Home

ModelMaker: A Tool for Rapid Modeling from Device Descriptions

1. This appendix shows a behavioral model created using the Synopsys Graphical Environment SGE shell and the initial VHDL model generated by the SGE shell The screen capture of the behavioral model is shown in Figure B 1 Symbol Editor dmaS237a BLOCK Figure B 1 The Behavioral Schematic Drawn in SGE The initial VHDL model file created from the behavioral schematic shown in Figure B 1 is shown in the next page 111 Appendix B Behavioral Schematic and Its Initial VHDL Model Basic VHDL File Generated for the Behavioral Schematic VHDL Model Created from SGE Symbol dma8237a sym 00t 5 23527523 1997 library IEEE use IEEE std_logic_1164 all use IEEE std_logic_misc all use IEEE std_logic_arith all use IEEE std_logic_components all entity DMA8237A is Generic DELAY TIME 0 NS Port CLK cs DREQO DREQ1 DREQ2 DREQ3 HLDA PINS READY RESET VCC VSS AO Al A2 A3 DBO DB1 DB2 DB3 DB4 DB5 DB6 DB7 EOP IOR IOW A4 A5 A6 A7 ADSTB AEN DACKO DACK1 DACK2 DACK3 HRQ MEMR MEMW end DMA8237A In In In In In In in In In In In In InOut InOut InOut InOut InOut InOut InOut InOut InOut InOut InOut InOut InOut InOut InOut Out Out Out Out Out Out Out Out Out Out Out Out Out STD_ STD STD STD STD_ STD STD STD STD STD STD STD_ STD STD STD STD STD STD STD STD STD STD_ STD_ STD STD_ STD STD STD STD STD_ STD_
2. desired by the modeler the search results of a multi word keyword are combined Similarly based on the Search Mode desired by the modeler the final search results are sorted before they are returned to the modeler The combining and the sorting processed will be discussed in more detail in the Keyword Searches section see section 4 4 4 3 2 2 Main Member Functions of CChart Class There are four important member functions in CChart class they are AddNode ParseNP ParseSentences and ClearChart The functionality and internal behavior of these functions are discussed in this section 68 Chapter 4 ModelMaker Basic Design AddNode m_pToken The AddNode public member function is needed to add a node to the CChart parse tree This function takes one argument the token to be added as the new node When the new node is the first node in the chart this member function will also create and initialize the chart If the chart exists a new node is created and added to the end of the node list of the existing chart This function returns a flag to specify whether it was run successfully Once the node is created the GetMeaning member function of the standard CDictionary object is called to determine the token s category part of speech If one of the parts of speech of the token is an asterisk sign however the token may be a part of a multi word token In this case the multi word dictionary object is consulted
3. DELAY gt 0 NS Port Map DO_D1 gt N_9 INTERNAL_DATA_BUS gt N_4 DBO_DB7 gt DBO_DB7 I_8 COMMAND_CONTROL Generic Map DELAY gt 0 NS Port Map DO_D1 gt N_9 X04 gt N_8 I9 OUTPUT_BUFFER Generic Map DELAY gt 0 NS Port Map X03 gt N_7 A4_A7 gt A4_A7 I_10 I_O_BUFFER Generic Map DELAY gt 0 NS Port Map X03 gt N_7 X04 gt N_8 AO_A3 gt A0_A3 IIL INC_DECREMENTOR Generic Map DELAY gt 0 NS Port Map INTERNAL_DATA_BUS gt N_4 X01 gt N_6 X03 gt N_7 I 12 DECREMENTOR Generic Map DELAY gt 0 NS Port Map X01 gt N_6 k SR e READ_WRITE_BUFFER Generic Map DELAY gt 0 NS Port Map X01 gt N_6 X02 gt N_5 X05 gt N_13 X06 gt N_12 X07 gt N_11 X08 gt N_10 I_14 WRITE_BUFFER 117 Appendix C Structural Schematic and Its Initial VHDL Model Generic Map DELAY gt 0 NS Port Map INTERNAL_DATA_BUS gt N_4 X05 gt N_13 X06 gt N_12 allen Us READ _BUFFER Generic Map DELAY gt 0 NS Port Map INTERNAL _DATA_BUS gt N_4 X07 gt N_11 X08 gt N_10 I_16 READ_WRITE Generic Map DELAY gt 0 NS Port Map INTERNAL_DATA_BUS gt N_4 I_17 STATUS Generic Map DELAY gt 0 NS Port Map INTERNAL_DATA_BUS gt N_4 I_18 BASE_READ_BUFFER Generic Map DELAY gt 0 NS Port Map X02 gt N_5 X05 gt N_13 X06 gt N_12 I_6 TEMPORARY Generic Map DELAY gt 0 NS Port Map INTERNAL_DATA_BUS gt N_4 ia REQUEST Generic Map DELAY gt 0 NS
4. Port Map INTERNAL_DATA_BUS gt N_4 X11 gt N_2 I_4 MASK Generic Map DELAY gt 0 NS Port Map INTERNAL_DATA_BUS gt N_4 X10 gt N_3 I_5 COMMAND Generic Map DELAY gt 0 NS Port Map INTERNAL_DATA_BUS gt N_4 X09 gt N_1 Tl TIMING_AND_CONTROL Generic Map DELAY gt 0 NS Port Map ADSTB gt ADSTB_DUMMY AEN gt AEN_DUMMY CLOCK gt CLOCK CS gt CS EOP gt EOP IOR gt IOR IOW gt IOW MEMR gt MEMR_DUMMY MEMW gt MEMW_DUMMY READY gt READY RESET gt RESET X09 gt N_1 L2 PRIORITY_ENCODER_AND_ROTATING_PRIORITY_LOGIC Generic Map DELAY gt 0 NS Port Map DACKO_DACK3 gt DACKO_DACK3_DUMMY end SCHEMATIC DREQO_DREQ3 gt DREQO_DREQ3 X10 gt N_3 X09 gt N_1 HLDA gt HLDA HROQ gt HRQ_DUMMY X11 gt N_2 configuration CFG_DMA8237A_SCHEMATIC of DMA8237A is for SCHEMATIC For I 7 I O_BUFFER_2 end for For I 80 use con end for For 19 0 use con end for for I_10 use con for I_ll use con end for for 12s use con end for For I 13 use con for I l4 use con for ToT 5 use con end for for I_16 use con end for end for end for use con figuration WORK OMMAND_CONTROL figuration WORK UTPUT_BUFFER figuration WORK I_O_BUFFER figuration WORK INC_DECREMENTOR figuration WORK DECREMENTOR figuration WORK figuration WORK WRITE_BUFFER figuration WORK READ_BUFFER figuration WORK READ_WRITE figuration WOR
5. ModelMaker also provides the search capability to allow modelers to work on any particular part of the device Moreover ModelMaker allows the modeler to view both the information and the VHDL code at the same time This is important for a rapid modeling process The lists of the currently available EDA tools are obtained from the HP web site and from the EE Times web site The latter can be found at http techweb cmp com eet eda edalinks html As in the case with the list provided by HP this URL can be changed at any time Consult the EE Times web master for more information 19 Chapter 3 Modeling with ModelMaker CHAPTER 3 MODELING WITH MODELMAKER This chapter describes the operation of ModelMaker An example is used here to show how a modeler can construct a device model using ModelMaker At the same time a description of how the internal classes are managed created deleted and used by ModelMaker is also presented in this chapter The primary objective of this chapter is to present an overall view of how ModelMaker works and to show how ModelMaker can be used by hardware modelers The product data sheet for the Intel 8237A direct memory access controller chip 15 is used as an example source document In creating this model the modeler is assumed to have access to a scanner Optical Character Recognition OCR software and the Synopsys Graphical Environment SGE shell software The scanner and software are needed t
6. Port Xoe in BIT Y SLN BIT Z 2 Out BIT end component component CHIPABC Generic DELAY TIME 0 NS Port A In BIT In Bit In BIT BIT BIT end component begin OUT3 lt OUT3_DUMMY I_1 CHIPXYZ Generic Map DELAY gt 3 NS Port Map X gt IN3 Y gt IN4 Z gt OUT3_DUMMY I_2 CHIPABC Generic Map DELAY gt 0 NS Port Map A gt IN1 B gt IN2 C gt OUT3_DUMMY D gt OUT1 3 end SCHEMATIC configuration CFG_STRUCTURAL_SCHEMATIC of STRUCTURAL is for SCHEMATIC for I_1l CHIPXYZ use configuration WORK CFG_CHIPXYZ_BEHAVIORAL end for for I_2 CHIPABC use configuration WORK CFG_CHIPABC_BEHAVIORAL end for end for end CFG_STRUCTURAL_SCHEMATIC Figure 3 6 The Initial Structural VHDL Model File 26 Chapter 3 Modeling with ModelMaker Once the modeler has both the input files the specification document text file and the initial VHDL file he she can start the modeling process using the ModelMaker tool It is suggested that both of these files be placed in the same directory with the ModelMaker executable file The modeler starts the ModelMaker tool by pressing the ModelMakerlIII which is linked to the ModelMakerlIIl exe file on his her Windows Explorer window This action executes the ModelMaker tool At this point the modeler must specify the location of the input files He she can do this by simply typing the file names
7. STD_ STD STD STD STD STD_ STD STD STD LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC LOGIC architecture BEHAVIORAL of DMA8237A is begin end BEHAVIORAL configuration CFG_DMA8237A_BEHAVIORAL of DMA8237A is for BEHAVIORAL end for end CFG_DMA8237A_BEHAVIORAL 112 Appendix C Structural Schematic and Its Initial VHDL Model APPENDIX C STRUCTURAL SCHEMATIC AND ITS INITIAL VHDL MODEL This appendix shows a structural model created using the Synopsys Graphical Environment SGE shell and the initial VHDL model generated by the SGE shell The screen capture of the structural model is shown in Figure C 1 and the initial VHDL model generated from it is shown in the next page 113 Appendix C Structural Schematic and Its Initial VHDL Model b aw PT HH JOYLNOS ONF NIWIL Figure C 1 The Structural Schematic Drawn in SGE 114 Appendix C Structural Schematic and Its Initial VHDL Model Basic VHDL File Generated for the Structural Schematic VHDL Model Created from SGE Schematic dma8237a sch Jan 30 16 27 10 1998 library IEEE use use use use entity Port end DMA8237A DMA8237A is IEEE std_logic_1164 al
8. X02 InOut BIT_VECTOR 0O TO 15 X05 InOut BIT_VECTOR O TO 15 X06 InOut BIT_VECTOR O TO 15 end component component TEMPORARY Generic DELAY TIME 0 NS Port INTERNAL_DATA_BUS InOut BIT_VECTOR 0O TO 15 end component component REQUEST Generic DELAY TIME 0 NS Port INTERNAL_DATA_BUS InOut BIT_VECTOR 0O TO 15 X11 InOut BIT end component 116 Appendix C Structural Schematic and Its Initial VHDL Model component MASK Generic DELAY TIME 0 NS Port INTERNAL_DATA_BUS X10 InOut BIT end component InOut BIT_VECTOR O TO 15 3 component COMMAND Generic DELAY TIME 0 NS Port INTERNAL_DATA_BUS X09 InOut BIT end component InOut BIT_VECTOR 0O TO 15 component TIMING_AND_CONTROL Generic DELAY TIME 0 NS Port ADSTB In BIT AEN In BIT CLOCK In BIT CS In BIT EOP In BIT IOR In BIT IOW In BIT MEMR In BIT MEMW In BIT READY In BIT RESET In BIT x09 Inout BIT end component component PRIORITY_ENCODER_AND_ROTATING_PRIORITY_LOGIC Generic DELAY TIME 0 NS Port DACKO_DACK3 In BIT_VECTOR 0 TO 3 DREQO_DREQ3 In BIT_VECTOR 0O TO 3 HLDA In BIT HRQ In BIT X09 InOut BIT X10 InOut BIT X11 InOut BIT end component begin DACKO_DACK3 lt DACKO_DACK3_DUMMY HRQ lt HRQ DUMMY MEMW lt MEMW_DUMMY MEMR lt MEMR_DUMMY ADSTB lt ADSTB_DUMMY AEN lt AEN_DUMMY I_7 I_O_BUFFER_2 Generic Map
9. can be seen in Figure A 2 B mdictbit result in verb _ vinf _ results in verb _ consist of verb _ consists of verb _ each of pdet _ in spite of prep _ instead of scon _ pair of _ regardless of prep _ depend on verb _ vinf _ depending on prep _ ving _ depends on verk _ synthetic aperture radar noun _ farther than adj k faster than adj fewer than adj _ greater than pdet _ less than pdet _ according to prep _ greater than or equal to pdet less than or equal to pdet _ a Figure A 2 Captured Multi Word Dictionary File 96 Appendix A ModelMaker v3 8 User Manual Gram txt ASCII file This file contains the grammar rules used by the ModelMakerlIII application Editing this file is not recommended as this can cause an abnormal operation of the ModelMakerlII tool Add rules to this grammar file only after a very careful consideration More information on this grammar file can be found in section 4 2 3 3 of the thesis A captured image of part of the grammar file is presented in Figure A 3 This grammar file contains five main fields on each line Each of the fields is separated by a space character The first field is called the rule ID field This field contains the grammar rule identifier for this entry The second field is the constituent count field This field specifies the number of constituents that are needed to create the product of
10. it generates several temporary files The main temporary file is the file called suite pl This file is written automatically by ModelMaker during initialization This file contains the parse trees of sentences in a Prolog format that will be used by the Template Constructor as input This file is deleted automatically once ModelMaker is terminated ModelMaker also generates files called _out pl _done pl and _runsemanal pl The last two files are deleted automatically by ModelMaker If ModelMaker is terminated abnormally and some of these two temporary files were not deleted it is recommended that these files be oe deleted manually The _out pl file however is the output file generated by the Template Constructor If the modeler is planning to use the Template View capability in ModelMaker he she needs to keep this file in this ModelMaker directory This file is generated when the modeler presses the Template Re Analyze button Though ModelMaker does not create any permanent output files it provides the user with the ability to save the VHDL code back to the VHDL input file 4 3 Object Oriented Description ModelMaker is implemented in C This is done to ensure maintainable and expandable code as ModelMaker is still in its prototype age The concepts of inheritance encapsulation member functions and member variables refer to the 55 Chapter 4 ModelMaker Basic Design Master
11. made point the mouse pointer anywhere in the Source Window and click the right mouse button This action will bring up a pop up editing menu window that allows the modeler to copy the selected information Model Window Customized Edit Window The Model Window placed in the bottom portion of the main working window is initialized with the basic VHDL model The Model Window is a read write edit window Information can be entered to and removed from the Model Window at any time To cut copy or paste information from and to this Model Window follow the same procedure with copying information from the Source Window First make the selection Second point the mouse pointer inside the window Third click the right mouse button to bring up the editing menu Keywords Window Editable Drop Down List This drop down list box is where the list of the identifiers or the list of the noun phrases will be presented to the modeler The modeler can either choose one of the keywords listed there or type his her own phrase in the keywords window This keyword is used by the ModelMaker tool as the search phrase Keywords Type Radio Buttons This radio control button lets the modeler choose whether to display the identifier keyword list or the noun phrase keyword list in the Keywords Window Search Type Drop Down List This drop down list allows the modeler to decide whether he she wants to perform a Boolean and search or a Boole
12. the structural dictionary and the behavioral dictionary The CGrammar and CMWDictionary objects are used to keep the information of the standard grammar and the multi word dictionary The pointer to these CDictionary CMWDictionary and CGrammar objects are then passed to the CChart constructor to create the CChart object 56 Chapter 4 ModelMaker Basic Design CDictionary ee ee a 3 Z CMWDictionary Load f CGrammar f Load Member i i Load 4 i LookUp GetMeaning E E RA i CTokenList CLexer 7 a a j T d Open O ae CSpecification i 1 7 GetToken gt tae CElement Ee LoadTxt CElement a LoadVhd FormHits F CChart AddNode o ClearChart 4 s ParseNP Figure 4 4 CSpecification Class Diagram The CSpecification uses a CLexer object to obtain the specification document token This task is primarily done by CSpecification LoadTxt section 4 3 2 1 Each token that the CSpecification object receives from the CLexer object is then parsed for its hierarchy type This is done to specify whether the token is a part of a paragraph a section header or a table Each token that is a part of a sentence is then passed to the CChart object by means of CChart AddNode This allows the CChart object to create a new node for the token and to parse it for possible new constituents The CChart object is also responsible for supplying the CSpecification
13. this CMWDictionary class is closer to the CGrammar class in its implementation 29 Chapter 3 Modeling with ModelMaker because this class is a static class that does not need any outside information There is no argument needed for the CMWDictionary constructor member function As in the case with the CGrammar class the CSpecification class constructor member function calls the CMwDictionary Load to initialize the CMWDictionary object The input file that contains all of the grammar knowledge is set to be the mdict txt file To illustrate this multi word dictionary object some portion of the aMDictionary object is presented in Figure 3 9 More information about this CGrammar class can be found in section 4 3 1 4 vd consist gt of verb gt infinitive 1 consists gt of verb l each gt of pre determiner aMDictionary 99 Pointer Figure 3 9 Part of the CMWDictionary Structure As soon as these grammar and dictionary objects are created an object of the CChart class called aChart is created The aChart object is used to store the constituents of the sentences The grammar and the dictionary objects need to be instantiated first as the CChart class needs to use these objects as its knowledge database Pointers to the aGrammar object the aDictionary object and the aMDictionary object are passed to the aChart object through the CChart constructor The aChart object created by the constructor is an
14. CGrammar Load function Constructor member functions are special member functions that are called when class objects are created 24 28 Chapter 3 Modeling with ModelMaker reads the input file and places the grammatical knowledge into its internal array of grammar rule objects The Noun Phrase Extractor needs this array to perform its parsing process Refer to section 4 3 1 5 for more information about this CGrammar class The aDictionary object is created to store the standard dictionary In calling the CDictionary constructor an argument specifies the file name of the dictionary file to be used This argument is needed since more than one Dictionary object is used in ModelMaker The standard dictionary file used to create aDictionary object is the rdict txt file To illustrate this dictionary object some portion of the aDictionary object is presented in Figure 3 8 Section 4 3 1 3 gives more information about this CDictionary class o J Q ca D 5 se Y break gt noun gt verb gt infinitive breaking gt gerund breaks gt verb briefly gt adverb Pointer t 4 y 4 4 4 4 4 y Figure 3 8 Part of the CDictionary Structure Similarly the CMWDictionary object is created by calling the CMWDictionary constructor member function The MW in the name of this class stands for Multi Word This object is created to store the multi word dictionary knowledge However
15. From 11 To 11 Parts 43 ed Figure 3 10 Sample of the CChart Structure The operation done by the CChart ParseNP function is performed just before the m_pSpec object clears the aChart object This function extracts all of the noun phrases from the aChart object These noun phrases are used by m_pSpec object as the noun phrase keywords More information about the CChart ParseNP public member function can be seen in section 4 3 2 2 Another operation done by the m_pSpec object before it clears the aChart object is to call the CChart ParseSentences public member function CChart ParseSentences extracts all of the sentence parse trees found in the existing chart and writes them to the Template Constructor input file This operation is essential 33 Chapter 3 Modeling with ModelMaker for the Template Constructor component which analyzes the categories of sentences based on the syntactic structure represented in the parse trees Once the CSpecification LoadVhd and the CSpecification LoadTxt private member functions are executed successfully m_pSpec is ready to perform the pre search to construct the search indexes With the CSpecification LoadVhd the m_pSpec object received the VHDL identifiers used in the initial VHDL model file Moreover with the CSpecification LoadTxt the m_pSpec object received an array of strings that contain all of the noun phrases from the source specificat
16. K CFG_I_O_BUFFER_2_BEHAVIORAL CFG_COMMAND_CONTROL_BEHAVIORAL CFG_OUTPUT_BUFFER_BEHAVIORAL CFG_I_O_BUFFER_BEHAVIORAL CFG_INC_DECREMENTOR_BEHAVIORAL CFG_DECREMENTOR_BEHAVIORAL READ_WRITE_BUFFER CFG_READ_WRITE_BUFFER_BEHAVIORAL CFG_WRITE_BUFFER_BEHAVIORAL CFG_READ_BUFFER_BEHAVIORAL CFG_READ_WRITE_BEHAVIORAL 118 Appendix C Structural Schematic and Its Initial VHDL Model end for for I_17 STATUS use configuration WORK CFG_STATUS_BEHAVIORAL end for for I_18 BASE_READ_ BUFFER use configuration WORK CFG_BASE_READ_BUFFER_BEHAVIORAL end for for I_6 TEMPORARY use configuration WORK CFG_TEMPORARY_BEHAVIORAL end for for I_3 REQUEST use configuration WORK CFG_REQUEST_BEHAVIORAL end for for I_4 MASK use configuration WORK CFG_MASK_BEHAVIORAL end for for I_5 COMMAND use configuration WORK CFG_COMMAND_BEHAVIORAL end for for I_1l TIMING_AND_CONTROL use configuration WORK CFG_TIMING_AND_CONTROL_BEHAVIORAL end for for I_2 PRIORITY_ENCODER_AND_ROTATING_PRIORITY_LOGIC use configuration WORK CFG_PRIORITY_ENCODER_AND_ROTATING_PRIORITY_LOGIC_BEHAVIORAL end for end for end CFG_DMA8237A_SCHEMATIC 119 Vita VITA Andreas Indra Gunawan was born in Surabaya Indonesia at the end of the year 1973 He came to the United States in June 1992 to attend West Virginia University WVU in Morgantown West Virginia He took an intensive Eng
17. Npoeta uan aae Figure 4 7 CMWDictionary Class Diagram 4 3 1 5 CGrammar Class As the CDictionary class is responsible for creating and holding the dictionary objects the CGrammar class shown in Figure 4 8 is responsible for creating and holding the grammar objects The data structure of the grammar entries is implemented in a constant array of objects This fact limits the current CGrammar object as it can only hold up to 150 rules this limit can be increased simply by changing a variable in the code The grammar size is expected to be moderately constant and the access speed of a constant array structure is faster than the access speed of a linked list structure r CGrammar qf CRule Figure 4 8 CGrammar Class Diagram The CGrammar Load member function loads the dictionary object from an external file This function needs to be called explicitly as an automatic loader is not yet implemented in the CGrammar constructor 61 Chapter 4 ModelMaker Basic Design To access any of the rules the CGrammar user can directly call the CRule object of the grammar structure by its index An integer public member variable of this CGrammar nRules is used to store the number of grammar rules that are successfully loaded by this class The CGrammar user can use this number as the upper limit in accessing the grammar rules 4 3 1 6 CElement Class CElement objects are used to store the source document inside the CSpecificati
18. The input for this function is the token returned from the CLexer object This function returns a flag to indicate whether it was run successfully or not A document inside the m_pSpec object is a collection of document elements These elements include sections paragraphs sentences tables and figures To maintain the document hierarchy these document elements are built from the CElement class In the current implementation a section leads paragraphs and a paragraph leads sentences tables and figures The current algorithm implemented in this LoadTxt function tries to recognize the document elements automatically for any input document A set of general rules to format a document for recognition is described in details in section 4 2 4 2 These rules are quite general and should include many document formats without requiring any adjustment The following two figures show the CSpecification document hierarchy built by CSpecification LoadTxt Figure 4 11 shows a sample document and Figure 4 12 shows how this document will be formed into a CSpecification object 65 Chapter 4 ModelMaker Basic Design 1 FUNCTIONAL DESCRIPTION The 3237a block diagram includes the major logic blocks and all of the internal registers The data interconnections paths are also shown Figure 3 8237A Internal Registers Name Size Number Base Address Registers 16 bits 4 Base Word Count Registers 16 bits 4 2 DIMA OPERATION The 8237A is d
19. be written manually or extracted automatically from a schematic capture program such as Synopsys Graphical Environment SGE Shell from Synopsys On initialization ModelMaker loads this file stores it as a local variable in the memory and displays it to the user in the Model Window During the loading process ModelMaker also parses this file to extract all of the VHDL identifiers to be used as search keywords More details on this Identifier Extractor can be found in section 4 3 2 1 Main Member Function of CSpecification Class 53 Chapter 4 ModelMaker Basic Design 4 2 4 2 Source Text File This file ingests the ASCII specification document text for the device that the modeler wishes to model In order to store this information in a document data structure ModelMaker attempts to recognize the format of this specification document At this time the ModelMaker document reader can recognize only a limited variation of formats ModelMaker separates the document into five element types section header paragraph sentence table and figure More detailed definitions of these element types are presented in the CSpecification Class section 4 3 1 1 In dividing the document into those five elements ModelMaker uses the following rules Elements are separated by two line feeds or two carriage returns Each section must have a section header this rule can be violated for the first section The section header must begin on
20. document with many sentences maintaining a complete chart for the whole specification document is not practical Each node can have as few as one constituent and as many as a thousand or more constituents The span of each of these constituents can be as small as a single word or as long as the whole sentence This function provides a way for the CSpecification object to delete all node and constituent objects when it detects the beginning of a new section or paragraph or the end of a sentence or the document If the value of Capture is true the existing chart is returned as a CString object Otherwise the return value of this function is an empty CString object 70 Chapter 4 ModelMaker Basic Design While doing the cleaning task the ClearChart function also searches the existing chart for any constituents representing a complete sentence by means of CChart ParseSentences function For each of such sentence constituents a parse tree is constructed and added to the Template Constructor input file suite pl These trees will be analyzed by the Template Constructor 4 3 2 3 Main Member Functions of CDictionary Class There are three main member functions of CDictionary class that are presented in this section These functions are the Load Member and GetMeaning functions Load FileName This private member function is used to load a dictionary input file for constructing a CDictionary object A file name is requir
21. for this public member function is the CNode object of the last word in the multi word in question This public member function provides a simple way for the user to access the categories of a particular multi word LookUp is performed by a simple linear search If an entry is found for the keyword that the user searched for a pointer to the matching CMWEntry object will be returned Otherwise a null pointer is returned 12 Chapter 4 ModelMaker Basic Design 4 3 2 5 Main Member Function of CGrammar Class The only member function of the CGrammar class is the Load function To access the rules of a grammar object the user can simply refer to the public member variable array of CRule objects used to store the grammar Load This public member function is similar to the CDictionary Load function This function loads the grammar file gram txt to construct a CGrammar object In the current implementation the CGrammar data structure can only store up to 150 grammar rules If the grammar file is loaded successfully the number of grammar rules loaded is returned to the caller The function requires no argument 4 3 2 6 Main Member Functions of CElement Class The CElement class is a generic class used to create the sentences paragraphs sections and documents element This function only has one main member function the FormHits function FormHits Token Hits BehaveHits StructHits pFirstHit aBehaviorDic aSt
22. must have a section header this rule can be violated for the first section e The section header must begin on a new line and be written in the format of Caption In this format represents digits from zero to nine and Caption can represent any string followed by two carriage returns The part inside the brackets can be omitted or repeated Examples of valid section headers would include 1 Introduction 2 1 Search Algorithm or 2 1 3 Multiple Search 100 Appendix A ModelMaker v3 8 User Manual e Tables and figures must be introduced by writing the word Table or Figure whichever is more appropriate as the first word An example of a valid table element can be seen below This is the last sentence from the last paragraph Table 1 1 Important Files Cartage Rei One cpp The first file is required in Two cpp The second file these locations Three cpp The third filet Retum L J This is the first sentence of the next paragraph en e Any other element group of sentences starting and ending with two carriage returns that is not recognized as a section table or figure will be called a paragraph e A period is used to declare the end of a sentence Each group of words that is inside a paragraph and ends with a period is called a sentence An exception to this rule can occur with the last sentence in the paragraph A paragraph can contain an unlimited number of sentences
23. object with the list of noun phrases that it has recognized These noun phrases will be used as the noun phrase search keywords The private member function CSpecification LoadVhd section 4 3 2 1 is responsible for loading the VHDL input file to the CSpecification object While CSpecification LoadVhd loads the input file it also extracts all of the VHDL 57 Chapter 4 ModelMaker Basic Design identifiers These identifiers are the words that will then be used by the CSpecification object as identifier keywords For efficiency both the identifier keywords and the noun phrase keywords are sorted and filtered for duplicates Once all noun phrases from the CChart object and the identifiers from the VHDL input file are compiled and filtered the CSpecification object re parses the document to search for occurrence of these keywords The results of these searches are indexes stored in CHitList objects A CHitList object is simply an object with two pointers and an integer One pointer is pointing to the CElement objects that contain the noun phrase or identifier and the other one is pointing to the next CHitList object The integer indicates the number of occurrences hits that a keyword has All of these processes and searches are done before the main window of ModelMaker is initialized Pre processing allows faster searches during the modeling activity 4 3 1 2 CChart Class The CChart class shown in Figure 4 5 is responsible fo
24. on the space provided or finding the files using the Browse buttons on the dialog box shown in Figure 3 7 In this example the specification document file is the dma8237a txt file and the basic VHDL file is the dma8237a vhd file ModelMakerIII Initiator dma8237a txt dma8237a vhd E lt Figure 3 7 The Dialog Box Used to Specify the Location of the Files Once the Continue button is pressed ModelMaker starts its actual processing Depending on the processing power of the machine used this initial processing may require some time During this initialization time ModelMaker creates the specification document object searches for keywords and compiles the search result database for each of the keywords Internally the first thing that ModelMaker performs during the initial processing is checking for the ModelMakerlI log setup file This file is created when 27 Chapter 3 Modeling with ModelMaker ModelMaker is executed for the first time In the current implementation of ModelMaker this configuration file is used only to store the path location to the SWI Prolog interpreter For future versions however this configuration file can be used to store configuration information If this file does not exist a dialog box is displayed to ask the modeler for the location of the SWI Prolog interpreter If the file exists it will be opened and used to configure the ModelMaker tool Next M
25. one shows that the operation in the data bus can be done only after the processor sent out the HLDA signal Based on this information the modeler can start adding his her process to the initial VHDL model The modeler can call it the data_bus process and use the HLDA signal in its sensitivity list This new process is illustrated in Figure 3 12 data_bus_process process HLDA begin if DMWA_CYCLE then DATA_BUS lt ADDRESS HL_8_BITS end if if READ_FROM_MEMORY then DATA_BUS lt DATA MEMORY end if end process Figure 3 12 VHDL Process Created Based on Search Phrase data bus The process shown in Figure 3 12 is certainly not a final VHDL process The DMA_CYCLE and READ_FROM_MEMORY activities need to be defined as signals and the HI_8_BITS of MEMORY address need to be specified This VHDL process however can serve as a starting point where the modeler can build his her model The modeler can continue his her modeling by performing searches on the signals that he she needs in this initial process The second illustration building a process based on information about a pin In this case the modeler is interested in modeling the system from the external pins The modeler can pick a pin and try to model the necessary process for the particular pin In 37 Chapter 3 Modeling with ModelMaker this illustration the modeler starts from the EOP pin Again the modeler can either choose this phrase from
26. only under the Microsoft Visual C environment For this reason this class will not be explained in details Lexer cpp amp Lexer h The only class in these files is the CLexer class This class is responsible for loading the input files and dividing their character strings into tokens A more descriptive definition of a token is presented in the section about the CLexer class 4 3 1 7 This class is built to support the specification document reader function of ModelMaker An object of this class is utilized by the CSpecification method to read the specification document By using this class ModelMaker is able to distinguish between a period that acts as a sentence terminator and a period that acts as part of a decimal number MMREditWin cpp amp MMREditWin h These files house the CMMREditWin class This class is inherited from the CRichEditCtrl class of the Microsoft Foundation Class Similar to the CInitiator class this class is platform dependent It will work only under the Microsoft Visual C environment ModelMaker uses two objects that are of 46 Chapter 4 ModelMaker Basic Design this class The first one is the source edit window and the second is the model edit window This edit window is customized as it has to be able to receive text with custom formatting tags created by the Template Constructor Using this class for the edit window ModelMaker is able to display text in italic bold and color ModelMakerlll c
27. or more multi word phrases that ended with the entry word are listed in the multi word dictionary Once the Noun Phrase Extractor finds this asterisk it invokes the multi word dictionary object to check for another possible category For example in Figure A 1 the asterisk sign for the entry word than refers to the multi words farther than faster than fewer than greater than and less than which are listed in the multi word dictionary e Mdict txt ASCII file This file contains the multi word dictionary knowledge for ModelMakerlIII The user is encouraged to make more entries at any time However each time a new multi word entry is added to this multi word dictionary the standard 95 Appendix A ModelMaker v3 8 User Manual dictionary file needs to be updated with the multi word tag See discussion on the rdict txt file above for more information about this multi word tag The entry field structure of the multi word dictionary file mdict txt is similar to the entry field structure of the standard dictionary file Each line of this multi word dictionary file also represents one multi word entry in the dictionary Each of the entry lines is also divided into two or more fields as in the standard dictionary file The only difference between the two entry field structures is that the multi word dictionary entry field structure allows more than one word in the first field A screen capture from a part of this dictionary
28. returned In this implementation of CLexer class all tokens will be returned in lower case letters No argument is needed in calling this GetToken function 4 4 Keyword Searches To explain the basic design of ModelMaker some of its main algorithms are discussed here Since the occurrences of the keywords are pre searched during ModelMaker initialization most of the discussions in this section deal with how ModelMaker returns this search result back to the modeler The pre search is done by the CSpecification class The pre search is performed just after the LoadVhd and LoadTxt functions execute Pre search fills the results index with the search results hit lists of the keywords in the keywords index For each of the keywords the CElement FormHits function refer to section 4 3 2 6 is called using the keyword tokens as the search keys This is done for both the VHDL identifier keywords index and the noun phrase keywords index For the VHDL identifiers the pre search processing is straightforward as they are single tokens Most of the noun phrases however are composed of two or more tokens Each word from these multiword noun phrases needs to be searched separately The results are then combined 74 Chapter 4 ModelMaker Basic Design 4 4 1 AND Search Mode The AND search mode specifies how ModelMaker builds the hit lists for multiple token keyword searches In this mode ModelMaker searches each paragraph for any occ
29. the ModelMaker directory with a file name out txt By pressing the Template View button ModelMaker will read this pre analyzed result and present it in the Source Window If the modeler does want to re analyze the document he she can use the Template Re Analyze button When this button is pressed ModelMaker will invoke the Template Constructor and present the output in the Source Window ModelMaker also gives the option to the modeler to view the document parse trees This can be done by pressing the Show Tree button This capability is used mainly for debugging purposes By viewing the tree one can see how the sentences are parsed within ModelMaker As the process of constructing these trees also requires a great amount of time a copy of preprocessed trees is stored in the ModelMaker local directory By default pressing this button will cause ModelMaker to read from the local file If the modeler wanted to reconstruct the parse trees he she can select this option during ModelMaker initialization Chapter 1 Introduction The last button the ModelMaker button functions as the about button By pressing this button the modeler can see the version and copyright information about this ModelMaker tool To illustrate the functionality of ModelMaker the search results for the identifier keyword hrq are presented below For this example the specification document used is the commercial 8237
30. the algorithms were written for functionality instead of performance Consequently some of these algorithms cause a long processing time A 486 based machine or higher should be able to run this ModelMaker application with no problem Naturally the higher the speed of the processor the better it is for this ModelMaker application 6 3 Future Enhancements As the research continues many enhancements can still be added to this ModelMaker tool These enhancements can be either in the performance direction or in the functionality direction Most of the ModelMaker code uses the object oriented programming method This fact should make the code modifiable and expandable In the performance direction one could try to tune each of the algorithms to increase the analyzing speed If each of the algorithms were optimized the overall performance of the ModelMaker tool could be increased considerably Another possible change to increase the performance would be to fully integrate the Template Constructor component inside ModelMaker Because all of the communication would be done 89 Chapter 6 Conclusions internally this integration could increase the performance substantially This integration is in progress under a different project To improve functionality one could modify ModelMaker to add more capabilities such as a VHDL code checker and a search engine for the Template Constructor results Having the VHDL code checker capability woul
31. the keyword list or simply type the phrase in the Keywords Windows Once this search is performed a result similar to the one shown in Figure 3 13 will be presented in the Source Window lt KEY eop gt each channel can be individually programmed to autoinitialize to its original condition following an end of process eop 0 pin description 0 12 eop eop in out end of process end of process is an active low bidirectional signal information concerning the completion of dma services is available at the bidirectional eop pin this is accomplished by pulling the eop input low with an external eop signal this generates an eop signal which is output through the eop line the reception of eop either internal or external will cause the 8237a to terminate the service reset the request and if autoinitialize is enabled to write the base registers to the current registers of that channel the mask bit and tc bit in the status word will be set for the currently active channel by eop unless the channel is programmed for autoinitialize during memory to memory transfers eop will be output when the tc for channel 1 occurs eop should be tied high with a pull up resistor if it is not used to prevent erroneous end of process inputs 2 dma operation 2 2 active cycle 2 2 2 block transfer mode in block transfer mode the device is activated by dreq to continue making transfers during the service until a tc caused b
32. the new viewer s executable file Notepad exe file name must be used for the viewer File main pl cannon pl roots pl and semrules pl are ASPIN components and are not distributed with ModelMakerlll 99 Appendix A ModelMaker v3 8 User Manual ModelMakerIII Input Files ModelMaker requires two main input files the document specification text file and the basic VHDL code file The modeler should be able to obtain these two input files directly from the device documentation Device documentation is assumed to consist of a behavioral and structural description of the device a high level block diagram and possibly a pin description table and other tables The source document input file A scanner with an Optical Character Recognition OCR program should be able to recover most of the text This ASCII text file forms the source document file that is needed by ModelMaker In order to store this information in a document data structure ModelMaker attempts to recognize the format of this specification document ModelMaker separates the document into five element types section header paragraph sentence table and figure More detailed definitions of these element types are presented in the CSpecification Class section 4 3 1 1 In dividing the document into those five elements ModelMaker uses the following rules e Elements are separated by two line feeds or two carriage returns e Each section
33. the token is a member of the dictionary the function will return a CMeaning pointer that points to the CMeaning object for that token Otherwise a null pointer is returned to the calling object 4 3 2 4 Main Member Functions of CMWDictionary Class There are two main member functions in the CMWDictionary class These member functions are the Load and LookUp functions Load No argument is needed for this public member function This function loads a multi word dictionary input file mdict txt for a CMWDictionary object A flag is returned to the caller object to specify whether it was run successfully The data structure used in this CMWDictionary class is a combination between the data structure of the CDictionary class and the CGrammar class Its division of entries and categories follows the CDictionary class style but fixed length variables are used store the data as in the CGrammar class This implementation was chosen since the number of entries for a multi word dictionary is expected to be a small constant For the current implementation the data structure can only store up to 100 entries Another limitation on the current implementation is the maximum number of characters in each word each category and each concept These numbers are set to 19 8 and 19 respectively All of these limitations can be easily modified by changing the numbers directly in the CMWDictionary source code LookUp pNode The argument needed
34. this noun phrase index list will be sorted and filtered for duplicates The actual search for this index list will be done by the CSpecification constructor right after the CSpecification LoadTxt and the CSpecification LoadVhd member functions are executed As soon as CChart ParseNP is done with its task the CSpecification object calls the CChart ClearChart function This is done to remove and reinitialize the entire chart LoadVhd FileName This private member function requires one argument the name of the VHDL model file to be loaded Similar to LoadTxt function this function is also an essential function in initializing the CSpecification class The main task of CSpecification LoadVhd function is to load the VHDL input file see section 4 2 4 1 into the CSpecification document object Another important task of this function is to parse the VHDL input file for identifiers These identifiers are used in the identifier index list after they are sorted and filtered for duplicates This function returns a flag to indicate whether it was run successfully or not Find Keyword And_Or Id_Np This public member function requires three arguments the keyword the search type and the keyword type The function provides a way for the caller object to search the specification document for a specific keyword This function can perform Boolean AND and OR searches on multiple word keys depending on the second argument
35. this rule The constituents are listed in the fifth and subsequent fields They are placed last to allow a variable number of constituents The current Noun Phrase Extractor limits this variable to the range of one to six constituents The third field is not used by ModelMaker The fourth field is the result field This field stores the new category of the combined constituents 97 Second field the constituents count First field the rule ID Third field the main concept Appendix A ModelMaker v3 8 User Manual B gramtxt si s3 ssi ss2 ss3 spl sp2 ps ae 1 1 1 3 2 4 3 2 4 1 1 1 1 1 1 1 1 1 1 1 t 1 1 1 e pred avs pred avs n pred avs d pred pvs pred pvs d peed pvs dad Gored d g d pred evs n pred evs dad pred evs n pred evs pred evs Le avs adv 2 avs verb ONN AONE ON KR ONHF WD A PNRFN amp a aws mad ads pred ss conj 2 avs mod vinf conj n ss Fifth field the constituents pred a 4 pred a a Fourth field the product winf Figure A 3 Captured amp Annotated Grammar File e Behavior dic ASCH file This file contains the list of the behavioral words used by the ModelMakerllI tool to rank the search results More information on how this file is used by the ModelMakerIII tool can be found in section 4 4 5 of the body of this thesis This dictionary file uses the same exact format as the standard dictionary file The only diffe
36. to terminate the ModelMakerlIII tool The close box X on the right top of the window is disabled ModelMaker Button This ModelMaker button is the about button for the ModelMakerlIl tool By pressing this button the modeler can view the version and copyright information about this ModelMakerII tool Info Box Static Text This info box is used by the ModelMakerlIII tool to present some information about the specification document back to the modeler The information that can be found in this static text box is the total number of noun phrases found in the text and the number of distinct sorted noun phrases that constitute the noun phrase index 109 Appendix A ModelMaker v3 8 User Manual Copyrights Trademarks and Licenses ModelMaker is an intellectual property of Virginia Tech Intel 8237A DMA controller data sheet is copyrighted by Intel Corporation Notepad is copyrighted by Rogier Meurs Intel and Pentium are registered trademarks of Intel Corporation Microsoft MS Microsoft Foundation Class MFC Windows Windows 95 Windows NT Windows Explorer and Visual C are registered trademarks of Microsoft Corporation Synopsys Synopsys Graphical Environment Shell and SGE are registered trademarks of Synopsys Inc SWI Prolog is a trademark of Jan Wielemaker University of Amsterdam 110 Appendix B Behavioral Schematic and Its Initial VHDL Model APPENDIX B BEHAVIORAL SCHEMATIC AND ITS INITIAL VHDL MODEL
37. word to the current phrase that the Noun Phrase Extractor is processing If any match is found this CMWDictionary object will return the category of the phrase back to the Noun Phrase Extractor To use this capability each of the last words in the multi word phrases listed in the multi word dictionary file mdict txt needs to be listed in the standard dictionary file rdict txt with a in its category 47 Chapter 4 ModelMaker Basic Design PrologFinder cpp amp PrologFinder h These files house the CPrologFinder class Similar to the CInitiator class this class is inherited from the CDialog class of the Microsoft Foundation Class An object of this class is used by ModelMaker to prompt the user for the location of the SWI Prolog executable file Once the SWI Prolog executable is found ModelMaker generates a log file in the local directory that contains the location information for that file For the next ModelMaker invocation it will use the information contained in this file instead of creating another object from this CPrologFinder class Specification cpp amp Specification h These files house the main ANSI C classes for the specification document reader CSpecification This CSpecification class is responsible for loading the specification document with a correct document hierarchy extracting the keywords out of the specification document and performing any other operations that deal with the specification document d
38. 2 Relevant Work the sci uses standard nrz format 1 use use gt gindx 1 gt agnt agnt gt 2 gt obj obj gt 3 id sci gt det the format format gt gindx 1 gt name null gt 4 gt attr adj gt 5 id nrz standard standard gt gindx 2 Figure 2 2 Semantic Analyzer Output By analyzing this Semantic Analyzer result a modeler who is familiar with the conceptual graphs could read the conceptual graph instead of the main sentence in building his her model However the process of preparing the sentence manually and running both the Parser and the Semantic Analyzer manually can be time consuming and may be irritating ModelMaker incorporates the Parser and the Semantic Analyzer to extract knowledge that will help the modeler produce models from English descriptions more rapidly VHDLCad and Modeler s Assistant also developed at Virginia Tech VPI amp SU as research projects are two other EDA tools to reduce design time for hardware designers These two applications capture information in the form of process model graphs PMGs These tools generate simple VHDL code from the PMGs These tools serve the same purpose with the ModelMaker tool help designers to reduce the design time These tools also use the same modeling target as the ModelMaker tool the VHDL language However neither of these tools has the natural language analyzer ca
39. 3 6 respectively 23 Chapter 3 Modeling with ModelMaker Figure 3 4 A Structural Schematic Drawn in SGE 24 Chapter 3 Modeling with ModelMaker VHDL Model Created from SGE Symbol behavioralchip sym Apr 18 14 21 25 1998 library use std_logic_1164 all use std_logic_misc all use IEEE std_logic_arith all use std_logic_components entity BEHAVIORALCHIP is Generic DELAY TIME 3 NS Port INI In BIT TN2 Shri BIT IN3 In BIT IN4 In BIT OUT1 Out BIT QUT2 gt Out BIT OUT3 Out BIT Ji end BEHAVIORALCHIP architecture BEHAVIORAL of BEHAVIORALCHIP is begin end BEHAVIORAL configuration CFG_BEHAVIORALCHIP_BEHAVIORAL of BEHAVIORALCHIP is for BEHAVIORAL end for end CFG_BEHAVIORALCHIP_BEHAVIORAL Figure 3 5 The Initial Behavioral VHDL Model File 25 Chapter 3 Modeling with ModelMaker VHDL Model Created from SGE Schematic structural sch Apr 18 13 41 57 1998 library E use std_logic_1164 all use std_logic_misc all use std_logic_arith all use std_logic_components all entity UCTURAL is R INE In BIT IN2 gt In BIT INS 332 Li BIT IN4 In BIT OUTI Out BIT OUT2 Out BIT OUT3 Out BIT end STRUCTURAL architecture SCHEMATIC of STRUCT signal OUT3_DUMMY BIT component CHIPXYZ Generic DELAY TIME 3 NS
40. 9 Chapter 4 ModelMaker Basic Design constructed with the grammar rules If a new constituent can be formed it will be added to the node s list of constituents This will continue until no more new constituents can be built from the existing constituents ParseNP No argument is required for this public member function This function is called just before the CChart ClearChart function is called The purpose of this function is to scan the CChart object for noun phrase constituents For each noun phrase found a noun phrase identifier is created and attached to the noun phrase identifier list To keep the identifiers short the words a all an each that the these and this are deleted from them This function returns all noun phrases in a single CString object In this CString oe 99 object all the white space characters are replaced by underscore signs and a pipe i sign l is placed at the end of each noun phrase ParseSentences This public member function has no argument and no return value and is invoked to scan a chart and to find all of sentence constituents in the existing chart For each sentence constituent a parse tree is constructed converted to a Prolog structure format and written to the file suite pl for analysis by the Template Constructor ClearChart Capture This public member function requires a Boolean argument Capture Since ModelMaker is intended to read a long specification
41. A DMA controller specification document by Intel These searches were done with the sequential mode search Figure 1 3 and the behavioral mode search Figure 1 4 In the current implementation ModelMaker reads all of the characters in lowercase Therefore search results displayed in the Source Window will all be in lowercase letters From these two different searches with the same keyword but different mode one can see how the results are ordered The mode selections primarily deal with how to order the search results Chapter 1 Introduction lt KEY hrq gt 0 pin description 0 15 hrq hrq out hold request this is the hold request to the cpu and is used to request control of the system bus dreq causes 8237a to issue the hrq 2 dma operation 2 2 active cycle when the 8237a is in the idle cycle and a non masked channel requests a dma service the device will output an hrq to the microprocessor and enter the active cycle 2 2 1 single transfer mode if dreq is held active throughout the single transfer hrq will go inactive and release the bus to the system 2 2 4 cascade mode the hrq and hlda signals from the additional 8237a are connected to the dreq and dack signals of a channel of the initial 8237a the 8237a will respond to dreq and dack but all other outputs except hrq will be disabled 2 3 transfer types 2 3 3 priority after completion of a service hrq will go inactive and the 8237a will wait for h
42. L model can be edited inside ModelMaker to add behavioral code and to insert source document fragments as comments TITLE ModelMaker A Tool for Rapid Modeling from Device Descriptions AUTHOR Andreas Indra Gunawan Copyright 1998 by Andreas Indra Gunawan TRADEMARKS The author has attempted throughout this thesis to distinguish proprietary trademarks from descriptive terms by following the capitalization style used by the manufacturer e Concept HDL and Cadence Design System are registered trademarks of Cadence Design System Inc Hewlett Packard and HP are registered trademarks of Hewlett Packard Company Intel and Pentium are registered trademarks of Intel Corporation i Logix and STATEMATE are registered trademarks of i Logix Inc Microsoft MS Microsoft Foundation Class MFC Windows Windows 95 Windows NT Windows Explorer and Visual C are registered trademarks of Microsoft Corporation e Sun and Sun Workstation are registered trademarks of Sun Microsystems Inc e Synopsys Synopsys Graphical Environment Shell and SGE are registered trademarks of Synopsys Inc Other product and company names mentioned herein might be the trademarks of their respective owners Acknowledgements I would like to thank Dr Walling R Cyre my academic and research advisor for all of his assistance and support that he has given to me I would also like to thank Dr James R Armstrong and Dr F Gail Gray for serving on my committees and fo
43. Maker will read the pre analyzed results of the Template Constructor and present it in the Source Window Template Re Analyze Button The Template View button deals with the external program Template Constructor By pressing this Template View button ModelMaker will invoke the Template Constructor and present the output in the Source Window If the Template Constructor is installed this button can be used to generate the semantic analysis of a new source document This process may require a lot of time 108 10 11 12 13 14 Appendix A ModelMaker v3 8 User Manual Show Tree Button ModelMaker also gives the option to the modeler to view the document parse trees This can be done by pressing the Show Tree button This capability is used mainly for debugging purposes By viewing the tree one can see how the sentences are parsed within ModelMaker As the process of constructing these trees also requires a lot of time a copy of a preprocessed tree is stored in the ModelMaker local directory By default pressing this button will cause ModelMaker to read from the local file If the modeler wanted to reconstruct the parse trees he she can select this option during ModelMaker initialization Save VHDL Button Use this button to save the VHDL model code in the Model Window back to the VHDL file Pressing this button will overwrite the old VHDL model file Exit Button This button is provided
44. ModelMaker A Tool for Rapid Modeling from Device Descriptions By Andreas Indra Gunawan gunawan vt edu Thesis submitted to the faculty of the Virginia Polytechnic Institute and State University in partial fulfillment of the requirements for the degree of Master of Science in Electrical Engineering Computer Engineering Option W R Cyre Chairman J R Armstrong F G Gray 20 May 1998 Blacksburg Virginia Keywords Tool ModelMaker VHDL Specification Copyright 1998 Andreas Indra Gunawan ModelMaker A Tool for Rapid Modeling from Device Descriptions Andreas I Gunawan ABSTRACT This thesis describes a tool that facilitates rapid modeling of devices from informal documents The ModelMaker tool facilitates the construction of models by analyzing the source specification document and presenting it based on the modeler s need ModelMaker analyzes and indexes the source document for the noun phrases and identifiers it contains When the modeler specifies the name of a pin or device that the modeler is working on ModelMaker will recover any behavioral or structural information about the particular pin or device ModelMaker can return this information based on the order of how relevant the information is to the model that the modeler is trying to build The modeling language that can be used for this tool is VHDL The initial VHDL model is derived from a block diagram of the source document using a schematic capture tool This VHD
45. More information on this template look form can be found in his 85 Chapter 5 ASPIN Components As soon as the modeler presses the Template Re Analyze button a system call to the SWI Prolog interpreter is executed This system call is done using the ShellExecute command in the Visual C language The protocol file _runsemanal pl is passed to the Prolog interpreter as its argument This protocol file will then initialize all of the knowledge databases and analyze each of the input files This protocol also redirects the outputs of the Template Constructor to the _out txt file Once all of the sentences are analyzed this output file will be closed by the Template Constructor As soon as ModelMaker sees that this output file is ready it reopens the file displays the contents to the Source Window and reformats the display based on the markup tags Master s thesis in progress This Master s thesis is expected to be published by December 1998 86 Chapter 6 Conclusions CHAPTER 6 CONCLUSIONS 6 1 System Capabilities The primary objective of this research was to create an Electronic Design Automation prototype application that combines the functionality of the ASPIN Parser and the ASPIN Semantic Analyzer and adds significant new functionality The purpose of ModelMaker is to help hardware designers improve their design productivity In the current version of ModelMaker an enhanced version of
46. Prolog rules to SWI Prolog Windows 95 based application e Template formatting Some Prolog rules were added to the Template Constructor to define the template formatting for ModelMaker e Test Case e DMA 8237A controller specification document A DMA 8237A controller specification document 15 was used as a test case This DMA 8237A specification document was scanned and converted automatically to a text format using commercial OCR software Two schematic drawings were made from the drawings inside the specification documents These schematics are the structural and the behavioral views of the device Two VHDL files were then generated from these schematics The current ModelMaker ModelMakerIII contains twenty nine C classes written in about 8000 lines of code The ModelMaker is compiled for the Windows 95 platform 15 Chapter 2 Relevant Work CHAPTER 2 RELEVANT WORK Electronic Design Automation EDA is used extensively by hardware designers Many educational institutions and commercial companies provide a variety of EDA tools which reduce design time and increase productivity The electronic format of designs allows designers to edit share or re use their designs easily Moreover synthesis tools help designers to examine test and change their designs Virginia Tech VPI amp SU has designed a number of EDA tools for research purposes Some of these tools are the ASPIN systems 12 VHDLCad 5 and Modeler s Assi
47. TOR SEMANTIC ANALYZER ccccccsssssssscecccccccssssscscececoes 81 iv 5 2 1 ASPIN Semantic Analyzer sicjcctisicianedjeiaaspiadvnsisdatsasiadeariecats ooeednaetads 83 5 2 2 Enhancement and Adaptations sais5 sci cers scctes cctsii acct ete Maced 84 5 2 3 Template Constructor and ModelMaker Communications e cesses 85 CHAPTER 6 CONCLUSIONS si ssssstssscasccedscsnsenssostsncdnesesdedeebensstenssesssnnsous 87 6 1 SYSTEM CAPABILITIES ecsscsousussassserastcoatessavadaudsasoussbeouenaunsssousasesnessvonsenapecnouesaavaseens 87 6 2 SYSTEM LIMITATIONS cc cdsnevehecsucesccevucsouceshauvaceyy Geis casnaxsucanveeeisasuuunasenseenveciocasncnnes 89 6 3 BPUTURE ENHANCEMENTS wisitessssisedescessscntenincesaceuasoscasissCesaaveasesavacehasnbectaasedasiatecncns 89 REFERENCES cass tcecassevanscnssbanssessonsconsenaseanadeusdonssae ge sotasseassouesessnedavedaasuans 91 APPENDIX A MODELMAKER V3 8 USER MANUAL osessessosssoesose 93 APPENDIX B BEHAVIORAL SCHEMATIC AND ITS INITIAL VHD EMO DED DA E T T 111 APPENDIX C STRUCTURAL SCHEMATIC AND ITS INITIAL VHDL MODEL ids ccisccas icccctsveseveusictnatecehbissestechscsceosaspadenesssosestoctesvnessvenessens 113 WTA E E E PEE EET E 120 LIST OF FIGURES FIGURE 1 1 MODELMAKER INTERFACE LAYOUT sssseseserererrrerrrrrrrrererererererererererererererere 3 FIGURE 1 2 CAPTURED MODELMAKER MAIN WINDOW cccccsceeeeeeeeeeeeeeeeeeeeeseeeeeeeeees 5 FIGURE 1 3 SEQUENTIAL MODE RESULT c
48. The VHDL model file The high level block diagram is equally important to the document specification text file By redrawing this diagram into a schematic capture application that is capable of generating a VHDL file the modeler can use this VHDL file as the framework for his her modeling using ModelMaker If schematic capture is not available the user can generate a text file containing an entity declaration with the name of the device and an empty architecture body Then give the file a vhd extension Test Files For test purposes two sets of sample input files are included in the distribution disk The ModelMakerlIII tool allows the user to specify the names of the input files Therefore these files can be replaced at any time by the modeler s own input files These sample files are listed below 101 Appendix A ModelMaker v3 8 User Manual Dma8237a vhd ASCII file This file is the VHDL file generated by the Synopsys Graphical Environment shell for the structural schematic of the 8237A DMA controller Dma8237a txt ASCII file This file is the text file scanned from the specification document of the 8237A DMA controller Other Files These are the list of the other files that may be found in the same directory with ModelMakerlIII These files are not as essential but support demonstration of some ModelMaker features Tree txt ASCII file This file contains preprocessed parse trees of the 8237A DMA control
49. The third argument specifies whether the keyword is a VHDL identifier or a noun phrase A public integer variable of the CSpecification class 67 Chapter 4 ModelMaker Basic Design m_Mode defines the sorting mode The caller object needs to set this variable to sort the results sequentially behaviorally structurally or relevantly All of these search types and modes will be discussed in more details in the Keyword Searches section section 4 4 In the current implementation of ModelMaker this Find function is the main engine behind the Do Search button The return value of this function can be sent directly to the Source Window of ModelMaker Since the occurrences of all keywords are searched on the initialization of ModelMaker CSpecification Find can find the occurrence of any keyword very quickly All the function needs to do is get the results based on an index and to return them This implementation allows ModelMaker to quickly respond to the modeler queries One disadvantage of this method is a slow initialization time for ModelMaker ModelMaker also allows the modeler to enter new search terms by typing in the keyword window or selecting a particular term inside the Model Window A new search is conducted for each new keyword that is not in the list of keyword index This new result will then be added to the database with its keyword added to the keyword index for future searches Based on the Search Type
50. This can be done using the CMWDictionary LookUp public member function For each category either from the standard dictionary or from the multi word dictionary a new constituent is created and attached to the node object for this particular token These constituents are called the basic constituents as their categories are taken from the dictionary Constituents are stored in CPart objects in the chart If the category of a word is not found either in the standard dictionary object or in the multi word dictionary object a category called identifier will be assigned for that word This is done to handle acronyms and identifiers such as device names and device types One disadvantage of this implementation is that every unknown word from the specification document will be recognized as an identifier which acts as a noun To avoid misclassification of new verbs the dictionary files should be updated for new specifications To help this process ModelMaker source files can be recompiled with its DEBUG definition set to 1 When this definition is set ModelMaker will keep a record of unknown words that it encounters This record will then be saved in the text file called unknown txt as soon as this application is terminated The user should review this file with new documents Once all of the basic constituents for a particular node are built the CChart AddNode function searches for any possible new constituents that can be 6
51. a new line and be written in the format of Caption In this format represents one or more digits from zero to nine and Caption can represent any string followed by two carriage returns The part inside the brackets can be omitted or repeated Examples of valid section headers would include 1 Introduction 2 1 Search Algorithm or 2 1 3 Multiple Search Tables and figures must be introduced by writing the word Table or Figure whichever is more appropriate as the first word An example of a valid table element can be seen below This is the last sentence from the last paragraph i Retum X J Table 1 1 Important Files Cari arriage Return One cpp The first file is required in Two cpp The second file Y these locations Three cpp The third file Return u J This is the first sentence of the next paragraph Gum 54 Chapter 4 ModelMaker Basic Design e Any other element group of sentences starting and ending with two carriage returns that is not recognized as a section table or figure will be called a paragraph e A period is used to declare the end of a sentence Each group of words that is inside a paragraph and ends with a period is called a sentence An exception to this rule can occur with the last sentence in the paragraph A paragraph can contain an unlimited number of sentences 4 2 5 Output Files ModelMaker does not create any permanent output files However
52. ad the input file character by character and compile a token to be given back to its user A token can be a word number or punctuation This class distinguishes a period at the end of a sentence from a decimal point in a number Using this CLexer class the string sentence will be separated as sentence and Also the section header 1 1 3 will be kept as one word instead of 1 1 and 3 CLexer Open GetToken Close Figure 4 10 CLexer Class Diagram To define the term token in more detail the rules that govern ModelMaker in demarking tokens are listed below e A Token can be formed from either e Letters or e a Number or e a Punctuation or e a Special punctuation or e a Blank space e Letters can be formed from either e a Letter followed by Letters or e a Letter followed by a Number or e a Period followed by Letters or a Hyphen or a Slash or an Underscore _ or any letter between A to Z or any letter between a to z 63 Chapter 4 ModelMaker Basic Design e A Number can be formed from either a digit followed by a Number or a Number followed by a Letter or a Special punctuation followed by a Number or any single digit number between 0 to 9 e A Punctuation can be formed from either a Colon or an Exclamation mark or a Left curly bracket or a Left parenthesis or a Left square b
53. alog class While ModelMaker is running at least three of these CDialog objects are created The first one is used to query the modeler about the input files This object is created right after the modeler clicks on the ModelMaker icon Once the modeler specifies the input files this CDialog object is destroyed and an object of the CSpecification class is created As the creation process of this CSpecification class is silent there is no visual indicator a second object from the CDialog class is created to inform the modeler that processing will take some time This CDialog object is destroyed as soon as the CSpecification object is created successfully The third CDialog object the main ModelMaker window is then created right after the second CDialog object is deleted The CSpecification object is the main object of the ModelMaker tool This object creates all of the other objects and brings them together to create a useful tool The creation process of this CSpecification object may require some time depending on the 42 Chapter 4 ModelMaker Basic Design size of the input files and the processing power of the machine used This is because the CSpecification object is responsible not only for creating the other necessary objects but also for loading the input files and analyzing them for the needed information A CLexer object is created by this CSpecification object to divide the input text file into tokens words punctuation and numb
54. an or search This Search Type drop down list is only useful for the multi word keywords Search Mode Drop Down List 107 Appendix A ModelMaker v3 8 User Manual This drop down list allows the modeler to decide whether he she wants to sort the search results sequentially relevantly behaviorally or structurally Do Search Button The modeler needs to press the Do Search button every time he she wants to view the information on a new keyword phrase When this button is pressed ModelMaker will show the returned information on the selected search key in the Source Window To perform the search the modeler can simply pick any of the keywords in the keywords list set the search type and the search mode and press the Do Search button Any information that is found by the search engine will be presented to the modeler in the Source Window The modeler can also select any of the words from the Model Window by highlighting them and press the Do Search button When any of the words in the Model Window is highlighted the search engine of the ModelMaker tool will use the words inside the selection as its keywords The ModelMaker tool also allows the user to type his her own words inside the Keywords Window This capability allows the modeler to perform the search on any keywords Template View Button The Template View button deals with the external Template Constructor By pressing this button Model
55. anned Document As described in section 4 2 4 2 the result from the OCR software may require some format adjustment to make it fully readable by the ModelMaker tool Depending on the quality of the document the result may also need to be checked by spell checking software The adjustment and checking need to be done manually before this source document file is given to ModelMaker ModelMaker attempts to automatically identify the hierarchical structure of a given source document Sending a source document with a format that does not follow the rules defined in section 4 2 4 2 to ModelMaker may cause some misidentification of the source document hierarchy Consequently ModelMaker may not perform at its best 22 Chapter 3 Modeling with ModelMaker The current version of ModelMaker v 3 08 contains lexical knowledge of about 4000 words in its standard dictionary This dictionary was constructed manually from words accumulated throughout the ASPIN project and the ModelMaker project Although this dictionary is adequate for research purposes it is far from being complete Consequently it is suggested that each of the new source documents be scanned for unknown words During the ModelMaker project a utility program was created to automatically extract all unknown words out of source document files The task of adding these words and their respective lexical categories however is still manual The second manual task for the modeler is to ge
56. arated into two different parts They are the graphical user interface GUTI part and the engine part The GUI part deals with the user interaction such as capturing user s inputs and displaying results back to the user This first part is written with the Microsoft Visual C 18 The Microsoft Foundation Class MFC library version 4 21 is used to implement this interface The second part the engine contains the actual code that is needed to perform the document analysis and searches This includes extracting noun phrases from the specification text extracting VHDL identifiers from the VHDL file and searching the specification text for keywords This engine is written under the ANSI C standard for portability This part can be re compiled for a different platform without changing much of its code The GUI will need a major re build to port ModelMaker to a different platform Some minor changes in the ModelMaker engine would be required since interfacing with the GUI required some Microsoft specific code The final standard is not available for public access as of April 1998 The final draft of this standard however can be obtained from the American National Standards Institute ANSI Standards Secretariat ITIC 1250 Eye Street NW Suite 200 Washington DC 20005 An Internet version of the final draft can be found in this URL http www maths warwick ac uk cpp pub wp html cd2 The URL of this document can be changed at any time by th
57. articular feature of ModelMaker refer to the body of this thesis About ModelMakerlIlI The ModelMakerIlI is the third generation of the ModelMaker tool It is a Windows 95 based application that facilitates device modeling by analyzing the specification document of the device and providing search facilities ModelMakerIII System Requirements The minimum system requirements for the ModelMakerlIII tool is a 486 based or better machine with a minimum of 10 megabyte free space on the hard drive and using the Windows 95 or higher operating system There is no minimum RAM Random Access Memory Naturally the more RAM the system has the faster the ModelMaker tool performs its processing ModelMakerlII can also be invoked under the Windows NT operating system However it has not been thoroughly tested in the Windows NT environment 93 Appendix A ModelMaker v3 8 User Manual To run the Template Constructor component of the ModelMakerHII tool the SWI Prolog interpreter is required The executable for this interpreter needs to be accessible from the computer where the ModelMakerlHII tool is installed This SWI Prolog interpreter is free for non commercial users More information on the SWI Prolog interpreter can be found in the SWI Prolog Information Sheet http swi psy uva nl usr jan S WI Prolog html Application Files Listed below are the needed application files of the ModelMakerIII tool All of these files are essentia
58. asses were rebuilt and modified to make them more compatible to the classes of ModelMaker The initial code which could handle only a relatively small input file was modified to allow very large input files Some other additions and enhancements were also done to the Parser classes during this integration process 5 1 1 ASPIN Parser The ASPIN Parser is a bottom up knowledge based sentence parser that constructs parse trees for input sentences The knowledge bases used by the Parser are a dictionary database and a grammar The following paragraphs contain a simplified discussion of the Parser The dictionary contains lexical categories for about 1200 vocabulary words The second database the grammar database contains about 120 context free grammar rules using about 48 non terminals A non terminal is a grammatical constituent that can be separated into two or more smaller constituents or lexical categories The Lexical categories are the terminals The terminals are adjective adverb conjunction determiner identifier modal negation noun time order pre determiner preposition pronoun relative conjunction subordinating conjunction past participle verb infinitive verb present participle infinitive be be to of and have Examples of non terminals are noun phrase sentence verb sequence subordinating clause and predicate Parsing is 78 Chapter 5 ASPIN Components performed by constructing a chart for each sentence Th
59. ata structure The other important class in these files is the CElement class The CSpecification class uses a CElement class for each of its structural elements The possible CSpecification elements are section paragraph and sentence More details on these elements and the CSpecification class are presented in the CSpecification Class section 4 3 1 1 StatusBox cpp amp StatusBox h These files contain the CStatusBox class This class is inherited from the CDialog class of the Microsoft Foundation Class An object of this class is invoked by ModelMaker only to give feedback to the user about its processing status StdAfx cpp amp StdAfx h These files are Microsoft Visual C specific files They are used to connect the application to the standard Microsoft Foundation Class library Resource h This file is a Microsoft Visual C specific file It is used to define all of the include lines for the application and to specify all of the other symbol definitions needed for the application 48 Chapter 4 ModelMaker Basic Design 4 2 2 Prolog Files The initial version of the ASPIN Semantic Analyzer SemAnal used in the ModelMaker application was written in Quintus Prolog Because converting the Semantic Analyzer from Quintus Prolog to the ANSI C environment was far beyond the objective of this research it was decided to modify this Semantic Analyzer to run on a PC with SWI Prolog This section presents some description of
60. cceessssscececececeessnsecececcesceesesssaeceeeeceseeensnteaeees 8 FIGURE 1 4 BEHAVIORAL MODE RESULT ccsesessececececeseesensecececceseeesensaaecececeeseeesenteaeees 9 FIGURE 2 1 SEMANTIC ANALYZER INPUT ccccsessscesccececsessnseceeececsceeesessaeceseceeseeesenteaeees 17 FIGURE 2 2 SEMANTIC ANALYZER OUTPUT cssssssscesececeesessnsececescesceessnseaeceseeeessnensaaees 18 FIGURE 3 1 SCANNED INTEL DMA 8237A CONTROLLER SPECIFICATION DOCUMENT CLAP AGE ETA EAA E ETE EEE E AE E ETEN 21 FIGURE 3 2 PARTIAL OCR RESULT OF THE SCANNED DOCUMENT 000c0csssessseeseeeeseeeees 22 FIGURE 3 3 A BEHAVIORAL SCHEMATIC DRAWN IN SGE 1 cccceecssseeseesessesesssssessseseseeees 24 FIGURE 3 4 A STRUCTURAL SCHEMATIC DRAWN IN SGE uuu ccceccceeeeeeessssesssssssessseseseeees 24 FIGURE 3 5 THE INITIAL BEHAVIORAL VHDL MODEL FILE c0c0ccsesessessseesssesesreees 25 FIGURE 3 6 THE INITIAL STRUCTURAL VHDL MODEL FILE 0ccccecessssssesssssssresereeees 26 FIGURE 3 7 THE DIALOG BOX USED TO SPECIFY THE LOCATION OF THE FILES 27 FIGURE 3 8 PART OF THE CDICTIONARY STRUCTURE cccsessssscecececsesessnseceeeeeeseeessnteaeees 29 FIGURE 3 9 PART OF THE CMWDICTIONARY STRUCTURE cccsccsssssscssscesssssssssssetsseeees 30 FIGURE 3 10 SAMPLE OF THE CCHART STRUCTURE ccccssessssecesececeesessnseceeecceseeessnseaeees 33 FIGURE 3 11 MODELMAKER SEARCH RESULT ON THE PHRASE DATA BUS sceeeeeee
61. checker was also added in each class This checker is only active during the debug mode Multi word recognition The current algorithm used in the Noun Phrase Extractor has the capability to recognize multi word tokens in a sentence This ability allows the Noun Phrase Extractor to understand multi word tokens such as consist of greater than or even greater than or equal to More discussion about this multi word recognition is presented in section 5 1 2 Rule skipping capability The ability to skip one or more words in using the grammar rules was implemented in the Noun Phrase Extractor This ability allows the Noun Phrase Extractor to recognize more sentences However due to slow performance this capability is not utilized in the current ModelMaker Noun Phrase Extractor algorithm A new function was developed to perform the noun phrase extraction from a chart built by the ASPIN Parser This function scans through all of the constituents of the existing chart and extracts all of the noun phrase constituents out of it This is the function that is used by the ModelMaker tool to extract all of the noun phrases out of the specification document object 14 Chapter 1 Introduction e Template Constructor e Quintus to SWI Prolog rules conversion Since the original Template Constructor code the ASPIN Semantic Analyzer was written in Quintus Prolog Unix based application some adjustments were needed to convert the
62. create its noun phrase index The current standard dictionary contains over 4000 words and the multi word dictionary contains 21 entries 5 2 Template Constructor Semantic Analyzer The Template Constructor is an adaptation of the ASPIN Semantic Analyzer The primary function of the Semantic Analyzer is to create conceptual graphs representing the meaning of the sentences in the input file The Template Constructor is an external component used by the ModelMaker to perform semantic analysis on sentences and fill behavioral templates from them The Template Constructor is implemented as a set of Prolog rules Whenever the Template Constructor is needed by the ModelMaker tool a system call to the Prolog interpreter is executed to interpret the Template Constructor rules ModelMaker writes to an input file in prolog rule style The Prolog interpreter will then load this input file interpret the rules and write the results to an output file This is not the most efficient way to do the communication but it is sufficient at this point because the primary objective of this research is to study the feasibility of the ModelMaker tool as a system An example of the Template Constructor output is shown in Figure 5 2 This result is based on the parse tree shown in Figure 5 1 The parse tree is taken from the chart of the Noun Phrase Extractor Prolog is an interpreted language refer to Programming in Prolog 7 Prolog source code con
63. ction in more detail That section also discusses the structure sections paragraphs and sentences of the source document object The CChart AddNode function is called for each token in the source document file When this function is called from the aChart object a new node is created inside the aChart object This CChart AddNode function is also responsible for creating new constituents that combine the new node with the existing constituents of earlier nodes This is done internally by the CChart class using the dictionary knowledge and the grammar knowledge built earlier for the CChart class For more information about the CChart AddNode refer to section 4 3 2 2 The CChart ClearChart public member function is needed to reset the aChart object Each time a new sentence is detected by the m_pSpec object it clears the aChart object Once this CChart ClearChart public member function is called the aChart object is reinitialized to an empty chart All of the existing node objects and constituent objects will be deleted and all of the memory used will be reclaimed This clean up process is needed to limit the memory space consumed by the CChart object ModelMaker can also save the aChart in an ASCII file The check box shown in Figure 3 7 with the caption re captured the sentence parse tree is provided to activate this capability This capability is set to active only when the modeler clicks on this check box When t
64. ctscsccooccascecossadaccocsdecssosssdccooscducceosendacoscsdecssosssdaceosctacesosensacoos 44 4 2 1 C Header and Source Files 0 na a a a a a a ai 44 422 Prolog Piles nenian E osetia nage 49 7 EE Database a IAAT EE AE E EE A EEE EE 50 ADA ipt FUES inntin e E e A a E eine 53 42 3 Outp t Eesaun ne E E T A R 55 4 3 OBJECT ORIENTED DESCRIPTION ssseseseoeeesssosooeoecesssooooeoceeessosoooocoeeesosooooceeeesssoooo 55 43l Main Class Desenptions seseo n see ial ee eileen aes 56 4 3 2 Main Member Functions 0 cccccceccccccccccccsssssscscccccessessssescceccessuauesscecseseeees 64 4 4 KEYWORD SEARCHEG cccccsscssessossccccsscessoscecceccessecscnscccoescessescenscscessccscoscescnscesoes 74 4 4 1 AND Search Mode ccccccsssccecccsecceccnneccecconedeccennceecennedeecensaeecennedeccenss 75 AAD OR Search MoOde s ccc r a AE a Oa aiaa NETE 75 44 3 Sequential Sortere esien ei nn NE eiaeia seai Eiei 75 4 44 Releyance Soit ennn a EA TE TAT E A U EA E RE 76 4A 5 Behavioral Sort EE E EEEE E ET EIE 76 AAG Strict ral Sort ta a a aa e oie dod a E AA T 76 CHAPTER 5 ASPIN COMPONENTS ccccsscsscccccssccsccccescccccccceccees 78 5 1 NOUN PHRASE EXTRACTOR PARSER eessessseeeccceccecccccececccceeccecceccececceecceccececececee 78 DA ASPIN Parser nonae te Mis tus nce lets EAE N Bac ole DN ved Rte Sie es 78 5 1 2 Enhancements and Adaptations 0 cesccesscecssececeecceceeececeeeeeceseeecseeeeeeaeees 79 5 2 TEMPLATE CONSTRUC
65. d allow the modeler to check his her model code while modeling Moreover searchable Template Constructor results would help when the modeler wished to focus on a particular part or section of the device 90 References References 1 Alfred V Aho and Jeffrey D Ullman The Theory of Parsing Translation and Compiling Prentice Hall Inc New Jersey 1973 2 James R Armstrong and F Gail Gray Structured Logic Design with VHDL Prentice Hall PTR New Jersey 1993 3 Jim Armstrong Process Level Modeling with VHDL Proc of 1998 International Verilog HDL Conference and VHDL International Users Forum Santa Clara California Mar 16 19 1998 4 Grady Booch Object Oriented Analysis and Design with Applications Second Edition Addison Wesley Publishing Company California 1994 5 David G Burnette A Graphical Representation for VHDL Models Master s Thesis Virginia Polytechnic Institute and State University Blacksburg VA 1988 6 Cadence Design System Inc Cadence Softwares Cadence Design System Inc http www cadence com software software html 7 W F Clocksin and C S Mellish Programming in Prolog Fourth Edition Springer Verlag Heidelberg 1994 8 Walling R Cyre Mapping Design Knowledge from Multiple Representations Proc Of 1991 IEEE International Conference on Computer Design ICCD 91 Boston Massachusetts October 1991 9 Walling R Cyre A Requi
66. dictionary file mdict txt and the grammar file gram txt These files are stored as ASCII space delimited files for portability reasons 4 2 3 1 Standard Dictionary File The standard dictionary file rdict txt contains a list of words along with each word s category An annotated screen capture from a small part of this dictionary can be seen in Figure 4 1 terminations noun _ The word s categories terminology nil X terms noun _ testing ving a tests verb _ text An to denote a The entry word multi word entry theorem nil _ theoretic nil theory nil Figure 4 1 Captured and Annotated Standard Dictionary File 50 Chapter 4 ModelMaker Basic Design Each line in this file represents one entry in the dictionary Each entry consists of a vocabulary word and a list of one or more lexical categories parts of speech Each oe category consists of a pipe followed by a category indicator followed by an GON 99 underscore A grave accent mark marks the end of an entry An asterisk in the category field denotes that one or more multi word phrases that ended with the entry word are listed in the multi word dictionary Once the Noun Phrase Extractor finds this asterisk it invokes the multi word dictionary object to check for another possible category For example in Figure 4 1 the asterisk sign for the entry than refers to the multi words farther than faster than fewe
67. ds combo box is initialized with the keywords list Moreover the search type is set for Boolean And search and the sort mode is set for 34 Chapter 3 Modeling with ModelMaker Sequential sort by default Once this is done the ModelMaker tool is ready to be used by the modeler At this time the window shown in Figure 1 2 is presented to the modeler From this point there are many ways that the modeler can perform his her modeling process If a structural VHDL file is used the modeler can start the modeling process by focusing on one component at a time For each of these components he she could perform keyword searches that bring back information about that particular component If a behavioral VHDL file is derived instead the modeler can start by searching on the input and output identifiers of the chip Each time a process or a function is suggested by the sentences returned the modeler can model the process or function inside the behavioral model Whether the modeler is using the structural VHDL file or the behavioral VHDL file the modeler can always perform searches on any keyword The keyword can be chosen from the keyword list selected from the Model Window or typed in the Keywords Window Because the search results are displayed at the same level with the VHDL file the modeler can do his her modeling in the model window while referring to the information displayed in the source window Some methods used in the m
68. e 36 FIGURE 3 12 VHDL PROCESS CREATED BASED ON SEARCH PHRASE DATA BUS 37 FIGURE 3 13 MODELMAKER SEARCH RESULT ON THE WORD EOP eere 38 FIGURE 3 14 MODELMAKER SEARCH RESULT ON THE WORD AUTOINITIALIZE 39 FIGURE 3 15 VHDL PROCESS CREATED BASED ON SEARCH PHRASE AUTOINITIALIZE 40 FIGURE 3 16 SAMPLE RESULT FROM THE TEMPLATE CONSTRUCTOR cccsssssesssesesseees 41 FIGURE 4 1 CAPTURED AND ANNOTATED STANDARD DICTIONARY FILE 0000seeeeeeee 50 FIGURE 4 2 CAPTURED MULTI WORD DICTIONARY FILE ccccccecssssssssssssscsssssssssseeees 52 FIGURE 4 3 CAPTURED amp ANNOTATED GRAMMAR FILE ccccccscssscececssscecscssssssssessseeees 53 FIGURE 4 4 CSPECIFICATION CLASS DIAGRAM cccccccccccessssssscecesececeesessseceseceeeeeesenteaeees 57 FIGURE 4 5 CCHART CLASS DIAGRAM ccccccccssssssecececececeensnseaecesecseeesessaaeceeeseeseeesenaaeees 58 FIGURE 4 6 CDICTIONARY CLASS DIAGRAM cccesssssccseceeessensececeseccceesessaeceeeseeseeesenteaeees 60 FIGURE 4 7 CMWDICTIONARY CLASS DIAGRAM uu ccccccesesssssececesececeesensnseceeeceesesensnteaeees 61 FIGURE 4 8 CGRAMMAR CLASS DIAGRAM csssesssceeesececeesensececeseceeeesensaaeceseseeseeeesnteaeees 61 FIGURE 4 9 CELEMENT CLASS DIAGRAM cccessessececeseceeeesensececececcesesessaeceeeseeseneesnteaeees 62 FIGURE 4 10 CLEXER CLASS DIAGRAM ccccceesessececececeeeessnseaeceseceeeesensaaecesececseseeenteaeee
69. e chart includes a node for each word of the sentence Each node has at least one constituent This constituent is the category of the word The category for the first constituent is obtained from the dictionary Once this constituent is added successfully the grammar rules are applied All rules are considered to determine if the most recent constituent can be combined with any constituents of previously analyzed nodes to form new constituents This process is repeated until there are no more rules that can be used to combine any of the existing constituents The primary classes of the Parser are CChart CDictionary and CGrammar The Parser employs a CParser class to use these classes In ModelMaker however this task is performed by the CSpecification class For more detailed information on how the initial version of the ASPIN Parser works consult 12 5 1 2 Enhancements and Adaptations Many enhancements and adaptations were needed in creating the Noun Phrase Extractor from the ASPIN Parser Some of these enhancements include a memory leak checker a multi word recognition ability a rule skipping capability and output constituent selector Some of the adaptations add robustness and error detection A new dictionary class was implemented with an enhanced search algorithm The memory leak checker was implemented to avoid any memory leaks Memory leaks which can crash an application or even an operating system are usually caused by po
70. e improvements The initial layout of the ModelMaker interface was a simple dialog based application with no functional control attached to it The release version of ModelMaker uses this dialog based application as its template The layout was modified and more control buttons were added Most of the contributions made in this version are on the actual engines behind the controls 11 Chapter 1 Introduction e ModelMaker II Integrated Noun Phrase Extractor This version of ModelMaker incorporates the modified ASPIN Parser now called the Noun Phrase Extractor Using this ability ModelMaker can extract most of the noun phrases that occur in a specification document Similarly to the identifiers these noun phrases are then used as search indexes Search capabilities for noun phrases In this version the ability to search the specification document data structure based on the noun phrase list was implemented From the noun phrases found by the Noun Phrase Extractor ModelMaker searches its specification document for any occurrence of these noun phrases during its initialization Similarly to the identifier case these search results are also placed in the search index e ModelMaker III Sentence Extractor The ability to extract sentence parse trees from the ModelMaker document specification data structure was implemented in this version This ability is needed by ModelMaker to be able to generate input files for the Template C
71. e web master 44 Chapter 4 ModelMaker Basic Design The necessary C header and source files to rebuild ModelMaker are listed in Table 4 1 and are explained in alphabetical order below Table 4 1 C Header and Source Files Program Files Header Files Chart cpp Chart h Dictionary cpp Dictionary h Grammar cpp Grammar h Initiator cpp Initiator h Lexer cpp Lexer h MMREditWin cpp MMREditWin h ModelMakerllI cpp ModelMakerIlI h ModelMakerlIIDlg cpp ModelMakerlIIDlg h MultiWord cpp MultiWord h PrologFinder cpp PrologFinder h Specification cpp Specification h StatusBox cpp StatusBox h StdAfx cpp StdAfx h Resource h e Microsoft specific code platform dependent code e All others ANSI C code Chart cpp amp Chart h These files contain classes needed to build the parse tree of each parsed sentence Some of the important classes included in these files are CChart CPart and CNode The CChart class is the main class in this file It uses the other classes to store its information The information from the CChart class is used by ModelMaker to extract the noun phrase segment of the sentence that will be used to build the noun phrase keyword list The classes included in these files are the enhanced version of the classes used in the ASPIN sentence parser the Parser In ModelMaker they are called the Noun Phrase Extractor Dictionary c
72. ectional three state signals connected to the system data bus during dma cycles the most significant 8 bits of the address are output onto the data bus to be strobed into an external latch by adstb in memory to memory operations data from the memory comes into the 8237a on the data bus during the read from memory transfer in the write to memory transfer the data bus outputs place the data into the new memory location 2 dma operation idle cycle commands do not make use of the data bus active cycle transfer types 3 5 address generation register description 10 software command hey do not depend on any specific bit pattern on the data bus 5 application information note 1 when the processor replies with a hlda signal the 8237a takes control of the address bus the data bus and the control bus the address for the first transfer operation comes out in two bytes the least significant 8 bits on the eight address outputs and the most significant 8 bits on the data bus the contents of the data bus are then latched into an 8 bit latch to complete the full 16 bits of the address bus Figure 3 11 ModelMaker Search Result on the Phrase data bus 36 Chapter 3 Modeling with ModelMaker The first sentence shows the data that needs to be sent to the data bus during DMA cycles Similarly the second sentence shows which signals is in the data bus during the read from memory transfer The third sentence the most important
73. ed as its argument This function is a private function since the only place it is used is in the CDictionary class constructor A flag is returned by this function to specify whether it was run successfully This function is an enhanced version of the CDictionary Load member function used in the ASPIN Parser The main enhancements for ModelMaker include better error detection and error handling unlimited dictionary entries and variable length words A new algorithm avoids various types of error in dictionary formatting Member Word An argument Word is needed for this public member function This function provides a quick way for the user to check whether a token is listed in the dictionary The argument specifies the token to be searched If the token is a member of the dictionary the function will return a CEntry pointer that points to the CEntry object for that token Otherwise a null pointer is returned to the caller object A 2 character index table is used to accelerate the search process GetMeaning Word This GetMeaning public member function is used to obtain the category or part of speech for a particular token stored in the dictionary object Similar to the Member function an argument is needed for this function This argument 71 Chapter 4 ModelMaker Basic Design specifies the token to be searched Internally this function uses the Member function to obtain the CEntry object of the token in question If
74. em memory Memory to memory transfer capability is also provided The 8237A offers a wide variety of programmable control features to enhance data throughput and system optimization and to allow dynamic reconfiguration under program control 4 The 8237A is designed to be used in conjunction with an external 8 Dit address latch It contains tour indepen dent channels and may be expanded to any number of channels by cascading additional controller chips The three basic transter modes allow programmability of the types of DMA service by the user Each channel can be individually programmed to Autoinitialize to its original condition following an End of Process EOP Each channel has a full 64K address and word count capability The 82734 4 and 8237A 5 are 4 MHz and 5 MHz versions of the standard 3 MHz 8237A respectively gtteegeees geiiiaifuda EEEF f a JETADEENcaasdetaMies Ri EPSE eth iit Figure 3 1 Scanned Intel DMA 8237A Controller Specification Document 1 Page 21 Chapter 3 Modeling with ModelMaker 8237A HIGH PERFORMANCE PROGRAMMABLE DMA CONTROLLER 8237A 8237A 4 8237A 5 Features Enable Disable Control of Individual DMA Requests Four Independent DMA Channels Independent Autoinitialization of All Channels Memory to Memory Transfers Memory Block Initialization Address Increment or Decrement High Performance Transfers up to 1 6M Bytes Second with 5 MHz 8237A 5 Directly Ex
75. empty CChart object The task of the constructor is only to prepare this CChart object Nodes and constituents are added and deleted later by the m_pSpec object 30 Chapter 3 Modeling with ModelMaker The Noun Phrase Extractor s Parser which is used by the m_pSpec object is an extension of the ASPIN bottom up parallel chart parser 23 The parser algorithm is called a parallel algorithm because it considers all possible alternatives at once This Parser uses the standard and multi word dictionary objects to create a chart from the document tokens words numbers punctuation The Parser searches in the dictionaries for the lexical category of each token and builds new constituents as combination of the particular token and the existing constituents based on the grammar rules The grammar is an array of rules which governs the combining process for constructing new constituents After aChart is constructed two more dictionary objects of the CDictionary class are created for the m_pSpec object These objects are the aBehaviorDic and the aStructuralDic These dictionaries of behavioral and structural words are used to order the search results The entries for these objects are loaded from the behavior dic file and structure dic files respectively While performing its tasks the m_pSpec object checks for any unsuccessful functions If any functions returns an error code the m_pSpec object will display an error me
76. er s Guide Microsoft Press Washington 1997 19 Balraj Singh A Parametrized CAD Tool for VHDL Model Development with X Windows Master s Thesis Virginia Polytechnic Institute and State University Blacksburg VA 1990 20 John F Sowa Conceptual Structures Information Processing in Mind and Machine Addison Wesley Publishing Company Massachusetts 1984 21 Synopsys Inc Synopsys Online Documentation v1997 08 2 1 0 Synopsys Inc 1997 22 Jan Wielemaker SWI Prolog Information Sheet University of Amsterdam http swi psy uva nl usr jan S WI Prolog html 23 Terry Winograd Language as a Cognitive Process Vol 1 Syntax Addison Wesley Publishing Company Massachusetts 1982 24 Patrick Henry Winston On to C Addison Wesley Publishing Company Massachusetts 1994 25 Michael J Young Mastering Microsoft Visual C 4 Sybex Inc California 1996 92 Appendix A ModelMaker v3 8 User Manual APPENDIX A MODELMAKER V3 8 USER MANUAL Introduction This User Manual is designed to help device modelers in using the ModelMakerlll tool This discussion gives a short description on what the ModelMakerlII tool is and how the modeler can utilize it Following this discussion the system requirements of ModelMakerllI are given A listing of the needed files to run ModelMakerlIII follows these two topics Moreover an installation guide is also presented here For more detailed information about any p
77. ers Three CDictionary objects one CMWDictionary object and one CGrammar object are created as knowledge databases for the CSpecification object Moreover a CChart object is created to handle the noun phrase extracting task Finally the source specification document is stored in a hierarchy of CElement objects representing sentences paragraphs and sections of the source document The Semantic Analyzer component is not implemented as an object This is because the Semantic Analyzer component is implemented in Prolog When the Semantic Analyzer is needed by the ModelMaker tool a system call to the Prolog interpreter is executed The communication between the CSpecification object and the Semantic Analyzer is done by means of input and output files In this chapter a detailed object oriented description of ModelMaker is presented To give more insight about the system the main ModelMaker files will be listed and explained first Following the explanation of the files the main classes and member functions and the intercommunications among them will be discussed in a more object oriented way Because the Noun Phrase Extractor and the Template Constructor were based on existing research work prior to this research each of these components will be explained with more detail in a separate chapter 43 Chapter 4 ModelMaker Basic Design 4 2 File Descriptions 4 2 1 C Header and Source Files ModelMaker source files can be sep
78. ersion is able to parse a ModelMaker document object to extract all of the noun phrases that occur in the document Finally ModelMakerlII is the third version of ModelMaker This version involved a total remodeling of its graphical user interface This version is designed to have more options and controls that comply with the Windows standards Another major addition in ModelMakerllI is the Template Constructor component This version is able to communicate with the template constructor which is written in Prolog and present the results in the ModelMaker Source Window in a template format This thesis is an extension of the paper presented at the VC VIUF 98 10 1 2 Overview ModelMaker is an executable application that is targeted for the Windows 95 platform Its main working window is a dialog box that contains two edit windows and some controls The two edit windows are the Source Window and the Model Window The layout of this main working window is displayed in Figure 1 1 The Source Window placed in the top portion of the main working window is where the analyzed specification document elements will be displayed by ModelMaker This read only window allows the user to select and copy some or all of the text and paste it into the Model Window The Model Window placed in the bottom portion of the main working window is initialized with the basic VHDL model This window is editable and the text can be saved back to the original model t
79. es by analyzing the specification document and presenting it based on the modeler s need When the modeler specifies the name of a pin or device that the modeler is working on ModelMaker will recover any behavioral or structural information about the particular pin or device behavioral and structural descriptions are defined in the Structured Logic Design with VHDL 2 ModelMaker can return this information based on the order of how relevant the information is to the model that the modeler is trying to build The only input needed by ModelMaker is a text version of the specification document and a basic model file for the device In the current implementation VHDL is the only modeling language that can be read by ModelMaker The basic VHDL file can be generated automatically using a schematic capture tool such as Synopsys Graphical Environment SGE ModelMaker is a prototype application It should not be used as a final application During the duration of this particular project three main versions of ModelMaker have been released for prototype demonstration The first version of Chapter 1 Introduction ModelMaker is capable of analyzing a source specification document and storing it in an hierarchical document object This version also loads the VHDL file and parses the VHDL identifiers However it only provides search capability on identifiers The second version of ModelMaker includes the Noun Phrase Extractor component This v
80. esigned to operate in tyro major cycles These are called the Idle and Active cycles Figure 4 11 Sample Specification Document Text partial Type Document Root Type Section 1 functional description Type Paragraph Type Sentence the 2237 a block internal registers Type Sentence the data interconnection also shown Type Sentence Type Paragraph Type Table figure 3 237a internal registers name size number Lesend base address registers 16 bits 4 base word court register 16 bits 4 Constituent Type Section 2 dma operation J Next Type Paragraph Type Sentence the 8237a is major cycles Figure 4 12 Document Elements Built by CSpecification LoadTxt The CChart AddNode function is called every time a new token is received from the CLexer object This action is needed to let the CChart object create the parse 66 Chapter 4 ModelMaker Basic Design tree for the current sentence Refer to section 4 3 2 2 for more description on the CChart class member functions The CChart ParseNP function is needed to extract all of the noun phrases that exist in the current parse tree This function is called every time the CSpecification object finds a new section header or a new paragraph The return value of the CChart ParseNP function will then be attached to the noun phrase index list in the CSpecification object Once the end of the input file is reached
81. ext file The primary objective is to allow the modeler to view source document information and the Model Window at the same time Chapter 1 Introduction Search Controls Source Window Keywords Window Other Model Window Controls Figure 1 1 ModelMaker Interface Layout ModelMaker requires two main input files the document specification text file and the basic VHDL code file The modeler should be able to obtain these two input files directly from the device documentation Device documentation is assumed to consist of a behavioral and structural description of the device a high level block diagram and possibly a pin description table and other tables A scanner with an Optical Character Recognition OCR program should be able to recover most of the text This ASCII text file forms the source document file that is needed by ModelMaker Some formatting issues on this text file are discussed in section 4 2 4 2 The high level block diagram is equally important to the document specification text file By redrawing this diagram into a schematic capture application that is capable of generating a VHDL file the modeler can use this VHDL file as the framework for his her modeling using ModelMaker While ModelMaker loads this VHDL file it parses all of the I O pin identifiers the component names and other identifiers Similarly while ModelMaker loads the document specification text file it parses all of the noun phrases This inf
82. form other than Windows 95 all that the programmer needs to do is to redesign the interface in the new platform Some adjustments in the platform dependent parts of the code had to be added to the code during this process However that should not cause too many problems as the main engine of ModelMaker is written in ANSI C code This ANSI C code can be ported to a Unix machine or to any other machines that has the ANSI C compiler Pre analyzed results and pre captured parse trees As the template constructing process may take a very long period this version of ModelMaker allows the modeler to save and re access templates This implementation can save a lot of time for the modeler For each new specification document text the modeler is only required to run the Template Constructor once By saving the result of the Template Constructor locally ModelMaker can re use this result for its next run The same applies to the parse trees structure ModelMaker can use the pre computed parse trees instead of regenerating the trees each time it is initialized 13 Chapter 1 Introduction Noun Phrase Extractor Rebuilt classes and memory leak check ability Some of the classes for the Noun Phrase Extractor were rebuilt This was done to ease the communication between ModelMaker and the Noun Phrase Extractor Another reason was to improve the error handling mechanism and to add protection to the data structure of the classes A memory leak
83. his capability is active the entire contents of the aChart object are captured and appended into a string object called m_Tree just before the aChart object is cleared This object is captured using a chart format as shown in Figure 3 10 This capability is not active by default because the capturing process of the parse tree 32 Chapter 3 Modeling with ModelMaker requires a significant amount of time In processing the 8237A DMA controller specification document having 5258 words ModelMaker required approximately four hours to run on a 300 MHz Pentium processor When the check box is not clicked however ModelMaker allows the modeler to view a preprocessed parse tree To use this ability the modeler needs to save the previously analyzed parse tree as the tree txt file located in the same directory with the ModelMaker tool executable Node 7 may Part 35 le 0 mod _ From Node 10 Part Rule 0 bei _ From Node 11 panded Part Rule 0 ven _ From 11 To 1 ed Part Rule 44 pvs _ From 7 To 11 Parts 35 36 37 be expanded Part Rule 165 nvs _ From 11 To 11 Parts 37 ed Part Rule 20 pred _ From 7 To 11 Parts 40 be expanded Part Rule 140 cl _ From 11 To 11 Parts 41 ed Part Rule 7 sp _ From 1 To 11 Parts 3 23 25 42 it contains four independent channels and may be expanded Part Rule 70 n _ From 11 To 11 Parts 43 ed Part le 137 rest _
84. igure 5 2 is based on general templates The general template for action type concepts is shown in Figure 5 3 action an action agent the device that performs the action destination the device receiving data from the action source the device supplying data to the action instrument a device or carrier that is used by the action operand an input value to the action result an output value produced by the action if a state or condition that enables the action causes an action caused by the behavior generates an event generated by the action initiated the event that initiates the action begin the event that begins the action end the event that ends the action manner a description of how the action is performed temporal a temporal relation with the occurrence of another action or event Figure 5 3 General Template for Action The example shows the Template Constructor results on the sentence The device will output an HRQ to the microprocessor and enters the active cycle In the resulting template the bold words are the concepts represented in the sentences The modeler can observe that the output action is performed by the device Similarly the input value is the HRQ and the destination is the microprocessor The lt symbol indicates an is a relationship and the indented items show the relations between concepts 5 2 1 ASPIN Semantic Analyzer The initial version was implemented in Quintus Prolog t
85. ile The only access given to the CDictionary object user is the public member function CDictionary Member and CDictionary GetMeaning This limitation ensures that there is no sharing violation between CDictionary objects To speed up the search process for any words a two dimensional pointer array of CEntry type is used as a hash table The first dimension is used to index the first character of the words and the second dimension is used to index the second character of the words This hash table is needed as the dictionaries are getting larger and larger the current dictionary has more than 4000 entries 4 3 1 4 CMWDadictionary Class The class diagram of the CMWDictionary class is presented in Figure 4 7 The CMwDictionary class is similar to the CDictionary class in its function As the CDictionary object creates and holds the data of standard dictionaries a CMWDictionary object creates and holds the data of the multi word dictionary Its data structure however uses the constant array structure instead of the link list structure This choice is based on the same reasoning with the choice of data structure for the CGrammar class As the number of entries for the CMWDictionary object is expected to be quite small and 60 Chapter 4 ModelMaker Basic Design constant the constant array structure provides more efficiency The current limit for this array is 100 entries f CMWbDictionary 100 Load E i cMwenty
86. ing Microsoft Visual C 4 25 the On to C 24 and the Inside the C Object Model 17 were used to achieve this goal Using the object oriented paradigm ModelMaker is presented with more details in this section The main classes in ModelMaker are described in this section This discussion is followed by descriptions of the main member functions of ModelMaker and finally a discussion about the interactions within instances of these main classes 4 3 1 Main Class Descriptions The primary object used by ModelMaker engine is an object m_pSpec of the CSpecification class The CSpecification class is the class that uses and combines the other main classes such as CChart CDictionary CGrammar CLexer and CMWDictionary This CSpecification class and those other main classes are presented in this section Booch class diagrams 4 are used to illustrate this section In this notation a dotted irregular outline denotes a class A line from class C1 to C2 with a dot on the Cl end and a number n indicates that C1 uses n objects from class C2 4 3 1 1 CSpecification Class As mentioned earlier the CSpecification class is the main class in the ModelMaker As shown in Figure 4 4 this class uses multiple objects of the other classes On its instantiation a CSpecification object creates the CDictionary objects the CMwWDictionary object and the CGrammar object The CDictionary objects are used to keep the information of the standard dictionary
87. inters that are not managed properly Memory leaks create an unstable system which endangers other applications that are running on the system The memory leak checker is implemented as a simple counter that holds the number of objects of a particular class at any time Each time an object of a particular class is created the counter for that particular class is increased Similarly each time an object of a particular class is destroyed the counter for that particular class is decreased If the counter of a particular class is not equal to zero at the end of the lifetime for the particular class one 79 Chapter 5 ASPIN Components can predict that the application contains potential memory leaks If this counter goes below zero one or more destructors have been called one too many times When the memory leak checker finds either of these two situations it will warn the user through a pop up dialog box Since this process added more complexity to the class constructor and the class destructor this feature is activated only in the debugging code and not the release executable Using this feature most of the memory leaks in the Noun Phrase Extractor can be eliminated In the attempt to create a more stable application some of the classes of the Noun Phrase Extractor were reimplemented These new classes use the concepts of private and public members This implementation enforces the use of object oriented rules for interclass communicat
88. ion document file These noun phrases are also used by the m_pSpec object as the keywords for its pre search From each of the words that came from these keywords pointers to the element location where that particular word exists in the source document object are created by the m_pSpec object A function of the CSpecification class parses the source document object and creates pointers to point to the elements that use the word from the keywords These pointers allow the m_pSpec object to quickly find the existence of any keywords in the specification document data structure These searches are done by the CSpecification constructor to allow quicker response to the modeler s queries during the modeling time This pre search processing concludes the operation of the CSpecification constructor member function Once this CSpecification constructor member function is executed successfully the CSpecification object m_pSpec is ready for the ModelMaker tool If no error occurs during the constructing process of this m_pSpec object the ModelMaker tool will continue its initial processing The last thing that the ModelMaker tool performs during initial processing is updating the edit windows the keywords and options combo boxes and the info box The source edit window is initialized with a predefined value a clear screen that says eK Specification Text Window and the model edit window is initialized with the basic VHDL file The keywor
89. ions This is needed to ensure that the code is maintainable and modifiable Another enhancement for the Noun Phrase Extractor includes the multi word recognition capability Using a multi word dictionary the Noun Phrase Extractor can recognize multi words such as greater than and consist of as terms having a lexical category The last word from each of the multi words carries a special category flag in the standard dictionary Each time a word with this flag is found by the Noun Phrase Extractor in the standard dictionary it will check whether this word together with the earlier words can be combined as a multi word as well In the new Noun Phrase Extractor a rule skipping capability is also provided This capability allows the grammar rules to skip one or more words in matching constituents By utilizing this capability the Noun Phrase Extractor can skip words to recognize more sentences However this ability increases the processing time Therefore this feature is disabled in the current version of ModelMaker gt A destructor is a set of code that is supposed to dereference and to clean all of the memory used by a class 80 Chapter 5 ASPIN Components One of the most important additions to the Noun Phrase Extractor is the noun phrase constituent selector The selector scans the chart and finds all noun phrase type constituents All of the noun phrases found by this algorithm are then passed to ModelMaker to
90. ior of the Template Constructor Once the output file is written by the 87 Chapter 6 Conclusions Template Constructor ModelMaker will read this output file and present its contents to the modeler ModelMaker was designed to correctly identify standard document formats of the specification document without any markup symbols The current implementation of the ModelMaker tool is able to correctly identify a document hierarchy that uses decimal section numbering The ModelMaker tool maintains this hierarchy information in its internal document object This information is useful in maintaining traceability between the model and the source document The VHDL Identifier Extractor reads VHDL files generated by the Synopsys Graphical Environment Shell standard format For other hardware description language however this version of ModelMaker may not be able to collect all of the identifiers This is because the algorithm used in this implementation is based on the position of the identifiers relative to VHDL keywords This limitation can be eliminated by modifying the VHDL reader algorithm of the ModelMaker tool The search and sort capabilities in the current ModelMaker application are useful features Using the search ability the modeler can retrieve the most relevant information from the source document to the particular section or part that he she is working with The And and Or Boolean search types are useful for multi word
91. iptions such as written specification He wrote his thesis based on this research work He defended his thesis on May 20 1998 and received his degree Master of Science in Electrical Engineering from Virginia Tech His immediate future plan is to begin his employment with Texas Instruments Inc in Dallas Texas as an ASIC Software System Engineer in the semiconductor group Andreas Indra Gunawan 120
92. it for hlda to go low before activating hrq to service another channel Figure 1 4 Behavioral Mode Result 1 3 Research Contributions The idea of ModelMaker was initiated by Dr Walling R Cyre The main idea was to combine existing tools a Parser and the SemAnal from the ASPIN project 12 to build a better and more useful tool for device modelers The research contributions that were made by the author of this thesis to the ModelMaker tool are listed below e ModelMaker I e Main C classes for the data structure of ModelMaker The primary C classes that are used for the data structure of ModelMaker were built intentionally for ModelMaker The document element hierarchy sections paragraphs and sentences was built using object oriented Chapter 1 Introduction programming The program files that are used to define these classes are the Specification cpp and the Specification h files The Lexer A Lexer class is built to decompose input text into tokens Lexer differentiates between a period in the middle of a number and a period at the end of a sentence The tokenizing rules used in Lexer are given in section 4 2 4 2 Improved word indexing algorithm for faster performance The CDictionary class for the ASPIN Parser was used as a template in building the CDictionary class for ModelMaker Due to poor performance of the look up algorithm used in this CDictionary class a new algorithm was designed In the new
93. keyword searches The modeler can specify whether he she wants to view the available information for each of the words or all of the words at any given time Moreover the ability to sort the search results sequentially relevantly structurally and behaviorally provides useful options for the modeler The modeler can choose whichever option is best for his her current needs All of these capabilities should increase the modeler s productivity because he she can view the necessary information directly instead of going through all of the specification documentation As an overall system ModelMaker does fulfill its objective This ModelMaker tool shows that natural language analyzing capability can be a useful tool for hardware 88 Chapter 6 Conclusions designers Not only can such a tool help the designer to understand the real category carried by the specification document but it can also help the designer to modify his her designs based on modified document specifications 6 2 System Limitations Currently the executable version of ModelMaker can be run under the Windows 95 environment only This is because its graphical user interface was developed under Microsoft Visual C The main engine behind the ModelMaker tool however can be rebuilt for any other platform because it was written in ANSI C Standard Another minor limitation of this ModelMaker system is the processing time Because this is a prototype system many of
94. l IEEE std_logic_misc all IEEE std_logic_arith all IEEE std_logic_components all CLOCK In BIT CS In BIT DREQO_DREQ3 In BIT_VECTOR 0O TO 3 HLDA In BIT READY In BIT RESET In BIT AO_A3 InOut BIT_VECTOR 0 TO 3 DBO_DB7 InOut BIT_VECTOR 0 TO 7 EOP InOut BIT IOR InOut BIT IOW InOut BIT A4_A7 Out BIT_VECTOR 0O TO 3 ADSTB Out BIT AEN Out BIT DACKO_DACK3 Out BIT_VECTOR 0O TO 3 HRQ Out BIT MEMR Out BIT MEMW Out BIT architecture SCHEMATIC of DMA8237A is signal N_1 BIT signal N_2 BIT signal N_3 BIT signal N_4 BIT_VECTOR signal N_5 BIT_VECTOR signal N_6 BIT_VECTOR signal N_7 BIT signal N_8 BIT signal N_9 BIT_VECTOR signal N_10 BIT_VECTOR signal N_11 BIT_VECTOR signal N_12 BIT_VECTOR signal N_13 BIT_VECTOR signal DACKO_DACK3_DUMM BIT_VECTOR signal HRQ_DUMMY BIT signal MEMW_DUMMY BIT signal MEMR_DUMMY BIT signal ADSTB_DUMMY BIT signal AEN_DUMMY BIT component I_O_BUFFER_2 Generic DELAY TIME 0 NS Port DO_D1 TInOut BIT_VECTOR 0O TO 1 INTERNAL_DATA_BUS InOut BIT_VECTOR 0 TO 15 DBO_DB7 Out BIT_VECTOR 0O TO 7 end component component CO Generic Port end component IMAND_CONTROL DELAY TIME 0 NS DODI x04 InOut BIT_VECTOR 0 TO 1 InOut BIT component OUTPUT_BUFFER Generic DELAY TIME 0 NS Port X03 InOut BIT A4_A7 Out BIT_VECTOR 0O TO 3 end component 115 Appendix C Structural Schema
95. l to run the ModelMakerlIII tool properly It is important that the modeler place all of these files in one directory e ModelMakerlIIl exe binary executable This is the main executable file of the ModelMakerlIII application Click on the icon of this file to run ModelMakerII e Rdict txt ASCII file This file contains the standard dictionary for ModelMakerlIII Currently this dictionary contains lexical knowledge of about 4000 words This file can be edited at any time An annotated screen capture from a small part of this dictionary can be seen in Figure A 1 to illustrate the formatting of this dictionary file 94 Appendix A ModelMaker v3 8 User Manual B rdictbct terminations noun _ The word s categories terminology nil terms noun _ test tested a testing ving tests verk _ text An to denote a The entry word multi word entry theorem nil _ theoretic nil theory nil _ ere nn Figure A 1 Captured and Annotated Standard Dictionary File Each line in this file represents one entry in the dictionary Each entry consists of a vocabulary word and a list of one or more lexical categories ee parts of speech Each category consists of a pipe I followed by a category indicator followed by an underscore A single quotation mark marks the end of an entry These entries need to be sorted alphabetically An asterisk in the category field denotes that one
96. lda to go low before activating hrq to service another channel 4 programming the 8237a will accept programming from the host processor any time that hlda is inactive this is true even if hrq is active 5 application information note 1 the multimode dma controller issues a hrq to the processor whenever there is at least one valid dma request from a peripheral device Figure 1 3 Sequential Mode Result Chapter 1 Introduction lt KEY hrq gt the hrq and hlda signals from the additional 8237a are connected to the dreq and dack signals of a channel of the initial 8237a the 8237a will respond to dreq and dack but all other outputs except hrq will be disabled the multimode dma controller issues a hrq to the processor whenever there is at least one valid dma request from a peripheral device the 8237a will accept programming from the host processor any time that hlda is inactive this is true even if hrq is active if dreq is held active throughout the single transfer hrq will go inactive and release the bus to the system when the 8237a is in the idle cycle and a non masked channel requests a dma service the device will output an hrq to the microprocessor and enter the active cycle hrq out hold request this is the hold request to the cpu and is used to request control of the system bus dreq causes 8237a to issue the hrq after completion of a service hrq will go inactive and the 8237a will wa
97. le This text file contains information about the ModelMakerlII tool This file is not needed directly by the ModelMakerlllI tool and can be deleted at any time This file is just an information file for the ModelMakerlIII tool users 103 Appendix A ModelMaker v3 8 User Manual e Notepad License txt ASCII file This file holds the license agreement for the Notepad application The ModelMakerIII tool uses this Notepad application as an external ASCII file viewer This Notepad is used as it handles big files efficiently This Notepad application can be freely distributed as long as it is accompanied by this license file ModelMakerlIII Installation Guide To install the ModelMakerIII tool in a system create a new directory and copy all of the ModelMakerIII distribution files to this newly created directory The ModelMakerlIII tool is an independent application As long as it can find its own components it will run with no problem Once all of these distribution files are placed in one directory the ModelMakerIII tool is ready to be launched Using ModelMakerlIlI This section is designed as a quick reference for the modeler in using the ModelMakerlIII tool e Initiate To initiate the ModelMakerlII tool simply open the ModelMakerllI directory and click on the ModelMakerlIII icon shown in Figure A 7 ModelMakerlI exe Figure A 7 ModelMakerlIII Executable file 104 Appendix A ModelMaker v3 8 User Manual Pre
98. ler document When the user presses the Show Tree button the ModelMaker tool will search for this file read it and display its content in the Source Window This capability is added to the ModelMaker tool as the actual process of creating parse tree file can be time consuming To re create this file put a check on the Re capture the sentence parse tree check box when specifying the input files shown figure A 6 and explained in the Using ModelMakerlIlI section of this user manual 102 Appendix A ModelMaker v3 8 User Manual ModelMakerIII Initiator dma8237a txt dma8237a vhd E a Figure A 6 The Dialog Box Used to Specify the Location of the Input Files Once the main ModelMakerIII window is up press the Show Tree button Be aware that this process may take a very long time The current parse trees will be shown in the Source Window The content of Tree txt file is copied from this Source Window _Out txt ASCII file This file contains templates produced by the Template Constructor from the 8237A DMA controller document When the user presses the Template View button the ModelMaker tool will search for this file and display its contents to the Source Window ModelMakerlIII log ASCII file This file is created the first time the ModelMakerllII tool is executed in the system This file is used by the ModelMakerIII tool to track the user configuration ReadMe txt ASCII fi
99. lish program for six months before he finally joined WVU as a freshman in January 1993 Invited by the program he joined the WVU Honors Program from his second semester in WVU During his undergraduate studies he was an active member of the WVU Electric Formula Lightning Racing Car design team He served as the vice president of the Eta Kappa Nu electrical engineering honor society Beta Rho chapter from 1995 to 1996 He was also an active member of the Tau Beta Pi and the Golden Key honor society During his free time he was working part time as a computer consultant for the WVU On May 1996 he graduated magna cum laude from the West Virginia University with two degrees B Sc in Electrical Engineering and B Sc in Computer Engineering To continue his study he joined the Virginia Polytechnic Institute and State University or usually called Virginia Tech in August 1996 Again he joined the E E department with concentration in computer engineering In the first semester he worked as a graduate research assistant under Dr Scott F Midkiff to study the feasibility of a semi automatic system that gathers and presents information from the Internet on line services and CD ROM databases From the second semester until his last semester in Virginia Tech he worked as a graduate research assistant under Dr Walling R Cyre The research objective was to create a tool for rapid construction of behavioral engineering models from natural language descr
100. me ModelMaker presents this list of identifiers to the modeler as search keywords Search capabilities for the identifiers The ability to search the ModelMaker specification document data structure based on the identifiers is implemented in this version During its initialization process ModelMaker searches its document data structure for any occurrence of the identifiers to build a search index Using this index ModelMaker presents search results in its Source Window This implementation also allows the modeler to perform Boolean And and Or searches for multi word identifiers To add more flexibility ModelMaker allows the modeler to select any token in the Model Window as a keyword phrase If any word in the new keyword phrase is not found in the search index ModelMaker will perform a new search on this new keyword at run time The result will then be added to the search index for future reference Sorting capabilities for the search results The ability to sort the search results based on predefined modes is implemented in the ModelMaker tool The predefined modes are the sequential mode the relevance mode the behavioral mode and the structural mode Using this capability the modeler can view the search results in the order of how behavioral the information is or how structural the information is Each of these modes is discussed with more details in the algorithm section see section 4 4 Graphical user interfac
101. nerate the initial VHDL model file Depending on whether the modeler is interested in building a behavioral or structural model of the device he she needs to enter the appropriate schematic to the schematic capture software For behavioral modeling only the device outline with external pins is used For structural modeling a structural decomposition block diagram is used As a warning the components of block diagrams are not usually discussed in the source specification document so the structural information actually complicates the modeling task The behavioral schematic diagram the structural schematic diagram or even both of them can usually be found in the device specification document In the specification document shown in Figure 3 1 the structural schematic is shown in the bottom left corner block diagram section and the behavioral pin schematic diagram is shown in the bottom right corner pin configuration section Both the behavioral and the structural schematic of this source specification document are drawn using the Synopsys Graphical Environment SGE shell and two VHDL models were extracted from them These schematics and VHDL models are presented in Appendix B and Appendix C of this thesis For illustration small behavioral and structural schematics were developed using SGE shell and are shown in Figure 3 3 and Figure 3 4 The initial VHDL model files generated from these schematics are presented in Figure 3 5 and Figure
102. nnel can be individually programmed to autoinitialize to its original condition following an end of process eop 0 pin description 0 12 eop the reception of eop either internal or external will cause the 8237a to terminate the service reset the request and if autoinitialize is enabled to write the base registers to the current registers of that channel the mask bit and tc bit in the status word will be set for the currently active channel by eop unless the channel is programmed for autoinitialize 2 dma operation 2 2 active cycle 2 2 1 single transfer mode when the word count rolls over from zero to ffffh a terminal count tc will cause an autoinitialize if the channel has been programmed to do so 2 2 3 demand transfer mode only an eop can cause an autoinitialize at the end of the service 2 3 transfer types 2 3 2 autoinitialize by programming a bit in the mode register a channel may be set up as an autoinitialize channel during autoinitialize initialization the original values of the current address and current word count registers are automatically restored from the base address and base word count registers of that channel following eop the mask bit is not altered when the channel is in autoinitialize following autoinitialize the channel is ready to perform another dma service without cpu intervention as soon as a valid dreq is detected in order to autoinitialize both channels in a memory to memor
103. ntic Analyzer was enhanced as part of another effort to deliver human readable templates instead of conceptual graphs The Semantic Analyzer is called the Template Constructor in the ModelMaker tool Because Windows 95 is the target platform for the Template Constructor the first adaptation to the Semantic Analyzer was converting some of the Quintus Prolog specific rules to SWI Prolog rules 22 The public domain SWI Prolog interpreter was chosen because this interpreter is available in both the Unix platform and the Windows platform The Template Constructor input and output files were reformatted to the need of the ModelMaker application Unlike its parent application this Template Constructor only returns results when sentence analysis is successful Because the results of the Semantic Analyzer were in a conceptual graph format some adjustment was done in the 1 NSF Grant MIPS 9707317 84 Chapter 5 ASPIN Components Template Constructor to return the results in a template form An example of this result was presented in Figure 5 2 This formatting includes the output layout boldface italics and color This formatting is attached to the output of the Template Constructor as markup tags These markup tags are written in an HTML like style The current ModelMaker implements only three types of markup they are bold italic and color The markup rules used by ModelMaker are shown below e Bold lt b gt This will be
104. o prepare the two input files needed by the ModelMaker tool the specification document text source file and the initial VHDL model file The first task for the modeler is to scan the specification document The first page of the scanned Intel DMA 8237A specification document is reproduced in Figure 3 1 Once this is done the OCR software captures the text of the scanned document Some part of the OCR result is shown in Figure 3 2 for an illustration For more information on the input files refer to section 4 2 4 20 Chapter 3 Modeling with ModelMaker 8237A HIGH PERFORMANCE PROGRAMMABLE DMA CONTROLLER 8237A 8237A 4 8237A 5 Enable Disabie Control of Individual m Directly Expandable to Any Number of DMA Requests Channels Four Independent DMA Channels End of Process Input for Terminating Independent Autoinitialization of All Transfers Channels Software DMA Requests Memory to Memory Transfers Independent Polarity Control for DREQ Memory Block Initialization and DACK Signals Available in EXPRESS Address Increment or Decrement sar bans a nia High Performance Transfers up to 1 6M Bytes Second with 5 MHz 8237A 5 anaa le in 40 Lead Cerdip and Plastic Gee Packaging Spec Order 231369 The 8237A Multimode Direct Memory Access DMA Controller is a peripheral interface circuit for microproc essor systems It is designed to improve system performance by allowing external devices to directly transfer information trom the syst
105. o run on Sun workstations It accepts parse trees as Prolog structures and generates semantic analyses in the form of conceptual graphs 8 9 20 a form of semantic networks To invoke the Semantic Analyzer the main file main pl needs to be called from the Quintus Prolog environment This file loads the knowledge database files and 83 Chapter 5 ASPIN Components the input file The rule database files are the root database the semantic rule database and the canonical graph database The root database contains information about the root word of each word listed in the dictionary This information is used by the Semantic Analyzer to determine the root word for each word found in the sentence The semantic rule database contains the semantic rules used by the Semantic Analyzer to guide the attachment of canonical graphs These rules are based on the grammatical structures of the ASPIN Parser application The canonical graph database contains the canonical graphs for words that are currently covered by the Semantic Analyzer vocabulary Finally the input file needed by this application is the suite pl file This file contains lines of Prolog rules with each line representing a parse tree of a sentence These parse trees are generated manually from the result of the ASPIN Parser application For more information about the initial version of the Template Constructor see 12 5 2 2 Enhancement and Adaptations The Sema
106. o the CChart data structure a single constituent is also added This constituent is a terminal constituent The CGrammar object is then invoked to apply grammar rules that can combine this constituent with any other earlier constituents If a rule is found then a new CPart object is created and attached to the list of CPart objects for this node 4 3 1 3 CDictionary Class The CDictionary class as represented in Figure 4 6 is responsible for creating and maintaining the dictionary data structure Objects of this class are used to store the basic dictionary data the structural dictionary data and the behavioral dictionary data Each of these dictionary objects has a private pointer of type CEntry that holds the first entry of the dictionary Moreover each of these CEntry objects has a pointer to the next entry of the dictionary These structures create link list data structures with the first entry of each dictionary as their roots Each CEntry object contains a vocabulary word and each CMeaning object contains a category 59 Chapter 4 ModelMaker Basic Design LoadQ i Member s GetMeaning Figure 4 6 CDictionary Class Diagram The CDictionary object is initialized as soon as it is created The constructor of this CDictionary object requires the file name of the dictionary file as an argument This constructor is the one that will call the private CDictionary Load member function section 4 3 2 3 to load the dictionary f
107. odelMaker runs its internal initialization procedure This procedure includes creating the Model Window Source Window and other controls initializing all of the internal variables and pointers and establishing the default setup The second activity that ModelMaker performs during the initial processing is creating the specification document object Internally this specification document object is called m_pSpec and is an instance of the CSpecification class The constructor member function of the CSpecification class processes the source specification document and the initial VHDL model files First the CSpecification constructor creates the grammar and dictionary objects that are needed for the processing These objects are the aGrammar of CGrammar class object the aDictionary of CDictionary class object and the aMDictionary of CMWDictionary class object More information on how these objects are created and loaded can be found in section 4 3 1 5 4 3 1 3 and 4 3 1 4 The aGrammar object is created to store the grammar knowledge for the m_pSpec object The aDictionary object and the aMDictionary object are created to store the standard dictionary and the multi word dictionary knowledge for the m_pSpec object The input file that contains the grammatical knowledge is the gram txt file The CGrammar Load member function is called by the CSpecification constructor member function to initialize the CGrammar object This
108. odeling process of the behavioral 8237A DMA controller model are presented below to illustrate the modeling process with the ModelMaker tool The schematic and initial VHDL model file for this behavioral model is presented in Appendix B This is an illustration of the building process for a model of the data bus In this case the modeler is interested in modeling the data bus operation for the system The modeler can start the modeling process by performing a search on the phrase data bus The modeler can either choose this phrase from the noun phrase keyword list or simply type the phrase in the Keywords Windows Once this search is performed results similar to the one shown in Figure 3 11 will be presented in the Source Window 35 Chapter 3 Modeling with ModelMaker Within this search result the modeler may notice at least three important sentences These sentences are During DMA cycles the most significant 8 bits of the address are output onto the data bus to be strobed into an external latch by ADSTB In memory to memory operations data from the memory comes into the 8237A on the data bus during the read from memory transfer and When the processor replies with a HLDA signal the 8237A takes control of the address bus the data bus and the control lt KEY data bus gt 0 pin description 0 4 cs this allows cpu communication on the data bus 0 9 db0 db7 dbO0 db7 in out data bus the data bus lines are bidir
109. on object The CElement class is a generic class that can be used for section paragraph and sentence elements An element object will have a list of tokens a list of constituents other element objects below the level of this element object and a list of other elements of the same level Figure 4 9 presents the class diagram for the CElement class CElement p FormHits 1 A Figure 4 9 CElement Class Diagram The list of token objects of CTokenList class is used to store the section headers for the section elements and the sentences for the sentence elements The CElement object uses two pointers of the same class for the list of constituents and list of other elements Sentences are constituents of paragraph and paragraphs are constituents of section The list of other elements pointer is used to point to the next element of the same level The CElement class also keeps three integers m_Hits m_BehaviorHits and m_StructurHits to count the number of matched keyword behavioral word or structural word that is being searched These integers are used by the CElement FormHits functions More information about these integers is presented in the FormHits function discussion 62 Chapter 4 ModelMaker Basic Design 4 3 1 7 CLexer Class The CLexer class is a small but essential class As shown in Figure 4 10 this class does not use any object from other classes The only task of an object of this class is to re
110. onstructor These sentence parse trees are developed by the Noun Phrase Extractor Each of these sentences is then converted to a Prolog format Integrated template constructor This version of ModelMaker integrates the modified ASPIN Semantic Analyzer SemAnal now called the Template Constructor into the second version of ModelMaker Using this capability ModelMaker tries to match the structure of the sentences from the Noun Phrase Extractor parse tree to the pre defined templates 12 Chapter 1 Introduction Source and Model Window Modification New Source and Model Windows are implemented in this ModelMaker version These windows are based on the CRichEditCtrl class of the Microsoft Foundation Class library version 4 21 The text in these windows can be formatted with bold italic and colors This is done to allow ModelMaker to show the search results in a better format Some special markup rules were made to implement this formatting capability The Template Constructor attaches these markup tags into its output to control the formatting of the templates The new Source and Model Window support the standard editing tools such as cut copy and paste Graphical user interface total redesigned The graphical user interface of this ModelMaker version was redesigned This was done to clean up the source code and to satisfy Windows programming standards Moreover this improves ModelMaker portability To use ModelMaker tools in a plat
111. or For the search mode the possible options given are sequential relevance behavioral and structural More details on these search modes and types are discussed in section 4 4 Keyword Searches The rest of the controls are simple buttons In the current implementation of ModelMaker there are a total of seven control buttons They are the Do Search button the Save VHDL button the Template View button the Template Re Chapter 1 Introduction Analyze button the Show Tree button the Exit button and finally the ModelMaker button The modeler needs to press the Do Search button every time he she wants to view the information on a new keyword phrase When this button is pressed ModelMaker will show the information on the selected search key in the Source Window Once the modeler is done with the modeling he she can save the new VHDL code by pressing the Save VHDL button Pressing this button will overwrite the old VHDL file with the new VHDL code shown in the Model Window The Template View button and the Template Re Analyze button deal with the external program Template Constructor The Template Constructor is a set of Prolog rules that will analyze sentences using a set of predefined templates As the analyzing process can take a substantial amount of time ModelMaker can use a pre analyzed result This pre analyzed result can be saved in
112. or This grammar rule file was originally designed for the ASPIN sentence parser the Parser A captured image of part of the grammar file is presented in Figure 4 3 This grammar file contains five main fields on each line Each of the fields is separated by a space character The first field is called the rule ID field This field contains the grammar rule identifier for this entry The second field is the constituent count field This field specifies the number of constituents that are needed to create the product of this rule The constituents are listed in the fifth and subsequent fields They are placed last to allow a variable number of constituents The current Noun Phrase Extractor limits this variable to the range of one to six constituents The third field is not used by ModelMaker The fourth field is the result field This field stores the new category of the combined constituents 52 Chapter 4 ModelMaker Basic Design B gramtxt Second field the constituents count First field the rule ID n pred conj pred Fifth field the constituents Fourth field the Third field the product main concept a aws mad adw winf Figure 4 3 Captured amp Annotated Grammar File 4 2 4 Input Files ModelMaker requires two input files These input files are the VHDL model file and the source text file 4 2 4 1 VHDL Model File This file stores the VHDL code for the ModelMaker The initial VHDL code can
113. ormation is used by the ModelMaker tool to create the identifier and the noun phrase search indexes on the specification document Since the search indexes are constructed during the initialization Chapter 1 Introduction of ModelMaker searches are performed very rapidly This encourages the modeler to browse around the search phrases without any annoying time penalty The keywords window located between the Source Window and the Model Window contains a drop down list box and radio buttons The drop down list box is where the list of the identifiers or the list of the noun phrases will be presented to the modeler The modeler can either choose one of the keywords listed there or type his her own phrase in the keywords window This keyword is used by the ModelMaker tool as the search phrase ModelMaker will then search the specification document for behavioral and structural information about the keywords The radio control button lets the modeler choose whether to display the identifier keyword list or the noun phrase keyword list The identifier and the noun phrase lists are obtained from the identifier search index and the noun phrase search index respectively In modeling a device a modeler focuses on a single pin or component of the device When given this pin name or component name as the keywords ModelMaker will show the information about the particular pin or component in the Source Window The modeler can read the information co
114. pability This shows the need of the ModelMaker tool 18 Chapter 2 Relevant Work No commercial tool begins its design cycle directly with English text Most EDA tools use schematic diagrams state diagrams or data flow diagrams for design capture Moreover some EDA tools simply let the modeler to start their modeling process from the HDL level The STATEMATE 13 14 tool by i Logix Inc for example uses state charts activity charts and module charts to capture the modeler s design More information on the state charts and activity charts can be found in the Specification and Design of Embedded Systems 11 The schematic design editor subject is discussed in the Structured Logic Design with VHDL 2 Once a design is captured and validated STATEMATE can generate VHDL code from it Similarly Synopsys Graphical Environment Shell 21 and Concept HDL 6 allow modelers to design their devices in an initial stage before actually creating the VHDL or Verilog code In this case however the initial stage is the schematic diagram All of these applications again do help the designers in speeding up their design time Yet they also lack any natural language analyzer capabilities Compared to all of these other electronic design automation tools ModelMaker does have natural language analyzer capabilities It combines the two ASPIN natural language analyzers the Parser and the Semantic Analyzer in creating a better design environment
115. pandable to Any Number of Channels End of Process Input for Terminating Transfers Software DMA Requests Independent Polarity Control for DREQ and DACK Signals Available In EXPRESS Standard Temperature Range Available In 40 Lead Cerdip and Plastic Packages See Packaging Spec Order 231369 The 8237A Multimode Direct Memory Access DMA Controller is a peripheral interface circuit for microprocessor systems It is designed to improve system performance by allowing external devices to directly transfer information from the system memory Memory to memory transfer capability is also provided The 8237A offers a wide variety of programmable control features to enhance data throughput and system optimization and to allow dynamic reconfiguration under program control The 8237A is designed to be used in conjunction with an external 8 bit address latch It contains four independent channels and may be expanded to any number of channels by cascading additional controller chips The three basic transfer modes allow programmability of the types of DMA service by the user Each channel can be individually programmed to Autoinitialize to its original condition following an End of Process EOP Each channel has a full 64K address and word count capability The 8273A 4 and 8237A 5 are 4 MHz and 5 MHz versions of the standard 3 8237A respectively Figure 3 2 Partial OCR Result of the Sc
116. pp amp Dictionary h These files contain classes needed to hold the dictionary structure used in ModelMaker The main class in these files is the CDictionary class Other important classes that are used by this class are the CEntry class and the CMeaning class A CDictionary object can have an unlimited number of 45 Chapter 4 ModelMaker Basic Design entries pointers to CEntry objects which in turn can have an unlimited number of categories pointers to CMeaning objects Objects that are created from this CDictionary class include the main dictionary aDictionary the behavioral word dictionary aBehaviorDic and the structural word dictionary aStructuralDic which are all used in the CSpecification class see section 4 3 1 1 The classes included in these files are also the enhanced version of the classes used in the ASPIN sentence parser Grammar cpp amp Grammar h The CGrammar class is the main class of these files This class is used by ModelMaker to store the grammar rules that are used by the Noun Phrase Extractor These grammar rules are stored as CRule objects These classes are built upon the ASPIN sentence parser s classes Initiator cpp amp Initiator h These files house the CInitiator class This class is inherited from the CDialog class of the Microsoft Foundation Class An object of this class is used by ModelMaker to prompt the user for the input files names This class is platform dependent It will work
117. pp amp ModelMakerllI h These are the main files for the ModelMaker application in the Windows platform These files contain the CModelMakerllIApp class that is inherited directly from the CWinApp class of the Microsoft Foundation Class The only function of this class is to start the application under the Windows environment ModelMakerIIIDlg cpp amp ModelMakerIIIDlg h The only class in these files is the CModelMakerIIIDlg class When the ModelMaker application is launched an object of this class is initialized by the CModelMakerlIIApp class Once initialized this object becomes the main ModelMaker user interface All of the user interface control objects are governed by this CModelMakerlIIIDlg object MultiWord cpp amp MultiWord h These two files contain the CMWDictionary class the CMWEntry class and the CMWMeaning class All of these classes are ANSI C classes The CMWDictionary is the main class in these files and uses the other two classes as its elements This CMWDictionary class is responsible for holding the multi word dictionary structure This is to handle phrases such as consist of faster than greater than or equal to or any other multi word phrases The Noun Phrase Extractor will invoke this CMWDictionary object if it finds an asterisk symbol listed as a word s category from the standard CDictionary object This CMWDictionary object will then compare any available phrase that ends with that
118. printed in Bold lt b gt e Italic lt i gt This will be printed in Italic lt i gt e Color lt c XXXXXX gt This will be printed in XXXXXX Color lt c gt In the color markup XXXXXX the first two digit stands for hexadecimal digits 00 to FF of the blue intensity The second two digit stands for hexadecimal digits of the green intensity and the last two digit stands for hexadecimal digits of the red intensity 5 2 3 Template Constructor and ModelMaker Communications The Template Constructor application and the ModelMaker application use an input file and an output file to communicate with each other As soon as the ModelMaker tool acquires the chart of all sentences found in the specification document file by means of the Noun Phrase Extractor it creates the input file for the Template Constructor This file is suite pl file Once this file is created ModelMaker will also oe create the protocol file This protocol file _runsemanal pl is a Prolog file that is needed to control the Template Constructor component This protocol file contains the rules to redirect the output the rules to invoke the Template Constructor and the rules to analyze each of the input sentences The latter is needed because the Template Extractor component can only analyze one sentence at a time 8 Part of the development of this template look form was done by Jeff Hess Virginia Tech Electrical and Computer Engineering Department
119. py selections to the Model Window or use the information to complete the model being built The latest implementation of the ModelMaker s main window is shown in Figure 1 2 Chapter 1 Introduction ModelMakerIII Eile FS each channel can be individually programmed to autoinitialize to its original condition following an end of process eop 0 pin description 0 12 feop feop in out end of process end of process is an active low bidirectional Signal information concerning the completion of dma services is available at the bidirectional eop pin this is accomplished by pulling the eop input low with an external eop simal VHDL Model Created from SGE Schematic dma8Z37a sch Oct 12 21 21 07 1997 library IEEE use IEEE std_logic_1164 all use IEEE std_ logic misc all use IEEE std_logic_arith all use IEEE std_logic_components all entity DMA8Z37A is Port i CLOCK In BIT cS In BIT DREQO_DREQ3 In BIT_VECTOR O TO 3 HLDA In BIT READY In BIT RESET In BIT AQ_A3 InOut BIT_VECTOR O TO 3 DBO_DB InOut BIT_VECTOR O TO 7 Figure 1 2 Captured ModelMaker Main Window The search control section includes the search type and the search mode drop down lists The search type drop down list deals with how ModelMaker combines the search results of multi word identifiers The possible options given in this search type 29 drop down list are the Boolean and and
120. r all of their assistance and support for this project I would also like to thank Ritesh Sojitra and Jeff Hess graduate research assistants under Dr Walling R Cyre for any help that they have provided to me throughout this project And Finally I would like to thank my parents sisters brothers in law friends and colleagues for their continuous support and encouragement iii CONTENTS ACKNOWLEDGEMENTS ssccssscsssssisstesesesedeccscsecsdecsevanecsecsdccceceedssseosessdecvans iii MOON TEIN US TET T E AA AIETE EA EEE iv LIST OF FIGURES oooi oee ssoiessreositsessoa sisser eisso ao isi seoor ea aioa peaosas es os vi TES OETA BUSES sestccicnvcscdenctenceancuahsatevenusersenchcdsaonetusinduclsovauncageatencncdatoavens vi CHAPTER 1 INTRODUCTION 4 cccccccsssccccccccccccccccccccccccceecccccees 1 TLD MOTIVATION ruunaa a A E AEE EN 1 1 2 OVERVIEW ici ceccceieccecsdicsscscssrscecseacessscessdnceteceedscsecedccecscosscasescecsceecdcsetacseneecceecesessse 2 1 3 RESEARCH CONTRIBUTIONS e esessssseseocssssssseccoosssssessceoosssssssscsocssssssssceosossssssssesssss 9 CHAPTER 2 RELEVANT WORK 2 cccccssscscccccsccscccccescccccccceecces 16 CHAPTER 3 MODELING WITH MODELMAKER 0000 20 CHAPTER 4 MODELMAKER BASIC DESIGN 0 ccccccccsssceee 42 AT INTRODUCTION tesssvicsessssaiisse ccacddovcssudcedccasdessesaddseecctccdecsdbudebecdesdetcessdoecchacdeccssates s 42 4 2 FILE DESCRIPTIONS vics
121. r building the parse tree from each of the parsed sentence This class implements the Noun Phrase Extractor EE 7 ye CChart 2 CDictionary 1 AddNode 3 200000 eT Load _ o ClearChart CNode Member w ie yy GetMeaning d Tt eani A 4 7 CMWDictionan 3 CGr mm r Load 2 Load Se LookUp Figure 4 5 CChart Class Diagram 58 Chapter 4 ModelMaker Basic Design The CChart class uses the CNode class CPart class CDictionary class CMwDictionary class and the CGrammar class A node is an object that carries information about one word and a constituent a CPart object is a sequence of one or more nodes that forms a grammatical constituent according to the rules of the CGrammar object Two pointers of CPart class and three pointers of CNode class are used to construct a CChart object These pointers point to objects such as the first and the last node or constituent objects on the chart The links to the CDictionary CMWDictionary and CGrammar classes are also implemented as pointers and are initialized upon the instantiation of CChart object The objects of these classes are used by CChart to store the dictionary entries multi word dictionary entries and the grammar rules An object of CChart starts with empty pointers to the node and the constituent objects For each word of a new sentence a node object is created by CChart AddNode function see section 4 3 2 2 Each time a node is added t
122. r than greater than and less than which are listed in the multi word dictionary 4 2 3 2 Multi Word Dictionary File The entry field structure of the multi word dictionary file mdict txt is similar to the entry field structure of the standard dictionary file Each line of this multi word dictionary file also represents one multi word entry in the dictionary Each of the entry lines is also divided into two or more fields as in the standard dictionary file The only difference between the two entry field structures is that the multi word dictionary entry field structure allows more than one word in the first field A screen capture from a part of this dictionary can be seen in Figure 4 2 51 Chapter 4 ModelMaker Basic Design B mdict bt result in verb _ vinf results in verk _ consist of verb _ vinf _ consists of verb _ each of pdet _ in spite of prep _ instead of scon je pair of _ regardless of prep _ depend on verb _ vinf _ depending on prep _ ving _ depends on verb _ synthetic aperture radar noun _ farther than adj _ faster than adj _ fewer than adj _ greater than pdet _ less than pdet _ according to prep _ greater than or equal to pdet less than or equal to pdet _ a Figure 4 2 Captured Multi Word Dictionary File 4 2 3 3 Grammar File The grammar file gram txt contains the grammar rules needed by the Noun Phrase Extract
123. racket or a Question mark or a Right curly bracket or a Right parenthesis or a Right square bracket or 66499 a Semi colon e A Special punctuation can be formed from either a Period or oo 99 a Comma e Any other character string is recognized as a Blank 4 3 2 Main Member Functions As each ModelMaker class has many member functions not all of them will be discussed in this section Only member functions that are crucial for the discussion of ModelMaker intercommunication is presented here 4 3 2 1 Main Member Functions of CSpecification Class The main member functions of the CSpecification class include LoadTxt Q LoadVhd and Find This section discusses the functionality and the internal behavior of these functions 64 Chapter 4 ModelMaker Basic Design LoadTxt FileName aChart This private member function requires two arguments the name of the source document file and the pointer to the CChart object This function scans the source document file constructs the document inside the m_pSpec object and communicates with the CChart object Communication to the CChart object is done by calling the CChart AddNode CChart ParseNP and CChart ClearChart functions This communication is needed to let the CChart object create sentence parse trees The CSpecification LoadTxt function is called by the CSpecification constructor
124. rements Sublanguage for Automated Analysis International Journal of Intelligent Systems 10 7 665 689 July 1995 10 Walling R Cyre and Andreas Gunawan ModelMaker A Tool for Rapid Modeling From Device Descriptions Proc of 1998 International Verilog HDL Conference and VHDL International Users Forum Santa Clara California Mar 16 19 1998 11 Daniel D Gajski Frank Vahid Sanjiv Narayan and Jie Gong Specification and Design of Embedded Systems Prentice Hall PTR New Jersey 1994 12 Rob Greenwood Semantic Analysis for System Level Design Automation Master s Thesis Virginia Polytechnic Institute and State University Blacksburg VA 1992 13 David Harel and Amnon Naamad The STATEMATE Semantics of Statecharts i Logix Inc July 1996 91 References 14 David Harel and Michal Politi Modeling Reactive Systems with Statecharts i Logix Inc June 1996 15 Intel Peripherals Intel Corporation Mt Prospect Illinois 1990 16 Balraj Singh John Wicks Philip Wright and James R Armstrong The Modeler s Assistant A CAD Tool for Behavioral Model Development Proc of the 11 IFIP WG10 2 International Conference on Computer Hardware Description Languages and their Applications Ottawa Ontario Canada April 26 28 1993 17 Stanley B Lippman Inside the C Object Model Addison Wesley Publishing Company Massachusetts 1996 18 Microsoft Visual C Programm
125. rence is that the list of lexical category can be left empty nil as the value will not be used by ModelMaker To illustrate this a behavior dictionary file is shown in Figure A 4 By behavior dic handle nil _ handled nil _ handler nil _ handles nil _ handling nil _ handshake nil handshaking nil _ happens nil _ held nil _ help nil _ heterodyning nil _ a 1 HEL L1 FEF Figure A 4 Behavior Dictionary File 98 Appendix A ModelMaker v3 8 User Manual e Structure dic ASCII file This file contains the list of the structural words used by the ModelMakerIII tool to rank the search results Refer to section 4 4 6 of this thesis for more information on how the ModelMakerlIII tool utilizes this file Similar to the behavioral dictionary file this dictionary file uses the same exact format as the standard dictionary file A structural dictionary file is shown in Figure A 5 B structure dic assembly nil _ consist nil _ contain nil _ form nil format nil _ a has nil inii nri S include nil _ part nil _ Figure A 5 Structural Dictionary File e Notepad exe binary executable This is an ASCII viewer application a free extension of the Windows Notepad application This external application is required since some commands in ModelMakerlII require an ASCII viewer If a different ASCII viewer is preferred simply replace this file with
126. rocessor and enter the active cycle The resulting template shows that the device will perform two operations The operations are enter and output The whole template can be interpreted in two sentences the device will enter the active cycle and the device will send an HRQ signal to the microprocessor The modeler can then use these sentences to verify his her model 41 Chapter 4 ModelMaker Basic Design CHAPTER 4 MODELMAKER BASIC DESIGN 4 1 Introduction ModelMaker integrates a search engine and two components from a natural language understanding system with a graphical user interface providing an editor and search controls The natural language processing components are the Parser and the Semantic Analyzer These components are taken from the ASPIN system 12 They were enhanced and customized from the original versions to serve in ModelMaker The new versions are called the Noun Phrase Extractor and the Template Constructor respectively ModelMaker provides a common environment for both of these tools and adds more functionality to build a new tool for rapid modeling of a device from informal specifications This new functionality includes the Document Reader the VHDL Identifier Extractor and a search ability based on the knowledge it builds during initialization and run time In object oriented terms ModelMaker can be separated into many objects The graphical user interfaces are objects from the Microsoft CDi
127. ructuralDic CheckBehavior CheckStructural This public member function is a recursive function used to find all occurrences of keywords in the document elements This function requires nine arguments The first one Token specifies the token to be searched Hits BehaveHits and StructHits are call by reference arguments 24 These arguments are used to pass the number of hits matched tokens from the last level to the first level of the recursion The values of these arguments need to be reset to zero as they are passed to the function The fourth argument pFirstHit points to the first matched element The fifth and sixth arguments are pointers to the behavioral and structural dictionary objects Finally the last two arguments are Boolean flags to specify whether behavioral and structural vocabulary checks need to be performed A flag is returned by this function to specify whether it was run successfully 13 Chapter 4 ModelMaker Basic Design For each of the keywords noun phrases and identifiers the m_pSpec object calls this function during its initialization This is done to prepare the search results index 4 3 2 7 Main Member Functions of CLexer Class GetToken This public member function is the main member function in the CLexer class When called this function returns the next token from the open document file in the form of a character pointer If the end of file is reached or any error is detected a null pointer will be
128. ry useful for modelers to find where that particular instance is in the actual document specification 75 Chapter 4 ModelMaker Basic Design text This allows modelers to go back to the actual document specification text for more detailed information 4 4 4 Relevance Sort Relevance sort mode sorts the search results based on the number of keywords from the entire keyword list that occur in the paragraph The more keywords that occur in the paragraph the higher its relevancy As the order of sentences in this mode may be different from the actual order in the source document text the section headers for these sentences are not displayed This mode is useful for modelers in finding any information that is more general to all the identifiers and the noun phrase This sort returns the same information returned by the other sorts The only difference is in the order of the results 4 4 5 Behavioral Sort In this behavioral sort mode the search result is sorted based on the number of behavioral words that each paragraphs has An object from the CDictionary class is used to hold the behavioral word dictionary While performing the initial search ModelMaker checks for any occurrence of behavioral words The more behavioral words the paragraph has the higher its behavioral content Based on a similar reasoning with the relevance sort the section headers for these paragraphs are also not displayed As behavioral words are usually used in de
129. s Figure 3 15 VHDL Process Created Based on Search Phrase autoinitialize The two examples presented above illustrate the capability of the ModelMaker tool with the Noun Phrase Extractor To show the usability of the Template Constructor an illustration is presented below Suppose the modeler is trying to verify his understanding of the sentences from the source specification document By pressing the Template View button the Template Constructor will return all of the analyzable sentences in the Source Window If the sentence in question is an analyzable sentence the modeler can use the information from the Template Constructor to decipher the sentence For example Figure 3 16 shows one of the analyzed sentences from the Template Constructor 40 Chapter 3 Modeling with ModelMaker Sentence 330 the device will output an hrq to the microprocessor and enter the active cycle 1 AND lt behavior and ENTER 2 and OUTPUT 3 2 ENTER lt call lt event enters CYCLE 4 agent DEVICE the device 4 CYCLE lt state lt behavior attribute ACTIVE 3 OUTPUT lt move lt action destination MICROPROCESSOR lt processor lt device the microprocessor operand HRQ lt id an hrq agent DEVICE the device modal WILL Figure 3 16 Sample Result from the Template Constructor The Template Constructor result shown in Figure 3 16 came from the sentence The device will output an HRQ to the microp
130. s 63 FIGURE 4 11 SAMPLE SPECIFICATION DOCUMENT TEXT PARTIAL cc scccccccceesesssseseees 66 FIGURE 4 12 DOCUMENT ELEMENTS BUILT BY CSPECIFICATION LOADTXT 0 0006 66 FIGURE 5 1 SAMPLE PARSE TREE ccssscccccccccssssssnsececesecsesessnseaecesecsesesessaeeeeeseesenenssaaeess 82 FIGURE 5 2 SAMPLE RESULT OF THE TEMPLATE CONSTRUCTOR ccccsssessssssessssetsseeees 82 FIGURE 5 3 GENERAL TEMPLATE FOR ACTION cccccccccsccecssssssceeeseceeeesensaeceeeceesesessnteaeees 83 LIST OF TABLES TABLE 4 1 C HEADER AND SOURCE FILES cccccccccsscscscscssssscseeeeeeeseeeeseeeeseeeseeeeeenees 45 FABLE 4222 PROLOG FILES 23 502 cseessthagense atone EE EE ET 49 vi Chapter 1 Introduction CHAPTER1 INTRODUCTION 1 1 Motivation In today s technology device modeling has become increasingly important Simulation and synthesis are two applications of device modeling Languages such as VHDL and Verilog are becoming more popular among device engineers Engineers also called modelers use device modeling to test or rebuild legacy systems to build libraries of commercial components or even to build new devices This device modeling concept however comes with a price Especially in dealing with legacy systems a considerable amount of expertise and effort is needed because most of these systems come with ambiguous and incomplete specification documentation The ModelMaker tool facilitates the modeling of these devic
131. scribing the behavior of a device this sorting mode is useful in finding more information about the device The behavioral word dictionary is saved in the Behavior dic file This sort returns the same information returned by the other sorts The only difference is in the order of the results 4 4 6 Structural Sort Similar to the behavioral sort this mode is also based on a CDictionary object the structural dictionary The structural word dictionary is saved in the structure dic file In this mode a paragraph that has more structural words on it will have a higher 76 Chapter 4 ModelMaker Basic Design structural content which will be displayed first ModelMaker also parses the input specification document for these structural words during its initial search Section headers for these paragraphs are also not displayed This search mode is useful to the modelers in building the structure of the device This sort returns the same information returned by the other sorts The only difference is in the order of the results 11 Chapter 5 ASPIN Components CHAPTER5 ASPIN COMPONENTS 5 1 Noun Phrase Extractor Parser The Noun Phrase Extractor is an adaptation of the ASPIN Parser The ASPIN Parser is a stand alone console application written in C object oriented programming The Parser is supported by a context free grammar and a dictionary giving the lexical categories of words To adapt the Parser some of the cl
132. sentences For any given string of tokens words numbers and punctuation the Parser generates zero or more parse trees Some of the trees may represent sentences Others represent lesser constituents such as noun phrases For example Figure 2 1 shows a parse tree for the sentence the sci uses standard nrz format and for one of its noun phrases The grammar used by the parser was developed in a study of the English used to describe microprocessors 1 s 2 c sl c ss2 c nl c np12 t det the c head2 t id sci c pred2 c avs2 t verb uses c nl c np19 c adjs1 t adj standard c head4 t id nrz c head1 t noun format J tC Figure 2 1 Semantic Analyzer Input Using the result of the Parser a modeler can use the ASPIN Semantic Analyzer 12 to analyze the information that is carried by the sentences The Semantic Analyzer attempts to extract the category of the sentence and represent it as a conceptual graph 8 9 20 To illustrate this process Figure 2 2 shows the conceptual graph formed by the Semantic Analyzer for the sentence the sci uses standard nrz format The input to the Semantic Analyzer was the parse tree of Figure 2 1 Current location of the list is at http hpcc923 external hp com csocat eda eda maineda htm This URL can be changed at any time by the Hewlett Packard web master Contact the Hewlett Packard web master for more information 17 Chapter
133. ssage for the modeler and stop all of its processing These errors include illegal dictionary format dictionary file not found and any other errors that deal with the input files At this point if no error is found the m_pSpec object starts to load the initial VHDL model file and the source document file This is done by calling the CSpecification LoadVhd and the CSpecification LoadTxt private member functions The task of the CSpecification LoadVhd function is to load the initial VHDL model file provided by the modeler into m_pSpec internal variable While performing this task this function also extracts all of the VHDL identifiers found in the VHDL file These identifiers are then stored This list is used by the m_pSpec object as the identifier keywords For more information on the CSpecification LoadVhd private member function see Main Member Function of CSpecification Class section section 4 3 2 1 31 Chapter 3 Modeling with ModelMaker The CSpecification LoadTxt function loads the source document file in the m_pSpec object This task is not simple for the function needs to identify the correct format of the specification document This function is also responsible for calling the CChart AddNode member function the CChart ClearChart member function and the CChart ParseNP member function for the aChart object serving the m_pSpec object Section 4 3 2 1 discusses the CSpecification LoadTxt private member fun
134. ssing this ModelMakerllII button will bring up the input file dialog box as shown in Figure A 8 ModelMakerlII Initiator dma8237a trt dma8237a vhd E Figure A 8 The Dialog Box Used to Specify the Location of the Input Files Specify the input files in that input file dialog box and press the Continue button If this is the first time the ModelMaker tool is executed another dialog box will pop up querying the user about the location of the SWI Prolog executable file This dialog box can be seen in Figure A 9 The ModelMakerlII tool saves the user configuration inside the ModelMakerlI log file Therefore the modeler only needs to specify this SWI Prolog location once SWI Prolog Finder C Program Files pI BIN PLWIN EXE eae Figure A 9 The Dialog Box Used to Specify the Location of the SWI Prolog Executable 105 Appendix A ModelMaker v3 8 User Manual Since the SWI Prolog is not needed with the ModelMakerlII distribution just press the I don t have SWI Prolog button This should initiate the ModelMaker tool and bring up the main ModelMaker window as shown in Figure A 10 each channel can be individually programmed to autoinitialize to its original condition following an end of process feop 0 pin description 0 12 feop feop in out end of process end of process is an active low bidirectional signal information concerning the completion of dma services is a
135. stant 3 16 19 Commercial companies such as i Logix Inc Synopsys Inc and Cadence Design System Inc are just a few of many EDA tool providers Hewlett Packard maintains in their web site a long list of currently available commercial EDA tools Of these many EDA tools none of them provide natural language analysis capability Instead many of these EDA tools start their design cycle from state diagrams or schematic diagrams Some of the other tools let the modelers to start from the hardware description language HDL level directly as some modelers prefer to create their own models The primary objective of this chapter is to discuss some of these currently available EDA applications and to justify the need of ModelMaker The Automated Specifications Interpreter ASPIN project is a research project being conducted by the Automatic Design Research Group ADRG at Virginia Tech The objective of this research project is to develop applications that improve designer productivity ASPIN accomplishes this objective by helping a user to develop good specifications and helping the designers understand the information behind the specification documents The two components of the ASPIN system of interest here are the Parser and the Semantic Analyzer For more information about the Automatic Design Research Group contact adrg vt edu 16 Chapter 2 Relevant Work The Parser component is an application that constructs parse trees for English
136. tains a list of rules instead of program code Prolog uses these rules as a knowledge database A Prolog interpreter is always needed to interpret Prolog rules because it can not be compiled to binary format 81 Chapter 5 ASPIN Components Sentence 330 the device will output an hrq to the microprocessor and enter the active cycle s 330 c s2 c sp2 c nl c npl2 t det the c headl t noun device c pred3 c avs3 t mod will t vinf output c ni c npl2 t det an c head2 t id hrg c d2 t prep to c ni c npl2 t det the c headl t noun microprocessor 1 t conj and c pred2 c avs2 t verb enter c ni c np11 t det the c adjs1 t adj active head1 t noun cycle EO ey es Figure 5 1 Sample Parse Tree Sentence 330 the device will output an hrq to the microprocessor and enter the active cycle 1 AND lt behavior and ENTER 2 and OUTPUT 3 2 ENTER lt call lt event enters CYCLE 4 agent DEVICE the device 4 CYCLE lt state lt behavior attribute ACTIVE 3 OUTPUT lt move lt action destination MICROPROCESSOR lt processor lt device the microprocessor operand HRQ lt id an hrq agent DEVICE the device modal WILL Figure 5 2 Sample Result of the Template Constructor 82 Chapter 5 ASPIN Components The template shown in F
137. the ASPIN Parser Noun Phrase Extractor application is tightly integrated in its binary code Naturally all of the functionality from the initial version of the Noun Phrase Extractor the Parser is inherited by the ModelMaker tool However because the ModelMaker tool was developed using object oriented programming these two components can be separated or modified at any time All of the communication between the ModelMaker component and the Noun Phrase Extractor component is done by message passing The Noun Phrase Extractor and ModelMaker combination works well in making the Parser useful to a modeler The enhanced version of the ASPIN Semantic Analyzer the Template Constructor is not yet integrated into the ModelMaker binary code ModelMaker is an executable application written in Microsoft Visual C and the Template Constructor is a set of SWI Prolog rules that will always require an SWI Prolog interpreter The ModelMaker tool communicates with the Template Constructor through temporary files Using this method the ModelMaker tool can evoke any of the functions of the Template Constructor For each sentence ModelMaker receives from the Noun Phrase Extractor an entry is added to the Template Constructor input file Once all of the sentences are received the Template Constructor control file is called from the SWI Prolog interpreter This control file contains the controlling rules prepared by the ModelMaker component to control the behav
138. the Prolog files which are listed in Table 4 2 Table 4 2 Prolog Files Files Information Contained Functions Canon pl Meanings of words Semrules pl Semantic Analysis rules Roots pl Mapping tokens to root form Main pl Semantic Analyzer engine Canon pl This file contains the conceptual type hierarchy for dictionary words as well as the conceptual graph meanings for words that are currently covered by the vocabulary of the Semantic Analyzer It also contains definitions for the routines to find the minimal semantic super type of a concept Roots pl This file contains information about the root form of each word listed in the dictionary The Semantic Analyzer uses this information to select semantics from the canon for a given word Semrules pl This file contains the semantic analysis rules used by the Semantic Analyzer to combine the meanings of words into sentence meanings Each grammatical rule of the parser has a semantic analysis rule associated with it 49 Chapter 4 ModelMaker Basic Design Main pl This file is the main file of the Semantic Analyzer Once invoked this file calls the rest of the Semantic Analyzer files This file contains the procedures invoked by the semantic rules to combine word semantics 4 2 3 Database Files ModelMaker uses three database files to support parsing by the Noun Phrase Extractor These files are the standard dictionary file rdict txt the multi word
139. tic and Its Initial VHDL Model component I_O_BUFFER Generic DELAY TIME 0 NS Port X03 INOUE BIT X04 InOut BIT AO_A3 Out BIT_VECTOR 0O TO 3 end component component INC_DECREMENTOR Generic DELAY TIME 0 NS Port INTERNAL_DATA_BUS InOut BIT_VECTOR 0O TO 15 X01 InOut BIT_VECTOR O TO 15 X03 InOut BIT end component component DECREMENTOR Generic DELAY TIME 0 NS Port X01 InOut BIT_VECTOR O TO 15 end component component READ_WRITE_BUFFER Generic DELAY TIME 0 NS Port X01 InOut BIT_VECTOR 0O TO 15 X02 InOut BIT_VECTOR O TO 15 X05 InOut BIT_VECTOR O TO 15 X06 InOut BIT_VECTOR O TO 15 X07 INOUE BIT_VECTOR O TO 15 X08 InOut BIT_VECTOR O TO 15 end component component WRITE_BUFFER Generic DELAY TIME 0 NS Port INTERNAL_DATA_BUS InOut BIT_VECTOR 0O TO 15 X05 InOut BIT_VECTOR O TO 15 X06 InOut BIT_VECTOR O TO 15 end component component READ_BUFFER Generic DELAY TIME 0 NS Port INTERNAL_DATA_BUS InOut BIT_VECTOR 0O TO 15 X07 InOut BIT_VECTOR O TO 15 X08 InOut BIT_VECTOR O TO 15 end component component READ_WRITE Generic DELAY TIME 0 NS Port INTERNAL_DATA_BUS InOut BIT_VECTOR O TO 15 end component component STATUS Generic DELAY TIME 0 NS Port INTERNAL_DATA_BUS InOut BIT_VECTOR 0O TO 15 end component component BASE_READ_BUFFER Generic DELAY TIME 0 NS Port
140. urrence of each tokens in the keyword Anytime a match is found in a paragraph the section header of the paragraph is added to the search result The sentences in that paragraph however will not be added to the search result unless all of the words in the keywords occur in that particular sentence In some cases the section header will be displayed without any sentences written under it This is just to show that one or more but not all of the tokens in the keywords occur under this section By reducing the number of tokens in the keyword the modeler may be able to find more results under this section header 4 4 2 OR Search Mode Similar to AND search mode OR search mode specifies how ModelMaker builds the search result of multi word keyword searches In this mode the search result from each of the word in the keywords is displayed separately under each of the keywords This feature allows the modeler to see the search result of each of the words independently This OR search mode is only available for the identifiers or user typed keywords This limitation was established since the noun phrase index list is supposed to contain the entire noun phrase found by ModelMaker 4 4 3 Sequential Sort The sequential sort mode is the default mode for any ModelMaker search In this mode search result is sorted based on its original order in the source document text All section headers will also be displayed to the modeler This mode is ve
141. vailable at the bidirectional eop pin this is accomplished by pulling eop input low with an e M nal eor simal OF VHDL Model Created from SGE Schematic dma8237a sch Oct 12 21 21 07 1997 library IEEE 2 use IEEE std_logic_1164 all use IEEE std_logic misc all use IEEE std_logic_arith all use IEEE std_logic_components all entity DMA8Z374 is Port CLOCK In BIT cs In BIT DREQO_DREQ3 In BIT_VECTOR O TO 3 HLDA In BIT READY BIT RESET In BIT A40_A3 InOut BIT_VECTOR O TO 3 DBO_DB InOut BIT_VECTOR O TO 7 Figure A 10 The ModelMaker Tool Main Window Depending on the size of the input documents the initial processing may require some time Listed below are the components of the ModelMakerIII tool main window as shown in Figure A 4 1 Source Window Customized Edit Window The Source Window placed in the top portion of the main working window is where the analyzed specification document elements will be displayed by ModelMaker This window allows the user to select and copy some or all of the text and paste it into the Model Window or any 106 Appendix A ModelMaker v3 8 User Manual other edit windows This window however is a read only edit window No information can be entered to or removed from the Source Window To copy information from the Source window the modeler need to select the portion of the text that he she wants to copy Once the selection is
142. version a hash table of pointers is used as an index into the CDictionary object This new algorithm made a significant improvement in the search performance of ModelMaker More information about the hash table is presented in section 4 3 1 3 Specification document reader algorithm An algorithm was built to transform the specification document text into the ModelMaker document data structure The first version of this algorithm required additional markup to be added to the specification document This was to ensure that ModelMaker read the document hierarchy correctly The markup style used was similar to the Hypertext Markup Language HTML The current version does not need this additional markup anymore A new algorithm was built to read the document specification file with standard formatting However the style of the documents that ModelMaker is able to read automatically is still limited by some rules These rules are explained with more details in section 4 2 4 2 These current rules are able to recognize the test documents with a small amount of adjustment in the specification document text formatting VHDL Identifiers Extractor algorithm A VHDL reader algorithm was created to read the VHDL input file into the ModelMaker tool This algorithm is capable of extracting most of the 10 Chapter 1 Introduction identifiers listed in the VHDL file An identifier is defined as a signal name a port name a component name or an entity na
143. y transfer both word counts should be programmed identically 3 register description 3 1 current address register it may also be reinitialized by an autoinitialize back to its original value autoinitialize takes place only after an eop 3 3 base address and base word count registers during autoinitialize these values are used to restore the current registers to their original values 3 7 mask register each mask bit is set when its associated channel produces an eop if the channel is not programmed for autoinitialize Figure 3 14 ModelMaker Search Result on the Word autoinitialize 39 Chapter 3 Modeling with ModelMaker From the autoinitialize search result the modeler may notice at least one crucial item The sentence that carries that information says During autoinitialize initialization the original values of the current address and current word count registers are automatically restored from the base address and base word count registers of that channel following EOP Based on this information the modeler can build his her autoinitialize process The sentence specifies the signal that triggers the actions and the actions that need to be done With this knowledge the modeler can build a VHDL process as shown in Figure 3 15 autoinitialize_process process EOP begin if EOP event and EOP 0 then CURRENT_ADDRESS lt BASE_ADDRESS CURRENT_WORD_ COUNT lt BASE_WORD_COUNT end if end proces
144. y word count going to ffffh or an external end of process eop is encountered 2 2 3 demand transfer mode in demand transfer mode the device is programmed to continue making transfers until a tc or external eop is encountered or until dreq goes inactive only an eop can cause an autoinitialize at the end of the service eop is generated either by tc or by an external signal 2 3 transfer types 2 3 1 memory to memory when the word count of channel 1 goes to ffffh a tc is generated causing an eop output terminating the service the 8237a will respond to external eop signals during memory to memory transfers 2 3 2 autoinitialize during autoinitialize initialization the original values of the current address and current word count registers are automatically restored from the base address and base word count registers of that channel following eop if interrupted externally eop pulses should be applied in both bus cycles Figure 3 13 ModelMaker Search Result on the Word eop 38 Chapter 3 Modeling with ModelMaker From the search result shown in Figure 3 13 the modeler can observe significant actions that are connected to the word autoinitialize To learn more about autoinitialize the modeler can then use this word as the next search word Using this new search word a result similar to the one shown in Figure 3 14 will be shown in the Source Window lt KEY autoinitialize gt each cha

ModelMaker: A Tool for Rapid Modeling from Device Descriptions

Contents

Download Pdf Manuals

Related Search

Related Contents