JP2007041805A - Program generation device, program generation method and compiler - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device for easily providing the program of a compact network device for reducing a variable memory capacity to be installed, and for achieving a high speed response. <P>SOLUTION: From an electronic file where the formats and configurations of packets are formally described in advance only by the number of packets to be processed, the packet format is analyzed, and a protocol stack for processing those packets is automatically generated by computer processing, and a virtual production result whose execution costs are the minimum such as a processing performance or a necessary memory size is selectively generated from among various virtual generation results so that a protocol stack as a much more higher performance and saving memory can be generated. A network message processing function, so called, the software of a protocol stack mounted on compact network equipment in a network system can be much more compact and easily mounted. Thus, it is possible to provide a compact network device by reducing a variable memory capacity to be installed, and achieving a much more high speed response. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a technique for more compactly and easily mounting software of a so-called protocol stack mounted on a small network device.

  2. Description of the Related Art A network device compliant with a standard for a method of connecting electronic devices by wire and wireless, such as Internet Protocol (IP), includes a software group for processing packet data, a so-called protocol stack. Install such network devices not only in large computers and personal computers, but also in electronic devices such as AV (Audio Visual) home appliances, white goods home appliances, equipment, sensors, portable information devices, and mobile phones. Is being considered.

  Among these electronic devices, network devices mounted on small electronic devices (portable information devices, mobile phones, AV home appliances, sensors, etc.) are small and have low power consumption, white home appliances, Network devices mounted on equipment are required to be inexpensive.

  By the way, for example, Patent Document 1 describes a technique for minimizing the size of generated object code in an object-oriented language processing system. According to this technology, the class name to which the method belongs is compared with the input file name during the semantic analysis of the class declaration. As a result, if both match, the entity and declaration of the method address storage table are If they do not match, only the declaration of the method address storage table is generated in the output object code. As a result, only one method address storage table entity is generated in the target object code, so that the size can be reduced.

JP-A-6-266562

  In order to respond to the above-described requirements, restrictions on the memory size and processor performance of the network device are imposed on the network device to be mounted on the electronic device as described above. Under such restrictions, in order for a network device to exhibit a sufficient function, software incorporated therein is required to be able to operate at higher speed on a smaller size memory.

  However, it is difficult to describe such software in a general-purpose programming language (for example, C language). For example, when a runtime determination function realized by using the method address storage table is used, a method address storage table and unnecessary instances are generated when an execution code is installed. For this reason, the size of the execution code becomes large, and it is difficult to execute it with a small network device having the above-described restrictions. Dynamic parsing mechanisms also increase executable code.

  Therefore, an object of the present invention is to more easily implement protocol stack software that can be executed more compactly and at higher speed.

The present invention
A program generation method for causing an information processing apparatus having a storage unit and a calculation unit to generate a program for causing a network device that accepts a series of data columns including a plurality of data elements to process the data sequence via a network,
The storage means stores a definition file in which item information including size information of each data element in the data string received by the network device is described according to the order of the data elements in the data string,
The program generation method is as follows:
A process in which the computing means refers to the definition file;
The arithmetic means, as the program, a program for processing each data element in the data string in a predetermined size according to the order from the top of the data string, in the definition file Processing based on the description order of the item information of each data element and the size information of each data element;
The program generation method characterized by including this is provided.

  According to the present invention, software of a protocol stack that can be executed more compactly and faster can be generated more easily.

  Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

  First, the configuration of an information processing system that generates a protocol stack according to the present embodiment will be described with reference to FIGS. 1 and 2.

  The information processing system according to the present embodiment has a hardware configuration that can execute a predetermined program in accordance with an instruction from a user. For example, as shown in FIG. 2, a program (protocol stack compiler, general-purpose compiler, OS, etc.) for executing a protocol stack code generation process, which will be described later, and data (network message format and structure to be processed by the protocol stack are predetermined). An external storage device 201 such as a hard disk in which a text-format computerized file such as ASCII (American Standard Code for Information Interchang) code described in a format is stored; A processor 202 that executes a program loaded on the main storage device 203 from the main storage device 203, the external storage device 201, etc., a drive 204 that executes data read / write of a portable storage medium, and an input device that receives data input from a user ( Keyboard, ma , Etc.) 205, an output device 206 such as a display for outputting information presented to the user (for example, the execution state of the program).

  Note that the protocol stack compiler and the general-purpose compiler may be installed from the portable storage medium to the external storage device 201 via the drive 204, or in an environment where the information processing system can be connected to the network, It may be installed in the external storage device 201 via The protocol stack definition file may be copied from a portable storage medium to the external storage device 201 via the drive 204, or may be created by the user using the input device 205 and the external storage device 201. It may be stored in.

  By executing the above-described program on such a hardware configuration, the information processing system realizes a functional configuration for automatically generating a protocol stack from a protocol stack definition file. That is, as shown in FIG. 1, the data storage unit 100 in which the protocol stack definition file 101 is stored, and the processing code of the protocol stack (automaton and this) based on the analysis result of the message data format described in the protocol stack definition file 101 The protocol stack compiler unit 102 for generating the automaton execution program (hereinafter referred to as protocol stack code) 103 and the general-purpose compiler unit 104 for generating the execution code 105 executable by the processor 115 of the network device 116 from the protocol stack code 103 are realized. Is done.

  With such a functional configuration, the execution code 105 that causes the network device 116 to execute processing related to a transmission / reception message with another device (for example, the control device 111) on the network 112, that is, software of the protocol stack is automatically performed. Generated. The generated execution code 105 is transferred to the memory 114 on the network device 116 via an appropriate medium for execution by the processor 115 of the network device 116.

  Next, referring to FIG. 3 and FIG. 4, the data structure of message data (hereinafter, processing target message data) processed by the protocol stack software generated by the information processing system and the description in the protocol stack definition file 101 And will be described. Here, as a specific example of processing target message data, packets exchanged between network devices, in particular, three types of packets for controlling and monitoring lighting fixtures (packets for turning on lighting fixtures, lighting fixtures turned off) And a packet for inquiring whether the device is a lighting device).

  As shown in FIG. 3, a packet (lighting instruction packet) 301 for turning on the lighting fixture, a packet (lighting off instruction packet) 302 for turning off the lighting fixture, and an inquiry as to whether or not the lighting fixture is used. The packet (inquiry packet) 303 includes a transmission source address storage field 310 that stores the network address (transmission source address) of the transmission source device, and a transmission destination address storage that stores the network address (transmission destination address) of the transmission destination device. A field 311, a command storage field 312 for storing a command to be given to the lighting device, and a reservation field 313 are provided.

  Here, “0001” is an example of the network address of the control device 111 for the lighting device in the transmission source address storage field 310, and an example of the network address of the network device 116 in the lighting device is in the transmission destination address storage field 311. "0002" is stored. The command storage field 312 stores one of the illumination lighting command “01”, the extinguishing command “02”, and the inquiry command “03” according to the type of packet.

  Further, the turn-on instruction packet 301 and the turn-off instruction packet 302 are provided with a device code storage field 314 in which the device code of the destination device is stored. Here, “0001” is stored in the device code storage field 314 as an example of the identifier assigned to the lighting device.

  FIG. 4 shows descriptions representing the data formats and data structures of these three types of packets 301 to 303, that is, text data described in the protocol stack definition file 101 according to a predetermined format.

The description data in the protocol stack definition file 101 is composed of the same number of elements (here, three elements 420, 421, and 422) as the types of packets to be processed. Each element 420, 421, and 422 includes a comment 405 that is a note with a comment symbol (//) and a message definition 402 that represents a formal definition of a packet that is a corresponding message. Here, a comment 405 is included in each element, but a comment 405 is not necessarily included in each element. In the following, for convenience of explanation, the message definition corresponding to the lighting on packet 301 is the first message definition, the message definition corresponding to the lighting off packet 302 is the second message definition, and the message definition corresponding to the inquiry packet 303 is the first message definition. This is called a 3-message definition.
Each message definition 402 includes a message class name 406, an item list 404 corresponding to each field in the corresponding packet, and an identifier (message identifier) 407 of the corresponding packet. In the list 404, items corresponding to each field in the corresponding packet are arranged according to the order of the fields in the corresponding packet.

  Each item includes a qualifier 408 indicating the data storage format of the corresponding field, an item name 409, an occupied size 410 of the corresponding field, and a constraint condition 411 for the data of the corresponding field.

  Here, as specific examples of the modifier 408, “bigendian” indicating that the contents of the field are stored in a big endian format, “littleendian” indicating that the contents of the field are stored in a little endian format, For example, “any” or the like indicating that the content is stored in an arbitrary format. Instead of including the qualifier 408 in each item, the qualifier 412 indicating the data storage format of the corresponding field of all items may be placed before the message class name 406.

  The occupied size 410 may be, for example, numeric data in decimal notation (“2”, “100”, etc.), numeric data in hexadecimal notation (“0x”, etc.) in accordance with a 16 base instruction, It may be a mathematical expression constituted by designation using other item names being used and combinations thereof.

  Specific examples of the constraint condition 411 include a condition “= specific value” for limiting the value of data in the corresponding field to a specific value, a condition specification using an inequality sign, a bit field constraint, and a mathematical formula that combines these elements. The conditions represented are mentioned.

  Note that, since a protocol stack is usually composed of a plurality of layers, a packet to be processed is composed of a plurality of nested elements. In the message definition regarding such a packet, an item type indicating that a part or all of the upper layer packet is stored in the payload portion of the lower layer packet may be used. In order to specify a field for storing part or all of the upper layer packet data, such a message definition is described in a message class having a data format of the upper layer packet, for example, and this message class name. An item with is described.

  In addition, here, a packet addressed to the lighting device is given as a specific example of the processing target message of the protocol stack, but the processing target message of the protocol stack is not limited to this. For example, home network devices (white goods, AV home appliances, etc.), network devices (personal computers, mobile phones, network control devices, etc.), equipment network devices (equipment equipment, monitoring devices, etc.), etc. It can be a message to be processed by the protocol stack. When these packets are set as messages to be processed, the format and structure thereof are described in the protocol stack definition file in a predetermined format, similar to the packets addressed to the lighting device.

  Next, a protocol stack code generation process executed by the protocol stack compiler unit 102 based on the description content of the protocol stack definition file 101 will be described with reference to FIG. Here, for convenience of explanation, an outline of the flow of the protocol stack code generation process will be described, and then a part of the process (code conversion process S603) will be described in detail.

  The protocol stack compiler unit 102 reads each message definition 402 from the protocol stack definition file 101 and places it in the memory (S601), and checks the syntax error of each message definition 402 (S602).

  As a result, if there is a syntax error in any of the message definitions 402, the protocol stack compiler unit 102 displays an error message including the detailed contents of the syntax error on the output device 206 (S605), and ends the processing. .

  On the other hand, if there is no syntax error in any of the message definitions 402, the protocol stack compiler unit 102 converts each message definition 402 into a general language code by a code conversion process described later (S603), and the conversion result Is output to the electronic file as the protocol stack code 103 (S604). As a result, a protocol stack code is generated.

  Here, as described above, details of the code conversion processing S603 will be described.

  The protocol stack compiler unit 102 creates a structure tree based on the description of each message definition 402 (S606). The structure tree generated at this time has a tree structure in which the message class name is a root, each item is a node, and the message identifier 407 is a leaf (terminal node).

  For example, the structure tree generated from the first message definition 402 includes a root element 701 and five items “bigendian sourceAddress [2]” “bigendian destAddress” generated from the message class name “LightControlMsg” 406 as shown in FIG. [2] = MYSELF ”,“ command [1] = 0x01 ”,“ any reserved [1] ”,“ bigendian objectCode [2] = 0x0a ”, and the message identifier as the setting value of the attribute“ function ” And a leaf 707 including “lightTurenOn” 407.

  The type of each node 702 to 706 is determined based on at least one of the qualifier 408 and the constraint condition 411 of the corresponding item. For example, the type of the node corresponding to the item including the modifier “any” is “arbitrary element”, the type of the node corresponding to the item including the constraint 411 is “constraint”, does not include the constraint 411, and The type of the node corresponding to the item that does not include the modifier “any” is “record”.

  Here, a structure tree having a partial tree is generated from a plurality of message definitions including the same message class name 406. For example, from the first to third message definitions 402 (see FIG. 4) including the common message class name 406, a structure tree as shown in FIG. 7 is generated. The first to third message definitions 402 have the first and second items including a common message class name and common constraint conditions. The structure tree generated from these message definitions 402 branches from the node 803 as shown in FIG. That is, based on the first message definition 402, the same structure tree (root element 801, nodes 802 to 806 and leaf 807) as in FIG. 6 is generated, and the third and subsequent items and messages of the second and third message definitions 402 are generated. Based on the identifier, two subtrees starting from the node 803 of the structural tree (ie, the nodes 808 to 810 and the leaf 811 corresponding to the second message definition 402, the nodes 812 to 813 corresponding to the third message definition 402, and A reel 814) is added.

  When the structure tree for each message class is generated based on the description contents of the protocol stack definition file 101 in this way, the protocol stack compiler unit 102 executes an optimization subroutine (tree integration subroutine) described later. By calling, the common subtrees included in this structural tree are integrated (S607). This optimization subroutine may be called only when a call is designated by an option.

  Then, an automaton is generated by converting the structure tree into a state machine, each node into a state, and each leaf into an external function read process. An automaton generated from the structural tree of FIG. 7 is shown in FIG.

  The automaton in FIG. 8 is characterized by the state tree in the structural tree in FIG. 7, the nodes 801 to 806, 808 to 810, 812, and 813 in the states 1001 to 1015 and the attribute “function” values in the leaves 807, 812, and 814. It is generated by converting into the read processing 1009, 1013, 1015 of the external function to be attached. Here, the conversion from the node to the state is performed according to the following rules (1) to (3), for example.

(1) Type = recording node This node is converted into a state “C” that is repeated by the number corresponding to the value of the attribute “length”. For example, a node 802 having a length attribute value “2” is converted into two consecutive states “C” 1001 and 1002. In these states “C”, the one-character code received from the upper layer is stored in a specific area of 2 bytes on the RAM, and the state is further advanced to the next state. Here, the storage area of the character code is characterized by the value of the attribute “name” of the conversion source node, and the holding method is determined according to the value of the attribute “type” of the conversion source node.

  For example, when the processing target system is bigendian, the first state 1001 of the two states “C” 1001 and 1002 generated from the node 802 is characterized by the name attribute value “sourceAddress” of the transformation source node 802. A code for holding the character code from the upper area is described in the upper part of the attached RAM area, and the second state 1002 is for holding the character code from the upper area in the lower part of the RAM area. Is written. Here, since the type attribute value of the transformation source node 802 is “bigendian”, the code is generated in this order. However, if the type attribute value of the transformation source node 802 is “littleendian”, Will generate code in the reverse order.

  Here, the case where the target processing system is bigendian is taken as an example, but if the target processing system is littleendian, the code is generated in the reverse order to the case where the target processing system is bigendian. In order to be able to deal with either bigendian or littleendan target processing systems, it may be optionally accepted that the target processing system is either bigendian or littleendan. In this case, a branch process corresponding to the contents of option designation may be executed.

(2) Type = restricted node This node is converted into a state “B” that is repeated by the number corresponding to the value of the attribute “length”. For example, each node 804, 808, 812 having a length attribute value “1” is converted into a state “B” 1005.

  In these states “B”, the next state is determined according to the one-character code received from the upper layer. For example, if the current execution state is state 1005, state 1006 if the character code from the upper layer is “01”, state 1010 if the character code from the upper layer is “02”, characters from the upper layer If the code is “03”, the state is shifted to each of the states 1014.

  In the state “B” corresponding to the node of Type = constraint, if the character code from the upper layer does not satisfy the constraint condition of the corresponding node (that is, the value of the attribute “value” of the corresponding node) If they do not match, the entire received packet must be treated as illegal. For this reason, for the state “B” corresponding to the node of Type = constraint, a special state “D” 1016 to be transitioned when the character code from the upper layer does not satisfy the constraint condition of the corresponding node is prepared. . For example, in the state “B” 1005 corresponding to Type = constrained nodes 805, 809, 813, the character code from the upper layer is the value attribute value “01” “02” “03” of each corresponding node 805, 809, 813. ”Is written so as to transition to a state“ D ”1016 where no processing is performed.

  Such a state “D” is generated in all message classes.

(3) Type = node of arbitrary element This node is converted into a state “A” that is repeated by the number corresponding to the value of the attribute “length”. For example, the nodes 805, 809, and 813 having the length attribute value “1” are converted into one state “A” 1001, 1010, and 1014, respectively. In these states “A”, processing for shifting the state to the next state is performed. For example, when the current execution state is the state 1001, the state only transitions to the next state 1002.

  Further, the protocol stack compiler unit 102 generates a general-purpose description language format program (automaton execution program) for executing the automaton generated in this way (S608). FIG. 9 shows a flowchart of the automaton execution program generated in common for each message class.

  Finally, the protocol stack compiler unit 102 optimizes the generated protocol stack code by calling an optimization subroutine (code integration subroutine, RAN area optimization subroutine) described later (S609). Note that the optimization subroutine may be called only when calling is specified by an option.

  The automaton is executed as follows by the automaton execution program generated by the above processing (see FIG. 9).

  First, position information of the first state 1001 of the automaton is set as execution state position information representing the current execution state of the automaton in a predetermined area (execution state position information storage area) for 1 byte on the RAM (S901). Thereafter, data of a certain size in the packet is acquired from the upper layer (S902). The size of data acquired from the upper state is arbitrary, but here, a one-character code (that is, 2-byte data) is acquired.

  Thereafter, based on the acquired data, the code of the state at the position indicated by the execution state position information is executed.

  For example, if the execution state position information indicates the position of the state “B” at this time, the execution state position information is updated with the position information of the state determined according to the character code acquired from the upper layer. For example, if the current execution state is state “B” 1003 and the character code is “00”, the execution state position information is updated with the position information of the next state “B” 1004, and the character code is another value. If so, the execution state position information is updated with the position information of the state “D” 1016. Also, when the current execution state is state “B” 1005, if the character code is “01”, it is position information of the next state “A” 1006, and if the character code is “02”, the next state “A”. If the position information of 1010 is the character code “03”, the position information of the next state “A” 1014, and if the character code is any other value, the execution state position information is the position information of the state “D” 1016. Is updated.

  If the execution state position information indicates the position of the state “A”, the execution state position information is updated with the position information of the next state of the current execution state. For example, when the current execution state is state “A” 1006, the execution state position information is updated with the position information of the next state 1007.

  If the execution state position information indicates the position of the state “C”, the character code is stored in a specific area on the RAM, and the execution state position information is updated with the position information of the next state of the current execution state. Is done. For example, when the current execution state is the state 1001, the character code is stored in the first byte of the RAM area characterized by “sourceAddress”, and the execution state position information is changed with the position information of the next state 1002. . If the current execution state is state 1002, the character code is stored in the second byte of the RAM area characterized by “sourceAddress”, and the execution state position information is changed with the position information of the next state 1003. .

  If the execution state position information indicates the position of the external function call processing, the external function is called. For example, when the current execution state indicates one of the processes 1009, 1013, and 1015, an external function (lightTurnOn, lightTurnOff, lightModeRequest) having a name corresponding to the value of the attribute “function” of the corresponding item is called. .

  When the state code is executed in this manner, the state transitions to the state at the position indicated by the updated execution state position information (S903), and the data acquired in S902 is the code of the last character in the packet. Is checked (S904). As a result, if it is not the code of the last character, the processing from S902 is repeatedly executed, and if it is the code of the last character, the processing ends.

  Thereafter, the protocol stack code 103 generated in this way is optimized by the general-purpose compiler unit 104 and converted into an execution code 105.

  According to such an automaton and an automaton execution program, for example, when the network device 116 receives the turn-off instruction packet 302, the state 1001 → the state 1002 → the state 1003 → the state 1004 → the state by repeating the steps S902 to S904. The state sequentially changes in the order of 1005 → state 1010 → state 1011 → state 1012 → process 1013. As a result, an external function characterized by “lightTurnOn” is finally called, and as a result, the lighting device is turned on. This function “lightTurnOn” needs to use the stored value of the transmission source address storage field of the illumination-off packet 302, that is, the network address of the packet transmission source in order to confirm and respond to the transmission source of the illumination-off packet 302. . The value can be used by this function because it is written to the RAM area characterized by sourceAddress by repeated execution of S902 to S904.

  When the network address “0103” other than the network address “0002” of the network device 116 is stored in the transmission destination address storage field 311 of the received packet, the execution state is 1001 → 1002 → 1003 → 1016 → Since 1016 → 1016 → 1016 → 1016, the processing corresponding to this packet (operation processing for the lighting device) is not executed.

  Thus, according to the automaton and the automaton execution program generated by the information processing system according to the present embodiment, the packet storage information is correctly determined, so that the packet can be processed.

  In normal packet reception processing of a network device, the entire received packet is processed after being copied to the RAM. For example, processing of the above three types of packets 301 to 303 requires a memory area of at least 8 bytes. . Further, since the process is started after the entire received packet is copied to the RAM, the process takes time.

  On the other hand, according to the automaton and automaton execution program that are automatically generated by the protocol stack code generation processing according to the present embodiment, an area for 3 bytes on the RAM (2 bytes characterized by the name attribute value of the node) It is sufficient if a RAM area for one minute and an execution state position information storage area for one byte are secured. In addition, when the reception of the packet is started, each time data of a predetermined size is received, the data is analyzed. Therefore, the processing is performed in comparison with the case where the analysis is started after the entire received packet is temporarily stored in the memory. It can be terminated early.

  Therefore, the RAM area can be saved and the program can be speeded up. That is, according to the present embodiment, it is possible to more easily implement software for a smaller protocol stack, that is, a memory-saving and high-speed protocol stack suitable for mounting on a small network device on a network system. it can.

  In the protocol stack definition file 101, only a necessary message definition (message definition related to a message to be processed by the network device) can be selected and described as appropriate from a plurality of message definitions belonging to the same message class. Therefore, it is possible to generate the minimum necessary code that supports only the packet that the network device should support. That is, since the message definition related to the message that does not need to be processed by the network device can be eliminated, the final code size can be reduced.

  For example, among the above-described three types of packets 301 to 303, when the processing target packet of the network device is only two types of packets (for example, 301 and 302), the protocol stack definition file 101 includes 3 shown in FIG. Of the two elements 420 to 422, only the two elements 420 and 421 corresponding to the processing target packets 301 and 302 need be described. As a result, the protocol stack code 103 is generated only from the message definitions included in the two elements 420 and 421. For this reason, the size of the protocol stack code 103 is reduced, and as a result, the size of the execution code 105 generated from the protocol stack code 103 is also reduced.

  Although the packet is cited as the processing target of the protocol stack, if it is a series of data strings that can be processed continuously by the processor, data other than the packet (higher communication data format data such as XML-format text data, table format) Data, etc.) can be processed by the protocol stack. Such a program for processing data other than packet data can also be automatically generated by the same processing as described above.

  In the above, the protocol stack compiler unit 102 generates the automaton and the automaton execution program compiled into the execution code by the general-purpose compiler unit 104 from the protocol stack definition file 101. The machine language format or binary format code (that is, the execution code 105) that operates on the network device may be directly generated from the protocol stack definition file 101. Alternatively, an intermediate execution code that is sequentially interpreted and executed in the network device may be generated.

  In the above description, the source code is generated from the message definitions for all packets. However, this is not always necessary. For example, after generating a structure tree from the message definitions for some packets, the same processing is performed for the message definitions for the remaining packets, and finally, the internally generated structure trees are merged into a graph. May be. If the generated partial tree structure is saved in an external storage memory in a storable format and is called from the external storage memory when the final source code is generated, the processing speed is increased. Can be achieved.

  Next, the optimization subroutine called in S609 will be described. Note that, as described above, whether or not to call each optimization subroutine may be specified by an option.

(1) Tree Integration Subroutine When the tree integration subroutine is called in S607, as shown below, the same subtree is integrated by the tree integration processing of FIG.

  The following processing is repeatedly executed by sequentially setting each node of all trees as the first processing target node a (S1101). That is, the following processing is executed for each node of all trees.

  The following processing is repeatedly executed by sequentially setting each node of all trees as the second processing target node b (S1102). That is, every time a new node is set as the first processing target node a, the following processing is executed with each node of all the trees as the second processing target node b.

  The order of the second processing target node b and the first processing target node is compared (S1103). As a result, if the order of the first processing target node a is not earlier than the order of the second processing target node b, S1103 is executed with the next node as the second processing target node b. On the other hand, if the order of the first process target node a is earlier than the order of the second process target node b, the following process is repeatedly executed by sequentially setting the next node of the first process target node a as the third process target node an. (S1104). That is, the following processing is executed for each of the next nodes of the first processing target node a.

  The following processing is repeatedly executed by sequentially setting the next node of the second processing target node b as the fourth processing target node bn (S1105). That is, every time a new node becomes the fourth processing target node bn, the following processing is executed for each of the next nodes of the second processing target node b.

  The subtree starting from the third processing target node an is compared with the subtree starting from the fourth processing target node bn. As a result, if the two match, the portion where the first processing target node a points to the next node an is changed to the fourth processing target node bn (S1107), and among the next nodes of the second processing target node b, The other node is set as the fourth processing target node bn, and S1106 is executed again. On the other hand, if the two do not match, S1106 is executed again with the other node among the nodes following the second processing target node b as the fourth processing target node bn.

  When all the same subtrees are identified by such processing, the unreferenced node is discarded (S1109).

  According to such processing, since the same subtree is integrated, the number of nodes is reduced, and as a result, the number of states of the generated automaton can be reduced. For this reason, the final output code amount can be reduced.

(2) Code Integration Subroutine When the code integration subroutine is called in S609, as shown below, the same code for executing the RAM memory operation processing existing between the states is integrated by the code integration processing of FIG. Is done. Here, a case where a code is generated based on the automaton in FIG. 8 is taken as an example. Here, for convenience of explanation, the position of the state 1001 is the number “1”, the state position of the state 1002 is the number “2”, the position of the state 1003 is the number “3”, and the identifier of the execution state position information storage area is “state”. The one-character identifier received from the upper layer is represented as “c”.

  In the state 1001, one code from the upper layer is stored in the first byte of the RAM area characterized by the name sourceAddress, and the execution state position information is changed to the next state number “2”. One of the three types of code A “sourceAddress [0] ← c” 1301, “sourceAddress [state-1] ← c” 1302, “sourceAddress [1-state] ← c” 1303 shown in FIG. It is composed of any combination of “state ← state + 1” 1304 and “state ← 2” 1305.

  In the state 1002, a code for storing one character from the upper layer in the second byte of the RAM area characterized by the name sourceAddress and changing the execution state position information to the next state number “3” is, for example, FIG. 11, three types of code A “sourceAddress [1] ← c” 1311, “sourceAddress [state−1] ← c” 1312, “sourceAddress [3-state] ← c” 1313, and two types of code B It is configured by a combination with either “state ← state + 1” 1314 or “state ← 3” 1315.

  Therefore, if the codes of the two states 1001 and 1002 are a common combination code “sourceAddress [state-1] ← c, state ← state + 1” (a combination of codes 1302 and 1304, a combination of codes 1312 and 1314), 2 Since the codes of the two states 1001 and 1002 can be integrated, the code size can be reduced.

  In order to integrate such codes, the protocol stack compiler unit 102 executes the following processing shown in FIG. First, the protocol stack compiler unit 102 generates all assignable code fragments for all nodes of all trees (S1401). For example, as code fragments in the state 1001, six types of combination codes, that is, “sourceAddress [0] ← c, state ← state + 1” (combination of codes 1301 and 1304), “sourceAddress [state-1] ← c, state ← state + 1 ”(combination of codes 1302, 1304),“ sourceAddress [1-state] ← c, state ← state + 1 ”(combination of codes 1303, 1304),“ sourceAddress [0] ← c, state ← 2 ”(combination of codes 1301, 1305),“ sourceAddress [state-1] ← c, state ← 2 ”(combination of codes 1302,1305),“ sourceAddress [1-state] ← c, state ← 2 ”(code 1303, 1305) is generated. As code fragments in the state 1002, six kinds of combination codes, that is, “sourceAddress [1] ← c, state ← state + 1” (combination of codes 1311 and 1314), “sourceAddress [state-1] ← c, “state ← state + 1” (combination of codes 1312 and 1314), “sourceAddress [3-state] ← c, state ← state + 1” (combination of codes 1313 and 1314), “sourceAddress [1] ← c, state ← 3 ”(combination of codes 1311 and 1315),“ sourceAddress [state-1] ← c, state ← 3 ”(combination of codes 1312 and 1315),“ sourceAddress [3-state] ← c, state ← 3 ”(code 1313 and 1315) is generated.

  The protocol stack compiler unit 102 calculates the cost of each code fragment generated at this time (S1402). The cost calculated here may be, for example, the length of the character string of the generated code, the binary size when the generated code is compiled, the estimated amount of calculation of the generated code, and the like.

  Further, the protocol stack compiler unit 102 counts the number of duplicate code fragments among all the generated code fragments (S1403). For example, since the code fragment “sourceAddress [state-1] ← c, state ← state + 1” generated for the state 1001 is also included in the code fragment group generated for the state 1002, the overlap number is 2 It is. Then, the protocol stack compiler unit 102 selects, for each node of all trees, a code fragment that minimizes a value obtained by dividing the cost value by a plurality of code fragments from the generated code fragment group, and this code fragment Is a code fragment assigned to the node (S1404).

  According to such processing, codes with lower costs can be preferentially assigned and overlapping codes can be integrated.

(3) RAM area optimization routine For example, when a tree branching from node A to node B and node C is generated, RAM area D used by the code generated from node B and its child nodes, and node C and its children The RAM area E used by the code generated from the node is not used at the same time. If these RAM areas D and E are made common, the RAM size of the execution code can be reduced. In this way, when code generated from several nodes uses a RAM area, a higher-performance and memory-saving protocol stack can be generated by optimizing the RAM area. Hereinafter, this process will be described.

  When the RAM area optimization routine is called in S609, the RAM areas that are not used at the same time are integrated by the RAM area sharing process of FIG. Here, the following definition is used, and the target class is S ← V (C) for C.

V (C) = {v1, v2 ...}: set of identification information of RAM areas used in class C D (C) = {d1, d2, ..., dm}: Class C message definition Set M (d) = {vi, vj, vk, ...}: set of identification information of RAM area used in message definition d

  First, the protocol stack compiler unit 102 checks whether or not the set S is an empty set φ (S1501), and as a result, if it is an empty set φ, the process is terminated. On the other hand, if it is not the empty set φ, a message definition d that satisfies the following formula is searched (S1502).

(M (d) ∩S) ∩ (M (d ′) ∩S) = φ, ∀d ′ ≠ d and
S∩M (d) ≠ φ and
S−M (d) S ≠ φ

  As a result, if the corresponding message definition d exists (S1503), the protocol stack compiler unit 102 divides S into S1 ← S∩M (d) and S2 ← SM (d), Considering each of these S1 and S2 as S, this algorithm is recursively called. When this recursive processing is completed, S1 and S2 obtained thereby are output as elements sharing the RAM area (S1504). For example, if there is a message definition d that satisfies the above conditions for S1 and S2, recursive processing with S1 and S2 and S2 as S obtained by recursive processing with S1, S2, and S1 as S, respectively. A total of four of S1 and S2 obtained by the above are elements sharing the RAM area. Of course, the elements finally obtained as the elements sharing the RAM area may be smaller than four or larger than four.

  On the other hand, if the corresponding message definition d does not exist (S1503), the protocol stack compiler unit 102 sets all the r satisfying the following formula for the identification information vεS of all RAM areas used in the class C. A set I (v) is obtained (S1505).

          ∀d∈D (C), r∈M (d) xor v∈M (d)

  Thereafter, the protocol stack compiler unit 102 minimizes the size of the shared RAM area of the set S1 ← (vi, vj,...) And the set S2 ← (I (vi) ∩I (vj) ∩ ...). In this way, a combination of elements belonging to S1 and S2 is selected. S1, S2, and S-S1-S2 are regarded as S, and this algorithm is recursively called. When this algorithm is completed, S1 and S2 are output as elements sharing the RAM area (S1506). For example, when S-S1-S2 is an empty set φ and has the same size as S1 and S2, the size of the required RAM area becomes half of the initial size by sharing S1 and S2.

  Among the optimization subroutines described above (tree integration subroutine, code integration subroutine, RAN area optimization subroutine), by changing the combination of subroutines to be read (for example, calling only one subroutine, etc.) A plurality of protocol stack codes are obtained as selection candidates, and a selection candidate suitable for the purpose (for example, one having the smallest execution cost such as processing performance and memory usage) is selected from the selection candidates. Try selecting it as a code. This makes it possible to generate a protocol stack with higher performance and smaller memory usage suitable for mounting on a small network device, for example.

  In the above description, the protocol stack code to be incorporated into the network device is generated on the information processing system, but it is not always necessary to do so. For example, the network device may have an execution function of the protocol stack code automatic generation processing, and the network device may generate the protocol stack code from the description contents of the protocol stack definition file regarding various packets processed by itself. .

  Furthermore, the one having the smallest execution cost (processing performance, required memory size, etc.) is selected from various virtual generation results, and is output as a protocol stack code. Thereafter, the general-purpose compiler unit 104 generates an execution code 105 that can be executed on the network device 116 from the protocol stack code.

It is a figure for demonstrating the function structure of the information processing system which concerns on embodiment of this invention. It is a schematic hardware block diagram of the information processing system which concerns on embodiment of this invention. It is a figure for demonstrating the example of a structure of the message data used as the process target of a protocol stack. It is a figure for demonstrating the description data of the protocol stack definition file which concerns on embodiment of this invention. It is a flowchart of the protocol stack code generation process which the protocol stack compiler which concerns on embodiment of this invention performs. It is the figure which showed the structure tree produced | generated from one message definition. It is the figure which showed the structure tree produced | generated from the several message definition which has the same message class name. It is the figure which showed the automaton produced | generated from the structural tree of FIG. It is a flowchart of a process of the automaton execution program which performs the automaton of FIG. It is a flowchart of the optimization process (tree integration process) which concerns on embodiment of this invention. It is an example of the generated code. It is a flowchart of the optimization process (code integration process) which concerns on embodiment of this invention. It is a flowchart of the optimization process (RAM area common processing) which concerns on embodiment of this invention.

Explanation of symbols

101: Protocol stack definition file, 102: Protocol stack compiler section, 103: Protocol stack code, 104: General-purpose compiler section, 105: Execution code

Claims (10)

A program generation method for causing an information processing apparatus having a storage unit and an operation unit to generate a data sequence processing program for causing a network device that receives a series of data sequences including a plurality of data elements via a network to process the data sequence. There,
The storage means stores a definition file in which item information including size information of each data element in the data string received by the network device is described according to the order of the data elements in the data string,
The program generation method is as follows:
A process in which the computing means refers to the definition file;
The definition file includes a program for the arithmetic means to process each data element in the data string in a predetermined size according to the order from the top of the data string as the data string processing program. Processing based on the description order of the item information of each data element and the size information of each data element in
A program generation method comprising: A program generation method according to claim 1, comprising:
The computing means is
A program in which the transition state is defined and the state of the program and the processing code in each state are defined, and according to the current state of the program using the data of the size in the data string A program for executing the processing code and transitioning the state of the program to the next state using the data of the size next to the data in the data string in accordance with the transition relation. A program generation method characterized by generating as a program. A program generation method according to claim 2, wherein
The computing means is
In association with each data element, the state of the program and the processing code are determined,
When there are a plurality of states having the same processing code among the states corresponding to the respective data elements, any one of the plurality of states is used as the plurality of states. , And determining the transition relation. A program generation method according to claim 2, wherein
The computing means is
A plurality of processing code candidates are generated for each state, and a processing code candidate common to a processing code candidate of another state is selected from the processing code candidates of the state as the processing code of each state. A program generator. A program generation device that generates a data sequence processing program for causing a network device that receives a series of data sequences including a plurality of data elements via a network to process the data sequence,
Storage means storing a definition file in which item information including size information of each data element in the data string received by the network device is described according to the order of the data elements in the data string;
As the data string processing program, a program for processing each data element in the data string by data of a predetermined size in the order from the top of the data string, the data in the definition file Calculation means for generating based on the description order of the item information of the elements and the size information of each data element,
A program generation device comprising: The program generation device according to claim 5,
The computing means is
A program in which the transition state is defined and the state of the program and the processing code in each state are defined, and according to the current state of the program using the data of the size in the data string A program for executing the processing code and transitioning the state of the program to the next state using the data of the size next to the data in the data string in accordance with the transition relation. A program generation device characterized by generating as a program. The program generation device according to claim 6,
The computing means is
In association with each data element, the state of the program and the processing code are determined,
When there are a plurality of states having the same processing code among the states corresponding to the respective data elements, any one of the plurality of states is used as the plurality of states. A program generating apparatus that defines the transition relationship. The program generation device according to claim 6,
The computing means is
A plurality of processing code candidates are generated for each state, and a processing code candidate common to a processing code candidate of another state is selected from the processing code candidates of the state as the processing code of each state. A program generator. A compiler for causing an information processing apparatus having a storage means and an arithmetic means to execute generation processing of a data string processing program for causing a network apparatus that receives a series of data strings including a plurality of data elements via a network to process the data string Because
The storage means stores a definition file in which item information including size information of each data element in the data string received by the network device is described according to the order of the data elements in the data string,
The compiler
A process in which the computing means refers to the definition file;
The definition file includes a program for the arithmetic means to process each data element in the data string in a predetermined size according to the order from the top of the data string as the data string processing program. Processing based on the description order of the item information of each data element and the size information of each data element in
A compiler that causes the information processing apparatus to execute. A compiler according to claim 9, wherein
In the calculation means,
A program in which the transition state is defined and the state of the program and the processing code in each state are defined, and according to the current state of the program using the data of the size in the data string A program for executing the processing code and transitioning the state of the program to the next state using the data of the size next to the data in the data string in accordance with the transition relation. A compiler characterized by generating as

Images