data, wherein the temporary data buffer area for buffering the burst data may include a single buffer area.

In the system, the data may be streaming data, wherein the temporary data buffer for caching the streaming data may include a circular buffer composed of multiple slices.

The system may further include: a message box for temporarily storing burst data messages, wherein the burst data messages may include type information indicating burst data transmission events.

In the system, the burst data transmission event may include at least one of the following events: a request for allocation of a temporary data buffer, allocation of a temporary data buffer, data to be read in a temporary data buffer, temporary data buffer Data in the zone has been read.

In the system, when the burst data transmission event is a request to allocate a temporary data buffer, the burst data message may further include a request parameter about the request for the allocation of the temporary data buffer, wherein the request for the allocation of the temporary data The request parameters of the cache area include the size of the requested single cache area and the identification of the application, the sending processor and the receiving processor; when the burst data transmission event is that the temporary data buffer has been allocated, the burst data message can Including allocation parameters about allocating temporary data buffer areas, wherein the allocation parameters about allocating temporary data buffer areas may include the size and location of a single buffer area actually allocated and the identification of the application, the sending processor and the receiving processor; When the burst transmission event is that there is data to be read in the temporary data buffer, the burst data message may further include parameters about the data to be read in the temporary data buffer, wherein, the data to be read in the temporary data buffer The parameters for reading data may include the address of a single buffer area, the length of the data to be read, and the application identification, identification of the sending processor and the receiving processor; when the sudden transmission event is that the data in the temporary data buffer area has been read When , the burst data message also includes a parameter that the data in the temporary data buffer has been read, wherein the parameter that the data in the temporary data buffer has been read may include the address of a single buffer As well as the application ID, the ID of the sending processor, and the ID of the receiving processor.

The system may further include: a message box for temporarily storing stream data messages, wherein the stream data messages may include type information indicating stream data transmission events.

In the system, the stream data transmission event may include at least one of the following events: a request to establish a stream connection, a stream connection established, a fragmentation state changed, a request to disconnect a stream connection, and a stream connection disconnected.

In the system, when the stream data transmission event is a request to establish a stream connection, the stream data message may further include a request parameter for requesting to establish a stream connection, wherein the request parameter for requesting to establish a stream connection may include a composition The number of fragments of the requested circular buffer area and the size of each fragment and the identification of the application identifier, the sending processor and the receiving processor; when the streaming data transmission event is that the streaming connection has been established, the streaming data message can also be Including the unique identification of the stream connection and the allocation parameters of the actual allocation of the temporary data buffer area, wherein the unique identification of the stream connection is used to globally uniquely identify the established stream connection; the allocation of the actual allocation of the temporary data buffer area Parameters may include the number of fragments that make up the allocated circular buffer and the size, location and number of each fragment, as well as the identification of the application, sending processor, and receiving processor; when the stream data transfer event is a request to disconnect the stream When , the stream data message may further include a request parameter for disconnecting the stream connection, wherein the request parameter for disconnecting the stream connection may include the unique identifier and application identifier of the stream connection, the sending processor and the receiving processor when the stream data transmission event is that the stream connection has been disconnected, the stream data message may further include parameters about the stream connection being disconnected, wherein the parameters about the stream connection being disconnected may include information about the stream connection The unique ID of the , as well as the ID of the application, the ID of the sending processor, and the ID of the receiving processor.

In the system, the message box may be a dedicated area divided from the physical memory, or the message box may be a dedicated hardware register set in the multiprocessor communication system.

In the system, the message box can also temporarily store resident messages, where the resident messages are suitable for scenarios where messages are frequently delivered.

In the system, the multiprocessor communication system may be integrated in a single chip system, or the multiprocessor communication system may be integrated in a multiprocessor system sharing multi-port physical memory.

According to another aspect of an exemplary embodiment of the present invention, there is provided a communication method for a multiprocessor communication system sharing physical memory, wherein the multiprocessor communication system includes a plurality of processors and physical memories, and the plurality of Processors send and receive data from each other, the method comprising: (a) dividing physical memory into a plurality of physical memory blocks such that each processor has a dedicated physical memory block; (b) by one of the plurality of processors a sending processor that is to send data sends data to a dedicated physical memory block of a receiving processor of the plurality of processors that is to receive data; and (c) being read by the receiving processor from its dedicated physical memory block Get the data.

In the method, between steps (a) and (b), may further include: (d) the sending processor requests the receiving processor to allocate a temporary data buffer area for buffering data in its dedicated physical memory block (e) the receiving processor allocates a temporary data buffer area in response to the request of the sending processor; wherein, in step (b), the sending processor sends data to the allocated temporary data buffer area; in step (c) , the receiving processor reads the data from the allocated temporary data buffer.

In the method, the data may be burst data, wherein the temporary data buffer area for buffering the burst data may include a single buffer area.

In the method, the data may be stream data, wherein the temporary data buffer for caching the stream data may include a circular buffer composed of multiple fragments.

In the multiprocessor communication system and communication method for sharing physical memory according to an exemplary embodiment of the present invention, there is no need to reserve a global memory shared area, and multiple processors can transmit data to each other through their respective dedicated physical memory blocks, thereby Improve memory usage efficiency and speed up data transfer.

Description of drawings

The above and other objects of exemplary embodiments of the present invention will become more apparent from the following detailed description in conjunction with the accompanying drawings exemplarily showing embodiments, in which:

1 shows a block diagram of a multiprocessor communication system sharing physical memory according to an exemplary embodiment of the present invention;

2 shows a flowchart of a communication method of a multiprocessor communication system sharing physical memory according to an exemplary embodiment of the present invention;

FIG. 3 shows an example of communication between multiprocessors sharing physical memory according to an exemplary embodiment of the present invention;

4 shows a flowchart of a communication method of a multiprocessor communication system sharing physical memory according to another exemplary embodiment of the present invention;

Figure 5 shows an example of a message box according to an exemplary embodiment of the present invention;

FIG. 6 shows an example in which multiprocessors sharing physical memory communicate for burst data according to an exemplary embodiment of the present invention;

FIG. 7A shows a flowchart of a method for transmitting processors in a multiprocessor communication system to communicate with respect to burst data according to an exemplary embodiment of the present invention;

FIG. 7B shows a flowchart of a method for a receiving processor in a multiprocessor communication system to communicate with respect to burst data according to an exemplary embodiment of the present invention;

FIG. 8 shows an example in which multiprocessors sharing physical memory communicate with respect to stream data according to an exemplary embodiment of the present invention;

FIG. 9A shows a flowchart of a method for transmitting processors in a multiprocessor communication system to communicate with respect to stream data according to an exemplary embodiment of the present invention;

FIG. 9B shows a flowchart of a method for receiving processors in a multiprocessor communication system to communicate with respect to stream data according to an exemplary embodiment of the present invention.

Detailed ways

Reference will now be made in detail to the exemplary embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like numerals refer to like parts throughout.

FIG. 1 shows a block diagram of a multiprocessor communication system sharing physical memory according to an exemplary embodiment of the present invention.

As shown in FIG. 1 , a multiprocessor communication system sharing physical memory (hereinafter referred to as “multiprocessor communication system”) according to an exemplary embodiment of the present invention includes: a plurality of processors 100 and a physical memory 200 . Here, as an example, the multiprocessor communication system may be integrated in a single chip system, or the multiprocessor communication system may be integrated in a multiprocessor system sharing multi-port physical memory. The processor 100 may be a processor such as a central processing unit, a microprocessor, a physical processor, or a digital signal processor. The physical memory 200 may be random access memory (RAM).

Here, the plurality of processors 100 (for example, processor 100-1, processor 100-2, ..., processor 100-N shown in FIG. 1 ) transmit and receive data to and from each other.

Additionally, physical memory 200 is divided into multiple physical memory blocks such that each processor has a dedicated physical memory block. As shown in FIG. 1, the physical memory 200 can be divided into: a dedicated physical memory block 200-1 of the processor 100-1, a dedicated physical memory block 200-2 of the processor 100-2, ..., processing Dedicated physical memory block 200-N of device 100-N.

Correspondingly, the sending processor that will send data among the plurality of processors 100 sends data to the dedicated physical memory block of the receiving processor that will receive data among the plurality of processors 100, and the receiving processor The processor reads the data from its dedicated block of physical memory. Here, in the following exemplary embodiments, for convenience of description, it is assumed that the processor 100-1 indicates a transmission processor that will transmit data, and the processor 100-2 indicates a reception processor that will receive data. Here, it should be understood that the sending processor and the receiving processor are not limited to the processor 100 - 1 and the processor 100 - 2 , and may also be other processors among the multiple processors 100 . Additionally, a single processor can be used to both send and receive data. That is, the processor is used as a sending processor when sending data and as a receiving processor when receiving data.

Specifically, the processor 100-1 may send data to the dedicated physical memory block 200-2 of the processor 100-2, and the processor 100-2 may send data from the dedicated physical memory block 200-2 of the processor 100-2. 2. Read the data.

In the above-mentioned multiprocessor communication system sharing physical memory according to the exemplary embodiment of the present invention, there is no need to reserve a global memory shared area, and multiple processors can pass data to each other through their respective dedicated physical memory blocks, thereby improving memory utilization. Use efficiency and speed up data transfer.

FIG. 2 shows a flow chart of a communication method of a multiprocessor communication system sharing physical memory (hereinafter referred to as "multiprocessor communication method") according to an exemplary embodiment of the present invention. Here, the multi-processor communication system includes multiple processors and physical memory, and the multiple processors send and receive data with each other. As an example, the processor may be a processor such as a central processing unit, a microprocessor, a physical processor, or a digital signal processor. The physical memory may be random access memory (RAM). Furthermore, as an example, the multiprocessor communication system may be integrated in a single chip system, or the multiprocessor communication system may be integrated in a multiprocessor system sharing multi-port physical memory.

As shown in FIG. 2, in step S100, the physical memory is divided into multiple physical memory blocks, so that each processor has a dedicated physical memory block.

In step S200, the sending processor that will send data among the multiple processors sends data to the dedicated physical memory block of the receiving processor that will receive data among the multiple processors. At step S300, the data is read by the receiving processor from its dedicated physical memory block. Here, as an example, a single processor can be used for both sending and receiving data. That is, the processor is used as a sending processor when sending data and as a receiving processor when receiving data.

In the communication method of the multiprocessor communication system sharing physical memory according to the above-mentioned exemplary embodiment of the present invention, there is no need to reserve a global memory shared area, and multiple processors can transmit data to each other through their respective dedicated physical memory blocks, thereby Improve memory usage efficiency and speed up data transfer.

FIG. 3 shows an example of communication between multiple processors sharing physical memory according to an exemplary embodiment of the present invention.

As an example, as shown in FIG. 3, a plurality of processors 100 may include a sending processor (i.e., processor 100-1) and a receiving processor (i.e., processor 100-2), wherein processor 100-1 requests Processor 100-2 allocates temporary data buffer area B for buffering data in its dedicated physical memory block 200-2, and processor 100-2 allocates temporary data buffer area B in response to the request of processor 100-1 b. It should be understood that the buffer area B is used to temporarily store data that the sending processor expects to send to the receiving processor.

Fig. 4 shows a flowchart of a communication method of a multiprocessor communication system sharing physical memory according to another exemplary embodiment of the present invention.

As shown in FIG. 4, in step S100, the physical memory is divided into multiple physical memory blocks, so that each processor has a dedicated physical memory block. In step S400, the sending processor may request the receiving processor to allocate a temporary data buffer area for buffering data in its dedicated physical memory block. In step S500, the receiving processor may allocate a temporary data buffer in response to the request of the sending processor. In step S200, the sending processor may send data to the allocated temporary data buffer. In step S300, the receiving processor may read the data from the allocated temporary data buffer.

Here, according to different actual application scenarios and different strategies for allocation and release of temporary data buffer areas, the data communicated by multiprocessors in a multiprocessor communication system may include burst data (Burst Data) and stream data (Stream Data). Data).

In addition, as an additional component, the multiprocessor communication system may further include a message box (Message Box), wherein the message box may be used to temporarily store several messages with complete semantics. Each processor in a multi-processor communication system according to an exemplary embodiment of the present invention has access to a message box. As an example, the message box may be a dedicated area partitioned from the physical memory (ie, a portion of the physical memory), or the message box may be a dedicated hardware register set in a multiprocessor communication system. It should be noted that in the exemplary embodiment shown in FIG. 3 , the message box 210 is a dedicated area divided from the physical memory (ie, a part of the physical memory 200 ).

Next, an example of a message box 210 according to an exemplary embodiment of the present invention will be described with reference to FIG. 5 .

Specifically, as shown in FIG. 5 , the message box 210 can be divided into X unit slots (SLOTs), wherein the unit slots can be designed to be fixed and of the same size. Generally, X can be 16, 32, 64 or other integers according to practical applications. Each unit slot can carry a message with complete semantics. Here, the message may be a command for one party of data communication to request the other party to perform an operation (for example, the sending processor requests the receiving processor to allocate a temporary data buffer area), or the message may be a command execution (for example, receiving processing The response after the processor allocates a temporary data buffer in response to a command from the sending processor). Here, as an example, each message includes the processor identifiers of both communicating parties, which are used to indicate the sender of the message and the receiver of the message. In addition, the unit slot can also carry a resident message, and the unit slot carrying the resident message will be occupied for a long time, and will not be set to an idle and available state, so that both parties to the communication will preferentially use the unit slot, In this way, the efficiency of message delivery can be improved.

FIG. 6 shows an example in which multiprocessors sharing physical memory communicate with respect to burst data according to an exemplary embodiment of the present invention.

As shown in FIG. 6 , as an example, when the data that the sending processor expects to send to the receiving processor is burst data, the temporary data buffer area B for buffering the burst data may include a single buffer area B1. It should be understood that bursty data may indicate data transmitted in bursts in data communications. In this case, the burst data message including information indicating the type of the burst data transmission event may be temporarily stored in the message box 210. Here, as an example, the burst data transmission event may include at least one of the following events : Request to allocate a temporary data buffer, the temporary data buffer has been allocated, the temporary data buffer has data to be read, and the data in the temporary data buffer has been read. In addition, as an example, when the burst data transmission event is a request to allocate a temporary data buffer, the burst data message further includes request parameters for requesting allocation of a temporary data buffer, wherein the request for allocating a temporary data buffer The request parameters include the size of the requested single buffer B1 and the application identifier (used to distinguish different requests on the same processor), the identifiers of the sending processor and the receiving processor; when a sudden data transmission event is a temporary data cache When the area has been allocated, the burst data message also includes allocation parameters for allocating a temporary data buffer area, wherein the allocation parameters for allocating a temporary data buffer area include the size and location of the actually allocated single buffer area B1 and the application Identification, the identification of the sending processor and the receiving processor; when the burst transmission event is that the temporary data buffer stores data to be read, the burst data message also includes parameters about the temporary data buffer area storing data to be read, Wherein, the parameters about the data to be read in the temporary data buffer area include the address of a single buffer area, the length of the data to be read, and the application identification, the identification of the sending processor and the receiving processor; when the burst transmission event is temporary When the data in the data buffer has been read, the burst data message also includes a parameter that the data in the temporary data buffer has been read, wherein the data about the temporary data buffer has been read The parameters taken include the address of a single buffer area and the identification of the application, the identification of the sending processor and the receiving processor.

A method for communicating burst data in a multiprocessor communication system according to an exemplary embodiment of the present invention will be described below with reference to FIG. 7A and FIG. 7B . Among them, FIG. 7A shows a flowchart of a method for transmitting processors in a multiprocessor communication system to communicate with burst data according to an exemplary embodiment of the present invention; FIG. 7B shows a multiprocessing method according to an exemplary embodiment of the present invention A flowchart of a method for a receiver processor in a communication system to communicate with respect to burst data.

Specifically, as shown in FIG. 7A, in step S101, the sending processor (ie, processor 100-1) performs initialization. As shown in FIG. 7B, in step S201, the receiving processor (ie, the processor 100-2) performs initialization.

Referring again to FIG. 7A , in step S102, the processor 100-1 constructs a burst data message, which may include type information indicating a burst data transmission event of "request to allocate a temporary data buffer area" (here, The type information is used to indicate the burst data transmission event) and corresponding request parameters (here, the request parameters may include information such as the size of the cache area requested for allocation, the application identification, the identification of the processor 100-1 and the processor 100-2) , and then write the constructed burst data message into a free slot in the message box. In step S103, the processor 100-1 sends an interrupt to the processor 100-2.

Referring again to FIG. 7B, in step S202, the processor 100-2 waits for an interrupt sent by the processor 100-1. When the processor 100-2 receives an interrupt from the processor 100-1, in step S203, reads the burst data message from the message box and acquires the type information. In step S204, it is determined which kind of burst data transmission event the type information is used to indicate. If it is judged according to the type information that the burst data transmission event is a request for allocation of a temporary data buffer, in step S205, request parameters for allocating a temporary data buffer are obtained. In step S206, a single cache area is allocated according to the request parameters. In step S207, the processor 100-2 constructs a burst data message, and the burst data message may include the type information indicating the burst data transmission event "temporary data buffer has been allocated" and corresponding allocation parameters (here, The allocation parameters may include the size and location of a single buffer actually allocated, the application identification, the identification of the processor 100-1 and the identification of the processor 100-2, etc.), and then write the constructed burst data message into the message box in a free slot. In step S208, the processor 100-2 sends an interrupt to the processor 100-1. Then the processor 100-2 returns to execute step S202.

Referring back to FIG. 7A again, in step S104, the processor 100-1 waits for an interrupt sent by the processor 100-2. When the processor 100-1 receives an interrupt from the processor 100-2, in step S105, the burst data message is read from the message box and the type information is acquired. In step S106, it is determined which kind of burst data transmission event the type information is used to indicate. If it is determined according to the type information that the temporary data buffer has been allocated for the burst data transmission event, in step S107, the allocation parameters of the allocated temporary data buffer are acquired. In step S108, data is written into a single buffer area according to the allocation parameters. In step S109, the processor 100-1 constructs a burst data message, and the burst data message includes indicating the type information of the burst data transmission event "temporary data buffer stores data to be read" and corresponding parameters (here , the parameters can include the address of a single cache area, the length of the data to be read, and the application identification, the identification of the processor 100-1 and the identification of the processor 100-2, etc.), and then write the constructed burst data message into the message box in one of the free slots. In step S110, the processor 100-1 sends an interrupt to the processor 100-2. Then the processor 100-1 returns to execute step S104.

Referring back to FIG. 7B again, when the processor 100-2 receives an interrupt from the processor 100-1, in step S203, reads the burst data message from the message box and acquires the type information. In step S204, it is determined which kind of burst data transmission event the type information is used to indicate. If it is determined according to the type information that the data to be read is stored in the temporary data buffer, in step S209, the data written by the processor 100-1 is read from the single buffer. In step S210, a single buffer is released. In step S211, the processor 100-2 constructs a burst data message, which may include type information indicating that "the data in the temporary data buffer area has been read" and corresponding parameters (here, the parameters include a single The address of the cache area, the application identification, the identification of the processor 100-1 and the identification of the processor 100-2, etc.), and then write the constructed burst data message into a certain free slot in the message box. In step S212, the processor 100-2 sends an interrupt to the processor 100-1. In step S213, it is judged whether to exit the current data communication, if it is determined to exit the current data communication, then end the current data communication; if it is determined not to exit the current data communication, then return to step S202.

Referring back to FIG. 7A again, when the processor 100-1 receives an interrupt from the processor 100-2, in step S105, the burst data message is read from the message box and the type information is acquired. In step S106, it is determined which kind of burst data transmission event the type information is used to indicate. In the case that the data in the temporary data buffer area has been read in the sudden data transmission event according to the type information, in step S111, it is judged whether all the data has been sent, and if all the data has been sent, the current data communication is ended; If not all data are sent, return to step S102.

When sending burst data through the above method, the transmission method is simple, and the single buffer area required by the burst data can be allocated as soon as it is used, and released when it is used up, which prevents memory leaks. Since the allocation and release of a single buffer requires certain system overhead, preferably, this transmission method can be applied to a small amount of occasional data communication scenarios.

FIG. 8 shows an example in which multiprocessors sharing physical memory communicate with respect to stream data according to an exemplary embodiment of the present invention.

As shown in FIG. 8 , as another example, in the case that the data that the sending processor expects to send to the receiving processor is streaming data, the temporary data buffer area B for buffering the streaming data may include a circular buffer area B2, wherein, The circular buffer B2 may be composed of multiple slices (for example, B2-1, B2-2, . . . , B2-K). It should be understood that streaming data may indicate a continuously transmitted sequence of data. In this case, the message box 210 may temporarily store a stream data message including information indicating the type of the stream data transmission event. Here, as an example, the stream data transmission event may include at least one of the following events: request to establish Stream connection, stream connection established, shard status changed, request to disconnect stream connection, stream connection disconnected. In addition, as an example, when the stream data transmission event is a request to establish a stream connection, the stream data message may further include a request parameter for requesting to establish a stream connection, wherein the request parameter for requesting to establish a stream connection may include a composition request The number of fragments and the size of each fragment of the circular buffer area B2 and the identification of the application identifier, the sending processor and the receiving processor; when the streaming data transmission event is that the streaming connection has been established, the streaming data message can also be Including the unique identification of the stream connection and the allocation parameters of the actual allocation of the temporary data buffer area, wherein the unique identification of the stream connection is used to globally uniquely identify the established stream connection; the allocation of the actual allocation of the temporary data buffer area Parameters may include the number of fragments constituting the allocated circular buffer area B2 and the size, location and number of each fragment, as well as the identification of the application, the sending processor and the receiving processor; When connected, the stream data message may further include a request parameter for disconnecting the stream connection, wherein the request parameter for disconnecting the stream connection may include the unique identifier and application identifier of the stream connection, the sending processor and the receiving processor when the stream data transmission event is that the stream connection has been disconnected, the stream data message may further include parameters about the stream connection being disconnected, wherein the parameters about the stream connection being disconnected include the unique ID and ID of the application, ID of the sending processor, and ID of the receiving processor. In addition, as an example, the message box 210 also temporarily stores resident messages, wherein the resident messages may indicate the state of each fragment of the circular buffer area B2 (for example, may indicate the status of each fragment in the circular buffer area B2). Whether there is data to be read and the length of the data to be read), in other words, the resident message can save information indicating whether there is data cached in each fragment of the circular buffer area B2 (for example, "has data" or "No data") and the length of the data to be read when "with data".

A method for communicating stream data in a multiprocessor communication system according to an exemplary embodiment of the present invention will be described below with reference to FIGS. 9A and 9B . Wherein, FIG. 9A shows a flow chart of a method for communicating stream data by a sending processor in a multiprocessor communication system according to an exemplary embodiment of the present invention; FIG. 9B shows a multiprocessor according to an exemplary embodiment of the present invention A flowchart of a method of communicating streaming data for a receive processor in a communication system.

Specifically, as shown in FIG. 9A, in step S301, the sending processor (ie, the processor 100-1) performs initialization. As shown in FIG. 9B, in step S401, the receiving processor (ie, the processor 100-2) performs initialization.

Referring to FIG. 9A again, in step S302, the processor 100-1 constructs a stream data message, which may include type information indicating a stream data transmission event of "request to establish a stream connection" (here, the type information is used to indicate stream data transfer event) and corresponding request parameters (here, the request parameters may include the number of fragments and the size of each fragment of the requested circular buffer area and the identification of the application, processor 100-1 and processor 100-2 and other information), and then write the constructed stream data message into a free slot in the message box. In step S303, the processor 100-1 sends an interrupt to the processor 100-2.

Referring again to FIG. 9B, in step S402, the processor 100-2 waits for an interrupt sent by the processor 100-1. When the processor 100-2 receives an interrupt from the processor 100-1, in step S403, reads the stream data message from the message box and acquires the type information. In step S404, it is determined which type of streaming data transmission event the type information is used for. If it is judged according to the type information that the stream data transmission event is a request to establish a stream connection, in step S405, a request parameter related to the request to establish a stream connection is acquired. In step S406, a circular buffer including multiple fragments is allocated according to the request parameters. In step S407, the processor 100-2 constructs a stream data message, which may include a unique identifier about the stream connection, type information indicating "the stream connection has been established", an application identifier, the processor 100-1 and the processor 100-2 information such as identification and corresponding allocation parameters (here, the allocation parameters may include the number of fragments of the circular buffer actually allocated and the size, position and number of each fragment), and then the constructed stream data The message is written to one of the free slots in the message box. In step S408, the processor 100-2 sends an interrupt to the processor 100-1, and returns to step S402.

Referring back to FIG. 9A again, in step S304, the processor 100-1 waits for an interrupt sent by the processor 100-2. When the processor 100-1 receives an interrupt from the processor 100-2, in step S305, the stream data message is read from the message box and the type information is acquired. In step S306, it is determined which type of streaming data transmission event the type information is used to indicate. When judging from the type information that the streaming data transmission event is that the streaming connection has been established, in step S307, obtain the unique identifier of the streaming connection, the application identifier, and the allocation parameters about the actual allocation of the temporary data buffer area, and obtain the fragmentation according to the allocation parameters and the size, location, and number of each shard. The processor 100-1 constructs a resident message again, wherein the resident message includes the state information of each fragment of the circular buffer, and then the processor 100-1 writes the constructed resident message into a message box. In a free slot (the resident message is not necessary, but optional, it always stays in the slot where it is located during the entire data transmission cycle of the stream connection, by means of the resident message, the processor 100-1 and The processor 100-2 can quickly exchange state information of each segment of the circular buffer during data transmission). In step S308, the processor 100-1 waits for data to be sent. In step S309, it is judged whether all the data has been sent. If not all the data has been sent, in step S310, the status information of the fragments is sequentially read from the resident message, and it is determined whether there is a "no data" fragment. If there are "no data" fragments, in step S311, data is sequentially written into the "no data" fragments of the circular buffer. In step S312, update the state of the fragment whose state has changed (including the length of data to be read in the fragment) and the type information indicating "the state of the fragment has changed" into the resident message in the message box, to indicate that the shard has data to be read. In step S313, the processor 100-1 sends an interrupt to the processor 100-2, and returns to step S309. On the other hand, if there is no "no data" fragment, return to step S304.

Specifically, in step S311, after the processor 100-1 writes data to a "no data" slice B2-1 in the circular buffer, in step S312, the processor 100-1 can write the data in the resident message The status information of the temporarily stored fragment B2-1 is changed to "has data" status (including temporarily storing the data length to be read in this fragment), and the processor 100-1 sends the data to the processor 100-2 in step S313 simultaneously. interruption. Wherein, in steps S309 and S310, if the processor 100-1 still has data to be written to the slice and there is a "no data" slice, it can continue to write data to the slice. If the processor 100-1 still has data to be written to the slices but all the slices are "with data", the process of writing data to the slices by the processor 100-1 is terminated, and returns to step S304, waiting for the processor to Interrupt sent by 100-2 to processor 100-1.

On the other hand, in the case that all data has been transmitted, in step S314, processor 100-1 determines whether to disconnect the stream connection. In the case of disconnecting the stream connection, in step 315, the processor 100-1 will construct a stream data message, which may include type information indicating "request to disconnect the stream connection" and corresponding request parameters (here, The request parameters may include information such as the unique identifier of the stream connection, the application identifier, the identifiers of the processor 100-1 and the processor 100-2), and then write the constructed stream data message into a certain free slot in the message box ( If a resident message is used during data transmission, the resident message is no longer needed at this time, and the slot where the resident message is located can be directly used to temporarily store the message to be sent). In step S316, the processor 100-1 sends an interrupt to the processor 100-2, and returns to step S304.

Referring back to FIG. 9B again, in step S402, the processor 100-2 waits for an interrupt sent by the processor 100-1. When the processor 100-2 receives an interrupt from the processor 100-1, in step S403, reads the stream data message from the message box and acquires the type information. In step S404, it is determined which type of streaming data transmission event the type information is used for. When judging from the type information that the stream data transmission event is a fragmentation state change (that is, the fragmentation stores data to be read), in step S409, it is determined from the resident messages in the message box in turn whether there is a fragmentation state to be read. Fragmentation of the data fetched, here, when the processor 100-2 determines whether there is a fragmentation of the data to be read from the resident message, it can determine the state of the subsequent fragmentation from the fragmentation read last time, and The state of each shard is not determined from scratch. If there is a "data" shard, at step S410, read data from the "has data" shard and clear the shard. In step S411, update the type information indicating "shard status changes" and the status of the segment whose status has changed to the resident message in the message box to indicate that there is no data in the segment. In step S412, the processor 100-2 sends an interrupt to the processor 100-1, and returns to step S409. That is to say, whenever the processor 100-2 receives an interrupt from the processor 100-1 for notifying the fragment that there is data to be read, the processor 100-2 will sequentially read from the "has data" fragment. Fetch data until there is no data that can be read (that is, read the "no data" fragment), in this case, the processor 100-2 stops reading data, and returns to step S402 to wait for the processor 100 Send interrupt again after -1.

On the other hand, in the case that the streaming data transmission event is determined to be a request to disconnect the streaming connection according to the type information, in step S413, the processor 100-2 acquires request parameters for disconnecting the streaming connection. In step S414, the allocated circular buffer area is released according to the request parameter about disconnecting the stream connection. In step S415, the processor 100-2 constructs a stream data message, which may include an indication that "the stream connection has been disconnected." " type information and corresponding parameters (here, the parameters include information such as the unique identifier of the stream connection, the application identifier, the identifiers of the processor 100-1 and the processor 100-2), and then write the constructed stream data message into the message In a certain free slot in the cartridge, the processor 100-2 sends an interrupt to the processor 100-1 in step S416. Finally, end this data communication.

Referring back to FIG. 9A again, in step S304, the processor 100-1 waits for an interrupt sent by the processor 100-2. When the processor 100-1 receives an interrupt from the processor 100-2, in step S305, the stream data message is read from the message box and the type information is acquired. In step S306, it is determined which type of streaming data transmission event the type information is used to indicate. When it is judged according to the type information that the stream data transmission event is that the data in the segment has been read, step S309 is executed. On the other hand, if it is judged according to the type information that the stream data transmission event is that the stream connection has been disconnected, the current data communication is terminated.

When streaming data is sent through the above method, the circular buffer area can be recycled and reused, and the allocation and release operations of the buffer area do not need to be performed frequently, which improves the communication efficiency. Therefore, this transmission method is suitable for data communication scenarios where a large amount of data is transmitted frequently. . The management strategy of the stream connection can effectively prevent memory leaks. For example, when one party of the communication does not need to communicate, the other party of the communication can be notified to disconnect the stream connection, so that the data receiving processor responsible for allocating the buffer can be released in time. The circular buffer area prevents memory leaks and avoids errors (such as system crashes) that cause data receiving processors.

In summary, in the multiprocessor communication system and communication method for sharing physical memory according to the exemplary embodiments of the present invention, there is no need to reserve a global memory shared area, and multiple processors can use their own dedicated physical memory blocks Transfer data to each other, thereby improving the efficiency of memory usage, and reducing the number of data copies during data transmission, thereby speeding up the speed of data transmission.

It should be noted that the above respective embodiments of the present invention are merely exemplary, and the present invention is not limited thereto. It should be understood by those skilled in the art that changes may be made in these embodiments without departing from the principles and spirit of the present invention, wherein the scope of the present invention is defined in the claims and their equivalents.

Claims (9)

1. A multiprocessor communication system sharing physical memory, comprising: a plurality of processors, wherein the plurality of processors send and receive data from each other; Physical memory, divided into multiple physical memory blocks such that each processor has a dedicated physical memory block; Wherein, the sending processor that will send the data among the plurality of processors sends the data to the dedicated physical memory block of the receiving processor that will receive the data among the plurality of processors, and the receiving processor receives the data from its dedicated block of physical memory to read said data, Wherein, the sending processor requests the receiving processor to allocate a temporary data buffer area for buffering data in its dedicated physical memory block, and the receiving processor allocates the temporary data buffer area in response to the sending processor's request, Wherein, the system further includes: a message box, wherein the message box is used to temporarily store several messages with complete semantics, wherein the sending processor and the receiving processor perform message interaction through the message box , Wherein, the temporary data buffer area is released after use. 2. The system of claim 1, wherein the data is burst data, wherein the temporary data buffer area for buffering the burst data comprises a single buffer area. 3. The system according to claim 1, wherein the data is streaming data, wherein the temporary data buffer area for buffering the streaming data comprises a circular buffer area composed of a plurality of fragments. 4. The system of claim 2, wherein the message box is used to temporarily store burst data messages, wherein the burst data messages include type information indicating a burst data transmission event. 5. The system of claim 3, wherein the message box is used to temporarily store streaming data messages, wherein the streaming data messages include type information indicating streaming data transmission events. 6. The system according to claim 4 or 5, wherein the message box is a dedicated area divided from the physical memory, or the message box is a dedicated hardware register set in a multiprocessor communication system. 7. The system according to claim 6, wherein the message box also temporarily stores resident messages, wherein the resident messages are suitable for scenarios where messages are frequently delivered. 8. The system according to claim 1, wherein the multi-processor communication system is integrated in a single-chip system, or the multi-processor communication system is integrated in a multi-processor system sharing a multi-port physical memory. 9. A communication method for a multiprocessor communication system sharing physical memory, wherein the multiprocessor communication system includes a plurality of processors and physical memory, wherein the physical memory also includes a message box, wherein the The message box is used to temporarily store several messages with complete semantics, the multiple processors send and receive data with each other, and the multiple processors perform message interaction through the message box, and the method includes: (a) dividing the physical memory into a plurality of physical memory blocks such that each processor has a dedicated physical memory block; (b) sending data, by a sending processor of the plurality of processors that is to send the data, to a dedicated block of physical memory of a receiving processor of the plurality of processors that is to receive the data; and (c) reading said data by said receiving processor from its dedicated block of physical memory, Wherein, between steps (a) and (b), also include: (d) The sending processor requests the receiving processor to allocate a temporary data buffer for caching data in its dedicated physical memory block; (e) the receiving processor allocates a temporary data buffer in response to the sending processor's request; Wherein, in step (b), the sending processor sends data to the allocated temporary data buffer; in step (c), the receiving processor reads the data from the allocated temporary data buffer, Wherein, the temporary data buffer area is released after use.