Home

PROMULA FORTRAN to C Translator User`s

1. If Line Nl Source 1 SUBROUTINE STRUC6 2 INCLUDE STRUC6 INC 1 STRUCTURE DATE 2 d INTEGER 4 DAY MONTH 3 1 INTEGER 4 YEAR 4 1 END STRUCTURE 5 G 6 C STRUCTURE TIME APP TIME 2 7 C LOGICAL 1 HOUR MINUTE 8 C END STRUCTURE 9 C 10 C This is the same as 11 C 12 C STRUCTURE TIME 13 C LOGICAL 1 HOUR MINUTE 14 G END STRUCTURE 15 C RECORD TIME APP TIME 2 16 17 STRUCTURE TIME 18 1 OGICAL 1 HOUR MINUTE 19 END STRUCTURE 20 21 STRUCTURE APPOINTMENT 22 RECORD DATE APP DATE 23 24 RECORD TIME APP TIME 2 25 26 1 CHARACTER 20 APP 4 27 LOGICAL 1 APP_FLAG 28 END STRUCTURE 29 RECORD APPOINTMENT NEXT_APP APP_LIST 7 30 RECORD DATE TODAY 3 WRITE 6 STRUC6 OUT 4 10 1 7 5 1 CALL GET DATE I TODAY 6 1 WRITE 6 TODAY DAY TODAY MONTH TODAY YEAR 7 1 APP LIST I APP DATE TODAY 8 1 TODAY DAY TODAY DAY 1 9 1 10 END DO 10 5 FORMAT 315 11 EXT APP APP LIST 1 12 OPEN 1 FILE STRUC6 BIN FORM UNFORMATTED STATUS UNKNOWN 13 WRITE 1 NEXT APP 14 WRITE 1 APP LIST 15 END The heading which is printed at the top of each page contains the name of this compiler along with its current version number on the left hand side of the page The right hand side normally contains the name of the
2. SUBROUTINE DEMO1 I J CHARACTER NAME 12 INITIAL 4 J INDEX NAME INITIAL I JINDEX NAME INITIAL RETURN END SUBROUTINE DEMO2 I J INTEGER NAME 3 INITIAL J INDEX NAME INITIAL I JINDEX NAME INITIAL RETURN END SUBROUTINE DEMO3 I J CHARACTER NAME 12 INTEGER INITIAL J INDEX NAME INITIAL I JINDEX NAME INITIAL RETURN END In this example the system function index and the user function jindex are both called with CHARACTER arguments and with INTEGER arguments in different places in the program The CHs treatment of the above looks as follows void demol i 3 int 1 j extern int jindex static char name 12 initial 4 xj fifindex name 12 initial 4 i jindex name 12 initial 4 return void demo2 i 3 int Tu extern int jindex static int name 3 initial j fifindex name 12 amp initial 4 xi jindex name amp initial return void demo3 i j lint iyt extern int jindex static char name 12 Static int initial j fifindex name 12 amp initial 4 xi jindex name 12 amp initial return 29 PromulaFortran Translator User s Manual Note that no matter how the jindex subprogram is treated the above cannot hope to execute properly since the ordering of the arguments is destroyed via the placement of lengths following t
3. PROGRAM VARI2 INTEGER 2 I1 12 13 14 INTEGER 2 ISUM IPROD IQUOT INTEGER 4 JSUM JPROD JQUOT REAL 4 RSUM RPROD RQUOT REAL 8 DSUM DPROD DQUOT 1 20000 I2 30000 I3 200 I4 300 ISUM I1 I2 Note addition overflow IPROD I3 I4 Note multiplication overflow IQUOT I1 I2 IA lt Note intermediate overflow JSUM I1 I2 JPROD I3 I4 JQUOT I1 I2 14 RSU I1 I2 RPROD I3 I4 RQUOT I1 I2 I4 DSU I1 I2 DPROD I3 I4 DQUOT I1 I2 14 WRITE 24H Short Integer Results 3113 ISUM IPROD IQUOT WRITE 24H Long Integer Results xkld 34 5j JSUM JPROD JQUOT WRITE 24H Short Real Results 29 P RSUM RPROD RQUOT WRITE 24H Long Real Results SE ESS NE DSUM DPROD DQUOT SIOP END In running this example with various FORTRAN compilers always on machines with 16 bit short integers VAX IBM mainframe IBM PC we have obtained the following three results 1 Universal promotion to long Short Integer Results 15536 5536 166 Long Integer Results 50000 60000 166 Short Real Results 50000 00000 60000 00000 166 00000 Long Real Results 50000 00000 60000 00000 166 00000 2 No automatic promotion to long Short Integer Results 15536 5536 o Long Integer Results 15536 60000 oL Short Real Results 15536 00000
4. PURPOSE General purpose initialization for MODI routines l PARAMETERS SWAB FLAG R jOGICAL XLIT FLAG R 1 RETURN VALUE None SIDE EFFECTS None Incorporates features to support portable LIO file handling 8 1986 AG long arg_swab_flag arg_xlit_flag extern void trlito 109 PromulaFortran Translator User s Manual static long K1 4 static long modi noninpbase hd_qpis hd_qpsupis firstnoninp hashmax hd_hashval swab_flag xlit_flag blanks_xlit_yet blank_bytes Initialization entry Initialization calculate some useful things noninpbase hd qpis hd qpsupis firstnoninp noninpbaset1L First non input node Determine practical maximum for no of strings we can hash using PUTSTR entry Knuth s vol 3 says that the avg no collisions 1 2 1 1 R where is HASH table density it might be nice to stop at 98 full R 0 98 since this translates to an average of 25 5 collisions per string entered X hashmax hd hashval 2L hd hashval 100L Save local values of SWAB and XLIT flags and convert blank string to match the language EBCDC or ASCII of the model file CREATOR E swab flag arg swab flag xlit flag arg xlit flag if xlit flag amp fifi4log blanks xlit yet trlito amp blank bytes amp K1 blanks xlit yet fifi
5. 101 2 48 UNFORMATTED WRITE OPTIMIZATION XU nasse eret 104 2 49 SUBPROGRAM CALL BY V ALUE ARGUMENTS XV 105 2 50 DOLLAR SIGNS AS INITIAL SYMBOLS IN IDENTIFIERS X Y cccccccccccccsssccescccesscesseecssecesssecssecesseecsseesessecsaseeseeees 105 2 51 LOCATION OF FORTRAN FILES TO BE INCLUDED ZNAME ccce enhn ennt nnns nnns esee nete rena nennen 105 2 52 PROJECT PROCESSING SEPROJECT iieri crees iste be hte esee iere ge eot e iege eO ERR e ee esie decade un 105 CONFIGURATION ELLE Rc 107 3 1 THE CONFIGURATION SWITCHES STATEMENT ccccccessessscecececeeseaececccecsenssaececececeeneseaeeeceesenessaeseeeeeesenseaeeeeeeeenes 107 iv PromulaFortran Translator User s Manual 3 2 THE CONFIGURATION COMMENTS 108 3 3 THE CONFIGURATION PATHNAMES 112 3 4 THE CONFIGURATION RESTRUCTURE entra sanis setenta 114 3 5 THE CONFIGURATION KEYWORDS 116 3 5 1 Simple Keyword Replacement ioci c t evades e E ER RR Ee E e ERN Ee ERSTE ets 119 3 5 2 Pattern Strings for COMMON blocks e rt ee t E Ne i a eR reli eene ed
6. 257 VMSWRT WRITE A VIRTUAL LONG TRUTH VALUE VECTOR 258 VMSWRU WRITE FORTRAN VIRTUAL UNSIGNED CHARACTER VECTOR nnn 258 VMSWVB WRITE amp VIRTUAL bee aee ttes a tete cete 259 ix PromulaFortran Translator User s Manual PLEASE READ THIS SECTION There are three fundamentally different ways of using PFC 1 Asa FORTRAN compiler 2 As a tool to produce maintainable C source code from a FORTRAN source which corresponds to the original as closely as possible so that it can be maintained by the original authors 3 As a tool to produce a maintainable C source code from a FORTRAN source which is logically equivalent to the original but which uses conventional C notation and standard C functions as much as possible These three views are referred to as the optimized FORTRAN and C biases respectively The default bias for is the FORTRAN bias The C and optimized biases are however fully supported by PFC and may be activated by the Bc and Bo command line switches In addition the PFC configuration file can be easily changed to make either the C or the optimized bias the default Please evaluate your own reasons for applying PFC to your FORTRAN codes and make the appropriate bias selection If you are still not satisfied with the output from PFC please look at the chapters in this manual which discuss the c
7. e eI 242 PDPASSN PDP FORTRAN SUBROUTINE 81 243 PDPCLOSE PDP FORTRAN SUBROUTINE 243 PDPCVTIM PDP FORTRAN EXTERNAL FUNCTION CVTTIM eeeeeeeeeenen e n n emen nn 244 PDPGTIM PDP FORTRAN EXTERNAL FUNCTION GTIM eeeeeeeeeenee e ene emen nsns 244 VMSCLS GEOSE VIRTUAE FIEE a a fe eet ee eee reve ee 245 VMSDEL CHANGE VIRTUAL INFORMATION eere e nsns i asi 245 VMSGLOB VIRTUAL GLOBAL ACCESS eeeeeeeeeeeeeeennenene iett e ese ese e esas e esas a ene setas asas e 246 VMSLOAD LOAD A VIRTUAL 246 VMSOPN OPEN A VIRTUAL MEMORY e ene n emen nnn ennt nnns nne nsus ese ese esa nsns ases n enean 246 VMSPTR GET VIRTUAL BYTE 247 VMSRBL REMOVE VIRTUAL BLOCK 248 VMSRDB READ FORTRAN VIRTUAL BOOLEAN VECTOR ssseserererererererererererererererererererererererererererererererere 248 VMSRDC READ FORTRAN VIRTUAL CHARACTER VECTOR eee ene e emen nn ennn nnns nnns ese nsns 249 VMSRDD READ FORTRAN VIRTUAL DOUBLE PRECISION VECTOR eee eene emen nn ener nn nsns n 249 VMSRDF READ FORTRAN VIRTUAL FLOATING POINT VALUES
8. e ene 173 vi 6 138 PromulaFortran Translator User s Manual FIFIPOW FORTRAN INTRINSIC FUNCTION 174 FIFISHF FORTRAN INTRINSIC FUNCTION ISHFT eee ener nnns nsns nsns nnn 174 FIFISIGN FORTRAN INTRINSIC FUNCTION nsns nsns esas nsns 175 FIFMAXO FORTRAN INTRINSIC FUNCTION MA XO 175 FIFMAXI FORTRAN INTRINSIC FUNCTION MAX esas nsns nsa 175 FIFMINO FORTRAN INTRINSIC FUNCTION MINO eese e emen 176 FIFMIN1 FORTRAN INTRINSIC FUNCTION MINI nnns nsns nsns nsns 176 FIFMOD FORTRAN INTRINSIC FUNCTION enn eset enisi nsn ese nsns nsns nsu 176 FIFNEF FORTRAN INTRINSIC FUNCTION NEF eeeeeeeeeeeeee ee e en eene nnne nnne nnn nnne rennen nnn 177 FIFNINT FORTRAN INTRINSIC FUNCTION 177 FIFRAD50 FORTRAN EXTERNAL FUNCTION 50 178 FIFRBIT FORTRAN INTRINSIC FUNCTION 178 FIFSIN FORTRAN INTRINSIC FUNCTION SIN eeeeeeeeeeeeeeennee nsns nsn e enn n nsn e 179 FIFSNCS FORTRAN SINGLE PRECISION SINE COSINE eren esses nre ese esses nsns 179 FIFSTRGV FORTRAN STRING VALUE CONVERSION eee e e 179
9. SUBROUTINE ALPHA IA GVAL PARAMETER AVAL 99 PARAMETER IVAL 10 2 IA AVAL GVAL IVAL RETURN END Most compilers type AVAL as a float constant and IVAL as an integer constant Adjustments are made to the constants themselves to obtain the correct type To correspond to this practice the PROMULA FORTRAN translation of the above is as follows void alpha ia gval long ia float gval define aval 99 0 lt define ival long 10 2 lt ia aval gval ival return fundef aval undef ival Note that the constants themselves have changed type to accommodate the typing of the PARAMETER symbols The P512 flag gives the opposite interpretation as can be seen in the following translation void alpha ia gval long ia float gval define aval 99 lt define ival 10 2 lt ia aval gval ival return undef aval undef ival 70 PromulaFortran Translator User s Manual Under this translation the constants retain their original types and the PARAMETERS are assigned the types of those constants 2 27 10 P1024 Exclude undefs From the Translation Examining the translations of the preceding part brings another fundamental difference between FORTRAN and C to light In FORTRAN PARAMETERs have scope only within the subprogram in which they are defined while in C the corresponding defines have file scope Consequently whenever PARAMETERS are translated
10. The default C intermediate output for this subprogram looks as follows void demo static int ival 10 jval 10 i for i 1 i lt 10 i ival i 1 10 jval i 1 10 if i 10 WRITE OUTPUT LISTIO STRG I equals 10 11 0 if i 11 WRITE OUTPUT LISTIO STRG I equals 11 11 0 return In this translation only the simple IF statement is treated as a compound statement in C The default C output attempts to maintain a one to one correspondence between the FORTRAN original and the C Processing the above subprogram with the Yb flag produces the following C output void demo static int ival 10 jval 10 i for i 1 i lt 10 i ival i 1 10 jval i 1 10 if i 10 WRITE OUTPUT LISTIO STRG I equals 10 11 0 if i 11 WRITE OUTPUT LISTIO STRG I equals 11 11 0 return Now the block IF in FORTRAN has also been collapsed to a compound if in C The Yb flag looks for every opportunity to remove unneeded braces The DO loop is not collapsed since it contains two statements There is another switch however Ya which allows multiple assignments to be collapsed into a single statement This switch is discussed fully in another section of this manual The C output using both Ya and Yb produces the following void demo static int ival 10 jval 10 i for i 1 i 10 i ival i 1 jval i 1 10 if i 10 WRITE OUTPUT LISTIO ST
11. eee ene emen nennen 250 VMSRDI READ FORTRAN VIRTUAL SHORT INTEGER VECTOR eee ee e eene nnnm ennn nnns nnns nsns 250 VMSRDL READ FORTRAN VIRTUAL LONG INTEGER VECTOR eceeeeree ee e e e emen nnn nnns nsns nsns 251 VMSRDS READ FORTRAN VIRTUAL STRING eeeeeeeeeeeeeenen nene ene ennnnnnn nnns nsns 251 VMSRDT READ FORTRAN VIRTUAL TRUTH V ALUE VECTOR eee 252 VMSRDU READ FORTRAN VIRTUAL UNSIGNED CHARACTER VECTOR ccccccccccececececececececececeeesesesseeseeeeeeees 252 VMSSAVE SAVE A VIRTUAL e reve deed 253 VMSUSE USE VIRTUAL a suse e esa ese ese ese esas uses ese esee a 253 VMSVECT VIRTUAL VECTOR INPUT OUTPUT eee enn 254 VMSWRB WRITE FORTRAN VIRTUAL BOOLEAN VECTOR eese e eme 254 VMSWRC WRITE FORTRAN VIRTUAL CHARACTER VECTOR eeeeeeeeteee e emen nsns ese ines nsns ue 255 VMSWRD WRITE FORTRAN VIRTUAL DOUBLE PRECISION VECTOR eee eee eene emere nnne 255 VMSWRF WRITE FORTRAN VIRTUAL SINGLE PRECISION VECTOR eeeeeenee e ene 256 VMSWRI WRITE FORTRAN VIRTUAL SHORT INTEGER VECTOR eee e e en nennen nnns eren esses nsns 256 VMSWRL WRITE FORTRAN VIRTUAL LONG INTEGER VECTOR e en ee 257 VMSWRS WRITE FORTRAN VIRTUAL VECTOR OF STRINGS 22
12. eeeeeeeeeeeeeeennne e ene emen nsns nsns nsns nsns 162 FIFDMAX1 FORTRAN INTRINSIC FUNCTION DMAX1 eeeeeeeeeeeeeene e e e ene nnne nnnnnn ener nnns nsns nsns nsns nsns nsns nna 163 FIFDMIN1 FORTRAN INTRINSIC FUNCTION eene 163 FIFDMOD FORTRAN INTRINSIC FUNCTION eee e e e e enennnnnnnnn nsus nsns ases ese ase esas asa 163 FIFDNINT FORTRAN INTRINSIC FUNCTION DNINT eeeeeeeeeeeeen 164 FIFDSIGN FORTRAN INTRINSIC FUNCTION 164 FIFEQF FORTRAN INTRINSIC FUNCTION 2 165 FIFEXIT FORTRAN EXIT SUBROUTINE e ese ese ese esas 165 FIFGETAR FORTRAN GET COMMAND LINE ARGUMENTS 22 165 FIFGETCL FORTRAN GET COMMAND LINE SUBROUTINE 222 166 FIFGETENV FORTRAN GET VALUE OF ENVIRONMENT V ARIABLES eee eene ene n ene ener nnne n nnns nsns nnns 166 FIFHBIT FORTRAN HIGH BIT MANAGEMENT e e e e enne 167 FIFIZABS FORTRAN INTRINSIC FUNCTION I2ABS eeeeeeeeeeeeeeenene e e e emen nennnnnnnnn nnns nsns n inns nsn e nsns nsns 167 2 FORTRAN EXTERNAL FUNCTION DDDATE
13. INTEGER I J K PRINT I J K I J K STOP END Include files used in unit Seq Filename Symbols defined in unit Base Identifier Obj Flag Inf2 Incf Ref Normal Cref Tflags Adr Valsize 278 I 1 0 0 0 5 0 0 1 18 0 4 286 J 1 0 0 0 5 0 0 1 18 0 4 294 K 1 0 0 0 5 0 0 1 18 0 4 Pcode generated by unit Location Operation 00000 nop 00001 sto 00002 lst 00003 lsc T J K 8 00006 wry 17 00009 lda d 278 L 00012 ldr Ls 278 1 00015 lda j 286 1 00018 ldr j 286 1 00021 adi 00022 lda k 294 1 00025 ldr k 294 1 00028 dvl 00029 wrv 5 00032 wln 00033 ner 00034 nop 00035 stn 00036 nop 45 00037 PromulaFortran Translator User s Manual These tables show the detailed information generated to form the C translation A detailed discussion of the actual symbol conventions and pseudo code is beyond the scope of this manual Individuals desiring additional information may contact Great Migrations LLC 2 9 5 Annotated Listing of C Output In addition to the annotated listing of the source code a similar listing can also be obtained for the C output produced The following is the listing produced with the ET flag during the processing of the code used earlier If Line N1 Translation ppm ng RUNE app flag 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 define SPROT include for void struc6
14. PROMULA FORTRAN to C Translator User s Manual Copyright 1988 2007 Great Migrations LLC ALL RIGHTS RESERVED COPYRIGHT NOTICE for PROMULA FORTRAN Version 9 38 Released May 2007 Published by Great Migrations LLC 7453 Katesbridge Ct Dublin Ohio 43017 614 761 9816 This User s manual for PROMULA FORTRAN is the property of Great Migrations LLC It embodies proprietary confidential and trade secret information The User s manual and the files of the PROMULA FORTRAN machine readable distribution media are protected by trade secret and copyright laws The use of PROMULA FORTRAN is restricted as stipulated in the Great Migrations LLC License Agreement which came with the PROMULA FORTRAN product and which you completed and returned to Great Migrations LLC The content of the machine readable distribution media and the User s manual may not be copied reproduced disclosed transferred or reduced to any electronic machine readable or other form except as specified in the License Agreement with the express written approval of Great Migrations LLC The unauthorized copying of any of these materials is a violation of copyright and or trade secret law DISCLAIMER OF WARRANTIES AND LIMITATIONS OF LIABILITIES THIS USER S MANUAL IS PROVIDED ON AN AS IS BASIS EXCEPT FOR THE WARRANTY DESCRIBED IN THE GREAT MIGRATIONS LLC LICENSE AGREEMENT THERE ARE NO WARRANTIES EXPRESSED OR IMPLIED INCLUDING BUT NOT LIMITED TO IMPLIED WARRAN
15. rename f modxx lio model 1io 9 9 UNIX x rename f modxx lio model 1io 9 9 ENDIF rename f modxx lio model 1io 9 9 STOP NULL Now we have gone too far the other way The conditional statement will always be executed We must tell PROMULA to translate the and ENDIF into if and endif The following configuration file does this 124 PromulaFortran Translator User s Manual SWITCHES Yd PRAGMA IF ifdef ENDIF endif Syntax PRAGMA nen tn Where c isthe comment character introducing a special comment si isakeyword in the source language ti is the corresponding pattern string in the target language where the represents anything on the comment following the keyword In the PRAGMA statement above then conditional statements begin with a 4 The IF conditional is translated into ifdef conditional and ENDIF whatever is translated into endif The resultant translation of the code above is as follows void test6 extern void rename f rename f modxx lio model 1io 9 9 ifdef UNIX rename f modxx lio model 1io 9 9 endif rename f modxx lio model 1io 9 9 STOP NULL 3 7 The Configuration Statement The configuration statement simply specifies the replacement character for the dollar sign Syntax c Where c is replacement character By default all symbols in the source c
16. typedef struc long day Sdate typedef struc unsigned Stime typedef struc Sdate Sappointmen OTYPE tran h 1 month year t char hour minute E d app date Stime app time 2 char Ly STRUCTURE TIME APP_TIM E m LOGICAL 1 HOUR MINUTI END This is STRUCTURE the same as STRUCTURE TIME LOGICAL 1 HOUR MINUTE END STRUCTURE REC x extern void g static Sappoi static Sdate static long i ORD TIME APP TIME 2 et date ntment next app app list 7 today D2 WRITE 6 L for i 1 D 1 get d WRITE 6 LISTIO INTA4 today day IN 34 35 36 37 38 39 40 41 1 app 1l 1 today next app app memo 80 unsigned char ISTIO STIRG w xww STRUCG Q0UI 20 0 2 7 1 1 D2 gt 0 D2 it 1 ate amp 1 amp today ist i 1 app date today day today day 1L app list 0 OPEN 1 FI WRITE 1 B X WRITE 1 B X r4 today month INT4 today year 0 LEN STRUC6 BIN 10 STATUS UNKNOWN FORM UNFORMATTED 0 E amp next app int sizeof Sappointment 0 E app list int 7 sizeof Sappointment 0 2 10 Treatment of Syntax Errors ERO 2 4 46 PromulaFortran Translator User s Manual PROMULA FORTRAN is like any other FORTRAN compi
17. SUBROUTINE DEMO DIMENSION A 10 B 10 10 C 10 10 10 D 10 10 10 10 10 I 1 10 C I I I D I I I 4 B I I C I I 1 A I B I 3 10 CONTINUE RETURN END This example contains a variety of arrays having from 40 to 40 000 bytes The following is the default translation void demo static long i static float a 10 b 10 for i OL 1 lt 101 i 10 c 10 10 10 d 10 10 10 10 1111111 413111111111 b il i 0 1 1 a i b 2 i return On a PC platform this function would not compile because the d array is too large Compiling the same example with SD30000 produces the following result void demo static long i static float a 10 b 10 10 c 10 10 10 auto float d d float ftnalloc 40000L for 01 i 10L i c i i i dtit it it 30 10L 101 b i i clO il lil afi b 21 Iil goto ELP ELP ftnfree char d The three variables with less than 30 000 bytes have been declared static while the large array has been declared dynamic To define this dynamic array logic has been added to the beginning and exit of the function to allocate and free the storage for the d array Note that PROMULA adds this logic for you Now assume that we want everything to be dynamic To do this use SS to turn off static and SDO to make everything dynamic The following is the result void demo auto
18. ISET BVALUE return idouble long idouble a long a _BVALUE b return E0000 0 a NULL NULL long multiply a a r 90 PromulaFortran Translator User s Manual long a r return E0000 1 a r NULL long iset bvalue b long b return E0000 2 NULL NULL b This translation treats the individual entry points as if all argument values were reinitialized at each entry point It is correct for most FORTRANS but is not correct for the assumptions made in the example above Using the Y1 flag produces the following void main argc argv int argc char argv extern void iset bvalue extern long idouble multiply a static long K1 10 static long K2 5 static long K3 2 static long K4 4 ftnini argc argv NULL iset bvalue amp K1 WRITE OUTPUT LISTIO INT4 idouble amp K2 INTA4 multiply a amp K3 amp K4 0 static long EO0a EOr E0b static long EO IENTRY int IENTRY static long idouble switch IENTRY case 0 goto IDOUBLE case 1 goto MULTIPLY_A case 2 goto ISET_BVALUE IDOUBLE idouble E0a EOb return idouble MULTIPLY_A EOa 0 idouble E0a EOb ISET BVALUE return idouble long idouble a long a
19. eene e e e en nn nnenennnnnnn nsns nsn use uses 215 FIOWHEXO WRITE HEXADECIMAL OR OCTAL CONSTANT eee e ene ene esu uses 216 FIOWLN WRITE FORTRAN END OF LINE eee e n ene nnnm nn nnnm nn nnns unns ise ese esa n nsa ese nsn aa sns u 216 FIOWNL PROCESS FORTRAN WRITE DATALIST STATEMENT eeeeeeee eene 216 FIOWRB WRITE FORTRAN BOOLEAN 217 FIOWRC WRITE FORTRAN CHARACTER 218 FIOWRD WRITE FORTRAN DOUBLE PRECISION VECTOR eee ere e ene 218 FIOWRF WRITE FORTRAN SINGLE PRECISION VECTOR eeeeeeeeeetene eee e eme nne ener ener nnns es eses i nsn nsn esas n sns 219 FIOWRI WRITE FORTRAN SHORT INTEGER VECTOR eeeeeenee e 219 FIOWRL WRITE FORTRAN LONG INTEGER 220 FIOWRS WRITE FORTRAN VECTOR OF STRINGS eeeeeeeeeeeeeene 220 FIOWRT WRITE FORTRAN TRUTH VALUE VECTOR ececeeeeee 221 FIOWRU WRITE FORTRAN UNSIGNED CHAR VECTOR emen rn en nnnn nnns 221 FIOWRX WRITE FORTRAN COMPLEX VECTOR eceeeeeeeenn e e emen nnnnnnnnnn nnns nn n ise e ese uses esi nsa e nsns esas esu 222 FIOWTAXCE SWRITE TEXT RECORD 222 FIOWVAL W
20. note that uses zero based subscripts This translation is relatively clean however it does introduce an additional pointer variable This translation is selected as the default because the symbol remains defined thus making debugging simpler An alternative translation can be produced via the P4096 flag This translation is shown below void main argc argv int argc char argv static long ia 10 define ib 3 ftnini argc argv NULL READ INPUT LISTIO DO 10 INT4 ia 0 WRITE OUTPUT LISTIO DO 2 INT4 ib 0 undef ib Notice that IB is now implemented with a define It is not an actual memory location in the final code Alternatively repeated uses of IB might generate inefficient code if its value must be computed upon each use 2 27 13 P8192 Use Parameter Identifiers in Equivalences As discussed in the section on the P4096 flag the processing of FORTRAN EQUIVALENCE into C is difficult and clumsy at best To complicate the situation PARAMETER values are often used within EQUIVALENCE declarations The following is the equivalent of the code presented above using PARAMETERs As ROGRAM DEMO PARAMETER IAOFF 5 INTEGER IA 10 IB 2 EQUIVALENCE IA IAOFF IB IAOFF 3 READ IA PRINT IB END In the default translation to ensure that the positions are correctly computed the values of the parameters are used in the C produced
21. return void demo3 i j 30 PromulaFortran Translator User s Manual int Es extern int jindex static char name 12 static int initial static string T1 NULL 0 j fifindex name 12 amp initial 4 1 name Tl n 12 i jindex amp T1 amp initial return M As with CHd this version at least keeps the argument positions correct 2 5 3 Substrings Once the CHd or CHv options are selected the relation between character variables and their character lengths becomes much more dynamic than with the other two approaches As a result various other minor problems associated with FORTRAN character management can be solved Consider the following code fragment SUBROUTINE DEMO1 CHARACTER NAME 12 INITIAL 4 INTEGER I J INITIAL NAME 5 8 INITIAL NAME I J INITIAL NAME INDEX NAME INITIAL J RETURN END In this example a substring of the name variable is being stored in the initial variable In the third statement a complicated expression is used to compute the start of the substring Under CHr or CHs the C looks as follows void demol static int i j static char name 12 initial 4 ftnsac initial 4 namet 4 4 ftnsac initial 4 1 1 1 1 ftnsac initial 4 name fifindex name 12 initial 4 1 j fifindex name 12 initial 4 1 return The basic problem is that to use a substrin
22. trlito amp blank bytes amp K1 blanks xlit yet fifi4log long TRUE return modi This sample is an actual fragment We did not create it to make this feature look good If the source has a consistent commenting style then an equivalent one for the target language can be created using this statement It is of course the user s responsibility to ensure that proper comments are being written 3 3 The Configuration PATHNAMES Statement A major problem faced in the migration of source codes from one environment to a new environment has to do with the accessing of include files There are three problem areas 1 Source pathnames often include dollar signs and possibly other characters such as slashes which cannot be easily used under target systems 2 The directory separation characters are not the same on different systems 3 The organization and search rules used on the source system might not be the ones needed for the new environment These problems are dealt with explicitly in PROMULA via the PATHNAMES statement in the configuration file Though it may not be feasible for all users to use the same PATHNAMES specifications no user need physically make any source changes to the INCLUDE or INSERT statements in his source codes This section describes how a particular pathname translation scheme can be established for a given set of conventions 112 PromulaFortran Translator User s Manual The PATHNAME st
23. 2 51 Location of FORTRAN Files to be Included Zname It is often desirable to have include files that are referenced from within FORTRAN programs stored in some other directories The Zname flag can be used to specify directories to be searched for include files The syntax of name varies from operating system to operating system but should be the same as the syntax used to construct the PATH variable used to locate the executable for PROMULA FORTRAN itself If the first character of name is a then the remainder of name is assumed to be the name of an environment variable which contains the include file search paths 2 52 Project Processing project In addition to processing independent single source files it is often desirable to process groups of related multiple files as a single project This is done via a project file which has the following syntax PROJECT project name SOURCE filel for 105 PromulaFortran Translator User s Manual file2 for fileN for where project name isthe name of the project fileI for are the names of the files that make up the project To process a project the following command line is used pfc project options where project 15 name of the project file If no extension is provided pr 3 is assumed options contains any other command line switches to be used The special character differentiates a project file from a single source file To avoid conflicts with other us
24. NU Format string location Current format position char fiofrmt NULL Current format address list int fiofstat 0 Current processing state int fioifmt 0 Number of current format entry int fioiresc 0 Number of current rescan entry int fiomaxc 0 Maximum characters in output line int fionfmt 0 Number of entries in format list int fionpren 0 Parenthetical nesting char fiorscan NULL Current rescan position int fiorscnt 0 Rescan count value char fiobfmt 40 Business formatting string char fiovfmt NULL Start of variable format int fiovflen 0 Length of variable format int fiofini fmt nfmt char fmt Pointer to the format or format pointers int nfmt Number of format pointers or type Description This function initializes the FORMAT environment needed by the FORTRAN style input output statements There is a minor problem associated with the processing of FORTRAN style FORMATS in C having to do with maximum string length In theory since a C string consists of a sequence of nonzero characters terminated by a zero byte it can be of any 188 PromulaFortran Translator User s Manual length desired Unfortunately all C compilers place a limit on the maximum length that a string constant may have typically 256 It is not at all unusual for FORMAT specification strings to be very long and they are typically defined as constants at compile time Fortunately though length of an individual stri
25. Size of compacted statement storage Size of the line number table Size of a data block Size of external information storage Size of include file information storage Word size of source platform Specify a configuration file Auto storage threshold 11 PromulaFortran Translator User s Manual Switch Options Section Characteristic effected by switch SD num 2 31 Dynamic storage threshold SS num 2 31 Static storage threshold SV num 2 31 Virtual storage threshold SZ num 2 31 Dynamic virtual threshold T 0 1 2 2 32 FORTRAN dialect DO loop assumption T 5 2 33 Treatment of internally generated temporaries UP num 2 34 PUNCH statement unit number UR num 2 34 READ or ACCEPT statement unit number UW num 2 34 WRITE or PRINT statement unit number W name 2 35 Name of file to receive prototype definitions Y 1 2 2 36 Miscellaneous control flags X Y a 2 37 Treatment of multiple assignments convention X Y b 2 38 Treatment of single statement nesting brace X Y 2 39 Constants reduction optimization X Y ch 2 40 Character optimization X Y d 2 41 Treatment of FORTRAN D debugging statements X Y f 2 42 Use of printf style formatting X Y 1 2 43 Initialization check for auto variables X Y 1 2 44 DO loop counter reduction optimization X Y p 2 45 Subprogram argument type checking X Y T 2 46 Single precision real arithmetic X Y S SV Sf 2 47 Subscript pointer notation X Y u 2 48 Unformatted write optimization X Y 2 49 Subprogram call
26. TERMINATION tc sname 1 tname 1 sname n tname n END Where dirchar is the directory separation character in the source pathname sltls2t2 a sequence of character pairs tname is a target language pathname or pathname prefix sname is a source language pathname or pathname prefix tc is a pathname termination character used in the source pathname The dirchar specification gives the character used to separate the pathname components in the original source codes By default this character is replaced by the appropriate separation character in the target pathnames The required dirchar specification specifies the character used to separate the pathname components in the original source codes This character is replaced by an equivalent character in the target pathnames For example moving from PRIME FORTRAN to UNIX lt character would be replaced by a character Moving from MS DOS to UNIX would replace a character with a character The optional REPLACE parameter specifies additional characters to be replaced in the source names As many pairs of characters as are needed may be included The standard PRIME language description for example contains the following specification for this option REPLACE _ This causes all dollar signs in the source pathnames to be replaced by underscores The mutually exclusive and optional UPPER LOWER parameters specify that all alphabetic characters in pathnames should be conv
27. Write FORTRAN String Value fiowvt Write FORTRAN Truth Value fiowvu Write FORTRAN Character Value fiowvx Write FORTRAN Complex Value fiovfini Initialize A Variable FORTRAN Format 6 209 FTNXCONS FORTRAN Exact Representation Constant Synopsis include fortran h PROMULA FORTRAN function declarations char ftnxcons base xcons xlen vlen int base Base of the exact representation constant char xcons Exact representation constant in string form int xlen Length of exact representation string int vlen Length of the resultant value Description Octal and hexadecimal character strings are physically stored as character strings and are treated as typeless numeric constants This function performs this conversion The parameter base specifies the base of the exact representation constant It may have a value of 8 or 16 only The xcons parameter points to the actual string for the constant The x1en parameter specifies the length of the string while the v1en parameter specifies the length of the resultant value It has a maximum setting or 32 Return value A pointer to the result of the conversion This pointer must always be cast a pointer to the type of the desired value 241 PromulaFortran Translator User s Manual See also fifibit Insert a bit fifrbit Reverse bit byte order fifxbit Extract a bit 6 210 P77GETU Prime FORTRAN 77 Function F77 GETU Synopsis finclude fortran h Define the platform
28. amp i amp j amp k In one place demo is called with a pointer to a long and in the other with a pointer to a float Alternatively with Yp active the following is produced 3 test for E247 The argument of type integer 4 has been entered where a call by reference argument of type real 4 is required Alternatively if the error checking level is set to 1 or above with the EL switch the following warning is produced without use of the Yp switch 3 test for W814 The argument of type integer 4 has been ntered wher an argument of type real 4 has been used 2 46 Single Precision Real Arithmetic Xr Yr Though the designers of C made a few strange decisions their only real mistake was to omit single precision real arithmetic Even now the major barrier to the acceptance of PROMULA as a pure compiler is that its benchmarks on single precision arithmetic programs are often slow Fortunately the newer ANSI C compilers do support single precision real arithmetic Unfortunately they require that single precision real constants have an F appended to distinguish them from double precision constants Most older C compilers consider this F to be a syntax error therefore you must tell PROMULA that your C compiler is capable of performing single precision real arithmetic The default Xr switch assumes that all single precision real arithmetic will be performed as double precision and that all single precision reals w
29. fiobfout Perform business format output fiodtos Convert floating point number 223 PromulaFortran Translator User s Manual fiorndv Round floating point value display fioshr Shift string right fiowtxt Write text record 6 176 FIOWVB Write FORTRAN Boolean Value Synopsis include fortran h PROMULA FORTRAN function declarations int fiowvb val unsigned short val Boolean value to be written Description Writes a Boolean short logical value to the current output file in accordance with the current format specification This function corresponds to the FORTRAN 77 convention in which values and not just l values can be written Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowbiv Write binary values fiowrb Write Boolean value vector 6 177 FIOWVC Write FORTRAN Character Value Synopsis finclude fortran h PROMULA FORTRAN function declarations int fiowvc c char c Character value to be written Description Writes a single character value to the current output file in accordance with the current format specification This function corresponds to the FORTRAN 77 convention in which values and not l values can be written Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing
30. in a static variable Multiple occurrences of the same constant use the same constant storage point thus the constant ends up behaving just like a variable There are many ways around this problem assigning the value to the variable K1 for example however most FORTRAN compilers do the equivalent of the above Though presumably no one would write code such as the preceding deliberately such code sequences do occur This is one of those gray areas of FORTRAN that can cause problems that appear when any sort of migration is attempted Programs that seem to work great under one FORTRAN implementation suddenly malfunction under another 2 18 Maximum Output Line Width Lnum A minor issue has to do with how long to make the C output when very long statements need to be written The default setting is 80 characters wide the width of a typical editor window If desired you may select any value This flag effects the look of the output only A setting of zero specifies that all output associated with a given executable source statement should be placed on a single line This setting is used with the Ln switch which places line number directives in the C output listing Note that some C compilers cannot process very long input lines For these use Lmax where max is the longest line width which can be accepted 2 19 Link Time Processing of COMMON Data Modules Lm Ls There is a real problem in processing C via FORTRAN having to do with the
31. static float a 10 b 7 6 c 5 4 3 d 5 4 3 2 static long i j k 1 a 2 0 3 1 4 3 1 01 1 1 j 1 1 2 0 102 PromulaFortran Translator User s Manual 1 6 d 1 1 k 1 1 i 1 7 Notice first in this example that though both the FORTRAN source and the Xs switch translation show multidimensional arrays the order of the array sizes is reversed in the translation This is necessary to maintain the same relative positioning of the elements within the arrays In the Ys translation arrays are all shown with a single dimension equal to the total size of each array Thus D shows 120 values which is BKB Wu Xx In both the translations the actual subscript values are reduced by one to reflect the fact that C subscripts begin at zero while FORTRAN begins at one In the Ys case the subscript array references are translated using pointer notation with the calculation reduced to its simplest form Thus C 3 4 5 becomes c 59 Also subscript calculations are performed using the FORTRAN convention This ensures that any games played in the programs with equivalencing or with varying COMMON layouts will work exactly as they worked in FORTRAN Using a C compiler which does constant reduction one would expect to have the same code produced by both of the above translations To demonstrate the subscript simplification process the listing below shows the same Ys bias translation but wi
32. 0 Ends the list of parameters 1 long Points to an error return variable 2 string Points to the file name 3 char Points to the status description 233 PromulaFortran Translator User s Manual 4 char Points to access type description 5 char Points to the form description 6 int Defines the record length 7 char Points to the blanks treatment description 8 Indicates that the file is read only 9 Indicates that the file is to be shared 10 char Points to the record type description Return value A zero if the open went well else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiofdata Establish file data fioerror Perform FORTRAN I O Error Processing fiolun Establish unit number fiointu Establish an internal file fioopen Open current file fiostatus Set I O Error Status 6 197 FTNPAUSE FORTRAN Pause Statement Synopsis include fortran h PROMULA FORTRAN function declarations void ftnpause message nmes char message Message int nmes Description Executes the pause operation of the FORTRAN pause statement This function assumes that some message has already been written to stdout In particular it writes the message PAUSE message to console and then waits for any new record from stdin If stdin and stdout are redirected then this function may not produce the desired result Return value None
33. 5 0 b 2 1 6 0 c 2 1 0 7 0 return The D statement is literally transformed into a comment in the C Using the Yd flag however produces the following intermediate C output void demo a b c float a b 7 6 c 5 41 3 WRITE OUTPUT LISTIO STRG Demo successfully entered 25 0 a 4 5 0 b 2 1 6 0 2 1 0 7 0 return In this treatment the statement is compiled as though it had been a normal statement 2 41 1 Treatment of Other Debugging Statements Ydstring The D debugging feature though it is very common among FORTRAN dialects is not part of the standard therefore it is fair game for alternative implementations The YSstring statement allows alternative ways of identifying debugging statements Consider the following FORTRAN code SUBROUTINE DEMO A B C DIMENSION 10 6 7 3 4 5 WRITE Demo successfully entered A 5 5 B 2 3 C1273 RETURN END l anor Note that the is used to shift the statement left as well commenting it out The default translation for this code is as follows void demo a b c float a b 7 6 c 5 41 3 WRITE Demo successfully entered 97 PromulaFortran Translator User s Manual 6 b 2 1 0 c 2 1 0 7 return Using the YD switch produces the following translation void demo a b c float a b 7 6 c 5 41
34. FIFSYSTM FORTRAN EXTERNAL FUNCTION SYSTEM eeeeeeeeeeeene nennen nsns 180 FIFTAN FORTRAN INTRINSIC FUNCTION enisi ases 180 FIFTIME FORTRAN EXTERNAL FUNCTION 181 FIFXBIT FORTRAN INTRINSIC FUNCTION 181 FIFXCREP FORTRAN EXTENDED CHARACTER REPRESENTATION 2 181 FIOBACK BACKSPACE A FORTRAN FILE eeeeeeeceeeeennne n e e eme n en nnnn enhn nn nn nnns nnns nsns e enis a ese usns 182 FIOBFOUT BUSINESS FORMAT OUTPUT ve desmorona e enn ener ener enhn nnn nsns nsn usus e sa ases e ese a ese se esa sese e 182 FIOCLOSE CLOSE CURRENT FORTRAN e nsi e ese assa ese e ananas a 184 FIOCPATH CONVERT PATHNAME eene e e e e emen 184 FIODTOS CONVERT DOUBLE VALUE TO STRING eeeeeeeeeeeeeeene e ese ese ese nsns esas 185 FIOERROR PERFORM FORTRAN I O ERROR PROCESSING eee e e ene enn nsu 185 FIOFDATA FORTRAN FILE DATA ee e en nnnm nmnmemnnnnnn nnn nn nn nnns nnn a ese e asas nsu 186 FIOFEND END FORMAT PROCESSING eeeeeeeeeene e e ene emen tetett erett e e ese ese ese a eset 187 FIOFFLD GET NEX
35. INT4 v 0 return Since many are not familiar with this notation the expression D2 D2 e2 v D1 D1 gt 0 D2 1 has the following value Using D2 as a temporary variable set it equal to the value of the iteration count as before If that temporary value is greater than 0 set D2 equal to it else set D2 equal to 1 Clearly this is the desired value Note that other approaches introduce an additional temporary variable which is set true initially and false on increment which forces execution of the loop at least once Clearly the approach here is cleaner Of course the same logic as above holds for loops with floating point controls The following is an example SUBROUTINE DEMO DO 10 V El E2 WRITE V 10 CONTINUE RETURN This is the same as before but 1 2 and v are all floating point The translation is as follows void demo static int D2 static float v el e2 e3 D1 for v el D1 e3 D2 e2 v D1 D1 D2 gt 0 D2 v D1 WRITE OUTPUT LISTIO REAL4 v 0 return 84 PromulaFortran Translator User s Manual The result is almost identical except that D2 the iteration count is typed as a fixed point value as is required by the specification On the other hand D1 which contains the increment is the same type as the iteration variable The above general translation is needed only if one o
36. If fmt is NULL and n mt is zero then an unformatted binary type operation is being performed If mt is NULL and nfmt is nonzero then a free form formatted operation is being performed If mt is not NULL and n mt is zero then a single string format is being processed else the specification consists of a sequence of n mt lines Once this function is called the format specification string remains in effect until either this function is called again or function iofend is called Return value A zero if the format appears well formed else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Performs standard error processing fiofwsp Skips over white space in the format 6 112 FIOFINP Formatted Input Synopsis include fortran h PROMULA FORTRAN function declarations void fiofinp context int context Context of call 0 value 1 end 189 PromulaFortran Translator User s Manual Description Performs nonvariable related formatted input functions until an end of format or a variable related specification is encountered The actual operations performed by this function are as follows Specification Code Description nH 1 Display Hollerith string cl cn 1 Display delimited string nX 2 Skip right n places TRn 2 Skip right n places Tn 3 Move to position n TLn 4 Skip left n places SS 5 Set the plus sign to a space SP 6
37. Message to be displayed at console i e stderr Description Exits to the operating system Return value None the function exits to the operating system and never returns See also None 6 60 FIFGETAR FORTRAN Get Command Line Arguments Synopsis finclude fortran h PROMULA FORTRAN function declarations void fifgetarg k cl ncl long k cl int ncl 165 PromulaFortran Translator User s Manual Description Copies the kth command line argument into a string variable arg The oth argument is the command name Return value None the function is void See also None 6 61 FIFGETCL FORTRAN Get Command Line Subroutine Synopsis include fortran h PROMULA FORTRAN function declarations void fifgetcl cl ncl char cl int ncl Description Copies the command line parameters as entered by the user into a string variable The individual parameters are separated by spaces Return value None the function is void See also None 6 62 FIFGETENV FORTRAN Get Value of Environment Variables Synopsis include fortran h PROMULA FORTRAN function declarations void fifgetenv ename lenin evalue lenout char ename Name of environment variable int lenin Length of name char evalue Returns value of environment variable int lenout Length available for variable value Description Searches the environment list for a string of the form ename value and returns value in evalue if such string i
38. None 6 69 FIFI2MINO FORTRAN Intrinsic Function IZMINO Synopsis include fortran h PROMULA FORTRAN function declarations short fifi2minO0 al a2 short al First value to be compared short a2 Second value to be compared Description Determines which of two short integer values is the smallest Return value The smallest short integer value See also None 6 70 FIFIZMOD FORTRAN Intrinsic Function IZMOD Synopsis include fortran h PROMULA FORTRAN function declarations short fifi2mod num dem short num The numerator for the calculation short dem The denominator for the calculation Description Computes the value of the remainder of num divided by dem If dem is zero then the result is zero Return value The value as computed above See also None 169 PromulaFortran Translator User s Manual 6 71 FIFIZNINT FORTRAN Intrinsic Function I2NINT Synopsis finclude fortran h PROMULA FORTRAN function declarations short fifi2nint a double a Argument containing the value Description Computes the nearest short integer to its double precision argument In particular fifnint a fifidint a40 5 if a gt fifidint a 0 5 if a lt where fifidint is the FORTRAN intrinsic function IDINT Return value The value as computed above See also fifidint FORTRAN intrinsic function IDINT 6 72 FIFIZPOW FORTRAN Intrinsic Function IZPOW Synopsis include fortran h PROMULA FORTRAN func
39. PROMULA FORTRAN function declarations double fifdnint a double a Argument containing the value Description Computes the nearest integer to its double precision argument In particular fifdnint a fifdint a 0 5 if a gt 0 0 fifdint a 0 5 if a lt 0 0 where fifdint is the FORTRAN intrinsic function DINT Return value The value as computed above See also fifdint FORTRAN intrinsic function DINT 6 57 FIFDSIGN FORTRAN Intrinsic Function DSIGN Synopsis include fortran h PROMULA FORTRAN function declarations double fifdsign mag sgn double mag The magnitude for the result double sgn The sign for the result Description Returns a value after the transfer of a sign The result is if sgn is at least zero else it 15 mag Return value 164 PromulaFortran Translator User s Manual The value as computed above See also None 6 58 FIFEQF FORTRAN Intrinsic Function EQF Synopsis finclude fortran h PROMULA FORTRAN function declarations int fifeqf a b float a First value to be compared float b Second value to be compared Description Compares two floating point numbers to determine if they would be equal in single precision floating point form Return value The short integer result of the comparison See also None 6 59 FIFEXIT FORTRAN Exit Subroutine Synopsis finclude fortran h PROMULA FORTRAN function declarations void fifexit numb long numb
40. Perform FORTRAN I O error processing fiofinp Get next formatted input specification fiostod Convert string to double fioffld Get next free form input field fioralph Read alphabetic information fiorchk Check current fixed input field 6 134 FIORDL Read FORTRAN Long Integer Vector Synopsis include fortran h PROMULA FORTRAN function declarations int fiordl value nval long value Vector of values int nval Number of values to be read Description Reads a vector of long fixed point values from the current input file in accordance with the current format specification Each value is assumed to have its own corresponding format specification if a formatted read is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Perform FORTRAN I O error processing fiofinp Get next formatted input specification fiostod Convert string to double fioffld Get next free form input field fioralph Read alphabetic information fiorchk Check current fixed input field 6 135 FIORDS Read FORTRAN String Synopsis finclude fortran h PROMULA FORTRAN function declarations int fiords str nstring nval char str Points to start of strings int nstring Length of each string int nval Number of strings to be read Description 202 PromulaFortran Translato
41. QHNUMB TI 2 29 1 QInumb Size of Compacted Statement Storage eese eene nennen eene tnter nennen 78 2 29 2 QEnumb Size of the Line Number Table cesses eene nennen nennen ren eene ennt 78 2 29 3 ODnumb Size of a Data Block et erae rare e iere iere erre 78 2 29 4 QXnumb Size of External Information Storage eese eee neret 78 2 29 5 QHnumb Size of Include File Information Storage eese enne 79 2 29 6 QWnumb Word Size of Source Platform eese eene trennen eene ener tenete nennen 79 2 30 SPECIFY A CONFIGURATION FILE RNAME cccsessccccececeesssaececececeessaascecececeeseaascesececeesesasaesececeeseuaaeseeececeessaeeeeees 79 2 31 STORAGE THRESHOLD VALUES SANUM SDNUM SSNUM SVNUM SZNUM isses enne 79 2 32 FORTRAN DIALECT DOLOOP ASSUMPTIONS TO T1 2 nnn 83 2 33 TREATMENT OF INTERNALLY GENERATED TEMPORARIES TA TS 87 2 34 SPECIFYING UNIT NUMBERS UR URNUM UP UPNUM UW UWNUM ccc ccccccceesessceceeececsenseaeceeececsensaaseeeees 89 2 35 FILE TO RECEIVE PROTOTYPE DEFINITIONS 89 2 36 MISCELLANEOUS CONTROL FLAGS Y1 Y2 cccccssssccccccecsessaececceecsessaececececsessnaececececceseseaecececseseseaeceeececeeensaeeeeees 89 2 36 1 The Treatment of En
42. Short Integer Results 13d 13d 13d n isum iprod iquot fprintf LUN 6 Long Integer Results 131d 131d 131ld n jsum jprod jquot fprintf LUN 6 Short Real Results 13 5f 13 5f 13 5f n rsum rprod rquot fprintf LUN 6 Long Real Results 13 5 13 5 13 5fMXn dsum dprod dquot exit 0 Note that in each case the left hand argument is converted to long thus causing the entire expression to be evaluated in that manner The result of running this version is shown below 19 Short Short Long Integer Results Long Real Results PromulaFortran Translator User s Manual 15536 50000 50000 00000 50000 00000 Integer Results Real Results 5536 60000 60000 00000 60000 00000 Note that it agrees with the universal promotion to long result shown above 166 166 166 00000 166 00000 The simpler alternative is shown below No conversions to long are made Thus overflows occur in every expression void main argc argv int argc char argv static int il i2 i3 i4 isum iprod iquot static double dsum dprod dquot static long jsum jprod jquot static float rsum rprod rquot ftnini argc argv il 20000 i2 30000 i3 200 i4 300 isum 11 12 iprod i3 i4 iquot il i2 i4 jsum 11 12 jprod i3 i4 jquot 11 12 14 rsum 11 12 rprod i3 i4 rquot il i2 i4 dsum 11 12 dprod 13 14 dquot il i2 i4 ftnopen 6 FILEN EX002
43. The address and length can become an ordered pair when the character information is passed to a subprogram 2 The character lengths can be saved and passed together at the end of each call 3 A separate entity termed a descriptor can be formed which contains the address and length The second approach above is the default one taken by PROMULA FORTRAN It is the most common approach used by other FORTRAN processors The CHr CHaracter raw command line switch tells PROMULA FORTRAN simply to translate FORTRAN character variables into C arrays of char When this switch is active references to variable length character strings in subprograms cause a syntax error The CHs CHaracter string command line switch tells PROMULA FORTRAN to treat all references to character variables as ordered pairs a pointer and a length These two values are always shown side by side This allows full use of variable length character strings but makes the translations more complicated and difficult to maintain The CHd CHaracter dynamic switch which is the default is a more complicated approach than the one above Though CHs is able to deal with all uses of variable length character strings it is not able to deal with arbitrary mixtures of character and non character subprogram arguments Using CHd character string lengths are collected at the back of subprogram calls rather than being paired with their sources directly In addition string concatenations
44. This can be seen in the following default translation of the above void main argc argv int argc char argv define iaoff 5 static int ia 10 static int ib iat3 lt ftnini argc argv NULL READ INPUT LISTIO DO 10 INT4 ia 0 WRITE OUTPUT LISTIO DO 2 INT4 ib 0 undef iaoff 74 PromulaFortran Translator User s Manual The pointer expression for the IB variable is the same as when constants were used In most cases the actual parameters forms can also be used The P8192 flag forces this use Using P8192 the following is produced void main argc argv int argc char argv define iaoff 5 static int ia 10 static int ib iatiaoff 2 ftnini argc argv NULL READ INPUT LISTIO DO 10 INT4 ia 0 WRITE OUTPUT LISTIO DO 2 INT4 ib 0 undef iaoff lt This form now uses the defined value for IAOFF however a different computation must be used 2 27 14 P16384 Display Include Files Separately Another issue which must be dealt with in translation has to do with include files Ideally FORTRAN include files should be translated into C include files The problems involved here are two fold First FORTRAN and C have very different scope rules In FORTRAN include files are typically included within subprograms whereas in C they are included at the front of files The second problem is that it is often necessary to reorder the symbols
45. b 7 6 c 5 41 3 The subtraction by one is now explicit in the C output This reduction process is even more obvious when pointer notation is used for array references The Ys flag discussed in another section of this chapter generates pointer style subscript expressions Using this flag alone produces the following void demo a b c float a b oe at4 5 0 b 13 6 c 27 7 return The values of 4 13 and 27 are now computed from fairly complex expressions which can be seen in the following produced using Xc and Ys void demo a b c float a b c a 5 1 5 0 95 PromulaFortran Translator User s Manual b 2 14 1 19 return t 3 1 6 F 2 1 3 1 0 6 0 4 3 It is suggested that the Yc flag always be left active however if you want to see where some values such as those shown above are coming from then Xc is the appropriate flag to use 2 40 Character Optimization Switches Xch Ych The FORTRAN CHAR and ICHAR functions convert between character representations and their integer values In most cases no actual function reference is needed simple assignments are sufficient The default Ych switch removes these unneeded calls to these functions Consider the following FORTRAN code SUBROU CHARAC I C END INE DEMO IV ER 1 C ICHAR 8 CHAR 56 with references to b
46. gt hd_numtab 0 WRITE 1 LISTIO INT4 modelhead gt hd_filrev INT4 modelhead gt hd_sysprim 0 return void test5 num INTEGER4 num rev num num rev 2L return char modelhead_area 16 3 5 3 Pattern String for External Functions Many UNIX FORTRAN compilers AIX and SUN OS for example append an underscore to external symbols This effect can be achieved via PROMULA by changing pattern string 172 The following configuration file KEYWORDS 172 END o ow produces the following translation Notice the underscores appended to the function symbols 121 PromulaFortran Translator User s Manual void test4 heading P1 char heading int P1 extern void test5 extern char Xmodelhead typedef struct long hd_hedsiz Length of header long hd_numtab No of tables in file long hd_filrev Model file revision date long hd_sysprim No of system primitive types Cmodelhead auto Cmodelhead Tmodelhead Cmodelhead Xmodelhead test5 amp Tmodelhead hd numtab amp Tmodelhead hd filrev Tmodelhead hd hedsiz Tmodelhead hd numtab Tmodelhead hd filrev WRITE 1 LISTIO STRG heading 10 0 WRITE 1 LISTIO INT4 Tmodelhead hd hedsiz INT4 Tmodelhead hd numtab 0 WRITE 1 LISTIO INT4 Tmodelhead gt hd_filrev INT4 Tmodelhead gt hd_sysprim 0 void test5 num rev lon
47. indicates that the calling function intends to change the information at the byte and or beyond Return value A pointer to the byte at the requested linear address See also vmsrbl Remove virtual block from chains vmswvb Physically write block to file 6 223 VMSRBL Remove Virtual Block Synopsis include fortran h PROMULA FORTRAN function declarations void vmsrbl 1 int vlb Block to be removed Description This function removes the indicated block from the most and least recently used chains so that the block can become the currently most recently used block Return value None the function is void See also None 6 224 VMSRDB Read FORTRAN Virtual Boolean Vector Synopsis include fortran h PROMULA FORTRAN function declarations int vmsrdb value nval long value Virtual address of values int nval The number of values to be read Description 248 PromulaFortran Translator User s Manual Reads a virtual vector of Boolean short logical values from the current input file in accordance with the current format specification Note that in this function each individual Boolean value is assumed to have its own corresponding format specification if a formatted read is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiordb Read FORTRAN Boolean
48. nfield char alpha Character string int nalpha Length of character string int nfield Width of field Description Write a character string left justified blank filled to the right into a fixed width field If the character string is longer than the field then only the left most characters are written Return value None the function is void The effect of its processing is reflected in the various global format control variables See also fiowhexo Write Hexadecimal or octal constant 6 158 FIOWBIV FORTRAN Write Binary Values Synopsis finclude fortran h PROMULA FORTRAN function declarations int fiowbiv value nvalue void value Points to values being written int nvalue Number of bytes to be written 214 PromulaFortran Translator User s Manual Description This function writes binary values to a file Return value A zero if all went well else an error code See the general discussion of the FORTRAN I O capabilities for a listing of the possible error codes See also None 6 159 FIOWDBL Write Double Precision Value Synopsis include fortran h PROMULA FORTRAN function declarations void fiowdbl value nsigdig double value Value to be converted int nsigdig Number of significant digits Description Converts a double precision value to free form display form and stores it at the current position in the coded communications record The parameter value contains the value to be converted whil
49. 1 STRUC6 INC 43 PromulaFortran Translator User s Manual Symbols referenced in SUBROUTINE STRUC6 Identifier Object Storage Type Comment APP DATE record APPOINTMEN DATE scalar APP_FLAG variable APPOINTMEN logical 1 scalar APP LIST record static APPOINTMENT ld array 679 APP_MEMO variable APPOINTMEN character 20 1d array 80 APP_TIME record APPOINTMEN IME ld array 4 APPOINTMENT structure DATE Structure DAY variable DATE integer 4 scalar GET_DATE subroutine 2 args integer 4 unspecified HOUR variable IME logical 1 scalar I variable static integer 4 scalar INUTE variable IME logical 1 scalar ONTH variable DATE integer 4 scalar EXT_APP record static APPOINTMENT scalar IME structure TODAY record static DATE scalar YEAR variable DATE integer 4 scalar Symbol references by line number in SUBROUTINE STRUC6 Identifier Line If u D defined M modified U used P passed APP_DATE 22 01 D 7 00 U APP FLAG 27 01 D APP LIST 29 01 D 7 00 M 11 00 U 14 00 U APP MEMO 26 01 D APP TIME 24 01 D APPOIN ENT 21 01 D DATE 1 01 D DAY 2 01 D 6 00 U 8 00 U GET DATE 5 00 U HOUR 18 01 D I 4 00 5 00 P 7 00 U INUTE 18 01 D ONTH 2 01 D 6 00 U EXT_APP 29 01 D 11 00 13 00 U IME 30 00 D 5 005 P 6 0
50. 129 PromulaFortran Translator User s Manual In the prototype the name as it appears in the FORTRAN comes first followed by the standard prototype information The decision to change these names is made entirely in the prototype file no changes are needed in the actual FORTRAN The C output looks as follows void demo extern void ftread extern void ftwrite static long a b 10 c 20 ftread 1 amp a 1 ftwrite 2 amp a 1 ftread 1 b 10 ftwrite 2 b 10 ftread 1 c 20 ftwrite 2 c 20 The old names have been completely replaced as a result of the prototype specification 4 4 Multiple Forms Another problem that comes up has to do with multiple forms Here a single function in FORTRAN has been translated in different ways either because of some weak typing convention or because the function is to be used in both virtual and non virtual mode As an example consider that you have a statistical analysis function which computes the mean and variance of a vector of values You have translated it twice once using virtual conventions and once using memory conventions Let us first see how these two versions of the following utility can be produced SUBROUTINE ANADAT VAL N XBAR VAR DIMENSION VAL N XBAR 0 0 VAR 0 0 DO 10 J 1 XBAR XBAR VAL J 10 CONTINUE XBAR XBAR DO 15 J 1 S VAL J XBAR VAR VAR 5 5 15 CONTINUE VAR VAR N 1 RETURN ND 52 T
51. 152 FIOSTIO Establish FORTRAN Standard I O Synopsis include fortran h PROMULA FORTRAN function declarations 211 PromulaFortran Translator User s Manual int fiostio action int action Action code for subsequent use Description Before any action can be performed on a standard I O file it must be associated with an existing FORTRAN file structure If there is no already existing structure for the standard unit then this function will attempt to create one The type of standard unit to be used is determined by the action code as follows Code Standard unit 1 console standard error 2 standard input 3 standard printer other standard output Return value A zero if all went well else an error code See the general discussion of the FORTRAN I O capabilities for a listing of the possible error codes See also fiortxt Read next text record fioerror Performs standard error processing 6 153 FIOSTOD Convert String to Double Synopsis include fortran h PROMULA FORTRAN function declarations double fiostod str nstr char str String to be converted int nstr Length of field Description Converts an alphanumeric string containing a number in scientific notation to a double precision floating point number The string can contain optional leading blanks an integer part a fractional part and an exponent part The integer part consists of an optional sign followed by zero or more decimal digits The fra
52. 3 WRITE OUTPUT LISTIO STRG Demo successfully entered 25 0 a 4 5 0 b 2 1 6 0 c 2 1 0 7 0 return 2 42 Use of Printf Style Formatting Xf Yf C and FORTRAN have very different ways of specifying coded write conversions For the FORTRAN bias all of the original FORTRAN machinery is maintained The actual WRITE statement is translated into a C WRITE statement which consists of a series of keywords followed by the parameters associated with those keywords In the C bias FORTRAN WRITE statements are translated into the C printf or fprintf functions whenever possible When not possible the C bias uses the same translation as the FORTRAN bias The Xf switch turns the printf conversion off while the Yf switch turns it on The following short code contains a formatted write statement and a list directed write statement SUBROUTINE DEMO 1 FORMAT 1X 215 3F10 2 WRITE 1 I J A B C WRITE I J A B C END The default intermediate C output for this code is as follows void demo static int i j static float a b c static char F1 1x 215 3 10 2 WRITE OUTPU EMT F1 1 INT4 i INT4 j REAL4 a REAL4 b REAL4 c 0 WRI E OUTPU LISTIO INTA4 i INTA4 j R AL4 a REA 4 b REA 4 0 0 In this translation the PROMULA FORTRAN runtime library is used to perform the writes The precise layout of the recor
53. 4 snumb t f v 9 sl name V S p r d a pc ps pl pd name sl a s num m s n o dialect num num 0 1 2 1 2 name m s numb numb name numb name numb numb numb numb numb numb numb name num Section Characteristic effected by switch Specify your C output Treatment of short arithmetic Casting level Details of C output format Treatment of character variables Appearance of comments in C output Treatment of data initializations Debugging information in C output Error message level Echo pseudo code produced Treatment of syntax errors Annotated listing of source code Annotated listing of C output Echo symbol references by line number Intermediate symbols table Input format used FORTRAN INTEGER type Specify a globals file COMMON variable convention Overall alignment control Inline functions file Target C int type Treatment of internally generated constants Maximum output line width Link time processing of COMMON initializations Inclusion of line numbers for debugging Source language dialect Nesting indentation to be used in output Inline comments C output margin width Upper braces placement convention in C output Lower braces placement convention in C output Name of file to receive the C output Splitting of C output into separate files Miscellaneous prototyping controls Additional prototyping control Listing append filename Listing file page height New listing filename Listing file page width
54. 6 2 CPXADD SHORT COMPEEX ADDITION 4 idea esse eccentric ee ee iro ren aea p E 139 6 3 CPXCJG COMPUTE THE SHORT COMPLEX CONJUGATE cccssccccccecsessssececececcesssesecececseneseeesececsesseseseseesesesessaeeeeees 140 6 4 CPXCMP SHORT COMPLEX COMPARISON cccessessscecececeesssecaeececseneseeesececeenesaaeseceesesesaaeceeeceesesusaeceesceecessaaeaeeess 140 6 5 CPXCOS COMPUTE THE SHORT COMPLEX COSINE cccsssessscecececseneececececeeseuacececececsesssaeceeeceesessaueceeseeeeessaaseeeecs 141 6 6 CPXCPX CONVERT TWO FLOATS TO SHORT COMPLEX esee 141 6 7 CPXDBL CONVERT DOUBLE PRECISION TO SHORT COMPLERX csssessssecececsesnscecceececsenssaeceeeceesesneaecececcesensaaseeeecs 142 6 8 CPXDIV SHORT COMPLEX DIVISION c sccccccesessessscecececcenssnesecececsenesecesecececesseseeeceeseeaaeseeeceeseaaaaseeseecsensaaeeeeess 142 6 9 CPXDPX CONVERT DOUBLE COMPLEX TO SHORT COMPLEX cccsessssescceceesessscecececeesenseaeeeeeceesessaeeeeceeceestaaeeeeess 142 6 10 CPXEXP SHORT COMPLEX EXPONENTIAL ccssssscccccecsesssssceccecesensaececececseseacecececceseaaaecececeesesaaeceecceceeensaeeeeecs 143 6 11 CPXIMA COMPUTE THE IMAGINARY PART OF A SHORT COMPLEX cccccceesessscececececsesseaeeeeececsesneaececeeeceensaaseeeees 143 6 12 CPXLOG SHORT COMPLEX NATURAL LOGARITHM cssssssccccecsesessscecceececseacececccecsesaeeeeecceseseaeceeceecsessaaeeeeess 144 6 13 CPXLOGI10 SHORT COMPLEX BASE 1
55. 90 Fortran 90 X3J3 S8 115 June 1990 The FORTRAN Technical Committee of the American National Standards Institute Mdge Data General FORTRAN 77 Reference Manual 093 000162 02 October 1983 Data FORTRAN General Corporation Mfn5 Cyber FORTRAN 5 FORTRAN Version 5 Reference Manual July 1983 Control Data Corporation Indicates partial support only This particular option should only be used if code is being moved directly from one of these compilers to the PROMULA compiler 2 22 Nesting Indentation to be Used in the Output N N0 Nn The N NO and Nn switches specify how nesting is to be indicated in the C output The N switch specifies that no nesting indentation is to be used in the output This is the default setting for the optimized bias The NO switch specifies that a tab character is to be used for each nesting level The advantage of this selection is that it minimizes the size of the output file while still showing nesting The disadvantage is that tab characters are generally expanded to 8 characters which makes listings containing multiple nesting levels difficult to read The Nn switch specifies the number of blanks to be used for each output nesting level explicitly The default indentation setting for the FORTRAN and C biases is 4 This gives a clean listing but makes the output file somewhat larger than would be achieved via NO 60 PromulaFortran Translator User s Manual 2 23 Inline Comments Output Margin Wi
56. EXCLUDE option All source structure would be excluded and then would be replaced by the desired target subdirectory name In some cases no generic translation scheme will work Certain names might have to be changed on an individual basis The final list of sname tname pairs achieves this end Each pathname is first translated using the generic specifications on the PATHNAME statement itself The resultant pathnames are compared with the snames in the list If the first n characters of a pathname match the n characters of an sname then those n characters are stripped and the associated tname is added to the front of the name As can be seen from the above it will be necessary to organize the INCLUDE files in the new environment Once that organization has been completed the PATHNAMES component of the language specification can be used to describe that structure No changes need be made in the FORTRAN source code INCLUDE and INSERT statements Return Value The pathname conversion information is stored in the parameter pak as follows Byte Description of content 0 Directory separation character 1 Exclude directories from pathname flag 2 Case conversion code 0 none 1 toupper 2 tolower 3 Prefix to be added to name length characters 4 Conversions list See also fiocpath Performs the path conversion specified here 6 146 FIORTXT Read Next Text Record Synopsis finclude fortran h PROMULA FORTRAN function declaratio
57. FIFIARGC FORTRAN Get Command Line Argument Count Synopsis 171 PromulaFortran Translator User s Manual include fortran h PROMULA FORTRAN function declarations long fifiargc Description Returns the index number of command line arguments i e the number of arguments minus one where the program name counts as one argument Return value The number of command line arguments minus one See also None 6 77 FIFIBIT FORTRAN Intrinsic Function IBIT Synopsis include fortran h PROMULA FORTRAN function declarations void fifibit bits ibit ival unsigned char bits Bit vector to receive new bit value int ibit Number of bit to receive value int ival Bit value to be inserted Description Set a specified bit in an arbitrary bit string equal to a specified value The left most bit position is numbered 1 with higher bits receiving higher values Return value None the function is void See also None 6 78 FIFICHAR FORTRAN Intrinsic Function ICHAR Synopsis finclude fortran h Standard C header file int fifichar cl char cl Character to be converted Description Converts a character code into its numeric display code or lexical value or collating weight The point of this function is that character values on the host processor are not necessarily the same as those on the machine for which a given FORTRAN program was written All character value references in a source FORTRAN program are pa
58. GTIM Synopsis finclude fortran h PROMULA FORTRAN function declarations void pdpgtim tic long tic Returns number of ticks since midnight Description Returns the current time of day The time is returned in a long variable and is given in terms of clock ticks past midnight though its actual resolution is only to the nearest second The routine assumes a 60 cycle clock therefore the value returned is actually only hour 60 min 60 sec 60 Return value 244 PromulaFortran Translator User s Manual None the function is void See also None 6 217 VMSCLS Close Virtual File Synopsis include fortran h PROMULA FORTRAN function declarations void vmscls Description This function scans the virtual blocks in memory and writes any that have been changed back to the virtual file This operation ensures that the virtual file can be used as a database after execution Then it physically closes the virtual file The normal virtual memory algorithm only writes a block when it has changed and when its memory slot is needed Return value None the function is void See also vmsglob Manages virtual global variables vmswvb Physically write block to file 6 218 VMSDEL Change Virtual Information Synopsis include fortran h PROMULA FORTRAN function declarations unsigned char vmsdel ioiad long ioiad Offset of information to be changed Description Returns a pointer to the byte at a sp
59. IA IFUNC GVAL RINT I R OP D UNCTION RFUNC I FUNC I I ETURN z ADD 0 ti Z 64 PromulaFortran Translator User s Manual contains references to three subprograms ALPHA RFUNC and IFUNC ALPHA is defined only RFUNC is both defined and referenced and IFUNC is referenced only Processing this code with the Wname switch will produce the following list of prototypes float rfunc long long ifunc float void alpha long float three subprograms are included Now using the P4 switch produces the following two only since IFUNC which was not defined is excluded void alpha 1long float float rfunc long Alternatively using the P8 switch produces the following list float rfunc long long ifunc float in which ALPHA is excluded since it was not referenced The effect of these switches can be most easily seen via the use of the Wname switch which actually writes out internally generated prototypes however remember that PROMULA FORTRAN can be asked to do extensive argument checking via other command line switches and via its warning message system The P4 and P8 flags control the contexts in which internal prototypes are retained and therefore effect the operation of all of these functions 2 27 4 P16 Ignore Prototypes for Definitions The P16 flag is used to modify the behavior of the Yp switch The Yp switch is described later in
60. INT4 ic 0 STOP NULL long ialpha ia gval P1 P2 char ia gval int P1 P2 static long ialpha WRITE OUTPUT LISTIO STRG ia P1 STRG gval P2 0 ialpha P1 P2 return ialpha Note that IA and IB are now treated as character variables Their initialization is quite natural Note however that the string lengths are allocated at their binary size which is not the desired value in this context 2 27 12 P4096 Define Equivalences via a define Another area in which C and FORTRAN differ is address aliasing Basically programs occasionally need to refer to the same area of memory in different ways In C pointers are used In FORTRAN EQUIVALENCE statements are used The following FORTRAN program contains a simple example PROGRAM DEMO INTEGER IA 10 IB 2 EQUIVALENCE IA 5 IB 2 READ IA P RINT IB Within this program the array IB is being used to refer to the fourth and fifth members of the array 1A The default translation for this example in C looks as follows 73 PromulaFortran Translator User s Manual void main argc argv int argc char argv static long ia 10 static long ib 3 ftnini argc argv NULL READ INPUT LISTIO DO 10 INT4 0 WRITE OUTPUT LISTIO DO 2 INT4 ib 0 In this translation the variable is defined as a pointer which points to the fourth member of the array
61. P2 0 ialpha P1 P2 return ialpha This C code will generate a fatal error from most contemporary C compilers since the status of ialpha is first as a void function and later as a long function To avoid this problem PROMULA FORTRAN must be told to delay the writing of any function declarations until the entire code has been read Once the best information about each function is acquired all prototypes can be written to a file which can then be included 68 PromulaFortran Translator User s Manual The P128 flag provides this service The translation of the above code using the P128 flag is shown below include test h lt void main argc argv int argc char argv lt ftnini argc argv NULL ialpha Hello World 5 6 STOP NULL long ialpha ia gval P1 P2 char gval int P1 P2 static long ialpha WRITE OUTPUT LISTIO STRG ia P1 STRG gval P2 0 ialpha P1 P2 return ialpha Note that an include has been added to the front of the code and that the code itself contains no extern declarations The file test h is as follows extern long ialpha This declaration of ialpha is correct and reflects the best information available 2 27 8 P256 Use ANSI C Function Declarations By default PROMULA FORTRAN is designed to produce a C output which will compile on as many different platforms as possible Consequently it uses the old style function de
62. PROMULA FORTRAN Compiler manual in its chapter on error messages Suffice to say that PROMULA FORTRAN does extensive syntax checks while it is processing the source code and extensive consistency checks after it has processed each subprogram The following is a sample listing produced which shows the type of messages that might be produced at the EL1 level 390 utest for W818 The argument ia is being defined with type integer 4 when it has been passed an argument of type character The W appended to the error number indicates a warning Messages at this level can be ignored but they typically indicate potentially serious problems At the EL2 level the following additional types of messages are issued 54 utest for C870 The array OBUF is being subscripted with less than 1 expressions 380 utest for C815 A data value of type character is being assigned to the variable IA of type integer 4 520 utest for C816 The binary type character is being used where type integer 2 is expected 558 utest for C861 Data is being allocated to common storage via the variable SEATRD 820 utest for C866 The real 4 type has previously been assigned to UC At this level in addition to warnings other usages are isolated that should either be checked or that represent potentially serious portation problems Finally at the EL3 level the following additional types of messages are issued 37 utest for N858 The identifier THERMAL with more t
63. PROMULA FORTRAN function declarations void ftnscopy sl nsl char sl int ns1 Description This function copies a variable number of strings into a destination string If at the end of that copy not all space in the destination is used it is filled with blanks Return value None the function is void See also None 6 204 FTNSLENG FORTRAN String Length Synopsis include fortran h PROMULA FORTRAN function declarations int ftnsleng sl nsl char sl int nsi Description This function computes the length of a character string or a sequence of strings being concatenated Return value The length of the result string See also None 6 205 FTNSUBS FORTRAN Substring Evaluation Synopsis finclude fortran h PROMULA FORTRAN function declarations void ftnssubs subs str slen ipos lpos string subs String specification to return substring data char str String which contains the substring int slen Specified length of the string int ipos Starting position of the substring int lpos Ending position of the substring Description 238 PromulaFortran Translator User s Manual This function creates a string structure for a substring reference in a FORTRAN program The operation it performs is minimal however it is needed in functional form to avoid evaluating the value of the substring bounds more than once Note also that there is a dialectal variant here Some allow the maximum string bound to exceed
64. Position a FORTRAN File on a Record Synopsis finclude fortran h PROMULA FORTRAN function declarations int fiorec irec int irec Record number desired Description This function merely records a record number at which the FORTRAN file about to be accessed is to be positioned Return value None the function is void See also None 6 141 FIOREW Rewind a FORTRAN File Synopsis include fortran h PROMULA FORTRAN function declarations int fiorew Description This function rewinds the file associated with the current FORTRAN file as specified in the global variable fiocurf Return value 205 PromulaFortran Translator User s Manual A zero if the rewind was successful else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also None 6 142 FIORLN Read FORTRAN End of Line Synopsis include fortran h PROMULA FORTRAN function declarations int fiorln Description Completes the current read operation by flushing the current format statement and by setting the current record controls to the end of the current record Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiofend End current format processing fioerror Perform error processing fiofinp Next input format specification 6 143 FIORNDV
65. PromulaFortran Translator User s Manual Obtains the current date from the operating system and stores the month day and year in the parameters Return value None the function is void See also None 6 66 FIFIZDIM FORTRAN Intrinsic Function I2DIM Synopsis include fortran h PROMULA FORTRAN function declarations short fifi2dim al a2 short al First value in difference short a2 Second value in difference Description Computes a positive difference Return value If al is greater than a2 then the value of a1 a2 is returned else zero is returned See also None 6 67 FIFI2DINT FORTRAN Intrinsic Function I2DINT Synopsis finclude fortran h PROMULA FORTRAN function declarations short fifi2dint a double a Argument containing the value Description Truncates its double precision argument to a short integer value In particular fifi2dint a 0 if lt 1 the largest integer with the same sign as a that does not exceed a if gt 1 Return value The value as computed above See also None 6 68 FIFIZMAX0 FORTRAN Intrinsic Function I2ZMAX0 Synopsis include fortran h PROMULA FORTRAN function declarations 168 PromulaFortran Translator User s Manual short fifi2max0 al a2 short al First value to be compared short a2 Second value to be compared Description Determines which of two short integer values is the largest Return value The largest short integer value See also
66. REPLACE 5 This causes all dollar signs in source pathnames to be replaced by underscores The TERMINATION character is provided for source systems such as VMS which allow disposition information to be appended to the filename preceded by a slash AII of this information must be ignored in most target systems The mutually exclusive and optional UPPER LOWER parameters specify that all alphabetic characters in pathnames should be converted to upper or lower case respectively Since some systems pathnames are case sensitive while others are not it is important to specify one of these options For example though most PRIME pathnames are shown in uppercase most transfer programs create lowercase names when transferring files to UNIX therefore the standard PRIME language description contains a specification of LOWER for this option For initial testing and use of PROMULA for particular small projects the simplest approach is simply to move all source files including the INCLUDE files into the user s local directory To do this PROMULA must be told to ignore all directory information in the source pathnames Under this alternative all characters up to and including the last occurrence of the directory component separations character are stripped from the source pathname This is achieved via the EXCLUDE option 113 PromulaFortran Translator User s Manual A possible alternative structure for the INCLUDE files for a UNIX impleme
67. Round Value Synopsis include fortran h PROMULA FORTRAN function declarations int fiorndv dspdig ndigit length char dspdig Digit string to be rounded int ndigit Number of rounded digits int length Length of digit string Description Truncates and rounds a numeric string of digits The parameter ndigit specifies the number of digits desired in the rounded result The parameter length specifies the total number of digits now in the string The dspdig string may contain only numeric characters Return value The function returns the carry value from the round If the input string consists of a sequence of 999 such that all become rounded to zero then the output string will contain 100 and the function will return a value of 1 else it will return a value of 0 See also None 206 PromulaFortran Translator User s Manual 6 144 FIORNL Process FORTRAN READ DATALIST Statement Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct char nmname Name of the variable void nmvalu Points to the variable values int nmtype Binary type of the variable int nmadr Points to the variable s dimensions namelist int fiornl name nname namelist name List of variables in this namelist int nname Number of variables in namelist Description This function reads a set of variable values from the current input file The namelist format is identical to that which may be used in specif
68. SAO SV10 The following is the result void demo auto long i static long a 32 b 72 c 472 d 4472 for i 0L i 10 10L 10L return L i d loat vmsdel c 4 loat vmsdel b loat vmsdel a i i Fi 4 i i 10L 10L 4 Ti l L 4 float vmsuse b i 20 4 x float vmsuse d i 1 1 30 float vmsuse c 1 1 101 4 Now the variables have become long static values and references to the variables have been replaced with function references 82 PromulaFortran Translator User s Manual 2 32 FORTRAN Dialect Doloop Assumptions T0 T1 T2 Below is the syntax for the FORTRAN DO statement It looks very simple however what you see is unfortunately not what you get FORTRAN compilers vary widely on how the DO statement is executed Syntax DO slab v el e2 e3 Where slab is the label of an executable statement called the terminal statement of the DO loop If slab is omitted the DO loop is terminated via an END DO statement v is an integer real or double precision control variable el isaninitial parameter e2 isa terminal parameter e3 isan optional increment parameter default is 1 el 2 and e3 are called indexing parameters they can be integer real double precision or symbolic constants variables or expressions The DO statement differs widely between FORTRAN 66 and FORTRAN 77 In FOR
69. The Rname switch may occur multiple times to allow the reading of multiple files If there is no extension supplied with the first file then an extension of cnf will be assumed If there is no extension supplied for additional files an extension of pro is assumed The reason for this is that typically when multiple files are read the first contains the configuration information while the additional files contain function prototype information It should be noted that configuration files may contain command line switch information thus a common use of a configuration file is to supply commonly used switches independently of the command line 2 31 Storage Threshold Values SAnum SDnum SSnum SVnum SZnum One of the realities that has to be faced in using FORTRAN on small machines is that FORTRAN is extremely naive in its use of memory Though the FORTRAN 77 standards committee gave the new language a SAVE statement which was to have helped this statement is rarely implemented In essence all variables in FORTRAN program are assigned a unique and fixed memory location If you do not have enough memory on your machine to do this then you are in real trouble PROMULA takes this problem head on since as consultants it is the one we have encountered most often in downsizing mainframe codes to the PC There are four types of memory made available to you via PROMULA 79 PromulaFortran Translator User s Manual First there is stat
70. a C WRITE statement which consists of a series of keywords followed by the parameters associated with those keywords In the C bias FORTRAN WRITE statements are translated into the C printf or fprintf functions whenever possible When not possible the C bias uses the same translation as the FORTRAN bias Note in the above that both list directed and FORMAT controlled statements are processed For formatted statements the equivalent printf specification is used For list directed conversions PROMULA FORTRAN uses a default set of specifications which may be modified by the user Note 7 that C allows multiple assignments to the same value to be written together Under the C bias the translator looks for such assignments and combines them whenever possible The FORTRAN bias does not by default perform these combinations since they destroy the correspondence between source and output statements This feature is also controllable via a command line switch Note 8 that in FORTRAN the default base for a subscript is 1 Thus all DO loops which generate subscripts tend to start at 1 In C however subscripts start at zero This fact makes for much more efficient code The translator looks for DO loops whose only purpose is to move through array subscripts and reduces their range to start at zero thus producing a very natural looking for statement and optimizing subscript expressions The DO loop reduction feature is on for all biases by default but it ma
71. a b c float a 15 10 b 20 static int i j k fora 1 i 10 i for j 1 j 2 3 1 11 1 for return In this form the loop variables have their original ranges however each reference to those variables must be decremented k 1 ped 0 j k lt 15 15 c 20 10 0 k egal ae c j 1 i 1 a k 1 li 1 b j 1 k 1 by one to compensate for the difference in C and FORTRAN subscripting conventions 100 PromulaFortran Translator User s Manual 2 45 Subprogram Argument Type Checking Xp Yp Though the practice is never desirable most FORTRAN compilers allow the user to pass variables of different types to the same subprogram parameter PROMULA follows the normal practice and does not perform checks to make certain that all parameters are consistent The YP switch turns this checking on If you have a program that is carefully written or if you want to check that all parameters are passed consistently then the Yp flag performs this service Consider the following simple subprogram in which the third parameter to function demo is REAL 4 in one instance and INTEGER 4 in another SUBROUTINE TEST CALL DEMO I J A CALL DEMO The default C output for this subprogram with Xp active is simply as follows void test extern void demo static long i j k static float a demo amp i amp j amp a demo
72. closest to looking like the original FORTRAN and produces relatively efficient code In addition there is no need to introduce extra symbols into the code The problem is that this is technically not valid code At best it would flunk an introductory course in C The common blocks are being defined in different ways in different places in the code The translation below has the same efficiency as the one above and is perfectly valid C programming It is produced with the Gr setting void demo extern char Xalpha Xbeta l float Xalpha float Xalpha 420 float Xalpha 400 4 float Xalphat 424 ftnscopy Xbeta 2 10 Xbetat 150 30 10 NULL return char Xalpha 428 char Xbeta 200 Here the problem is that the COMMON storage has been stripped of all its variable names thus the code is difficult to read What is beta 150 for example Use this flag if you are not interested in the intermediate C output or if you require special COMMON block layouts The simplest translation of this example is produced by the Gv setting and is shown below void demo extern float Xa 10 10 Xb Xc Xd extern char Xc1 Xc2 Xa 0 0 Xc Xb 4 Xd ftnscopy Xc1 2 10 Xc2 30 10 NULL return float Xa 100 Xb 5 Xc Xd char Xc1 150 Xc2 50 Here the COMMON variables themselves are the external symbols If the COMMON was done carefully in the FORTRAN or if you are intending to b
73. e nnns nsns asus nsns nnn 167 FIFI2DIM FORTRAN INTRINSIC FUNCTION I2DIM eeeeeeeeen e e emen nnnm nnns nsns nsns 168 FIFI2DINT FORTRAN INTRINSIC FUNCTION I2DINT eeeeeeeeeceeennne e eene nn nnn nnns nsns nsns nsns 168 FIFIZMAX0 FORTRAN INTRINSIC FUNCTION I2M A XO nne ene nnnm 168 FIFI2MINO FORTRAN INTRINSIC FUNCTION I2MINO eese 169 FIFIZMOD FORTRAN INTRINSIC FUNCTION 2 169 FIFIZNINT FORTRAN INTRINSIC FUNCTION I2NINT e e eene nnne rne rrr nnn nn 170 FIFIZPOW FORTRAN INTRINSIC FUNCTION 2 170 FIFDSHF FORTRAN INTRINSIC FUNCTION I2SHFT 170 FIFDSIGN FORTRAN INTRINSIC FUNCTION 125100 ee 171 FIFIABS FORTRAN INTRINSIC FUNCTION eee eene nnn ntt rne nnn 171 FIFIARGC FORTRAN GET COMMAND LINE ARGUMENT COUNT eene entere nnne nnn 171 FIFIBIT FORTRAN INTRINSIC FUNCTION IBIT een eene nnn nre rennen nnn 172 FIFICHAR FORTRAN INTRINSIC FUNCTION ICHAR ne e eene nennen nnn treten retten nnn 172 FIFIDIM FORTRAN INTRINSIC FUNCTION IDIM eee en n nennen nennen 173 FIFIDINT FORTRAN INTRINSIC FUNCTION 173 FIFINDEX FORTRAN INTRINSIC FUNCTION INDEX
74. eeceeetetee ee e ese nsn nsns nsns 156 DPXPOL DOUBLE COMPLEX CONVERSION TO POLAR 156 DPXPOW RAISE DOUBLE COMPLEX TO A POWER ecce e en 157 DPXREAL COMPUTE REAL PART OF DOUBLE COMPLEX eeeeeeeee e e e ene enne nnnnnn enhn nsns nsns se ese n nsns nsns nsu 157 DPXSIN COMPUTE THE DOUBLE COMPLEX 158 DPXSROOT COMPUTE DOUBLE COMPLEX SQUARE ROOT sees ener nennen nete nnne sese tenen 158 DPXSUB DOUBLE COMPLEX SUBTRACTION eeecececeeenene e ene ene enn ener ener nnns a senes eset ese ases e en e a Eea ases 159 FIFAMAXO FORTRAN INTRINSIC FUNCTION AMA XQ ieeseseseseeeeeeene e ene nsn nsns nsns nsu 159 FIFAMINO FORTRAN INTRINSIC FUNCTION nsns 160 FIFASC50 FORTRAN EXTERNAL FUNCTION 50 e ene ene ener ener nnn n eset es a ases e nsns 160 FIFCHAR FORTRAN INTRINSIC FUNCTION 161 FIFCOS FORTRAN INTRINSIC FUNCTION COS eeeeeseseseee e ene ene 161 FIFDATE EXTERNAL FUNCTION DATA ssseseeseeeeen n e nnns nnns nsns erete ete e e ere ere 161 FIFDDIM FORTRAN INTRINSIC FUNCTION DDIM eee nemen nsns nsns nsns nnn 162 FIFDINT FORTRAN INTRINSIC FUNCTION DINT
75. file being compiled the date the time and the page number relative to the file In this case the page width was set to be narrow so the name of the file is placed on a second line 42 PromulaFortran Translator User s Manual The annotated listing itself contains the include file number If the line number within the source file the nesting level of the statement and the actual source code statement The include file number is left blank for statements in the original source file The nesting level indicator is used with declaration statements when structures are being defined It indicates the level of nesting within the structure For executable statements the nesting level indicator indicates the degree of nesting within DO and or IF statements If there is no current nesting then the nesting level is left blank 2 9 3 Symbol Listing and Cross Reference Table The following symbol listing and cross reference table was produced using the Ex option The table consists of 4 sections Include files symbols referenced symbol references by line number and statement label types and references In general in these tables all user defined symbols are shown in uppercase while all descriptive symbols are shown in lowercase The include files section simply lists all include files encountered to date in the compilation along with the sequence number assigned to them Note that the base source file has a number of zero If there are no include f
76. ftnrecrd fiostrdi ftnstruc string mot n TRUE FALSE Ktrue Kfalse PromulaFortran Translator User s Manual Description of usein C 0 0 0 0 COMMON data module prefix type of initialized structures array platform type designator volatile keyword initialization string for descriptors replacement characters for and structure initializations of set array structure initializations data initialized structures array structure data initialization function call ftnfdata ftnrecrd type label for string descriptors element name of descriptor string length element name of descriptor string address symbol for logical TRUE constant value symbol for logical FALSE constant value symbol for logical TRUE constant variable symbol for logical FALSE constant variable The pattern strings transform simple strings generated into alternate forms The simple string is represented by a percent 96 sign within the pattern string The table below lists the available pattern strings Idn 64 153 155 156 157 172 174 262 263 264 265 267 268 269 include 96 pragma pfc 96 void ftnblkd 7610 x h endif ICF_ ifndef ICF_ define _ Keyword Description of use in C pattern string for writing includes pattern string for writing pragmas name of COMMON initializations function pattern string for COMMO
77. functions is named with the prefix vms The virtual memory system serves two very different yet complementary purposes The first is to allow programs with very large data needs to be implemented on systems with limited memory The second is to make the information within programs available to the PROMULA Application Development System and to other programs and tools constructed using the Promula system This topic is discussed extensively in the chapter on the PROMULA interface in the PROMULA FORTRAN Compiler Manual To achieve these two goals the virtual file manager uses the same data structure as the PROMULA Application Development System The hardest problem which must be faced by a virtual memory system is determining whether or not a current block is in memory or is on the file The approach taken by this implementation uses a block status byte The total virtual memory space is divided into n evenly sized blocks say N of them A vector N bytes long is then reserved in memory For any given block if the block status is zero then the block is not in memory If the status byte is nonzero then it specifies which memory block contains it The advantage of this approach is that it is very fast The disadvantages are as follows 1 The overall size of the virtual memory space needed must be known in advance This is no problem for this implementation because either a fixed PROMULA file is used with known size when it is opened or the translator has c
78. functions needed to initialize the actual data values NAMELIST is a much neglected and maligned feature of FORTRAN which solves this particular problem very nicely 2 7 2 The Initialization Switches As is discussed above C is needlessly weak in the area of data initialization Since we cannot change C we must accommodate this weakness in our C output conventions The D command line switch specifies when data initializations are to be performed and what storage type initialized variables should have Its individual settings are as follows Setting Meaning Dc This is the default setting despite its problems It requests that initialization produced by the DATA statement to local variables be implemented at compile time The initialization values are actually placed in the C source code 37 PromulaFortran Translator User s Manual This setting requests that data initializations be copied from the DATA statement into NAMELIST form into an internal character file so that they can be read into the actual variables at runtime This setting modifies the behavior of the Dc flag by specifying that variables initialized at compile time should be declared as being auto as opposed to the default of static A summary example is included here to show the effects of this switch The following program contains several data initializations 10 PROGRAM DEMO DIMENSION A 50 10 0 0 40 1 0 CHARACTER NAME 4 4 4HFred 2 Mary 3HJoe CALL DEMO
79. given module then C assumes that it is an int however many C compilers give a warning when this assumption is made Consider the following simple FORTRAN subroutine SUBROUTINE TEST J 99 I IFUNC J R RFUNC J PRINT I R STOP END which references two external functions IFUNC and RFUNC The default translation for this subroutine is as follows void test extern int ifunc extern float rfunc static int j i static float r j 99 i ifunc amp j r rfunc amp j WRITE OUTPUT LISTIO INT4 i REAL4 r 0 STOP NULL Notice that both ifunc and rfunc are explicitly declared By default the P1 flag is on Turning the P1 flag off produces the following void test extern float rfunc static int j i static float r j 99 ifunc amp j3 rfunc amp j WRITE OUTPUT LISTIO INT4 i REAL4 r 0 STOP NULL In this version the fact that i func is an int is assumed by the compiler therefore no explicit declaration is given 2 27 2 P2 Use ANSI Prototypes for Argument Functions The second flag associates a full ANSI C prototype with external function declarations passed as arguments when those functions contain arguments which are passed by value This flag is needed only in very specialized cases where detailed prototype checking is needed By default this flag is off Consider the following simple test program 63 PromulaFo
80. glb contains a list of program variables that are to be made global for use by the PROMULA Application Development System The content of this file and the general topic of the PROMULA interface are described in detail in the PROMULA FORTRAN Compiler Manual 2 14 Common Variables Convention Ga Gd Gp Gs Gr Gv That aspect of FORTRAN which has the highest likelihood of causing translation errors and or user readability objections is the COMMON statement Every storage trick and weak typing trick imagined gets used in the nuances of changing COMMON block definitions through a large program There is no ideal way to deal with common blocks PROMULA FORTRAN translates common blocks in one of six ways These are controlled via the G command switch The individual settings associated with this switch are mutually exclusive and are as follows Set Meaning Gp Define common blocks via local pointers and varying typedefs This is the default setting In particular COMMON block identifiers are declared simply as external char storage areas Locally the address of that storage area is assigned to a structure pointer whose members are defined in the same manner as the COMMON definition in the subprogram being translated This technique works in all cases except where byte alignment adjustments are needed The problem with this technique is that it is ugly and that it must use expressions of the form Cblock var to refer to members in the block These
81. has no real equivalent of the int type Its integers are either short or long In PROMULA FORTRAN therefore integer declarations are translated as either short or 1ong with int being used instead of the appropriate one The I command switch tells PROMULA FORTRAN whether int is short or long Its settings are mutually exclusive and are as follows Is Specifies that int is short Il Specifies that int is long The default setting for this switch depends upon the environment in which PROMULA FORTRAN is implemented Environments with short ints have Is as the default while environments with long ints have Il as a default This switch should be changed ONLY if cross compilation is intended Specifying the wrong default int type will produce incorrect results The following FORTRAN code contains three types of INTEGER declarations SUBROUTINE DEMO INTEGER 2 I J INTEGER 4 M 54 PromulaFortran Translator User s Manual NTEGER K I RITE I J K L M N ETURN ND HZH The C code below shows the translation which has an Is setting void demo static int i j static long m n k 1 WRITE OUTPUT LISTIO INT2 i INT2 j INT4 k INT4 1 INT4 m INT4 n 0 return In the translation the INTEGER 2 variables are shown as int since int is short The Is setting has no effect on the meaning of INTEGER as opposed to INTEGER 2 versus INTEGER 4 The binary type of IN
82. here T is our switch The problem is that control switches are rarely part of the actual code therefore when you have in front of you a code with logic like the above in it you have no idea what the expected result is There is no right translation Did the programmer do it intentionally or unintentionally Did he know that there were two standards If he did then which was he using If he did not then what did the JCL that someone probably gave him have in it At any rate if you are processing your own code then you probably know which dialect you are using If you are not translating your own code then make a guess The individual settings associated with this flag are as follows TO The minimum DO statement trip count is zero as specified in the FORTRAN 77 standard This is the default setting for this switch T1 The minimum DO statement trip count is one as specified in the FORTRAN 66 standard T2 The minimum DO statement trip count is zero as specified in the FORTRAN 77 standard Always use the formal FORTRAN 77 trip count computation when iterating through loops The DO statement is of course translated via the for statement in C The following is a simple translation to show the difference between TO T1 and T2 SUBROUTINE DEMO IVAL 0 N 0 DO 10 I 1 N IVAL IVAL 1 10 CONTINUE RETURN END The following translation was produced using the default TO setting void demo static int
83. hosting this code short p77getu lun short lun Returns logical unit number Description Scans the currently open files to find an available logical unit number Return value The unit number See also None 6 211 P77NLENA PRIME FORTRAN 77 Subroutine NLENSA Synopsis include fortran h PROMULA FORTRAN function declarations short p77nlena mes nmes char mes Message to be evaluated int nmes Length of message Description This function computes the length of a blank filled string by finding the last nonblank character Return value The length of the string See also None 6 212 P77TNOUA PRIME FORTRAN 77 Subroutine TNOUA Synopsis finclude fortran h PROMULA FORTRAN function declarations void p77tnoua mes nmes 242 PromulaFortran Translator User s Manual char mes Message to be written int nmes Length of message Description This function writes a message to the terminal screen with a newline Return value None the function is void See also None 6 213 PDPASSN PDP FORTRAN Subroutine Assign Synopsis include fortran h PROMULA FORTRAN function declarations void pdpassn lun name icnt mode int lun Logical unit number char name Name of file int icnt Character count for name char mode File mode information char cc File carriage control information Description This function assigns a file name and characteristics to a logical unit so that it may be defined
84. initialization of COMMON variables via DATA statements The FORTRAN standard clearly states that COMMON variables may only be initialized in BLOCK DATA subprograms however most FORTRAN dialects allow usages such as the following SUBROUTINE TEST COMMON BLOCK I J DATA I J 5 6 PRINT I J STOP END in which the variables I and J in the COMMON block BLOCK are initialized via a DATA statement The default translation for this subroutine looks as follows void test extern char Xblock typedef struct long i j Cblock auto Cblock Tblock Cblock Xblock static int FIRST 1 static namelist DATAVAR i Xblock 5 NULL j Xblock 4 5 NULL 57 static char SDATAVAR i 5 j 6 SEND if FIRST FIRST 0 fiointu PromulaFortran Translator User s Manual DATAVAL char DATAVAL 0 2 fiornl DATAVAR 2 NULL WRITE OUTPUT STOP NULL LISTIO INT4 1 gt 1 INT4 Tblock gt j 0 The details of this translation approach which uses an extension of the FORTRAN NAMELIST capability are discussed in the section on the and Dr command line switches In essence the effect of this approach is that the initialization values will be assigned to the variables I and J at the time that the function TEST is first called The problem addressed by the Lm and Ls switches is that initializing these two variables at the
85. integer value converted to single precision 159 PromulaFortran Translator User s Manual See also None 6 46 FIFAMINO FORTRAN Intrinsic Function AMINO Synopsis finclude fortran h PROMULA FORTRAN function declarations float fifaminO al a2 long al First value to be compared long a2 Second value to be compared Description Determines which of two long integer values is the smallest Return value The smallest long integer value converted to single precision See also None 6 47 FIFASCS50 FORTRAN External Function ASC50 Synopsis include fortran h PROMULA FORTRAN function declarations void fifasc50 icnt input output int icnt Number of characters to be converted void input Input Radix 50 characters to be converted void output ASCII characters return area Description The ifasc50 function converts a specified number of Radix 50 characters to ASCII If the input word contains an illegal Radix 50 character a is entered into the ASCII string Radix 50 is a special character representation in which up to 3 characters can be encoded and packed into 16 bits The Radix 50 characters and their corresponding code values are Character Radix 50 Value space 0 A Z 1 26 z 1 26 27 28 0 9 30 39 The radix values are stored three characters per word according to the formula cl 40 40 C2 40 c3 where the characters are numbered from left to right If the number of characters
86. iouwl which sets up the length value at the front of the record before it is written As a result repositioning of the file is not required The actual C output is of course more difficult to read therefore it is suggested that this flag only be used if unformatted output speed is critical 2 49 Subprogram Call by Value Arguments Xv Yv In FORTRAN all parameters are passed by address This convention generates inefficient code for simple values and makes passing constants and expressions messy insofar as the translation is concerned Nevertheless to ensure that PROMULA FORTRAN always produces correct results the default convention used is to pass all arguments by address The Yv switch can be used to allow for call by value Note that explicit function prototypes can always be provided via the configuration file These prototypes may specify call by value parameters on an individual function level which override the default setting of this switch 2 50 Dollar Signs as Initial Symbols in Identifiers X Y By default PROMULA accepts dollar sign as a valid character within identifiers They may occur anywhere that an alphabetic character may appear Unfortunately some dialects of FORTRAN use the to mark multiple return statement labels For these dialects the may not be allowed to occur in the initial position The X switch disallows in the initial position of identifiers The default Y switch allows them to occur anywhere
87. line switch though we will put the switch into a configuration file as discussed earlier in this chapter is below void demo i j long i j extern long jindex static char name 12 initial 4 static string Tl NULL 0 static string T2 NULL 0 i fifindex name 12 initial 4 Tl a name 1 12 T2 a initial T2 n 4 j jindex amp T1 amp T2 return the machinery needed for real VAX descriptors is here however the following configuration file is needed SWITCHES CHv KEYWORDS string DESCRIPTOR Dleng nm Dptr ftnsallo ftndallo NULL 0 0 0 0 NULL END This produces the following translation void demo i j long i j extern long jindex static char name 12 initial 4 static DESCRIPTOR T1 0 0 0 NU static DESCRIPTOR T2 0 0 0 NU i fifindex name 12 initial 4 Tl Dptr name Tl Dleng 12 T2 Dptr initial T2 Dleng 4 j jindex amp T1 amp T2 return 3 6 The Configuration PRAGMA Statement 123 PromulaFortran Translator User s Manual Sotfware developers have long faced the problem of nonportability with FORTRAN As a result many have written FORTRAN preprocessors that aid in making FORTRAN programs more portable A major feature of C that makes it portable is its ability to deal with conditio
88. minimum bound is always assumed to be zero As can be seen from the above there are considerable differences in the manner in which arrays are treated in the two languages For the C and optimized biases because C programmers are familiar with pointer notation and because many optimizations can be achieved when using pointer notation all FORTRAN arrays are defined in C as simple one dimensional vectors and all array references are done using the pointer notation Alternatively for the FORTRAN bias C brackets notation is used whenever possible however the order of the subscripts must be reversed in the C output In addition the s for subscript optimization switch is provided Using this switch the user may specify his own preference on top of his overall bias specification The following shows a sample set of fixed array definitions along with some basic subscript expressions as translated via the Xs and Ys switches LH SUBROUTINE DEMO DIMENSION A 10 B 6 7 C 3 4 5 Df 2 3 41 5 0 RSET Oey 0 aBSHPOSG Intermediate C output using the Ys switch is as follows void demo static int i j k l1 static float a 10 b 42 c 60 d 120 at2 0 b 19 1 59 2 37 1 1 4 6 1 3 1 6 5 IFIRAN AEN dt i 1 1 k 1 1 1 4 3 2 7 Intermediate C output using the default Xs switch is as follows void demo
89. nesting level The NL1 switch specifies that the lower brace is on the next line at the indentation point for the previous nesting level The NL2 switch specifies that the lower brace is on the next line at the indentation point for the current nesting level The default 15 NL1 and generates the following conditional statement 1 statement 2 The user may also select the following styles NUO NL2 NU1 NL1 NU1 NL2 conditional conditional conditional statement 1 statement 2 statement 1 statement 1 statement 2 statement 2 NU2 NL1 NU2 NL2 conditional conditional statement 1 statement 1 statement 2 statement 2 2 25 Name of the File to Receive the C Output Oname Normally PROMULA FORTRAN sends the intermediate C output produced to a file with the same name as the input with its extension changed to c If some other file is to receive the C intermediate output then use the O switch to specify its name There should be no whitespace between the O of the switch and the name of the file 61 PromulaFortran Translator User s Manual 2 26 Splitting of Output into Separate Files Os Om By default PROMULA FORTRAN writes all of its translated output to a file with the same name as the input file but with the extension changed to c In some instances if you are building a library or if you intend to do extensive editing of the output it is convenient to have the C code correspondin
90. of a pointer to the start of the string This location is updated to point immediately beyond the last character of the integer value Return value The converted value See also None 6 155 FIOUWL Establish FORTRAN Unformatted Write Length Synopsis include fortran h PROMULA FORTRAN function declarations void fiouwl recl long recl Record number desired Description This function merely records the length of the following unformatted record to be written Return value None the function is void See also None 6 156 FIOVFINI Initialize A Variable FORTRAN Format Synopsis include fortran h PROMULA FORTRAN function declarations 213 PromulaFortran Translator User s Manual int fiovfini fmt nfmt char fmt Pointer to the format int nfmt Length of FORMAT Description This function initializes the variable s format strings for use in the FORMAT environment needed by the FORTRAN style input output statements This function is needed to convert the format string into external display code which is used by the format system Return value A zero if the format appears well formed else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiofini Initialize FORTRAN format processing 6 157 FIOWALPH Write Alphabetic Information Synopsis include fortran h PROMULA FORTRAN function declarations void fiowalph alpha nalpha
91. of keywords or keyword strings These strings can be changed by the user although if they are changed the user must make certain that the result is still compilable Most KEYWORD changes require an equivalent change in the fortran h header file Each KEYWORD can be referenced either by its current form or by its identification number The keywords themselves are of two types 1 simple strings 2 pattern strings The table below shows the simple strings within the keyword table These appear in the output of PROMULA exactly as shown in the appropriate places Idn Keyword Description of use in C 1 void equivalent of SUBROUTINE 2 short equivalent of INTEGER 2 3 double equivalent of REAL 8 4 unsigned short equivalent of LOGICAL 2 5 char equivalent of INTEGER 1 6 long equivalent of INTEGER 4 7 float equivalent of REAL 4 8 unsigned long equivalent of LOGICAL 4 9 unsigned char equivalent of LOGICAL 1 10 double equivalent of REAL 16 11 dcomplex equivalent of DOUBLE COMPLEX 12 complex equivalent of COMPLEX 13 char equivalent of CHARACTER 14 static static keyword 15 auto auto keyword 16 extern extern keyword 17 typedef typedef keyword 18 struct struct keyword 19 ftnadr array that contains assigned FORMAT pointers 20 ftnalloc allocates memory for dynamic variables 21 ftndim array for dimension values for namelist 22 ftnlen number of lines in assigned FORMATs 23 ftnfree frees memory for dynamic variables 24 ftnsiz array for dimension values
92. of the possible error codes See also fiowbiv Write binary values fiowrc Write character vector 6 175 FIOWVD Write FORTRAN Double Value 224 PromulaFortran Translator User s Manual Synopsis finclude fortran h PROMULA FORTRAN function declarations int fiowvd val double val Double precision value to be written Description Writes a double precision value to the current output file in accordance with the current format specification This function corresponds to the FORTRAN 77 convention in which values and not just l values can be written Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowbiv Write binary values fiowrd Write double precision vector 6 179 FIOWVFE Write FORTRAN Float Value Synopsis finclude fortran h PROMULA FORTRAN function declarations int fiowvf val float val Single precision value to be written Description Writes a single precision value to the current output file in accordance with the current format specification This function corresponds to the FORTRAN 77 convention in which values and not just l values can be written Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowbiv Write binary values fiow
93. output correspond as closely as possible to FORTRAN original while still producing correct C Extensive use is made of defined symbols to achieve this goal In addition every effort is taken to keep the source and output statements in correspondence The objectives of the C bias on the other hand are to make the output as C like as possible No attempt is made to maintain a statement by statement correspondence and whenever possible FORTRAN I O statements along with associated FORMAT specifications are translated into traditional C I O statements The lines in the figure above are annotated with note numbers The notes are as follows Note 1 that the parameter n is not declared as a pointer since it is not changed within the routine PROMULA FORTRAN uses what are called prototypes of subprogram arguments so that it can generate optimal calling sequences These prototypes may be specified by the user or may be determined internally by the translator The above was internally determined Technically all parameters in FORTRAN are passed as pointers For FORTRAN codes that make use of this 13 PromulaFortran Translator User s Manual fact and mix types across calls PROMULA FORTRAN may be told to make all parameters pointers Alternatively the user can specify individual prototypes via a separate file Note 2 that ANSI FORTRAN requires that default integers and reals both occupy the same amount of memory typically 4 bytes Ther
94. possible error codes See also fioerror Perform FORTRAN I O Error Processing fiofini Initialize A FORTRAN Format fiointu Establish FORTRAN Internal Unit fiolun Establish FORTRAN Unit Number fiorbiv FORTRAN Read Binary Values fiordb Read FORTRAN Boolean Vector fiordc Read FORTRAN Char Vector fiordd Read FORTRAN Double Precision Vector fiordf Read FORTRAN Floating Point Values fiordi Read FORTRAN Short Integer Vector fiordl Read FORTRAN Long Integer Vector fiords Read FORTRAN String fiordt Read FORTRAN Truth Value Vector fiordu Read FORTRAN Unsigned Char Vector fiordx Read FORTRAN Complex Vector fiordz Read FORTRAN Double Complex Vector fiorec Position a FORTRAN File on a Record fiorln Read FORTRAN End of Line fiostatus Set FORTRAN I O Error Status fiostio Establish FORTRAN Standard I O 6 199 FTNREW FORTRAN Rewind Statement 235 PromulaFortran Translator User s Manual Synopsis finclude fortran h PROMULA FORTRAN function declarations int ftnrew lun int lun Logical unit number Description Executes the FORTRAN REWIND statement when translated via a non optimized user bias In addition to the logical unit number this function takes a variable number of parameters which specify the actual data to be supplied to the rewind function The parameter type codes passed to this function are as follows Code Parameter Description of Use Ends the lis
95. s Manual Computes the short complex base 10 logarithm of a short complex value Return value The complex base 10 logarithm See also cpxlog Computes complex natural logarithm 6 14 CPXLONG Convert Short Complex to Long Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex long cpxlong a complex a Contains the value to be transformed Description Computes the real part of a short complex value and converts it to long Return value The long result See also None 6 15 CPXMUL Short Complex Multiplication Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex complex cpxmul a b complex a The left hand value complex b The right hand value Description Multiplies two short complex numbers to form a third Return value 145 PromulaFortran Translator User s Manual The short complex result of the multiplication See also None 6 16 CPXNEG Compute the Short Complex Negative Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex complex cpxneg a complex a Contains the value to be transformed Description Computes the comp
96. select whether he wants universal promotion to long or no automatic promotion to long We have no provision for selective promotions Note that selective and universal promotion differ only in overflow conditions so users from such environments should use the universal promotion to long convention The default convention is universal promotion No automatic promotion is selected via the CS command line switch From a readability standpoint the best convention is unfortunately not the cleanest one In C the results of all short binary operators are short The only way to force C to produce a long result is to convert the arguments to long prior to the calculation The C output for the assignments in the above FORTRAN program under the universal promotion convention is as follows void main argc argv int argc char argv static int il i2 i3 i4 isum iprod iquot static double dsum dprod dquot static long jsum jprod jquot static float rsum rprod rquot ftnini argc argv il 20000 i2 30000 13 200 14 300 isum long il i2 long 13 14 iquot long il i2 i4 jsum long 11 12 jprod long 13 14 jquot long ilt i2 i4 rsum long il i2 rprod long i3 i4 rquot long il i2 i4 dsum long i1 i2 dprod long 13 14 dquot long il i2 i4 ftnopen 6 FILEN EX002 0UT 9 STATUS NEW 0 fprintf LUN 6
97. string the length of that string must be known since the storage technique for character variables itself does not have this information This is an especially difficult problem since the length of a given character string is defined not globally but rather by the context of its use The solution is difficult to accept but appears to be dictated by the needed generality of the translation If a given program does not use variable length characters CHARACTER in subprograms then there is no problem Programs using FORTRAN 66 conventions and programs using FORTRAN 77 conventions can both be handled straightforwardly by translating character variables into C arrays of char If a program does use the CHARACTER construct however then all references to character variables in the FORTRAN program are translated into a pair of values a pointer to the start of the characters being manipulated and an integer value which defines the length of the character string at that point in the code To distinguish these two cases the user must tell PROMULA FORTRAN what assumptions his program is making about the availability of character string length information In addition the user must specify whether he is mixing character and non character variables across subprograms 25 PromulaFortran Translator User s Manual To make life more difficult there are three alternative ways in which the character address and character length can be joined 1
98. that reproducible results can be achieved almost automatically without lengthy and expensive manual recoding work PROMULA FORTRAN translates FORTRAN code to C code which is then compiled via any standard C compiler and linked with the PROMULA FORTRAN runtime library to produce efficient executable code The resultant executable code produces the same results on a target platform as the original code does on the source platform In designing PROMULA FORTRAN we took the position that the only difference between a translator and a compiler should be that a compiler converts the source code into machine language while a translator takes it to a higher level language PROMULA FORTRAN compiles the FORTRAN source language into a low level pseudocode This pseudo code is much like the output produced by the first or second pass of contemporary compilers Second it optimizes that code using the same techniques as used during the optimization pass of a compiler Third it does code generation but the code generated is not machine code it is C A more detailed description of the design and methodology of PROMULA FORTRAN appeared in a series of three technical papers in the Journal of C Language Translation 1 Design of a FORTRAN to C Translator Fred K Goodman Vol 1 December 1989 and March 1990 2 FORTRAN to C Numerical Issues Fred K Goodman Vol 2 June 1990 3 FORTRAN to C Character Manipulation Fred K Goodman Vol 2 September
99. the number dcomplex dcomplex dpxcjg dcomplex a Contains the value to be transformed Description Computes the complex conjugate of a double complex value By definition c conj a a cr i a ci Return value The complex result See also None 6 26 DPXCMP Double Complex Comparison Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of the number dcomplex int dpxcmp a b 150 PromulaFortran Translator User s Manual dcomplex a The left hand value dcomplex b The right hand value Description Compares two double complex numbers Return value A zero if the numbers are the same else a one See also None 6 27 DPXCOS Compute the Double Complex Cosine Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of the number dcomplex dcomplex dpxcos a dcomplex a Contains the value to be transformed Description Computes the value of the double complex cosine of a double complex number By definition e a cos a Return value The double complex result See also None 6 28 DPXCPX Convert Short Complex to Double Complex Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float
100. the specified string length while others do not Return value None the function is void however the parameter subs is updated to contain a pointer to the start of the substring and its length See also None 6 206 FTNSTOP FORTRAN Stop Statement Synopsis include fortran h PROMULA FORTRAN function declarations void ftnstop message char message Message to be displayed at console i e stderr Description Executes the FORTRAN STOP statement by sending a message to the console stderr not the screen stdout and making a normal exit It should be pointed out that this is one of those conventions that reasonable people do disagree about Changing the output file to stdout is easily accomplished in the source for this function via a defined variable console Return value None this function does not return to the calling function See also None 6 207 FTNWEF FORTRAN End File Statement Synopsis finclude fortran h PROMULA FORTRAN function declarations int ftnwef lun int lun Logical unit number Description Executes the FORTRAN ENDFILE statement when translated via a non optimized user bias In addition to the logical unit number this function takes a variable number of parameters which specify the actual data to be supplied to the endfile function The parameter type codes passed to this function are as follows Code Parameter Description of Use Ends the list of parameters 1 long Points to an
101. this function is void See also None 6 198 FTNREAD FORTRAN Read Statement Synopsis include fortran h PROMULA FORTRAN function declarations int ftnread lun 234 PromulaFortran Translator User s Manual int lun Logical unit number Description Executes the FORTRAN READ statement when translated via a non optimized user bias In addition to the logical unit number this function takes a variable number of parameters which specify the actual data to be supplied to the detailed read functions The parameter type codes passed to this function are as follows Code Parameter Description of Use 1 long Points to an error return variable 2 Indicates that list directed I O is being performed 3 char int Points to a full FORMAT specification 4 char Points to a FORMAT string 5 int Specifies a record number 9 Specifies that more operations are to be performed 10 int Specifies that a series of operations are desired 11 Introduces a short integer value 12 Introduces a double precision value 13 Introduces a short logical value 14 Introduces a char value 15 Introduces a long value 16 Introduces a float value 17 Introduces a long logical value 18 Introduces an unsigned char value 19 Introduces a complex value 20 Introduces a string value 21 Introduces a C string value Return value A zero if the open went well else an error code See the general discussion of FORTRAN I O capabilities for a listing of the
102. time that TEST is first called may be incorrect Many of the dialects that allow local data initializations to COMMON variables assume that those initializations are performed at link time or at least that they are performed prior to the execution of any of the user s code For these implementations TEST need not even be explicitly called to obtain the I and J values Using the Ls switch produces the following translation for the above void test extern char Xbloc typedef struct long i j Cblock auto Cblock Tblo k 1 Cblock Xblock static int FIRST 1 static namelist i Xblock 5 NULL DATAVAR j Xblock s4 5 NULL static char DATAVAL SDATAVAR 5 3 6 SEND if FIRST FIRST 0 fiointu char DATAVAL 0 2 fiornl DATAVAR 2 NULL return lt 1 WRITE OUTPUT STOP NULL void ftnblkd test LISTIO INT4 1 gt 1 INT4 1 gt 0 There are two important changes in this translation relative to the previous one Note first that a return has been added to the block of code that controls the initializations of the variables This return allows us to perform an additional call to this function independently of any calls that may be made by the application itself Note second that a function called ftnb1kd has been added to the bottom of the code which
103. vector vmsdel Change a virtual value 6 225 VMSRDC Read FORTRAN Virtual Character Vector Synopsis include fortran h PROMULA FORTRAN function declarations int vmsrdc value nval long value Virtual address characters to be read int nval Number of characters to read Description Reads a virtual vector of character values from the current input file in accordance with the current format specification Note that in this function each individual character is assumed to have its own corresponding format specification if a formatted read is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiordc Read FORTRAN character vector vmsdel Change a virtual value 6 226 VMSRDD Read FORTRAN Virtual Double Precision Vector Synopsis include fortran h PROMULA FORTRAN function declarations int vmsrdd value nval long value Virtual address values to be read int nval Number of values to be read Description 249 PromulaFortran Translator User s Manual Reads a virtual vector of double precision floating point values from the current input file in accordance with the current format specification Each value is assumed to have its own corresponding format specification if a formatted read is being performed Return value A zero if there is no error flag set else an error
104. via defines they must also be undefined via undef at the end to simulate the FORTRAN scope In those cases where FORTRAN subprograms correspond to files in C the necessity for the undef is removed The P1024 flag turns off the undefs The following is the same translation as presented above with the P1024 flag turned on void alpha ia gval long ia float gval define aval 99 0 define ival long 10 2 ia aval gval ival return 2 27 11 P2048 Force Variables to Have Explicit Character Type Another issue having to do with typing is the mixture of CHARACTER and non character information within non integer variables The following FORTRAN code represents the sort of situation that can occur PROGRAM DEMO INTEGER IA 2 IB 2 DATA IA 5Hhello DATA IB 6H World IC IALPHA IA IB PRINT IC STOP END FUNCTION IALPHA GVAL CHARACTER IA GVAL PRINT IA GVAL IALPHA LEN IA LEN GVAL END A subprogram which explicitly assumes character variables has been intermixed with one that hides characters in INTEGER variables To make the situation clearer the lengths of the character arguments are explicitly used by the subprogram both in the free form PRINT and in the computation of the return value The default translation for this program does not produce correct results void main argc argv int argc char argv extern long ialpha sta
105. 0 LOGARITHM ccssccccccecsesessscececceeesensscecececeesenssaececececcessaeseecesesensaaseeeecs 144 6 14 CPXLONG CONVERT SHORT COMPLEX TO LONG 145 6 15 CPXMUL SHORT COMPLEX MULTIPLICATION c cccccccscssssscecececeecessecescceceeseseeecececeeseueaseeeececsesssaeeeeeesesesesaeaaeecs 145 6 16 CPXNEG COMPUTE THE SHORT COMPLEX NEGATIVE ccsscsccccccsessssscesececcesssecececececseaaeseeececeecssaeseeececsessaaeeeeees 146 6 17 CPXPOL SHORT COMPLEX CONVERSION TO 146 6 18 CPXPOW RAISE SHORT COMPLEX TO A 147 6 19 CPXREAL COMPUTE REAL PART OF SHORT COMPLEX esses eene 147 6 20 CPXSIN COMPUTE THE SHORT COMPLEX SINE c cccccccsssessscecececsesensecececececseseceseeececsesaaeceeececsesasaeseeseeceensaaeeeeees 147 6 21 CPXSROOT COMPUTE SHORT COMPLEX SQUARE ROOT eese eene 148 6 22 CPXSUB SHORT COMPLEX SUBTRACTION ccssssssscccecsessssesecceecsesesecesecececseuseeseeeecesesaaeceeececsesaaeceesceceeensaeeeeees 148 6 23 DPXABS COMPUTE THE DOUBLE COMPLEX ABSOLUTE VALUE eene 149 6 24 DPXADD DOUBLE COMPLEX ADDITION cccssssscccccecsessssececececsenesecesecececseuseeseeeceeseusaeceeececsesssaeceecesesensaaseeeees 149 PromulaFortran Translator Us
106. 0 U 7 00 U 8 00 YEAR 3 01 D 6 00 U Statement label types and references by line number in SUBROUTINE STRUC6 Label Line If u D defined U used A assigned 5 format 10 00 D 10 statement 4 00 U 9 00 D The statement label types and references table is a numerical listing of the statement labels along with their type statement Or format and a listing of the lines where they are referenced If the subprogram contains EQUIVALENCE statements then a fifth table type is generated the equivalence pairs table Given the code fragment below 48 SUBROUTINE ANA1 49 INTEGER BUF1 2048 BUF2 2048 BUF3 2048 50 BYTE OBUF 32767 51 EQUIVALENCE BUF2 1 OBUF 1 52 EQUIVALENCE BUF3 OBUF 62 END 44 PromulaFortran Translator User s Manual The following equivalence pairs table is produced Equivalence pairs Dependent Base Offset BUF2 OBUF 0 BUF3 OBUF 0 The base variable is the variable whose storage is being used to contain the dependent variable The offset is the byte offset within the base variable of the start of the dependent variable 2 9 4 Intermediate Compiler Tables The EP and EZ switches can be used to generate a listing of the intermediate compiler tables EZ lists the symbol tables while EP lists the pseudo code generated The following is a simple program along with the listing formed by the EP and EZ switches PROGRAM TES
107. 1 A STOP END SUBROUTINE DEMO A DIMENSION A 50 DATA VAL 55 6 DO 10 I 1 50 A R END I A I val E TURN These initializations are performed either as part of the declaration or in an explicit DATA statement The D switch treats both types of initializations in the same manner The following is the default translation of the example above void main argc argv int argc char argv extern void demo1 static float a 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 1 051 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 07 1 0 1201 0 1 0 120 T 0 1 07 1 0 120 1 0 120 1 0 120 120 T 0 1250 1 2305 7 0 static char name 16 apie gee herb d gim Me Ea tortue epe NEE mtu Vo Not Ven Y ftnini argc argv NULL demol a STOP NULL void demol a float al static float val 55 6 static long i for i OL i lt 50L i afi a i val return 38 PromulaFortran Translator User s Manual The two primary problems to note are that C has no equivalent of the n value notation and that a non null terminated sequence of characters can only be written as individual elements enclosed in single quotes Nonetheless this way of initializing variables is much more efficient than the runtime way and is the default The following shows the same translation
108. 1 Pointer to first string int nl Length of first string char s2 Pointer to second string int n2 Length of second string Description Compares two FORTRAN style strings to determine their lexical relationship The comparison proceeds character by character until either the end of both strings is reached or until a specific character difference is found If the strings are of unequal length the shorter string is treated as though it were padded with blanks to the length of the longer string Note that the lexical value of the character is obtained from the fifichar function Thus this function does not necessarily assume the display values of the host processor but rather can assume the display code values of another processor Return value 0 if no character difference is found n if a character in the first string is less than the corresponding character in the second string n if a character in the first string is greater than the corresponding character in the second string 231 PromulaFortran Translator User s Manual See also fifichar Convert character to display code 6 193 FTNFREE Free Dynamic Memory Synopsis include fortran h PROMULA FORTRAN function declarations void ftnfree ptr void ptr Start of memory to be freed Description Frees dynamic memory allocated by ftnalloc Return value None the function is void See also None 6 194 FTNINI Initialize FORTRAN Processing Synopsis inclu
109. 16 Introduces a float value 17 Introduces a long logical value 18 Introduces an unsigned char value 19 Introduces a complex value 20 Introduces a string value 21 Introduces a C string value Return value A zero if the write went well else an error code See the general discussion of FORTRAN T O capabilities for a listing of the possible error codes See also 240 PromulaFortran Translator User s Manual fioerror Perform FORTRAN I O Error Processing fiofini Initialize A FORTRAN Format fiointu Establish FORTRAN Internal Unit fiolun Establish FORTRAN Unit Number fiorec Position a FORTRAN File on a Record fiostatus Set FORTRAN I O Error Status fiostio Establish FORTRAN Standard I O fiowln Write FORTRAN End of Line fiowbiv Write FORTRAN Binary Values fiowrb Write FORTRAN Boolean Vector fiowrc Write FORTRAN Char Vector fiowrd Write FORTRAN Double Precision Vector fiowrf Write FORTRAN Single Precision Vector fiowri Write FORTRAN Short Integer Vector fiowrl Write FORTRAN Long Integer Vector Write FORTRAN Vector of Strings fiowrt Write FORTRAN Truth Value Vector fiowru Write FORTRAN Unsigned Char Vector fiowrx Write FORTRAN Complex Vector fiowvb Write FORTRAN Boolean Value fiowvc Write FORTRAN Character Value Write FORTRAN Double Value fiowvf Write FORTRAN Float Value fiowvi Write FORTRAN Short Integer Value Write FORTRAN Long Integer Value fiowvs
110. 1990 1 6 Rationale for Developing PROMULA FORTRAN Since the beginning of FORTRAN there has always been a problem FORTRAN is not transportable from machine to machine or even from one operating system to another In contrast the C language is unique in that it is available for almost every type of computer from home computers to supercomputers It is extremely efficient modular and portable It is presently the language of choice for many operating system programmers and compiler designers In addition to authoring PROMULA FORTRAN we have been developing FORTRAN development tools and FORTRAN applications for clients since 1967 In converting FORTRAN programs from one platform to another our typical problem was not that we could not find a good FORTRAN 77 compiler for the target platform but rather that we were confronted by FORTRAN programs that were almost always written in non standard non machine transportable FORTRAN dialects The typical mainframe FORTRAN program is not written in standard FORTRAN and makes assumptions about the machine and operating system for which it was originally written FORTRAN programs are just not portable PromulaFortran Translator User s Manual PROMULA FORTRAN was designed to deal with actual FORTRAN programs written by non structured FORTRAN programmers who took advantage of every possible special feature of their particular vendor s compiler and who had never known or cared that there was a
111. 27 2 P2 Use ANSI Prototypes for Argument Functions eese eene nennen reete 63 2 27 3 4 8 Exclude Referenced or Defined Prototypes essent nene ene een eene 64 2 27 4 1 Ignore Prototypes for Definitions eei e rere t e ot pe ib Re eere db etie eio o ub ovine 65 2 27 5 P32 Treat User Prototypes as System Functions eese 66 2 27 6 P64 Write PFC Style Prototypes not C Type esee nennen 67 2 27 7 P128 Write All Function decls to Header File eene hehehe nenne nenne nennen 66 2 27 8 P256 Use ANSI C Function Declarations eee eene 69 2 27 9 P512 Make Parameters Always Take Explicit Value Type eese 70 2 27 10 P1024 Exclude undefs From the Translation eese eene enne 7I 2 27 11 P2048 Force Variables to Have Explicit Character Type seen 7I 2 27 12 P4096 Define Equivalences via a fdefine eese eene retener trennen 73 2 27 13 P8192 Use Parameter Identifiers in Equivalences 74 2 27 14 P16384 Display Include Files Separately eese eene eene enne eene nennen 75 2 28 LISTING FILE CONTROL PANAME PHNUMB PNNAME PWNUMB esee enn enne nennen 77 2 29 QUANTITY CONTROL FLAGS QINUMB QENUMB QDNUMB QXNUMB
112. 39 PromulaFortran Translator User s Manual int argc char argv extern void demol auto float a 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 05 1 0 150 1 05 1 051 0 1 051 0 1 0 1 0 1 0 1 0 1 0 1 0 1 20 1 0 1 20 1 0 1 0 1 0 1 0 1 0 auto char name 16 Nat tmt Taye Tat tt ot to tet Y ftnini argc argv NULL demol a STOP NULL void demol a float a auto float val 55 6 static long i for i OL i lt 50L i a i a i val return This version declares the initialized variables as auto Note that many C compilers are unable to compile this version The effect of this auto declaration is that the variables are initialized at their specified values each time the function containing them is entered 2 8 Turn on Debugging Mode DB When using PROMULA FORTRAN as a compiler it is often convenient to use a debugger The DB switch makes this possible When used the debugger will refer to the original FORTRAN source code rather than to the intermediate C code This flag is actually only a macro for the following flag settings Turn comments off Ln Include line number information LO Do not generate newlines within statement output Qe32000 Allow 32000 bytes for line number information These flags are discussed in detail elsewhere in this chap
113. 4 D5 auto float T3 n 1 T2 m 1 for v 1 D4 inc D5 m 1 v D4 4 D5 gt 0 5 4 T3 atv 1 v 1 1 10 0 test amp T3 return As requested all variables including the temporary ones are now assigned to auto storage The default setting of the temporary storage flag Ts specifies that temporary variables have standard storage i e that they be assigned a storage class in the same manner as user defined variables The Ta flag requests that all temporaries be assigned to auto storage regardless of the rules being used to allocate other variables Using the Ta flag the following output is generated void inc int n m inc float a extern void test auto int T1 T2 D4 D5 auto float T3 static int v 1 1 2 mt 1 for v 1 D4 inc D5 m 1 v D4 D4 D5 gt 0 D5 v D4 atv 1 v 1 1 10 0 88 PromulaFortran Translator User s Manual test amp T3 return Under this convention the program variable v is assigned to static storage using the standard storage rules however the temporary variables generated internally are all assigned to auto storage 2 34 Specifying Unit Numbers UR URnum UP UPnum UW UWnum FORTRAN has a variety of contexts with I O statements in which no explicit unit number is provided PRINT READ WRITE PUNCH ACCEPT etc There are a variety
114. 4log long TRUE return modi Under this default the actual comment symbol is removed and the remainder of the comment is left alone Each block of comments is preceded by the C comment delimiter and is followed by the delimiter The resulting appearance of the comments is very different from that of the source The COMMENTS statement shown below COMMENTS 0 whim n xxm mim m xm 0 0 whim m xxm mim m xm 0 END tells PROMULA to treat DML and DDL comments in the same manner The first section of each specification says not to have a separate header for the comment blocks The second section establishes the following correspondence between source and target comment introducers Source Target The third section says not to have a separate trailer for each comment block Using this COMMENTS statement in a configuration file produces the following result long modi arg swab flag arg xlit flag Tum PURPOSE General purpose initialization for MODI routines 110 PromulaFortran Translator User s Manual PARAMETERS OGICAL SWAB FLAG R jOGICAL XLIT FLAG R s RETURN VALUE None EX SIDE EFFECTS None Incorporates features to support portable LIO file handling 8 Apr 1986 long arg swab flag arg xlit flag extern void trlito static long K1 4 static long modi noninpbas
115. 5536 00000 51 00000 Long Real Results 15536 00000 5536 00000 51 00000 3 Selective promotion to long Short Integer Results 15536 5536 166 Long Integer Results 50000 60000 166 18 PromulaFortran Translator User s Manual Short Real Results 50000 00000 5536 00000 166 00000 Long Real Results 50000 00000 5536 00000 166 00000 In reviewing these results a negative number means that a short integer overflow has occurred The typical FORTRAN result is the first one In this instance the output of all integer calculations is a long That result is then converted to the desired result type Notice that even the intermediate addition in the division example is calculated as a long In the second case the result of any integer calculation is also always short regardless of the surrounding context This type of result is unusual for mainframe FORTRANSs and is common for PC FORTRANS Note that Microsoft FORTRAN allows the user to select which type of convention is to be followed as a side effect of the WORDSIZE metacommand The third case is strange and difficult to deal with The particular result above can be gotten from VS FORTRAN Note that in an integer context universal promotion to integer is followed Also in all cases the intermediate addition result is promoted to long But for some reason the multiplication result is allowed to overflow while the addition result is not In designing PROMULA FORTRAN we allow the user to
116. A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowrf Write FORTRAN single precision vector vmsuse Use a virtual value 6 240 VMSWRI Write FORTRAN Virtual Short Integer Vector Synopsis include fortran h PROMULA FORTRAN function declarations int vmswri value nval long value Virtual address of values to be written int nval Number of values to be written Description Writes a virtual vector of short integer values to the current output file in accordance with the current format specification Note that in this function each individual value is assumed to have its own corresponding format specification if a formatted write is being performed Return value 256 PromulaFortran Translator User s Manual A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowri Write FORTRAN short integer vector vmsuse Use a virtual value 6 241 VMSWRL Write FORTRAN Virtual Long Integer Vector Synopsis include fortran h PROMULA FORTRAN function declarations int vmswrl value nval long value Virtual address of values to be written int nval Number of values to be written Description Writes a virtual vector of long integer values to the current output file in accordance with the current
117. C allows multiple assignments to the same value to be written together The translator looks for such assignments and combines them whenever possible Note 3 that in FORTRAN the default base for a subscript is 1 Thus all DO loops which generate subscripts tend to start at 1 In C however subscripts start at zero This fact makes for much more efficient code The translator looks for DO loops whose only purpose is to move through array subscripts and reduces their range to start at zero thus producing a very natural looking for statement and optimizing subscript expressions Note 4 that C has and operators which take advantage of the fact that most computers have increment and decrement operators The translator uses these operators whenever possible Note 5 that C has operators like z Z etc The use of these operators ensures that the address of the left hand side of the assignment will only be computed as often as necessary PROMULA FORTRAN uses these operators Note 6 that the DO loop running to statement 10 in the FORTRAN code is collapsed into a single compound statement and that the now unneeded statement label is removed Note 7 that though the DO loop statements in loop 15 cannot be reduced to a single statement the statement label can still be removed In summary PROMULA FORTRAN looks for every opportunity to simplify and optimize the translation and to make it look as natural as possible in t
118. C translator which converts FORTRAN code to clean portable and maintainable C code while allowing extensive control over the translation process PromulaFortran Translator User s Manual No matter how old or how extended your FORTRAN dialect is PROMULA will process it by first compiling it to the more versatile and more portable C language Your long established FORTRAN programs do not have to be maintenance burdens running inefficiently on old platforms with PROMULA you can give them new life on contemporary platforms where you can take advantage of new technology options including the option of program maintenance in either FORTRAN or C 1 3 Compiler Advantages As a FORTRAN compiler PROMULA FORTRAN offers a number of advantages over other FORTRAN compilers Portability Compile to C rather than machine code Maintain a single FORTRAN source code on multiple platforms Port and process your applications on almost any platform that supports a standard C compiler Multi Dialect Compile standard FORTRAN 66 and FORTRAN 77 dialects as well as various other extended Processing dialects such as VAX IBM VS PDP PRIME Honeywell and Data General FORTRAN Multi Platform Achieve reproducible results on multiple platforms without having to maintain separate source Availability codes on each platform When you migrate from one platform to another bring your FORTRAN applications with you including your FORTRAN compiler PROMULA FORTRAN is a
119. CONSTANTS KA KS ccccccccccecssssssececececsesssscecececeessaaececececsesenseaeeeeees 55 2 18 MAXIMUM OUTPUT LINE WIDTH LNUM u cccccccesesccccceceesessccececececsescaeceeececeeeaaesecececseaascecececeeseaaeaeseescsesentsaeeeeecs 57 2 19 LINK TIME PROCESSING OF COMMON DATA MODULES LM LS esee nennen 57 2 20 INCLUSION OF LINE NUMBERS FOR DEBUGGING LN LO sees 59 2 21 FORTRAN DIALECT SELECTION FLAGS MDIALECT ccccsscccccecsesssseceeceeceessaececececeesesssecececcesessaeeeecceeseneaaeeeeees 60 iii t2 PromulaFortran Translator User s Manual 2 22 NESTING INDENTATION TO BE USED IN THE OUTPUT N NO NN seeeeeeeeeeenee 60 2 23 INLINE COMMENTS OUTPUT MARGIN WIDTH NCNUM 61 2 24 UPPER AND LOWER BRACES CONVENTION IN C NUI NU2 1 NL2 eene 61 2 25 NAME OF THE FILE TO RECEIVE THE C OUTPUT ONAME cccccccccecsssssececececsesssececececeeseseaecececeeseseaeeeeeceesensaaseeeees 61 2 26 SPLITTING OF OUTPUT INTO SEPARATE FILES OS OM cssssssccccecsessssecececceseseaececececeeseaeaececceeecessaaeceeececeesssaeeeeees 62 2 27 MISCELLANEOUS PROTOTYPING CONTROL FLAGS PNUMB P NUMB 62 2 27 1 PI Include Definitions of int Functions det ert et te e EX Ie se 63 2
120. DX READ FORTRAN COMPLEX VALUES eeeeeeeeene e e ene en i ase ese ese esa ese ases ese esas 204 vii 6 139 6 140 6 141 6 142 6 143 6 144 6 145 6 146 6 147 6 148 6 149 6 150 6 151 6 152 6 153 6 154 6 155 6 156 6 157 6 158 6 159 6 160 6 161 6 162 6 163 6 164 6 165 6 166 6 167 6 168 6 169 6 170 6 171 6 172 6 173 6 174 6 175 6 176 6 177 6 178 6 179 6 180 6 181 6 182 6 183 6 184 6 185 6 186 6 187 6 188 6 189 6 190 6 191 6 192 6 193 6 194 6 195 PromulaFortran Translator User s Manual FIORDZ READ FORTRAN DOUBLE COMPLEX VALUES eee e e e emen nnenenen 204 FIOREC POSITION A FORTRAN FILE ON A e en emen nsn ise ese n nsn nsns nsa nsns 205 FIOREW REWIND A FORTRAN e 205 FIORLN READ FORTRAN END OF LINE eee ee ene n enne nnn nnns nnns nsns e ise sensn n esa ese esa nsns a 206 EFIORNDV ROUND VALUE veteres a E umet utere 206 FIORNL PROCESS FORTRAN READ DATALIST STATEMENT eeeeeenee ere 207 FIORPATH READ PATHNAME CONVERSION INFORMATION cette e e e emenennnnnnn 207 FIORTXT READ NEXT TEXT 5 5 e e e ene nsns ese sess uses e ese ese ese e esa ese nsns n sese 209 FIORWBV FORTRAN
121. E 1 LISTIO STRG heading 10 0 WRITE 1 LISTIO INT4 modelhead gt hd_hedsiz INT4 modelhead gt hd_numtab 0 120 PromulaFortran Translator User s Manual WRITE 1 LISTIO INT4 modelhead gt hd_filrev INT4 modelhead gt hd_sysprim 0 return void tes long num rev num return t5 num rev num rev t2L char modelhead_area 16 Of course the KEYWORDS of this section can be combined with those of the last to produce the following demo cnf file KEYWORDS long INTEGER4 13 CHARACTER Wy e 9 area os struct Tg ES END This configuration then produces the following translation void test4 heading P1 CHARACTER heading int P1 extern void test5 extern char modelhead area typedef struct INTEGER4 hd hedsiz Length of header INTEGER4 hd numtab No of tables in file INTEGERA hd filrev Model file revision date INTEGER4 hd sysprim No of system primitive types struct modelheag auto struct modelhead modelhead struct modelhead modelhead area test5 amp modelhead hd numtab amp modelhead hd filrev modelhead hd hedsiz modelhead hd numtab modelhead hd filrev WRITE 1 LISTIO STRG heading 10 0 WRITE 1 LISTIO INT4 modelhead gt hd_hedsiz INT4 modelhead
122. E0a a return 0 long multiply a a r long a r E0a a EO0r return E0 1 long iset bvalue b long b 91 PromulaFortran Translator User s Manual return 2 Using this approach each time the main entry point uses copies of the arguments for the other entries This approach produces the desired result for this interpretation 2 36 2 Output Form of Parameter Identifiers Y2 Normally PROMULA converts all identifiers to lower case Thus the following code SUBROUTINE DEMO IV PARAMETER GoodBye 0 Hello 1 IF IV EQ Hello WRITE Hello IF IV EQ GoodBye WRITE GoodBye END produces the following default translation void demo iv long iv define goodbye 0 define hello 1 if iv hello WRITE OUTPUT LISTIO STRG Hello 5 0 if iv goodbye WRITE OUTPUT LISTIO STRG GoodBye 7 0 undef goodbye undef hello Here the parameter identifiers have been converted to lower case It is occasionally desirable at least for parameter values to retain the original notation The Y2 flag produces the following C output void demo iv long iv define GoodBye 0 define Hello 1 if iv Hello WRITE OUTPUT LISTIO STRG Hello 5 0 if iv GoodBye WRITE OUTPUT LISTIO STRG GoodBye 7 0 undef GoodBye undef Hello In this translation the
123. F FUNCTIONS cccssessscecceeceesenececececseseueceseeececsesuaaesecececseuaaeceeccecsessaaeseeseeeseseaaeeeeess 133 5 2 GENERAL FORTRAN OPERATIONS cccssessscecececeesssecaeccecsenssececececsensseeesecsesesesseeeseeeceeseaaecececeecessaaececeescsessaaeeeeecs 134 3 3 INPUT OUTPUT OPERATIONS reete cree ipee ete RR EH cet rette ien Uie etd ewes 134 3 3 Runtime Error Messag s s sies e CE e ERE 134 5 4 NONCOMPLEX INTRINSIC FUNCTIONS ccsesscccccceceesssececceecsessaececececseseaeceeeeeccesesseseceeccecsessaaesecececesensaeceeececsensaaeaeeees 135 25 5 MIRTUAT2MEMOR YS YS TEM Nnm EA ad RO ERR 136 5 5 1 The Virtual Memory Management Algorithm 136 5 5 2 Virtual Memory Global Variables eanne aE e EEE enne nennen 137 5 6 SINGLE PRECISION COMPLEX ARITHMETIC cccccsessscecccecsessaececececseneaseceseeeceeneuaaececececseuaaeceeccecsesaueceeececsessaaeeeeecs 138 5 7 DOUBLE PRECISION COMPLEX ARITHMETIC ccccsessscecccecsesssesecececseseseceseeccscnesesecceccecsesaaeeeeececsessaaeseeeceesessaaeaeeess 138 6 RUNTIME LIBRARY FUNCTION DESCRIPTIONS ee eee eee ee eene seen eese esten sette setas eese ense esee tease eese tassa 139 6 1 CPXABS COMPUTE THE SHORT COMPLEX ABSOLUTE VALUE ccsessssssececsessnsececceecsensnececececeeseneaeceeeccesenseaeeeeees 139
124. Inquire about file record length 14 Inquire about file current record number 15 Inquire about file current blank convention Return value None this function is void See also None 6 114 FIOFMTV Compute FORMAT Value Synopsis include fortran h PROMULA FORTRAN function declarations int fiofmtv Description Computes the value of an integer constant in the FORMAT statement and updates the current format position so that it points to the first nonblank character beyond the end of the value Note that if the current character in the format is nonnumeric when this function is called then this function does not move the current position and returns a zero value Return value The value of the integer constant or a zero if there was no integer constant See also fiofwsp Skip white space in format 6 115 FIOFOUT Formatted Output Operations Synopsis include fortran h PROMULA FORTRAN function declarations void fiofout context int contex Context of call 0 value 1 end Description Performs nonvariable related formatted output functions until an end of format or a rescan or variable related specification is encountered The actual operations performed by this function are as follows Specification Code Description nH 1 Write characters from a Hollerith string cl cn 1 Write characters from a delimited string 191 PromulaFortran Translator User s Manual nX 2 Skip right n places TRn 2 Skip right
125. Multiplication Synopsis 155 PromulaFortran Translator User s Manual include fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of the number dcomplex dcomplex dpxmul a b dcomplex a The left hand value dcomplex b The right hand value Description Multiplies two double complex numbers to form a third Return value The double complex result of the multiplication See also None 6 38 DPXNEG Compute the Double Complex Negative Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of the number dcomplex dcomplex dpxneg a dcomplex a Contains the value to be transformed Description Computes the negative of a double complex value Return value The double complex result See also None 6 39 DPXPOL Double Complex Conversion to Polar Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of the number 156 PromulaFortran Translator User s Manual dcomplex dcomplex dpxpol a dcomplex a The value to be converted Description Converts a complex number into its polar form By definition r i in Cartesian form is exp 1 0 Return value The double complex result of the conversion See al
126. N initialization calls end of COMMON initialization function pattern string for external functions pattern string for argument surrogates pattern string for local COMMON variable pattern string for COMMON structure pattern string for COMMON external pattern string for translated include names pattern string for include endif pattern string for include ifndef pattern string for include define The KEYWORD statement in the configuration file begins with the command KEYWORD as a single line It is followed by a series of lines each containing a keyword string in quotes or a simple keyword identifier followed by the substitute string to be used in quotes The end of the replacement pairs is indicated by the word END on a single line The following subsections give several examples in which the look of the translation of the following code is altered using a KEYWORDS statement SUBROUTINE TEST4 INTE G INTEGER 4 hd_numtab INTEGER 4 hd_filrev INTEGER 4 hd_sysprim COMMO hd hedsiz W RETURN S ER 4 hd hedsiz Length of header No of tables in file Model file revision date No of system primitive types N MODELHEAD hd hedsiz hd numtab hd filrev hd sysprim CALL TEST5 hd numtab hd filrev hd numtab hd filrev RITE 1 hd hedsiz hd numtab hd filrev hd sysprim UBROUTINE TEST5 num rev 118 PromulaFortran
127. NI FORTRAN Intrinsic Function MINI Synopsis finclude fortran h PROMULA FORTRAN function declarations long fifminl al a2 float al First value to be compared float a2 Second value to be compared Description Determines which of two double precision values is the smallest Return value The smallest value converted to a long integer See also None 6 89 FIFMOD FORTRAN Intrinsic Function MOD Synopsis include fortran h PROMULA FORTRAN function declarations long fifmod num dem 176 PromulaFortran Translator User s Manual long num The numerator for the calculation long dem The denominator for the calculation Description Computes the value of the remainder of num divided by dem If dem is zero then the result is zero Return value The value as computed above See also None 6 90 FORTRAN Intrinsic Function NEF Synopsis finclude fortran h PROMULA FORTRAN function declarations int fifnef a b float First value to be compared float b Second value to be compared Description Compares two floating point numbers to determine if they would be unequal in single precision floating point form Return value The short integer result of the comparison See also None 6 91 FIFNINT FORTRAN Intrinsic Function NINT Synopsis finclude fortran h PROMULA FORTRAN function declarations long fifnint a double a Argument containing the value Description Computes the nearest integ
128. Natural Logarithm Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of the number dcomplex dcomplex dpxlog a dcomplex a The value to be transformed Description Computes the double complex natural logarithm of a double complex value Return value The double complex natural logarithm See also None 154 6 33 DPXLOG10 Synopsis finclude fortran h typedef struct double cr double ci dcomplex dcomplex dpxlogl10 a dcomplex a Description PromulaFortran Translator User s Manual Double Complex Base 10 Logarithm PROMULA FORTRAN function declarations The real part of the value The imaginary part of the number The value to be transformed Computes the double complex base 10 logarithm of a double complex value Return value The complex base 10 logarithm See also dpxlog 6 36 DPXLONG Synopsis include fortran h typedef struct double cr double ci dcomplex long dpxlong a dcomplex a Description Computes complex natural logarithm Convert Double Complex to Long PROMULA FORTRAN function declarations The real part of the value The imaginary part of the number Contains the value to be transformed Computes the real part of a double complex value and converts it to long Return value The long result See also None 6 37 DPXMUL Double Complex
129. ORTRAN may contain statements which require runtime support not included in this library In this case you can add the additional functions needed 1 12 Dealing with Common Blocks If there is any aspect of FORTRAN that can destroy the validity of a translation itis COMMON blocks especially when combined with EQUIV ALENCE statements Every storage trick ever conceived gets used in the nuances of changing COMMON block definitions through a large FORTRAN program There is no best way to translate COMMON blocks Some compilers for example insert extra bytes to achieve various types of alignments and programmers using these compilers will take that fact into account without any warning or comment for the user PROMULA FORTRAN translates COMMON blocks in one of four ways It is up to you to select the appropriate one depending upon your needs and preferences 1 The default is to declare the COMMON block identifiers simply as external void pointers and then to assign them locally to a structure pointer whose members are defined in the same manner as the COMMON definition in the routine This technique works in all cases except where alignments are needed 2 Incases where COMMON blocks are always defined in the same manner the common blocks are simply defined as external structures This gives more efficient code than the technique above 3 The most efficient technique can be used when no games are played with the common blocks at all Then th
130. OUT 9 STATUS fprintf LUN 6 Short Integer Results fprintf LUN 6 Long Integer Results fprintf LUN 6 Short Real Results fprintf LUN 6 Long Real Results exit 0 The result of running this translation is shown below EW 0 13d 13d 13d n isum iprod iquot 131d 131d 131d n jsum jprod jquot 13 5f 13 5f 13 5f n rsum rprod rquot 13 5 13 5 13 5f n dsum dprod dquot Short Integer Results 15536 5536 Long Integer Results 15536 5536 Short Real Results 15536 00000 5536 00000 Long Real Results 15536 00000 5536 00000 51 51 51 00000 51 00000 This result agrees with the no automatic promotion to long output from above and is probably wrong for most applications It should be noted that all other FORTRAN to C translators for the PC that we have reviewed produced only the no automatic promotion to long result If your program uses any short integer variables be certain to consider this semantics problem 2 4 Detailed C Output Format CF1 CF2 CF4 CF8 CF16 20 PromulaFortran Translator User s Manual Though the user has complete control over the actual content of the C output produced by PROMULA FORTRAN by using various other command line switches and by modifying the content of the configuration file see the chapter on the configuration file there are still a few miscellaneous aspects of the look of the C output that ca
131. PROMULA FORTRAN function declarations float fiftan x float x Argument containing the value Description 180 PromulaFortran Translator User s Manual Computes the tangent of a floating point argument Return value The value as computed above See also None 6 99 FIFTIME FORTRAN External Function TIME Synopsis include fortran h PROMULA FORTRAN function declarations void fiftime cl ncl char cl Returns the time in string form int ncl Length of the time return string Description Obtains the current time from the operating system Return value The current time in hr mi sec form See also None 6 100 FIFXBIT FORTRAN Intrinsic Function XBIT Synopsis include fortran h PROMULA FORTRAN function declarations int fifxbit bits ibit unsigned char bits Bit vector to receive new bit value int ibit Number of bit to receive value Description Extract a specified bit value from an arbitrary bit string The left most bit position is numbered 1 with higher bits receiving higher values Return value The function returns a one or a zero depending upon whether the bit specified is on or off See also None 6 101 FIFXCREP FORTRAN Extended Character Representation Synopsis include fortran h PROMULA FORTRAN function declarations 181 PromulaFortran Translator User s Manual char fifxcrep chrs char chrs Vector of characters Description In several dialects of FORTRAN especial
132. REWRITE BINARY VALUES eeeeeeeeeeee ee eene ese e nsns ase nsn esas nsu 210 FIOSHE SHIEE STRING A T ettet eter eee rete eet e aa eite bte deett 210 FIOSHR SHIFT STRING RIGHT ene e ene e nnn nnns nsns nsu te e e ete se ase ese e ese ese ases 210 FIOSPACE SKIP WHITE SPACE IN RECORD eeeeeeeene e ene 211 FIOSTATUS SET FORTRAN I O ERROR en e e esu nsa esa 211 FIOSTIO ESTABLISH FORTRAN STANDARD I O nsn nsns nsns asas 211 FIOSTOD CONVERT STRING DOUBLE n ene ener nnns nsns uses 212 FIOSTOI CONVERT STRING TO INTEGER eee ee e ene a a ae eE 213 FIOUWL ESTABLISH FORTRAN UNFORMATTED WRITE LENGTH eee 213 FIOVFINI INITIALIZE A VARIABLE FORTRAN FORMAT sesesesesererererererererererererererererererererererererererererererererere 213 FIOWALPH WRITE ALPHABETIC INFORMATION eere e e ene emen nn nnnm ener nnns nn sese sisi ese esi e esi n nsn ese asa nsns 214 FIOWBIV FORTRAN WRITE BINARY VALUES eeceeeeeeeene e e e 214 FIOWDBL WRITE DOUBLE PRECISION VALUE eee e e e ene emen nnnm ener 215 FIOWEF FORTRAN WRITE END OF FILE
133. RG I equals 10 11 0 if i 11 WRITE OUTPUT LISTIO STRG I equals 11 11 0 return Now the for loop is also reduced to a compound statement since it now contains a single multiple assignment 2 39 Constant Reduction Optimization Xc Yc 94 PromulaFortran Translator User s Manual The FORTRAN input processor often generates sequences of constant integer calculations which can be reduced prior to their output in the C The Yc switch allows this simplification to be performed and is the default for all biases The Xc switch excludes this optimization As an example of this constant reduction operation consider the following subprogram which contains various constant subscripts SUBROUTINE DEMO A B C DIMENSION A 10 B 6 7 C 3 4 5 A 5 5 0 B 2 3 6 0 C1727323 20 RETURN END The default intermediate C form for this fragment is as follows void demo a b c float a b 7 6 c 5 41 3 a 4 5 0 b 2 1 6 0 c 2 1 0 7 0 return Notice that in this output the values of the subscripts have been reduced by one over their FORTRAN original values This is because in FORTRAN subscripts have a default starting value of one while in C they always start at zero That this reduction is in fact being performed can be seen via the Xc switch which blocks constant reduction The C output with the Xc flag active looks as follows void demo a b c float a
134. RITE FLOATING POINT VALUE eee 223 FIOWVB WRITE FORTRAN BOOLEAN VALUE eeceeetete e e ene emen n nennen ener nnns nsns etes esse use e ese esu esa ese nsns nana un 224 FIOWVC WRITE FORTRAN CHARACTER VALUE eee e e ene e 224 FIOWVD WRITE FORTRAN DOUBLE 0 224 FIOWVF WRITE FORTRAN FLOAT VALUE eee ee e e enenemennmnnnnnnnnnn nnns nsus n ese senes isis sse uses ese ases 225 FIOWVI WRITE FORTRAN SHORT INTEGER 225 FIOWVL WRITE FORTRAN LONG INTEGER VALUE ee e e 226 FIOWVS WRITE FORTRAN STRING VALUE eeeeeeceeenne e ene e ese ese ese ese ese ese 226 FIOWVT WRITE FORTRAN TRUTH VALUE eee e e eene nnn nn enhn nn nn nnns nsns nsus esu i uses n esu ese isa nsn ese nsn n 227 FIOWVU WRITE FORTRAN CHARACTER VALUE ccccccccscscsccscececececececececesececesecsceseeesececesaseesseuesssessseseeseaeaes 227 FIOWVX WRITE FORTRAN COMPLEX VALUE 228 FIOWVZ WRITE FORTRAN DOUBLE COMPLEX VALUE 2 2 228 FINADS FORTRAN ADD STRINGS cte seteceexececeece eae stet cbe ete covets eene dp ehe e ea Cab deu 229 FTNALLOC ALLOCATE DYNAMIC MEMORY ececetenee e e e e E Or Neeb Ehee eSEE Erbi ase ese uses e e
135. ROMULA FORTRAN function declarations void fiobfout value bfmt double value Value to be converted char bfmt Business formatting descriptor string Description 182 PromulaFortran Translator User s Manual This utility function controls the conversion of a double precision value to string form under the control of a business formatting descriptor string The length of the string determines the field width of the final display If this width is too small for the number then the output will be a string of asterisks filling the field Valid characters for the string are ZS CR The use of these valid characters is explained below Plus If only the first character is then the sign of the number or is entered in the leftmost portion of the field fixed sign If the string begins with more than one sign they will be replaced by blanks and the sign of the number or will be printed in the field position immediately to the left of the first printing character of the number floating sign If the rightmost character of the string is then the sign of the number or will be printed in that field position following the number trailing sign Minus The minus sign behaves the same as a plus sign except that a space blank is entered instead of a if the number is positive plus sign suppression Dollar sign A dollar sign may at most be preceded in the string by an optional fixed sign A s
136. RTRAN function declarations void ftnalloc nbyte long nbyte Number of bytes to be allocated Description Returns a pointer to a storage area on the heap which is at least as long as the requested number of bytes If the bytes cannot be obtained then this function exits to the operating system Return value A pointer to the start of the storage area See also None 229 PromulaFortran Translator User s Manual 6 189 FTNBACK FORTRAN Backspace Statement Synopsis finclude fortran h PROMULA FORTRAN function declarations int ftnback lun int lun Logical unit number Description Executes the FORTRAN BACKSPACE statement when translated via a non optimized user bias In addition to the logical unit number this function takes a variable number of parameters which specify the actual data to be supplied to the backspace operation The parameter type codes passed to this function are as follows Code Parameter Description of Use 0 Ends the list of parameters 1 long Points to an error return variable Return value A zero if the backspace went well else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiointu Establish an internal file fiolun Establish unit number fioback Backspace current file fioerror FORTRAN error processing function fiostatus Set I O Error Status 6 190 FTNBLKD FORTRAN BLOCK DATA Synopsis
137. RTRAN runtime library is a set of approximately 250 functions designed for use with the C output of PROMULA FORTRAN It may also be used by those FORTRAN programmers who wish to program in C but who do not wish to give up the input output conventions formatting controls and intrinsic functions which they have grown used to This chapter is provided primarily for those user s of PROMULA FORTRAN who intend to maintain the C output independently of their FORTRAN originals Initially C codes using this library can be produced by producing C source code using PROMULA FORTRAN Once in C the programs may then be maintained by using these functions and their documentation The source code for all functions in this library is available If there is one certainty it is that no two FORTRANSs behave in the same way especially with regard to their runtime libraries Thus if your conventions differ from the ones used here or if you require some specialized behavior you may alter the library code Alternatively your version of FORTRAN may contain statements which require runtime support not included in this library In this case you can add the additional functions needed For more discussion of this topic see the chapter in this manual on the implementation of nonstandard FORTRAN dialects Another use of the material in this and the next chapter is to optimize the runtime behavior of the C output for special platforms or special needs Much of this customiza
138. Set the plus sign to a BN 7 Set blanks to null BZ 8 Set blanks to zero 9 Physically write the current line nP 10 Set the floating scale factor to n Return value Note the function is void however the global variable ioier may be set to an error code if a problem is encountered See also fionxtf Get next format specification fiorchk Check fixed input field fiortxt Read next text record 6 113 FIOFINQU Inquire About File Data Synopsis finclude fortran h PROMULA FORTRAN function declarations void fiofinqu option str ns int option Specifies data being inquired about char str Information to be returned int ns String length or integer information Description This function is used to inquire about the various file data options associated with the current FORTRAN file structure The particular data being inquired about is defined by the opt ion parameter as follows Option Description of data 1 Inquire about status of file 2 Inquire about existence of file 3 Inquire as to connection status of file 4 Inquire as to file s external unit number 5 Inquire whether file has a name 6 Inquire for name of file 7 Inquire as to files access method 190 PromulaFortran Translator User s Manual 8 Inquire if file can be accessed sequentially 9 Inquire if file can be directly accessed 10 Inquire if file is formatted 11 Inquire if file can be opened as a text file 12 Inquire if file can be opened as binary 13
139. T FREE FORM FIELD nnns nsns esi isis ese esu use ase ese esas nsu 188 FIOFINI INITIALIZE A FORTRAN 188 FIOFINP FORMATTED INPUT eeeeeeeeeeeeenn e enn e asa a ete ase ese ese uses 189 FIOFINQU INQUIRE ABOUT FILE DATA cesses enne 190 FIOFMTV COMPUTE FORMAT VALUE 191 FIOFOUT FORMATTED OUTPUT OPERATIONS eeeeeeeenene e e e e enne ese ese esa ese ese uses e nsn au 191 FIOFVINQ INQUIRE ABOUT FILE V ALUE cccccsessssecececsesessssecececscsesaeececccecsesnaaecceccecsenaaecececsesensaaeseseceeseneaaeea 192 FIOFWSP SKIP FORMAT WHITE SPACE 193 FIOINTU ESTABLISH FORTRAN INTERNAL UNIT 193 FIOITOS CONVERT INTEGER TOSTRING esas ases 194 FIOLREC POSITION A FORTRAN FILE ON A 2222 194 FIOLTOS CONVERT LONG INTEGER TO STRING eeeeeeeeeeeeene nene e ene 195 FIOLUN ESTABLISH FORTRAN UNIT e e e emen ese sanas nsu 195 FIONAME ESTABLISH FORTRAN UNIT BY NAME eee e e e eme enennnennn nn nnn nnns nsns ise enisi esu usn nsa ese esas nsns 196 FIONX
140. TE Variance VAR RETURN END Fa SUBROUTINE EX001 VAL N XBAR VAR DIMENSION VAL N WRITE 1x a ANALYSIS XBAR 0 VAR 0 0 10 J 1 N XBAR XBAR VAL J CONTINUE XBAR XBAR N WRITE 6H Mean F10 3 XBAR DO 15 J 1 N S VAL J XBAR VAR VAR 5 S CONTINUE IF VAR NE 0 0 VAR VAR N 1 WRITE Variance VAR RETURN END FORTRAN BIAS TRANSLATION void 001 1 xbar var long n float val xbar var static long j static float s WRITE OUTPUT VFMT 1x a CSTR ANALYSIS 0 xbar 0 0 var 0 0 for j 0 j n j xbar 1 xbar xbar n WRITE OUTPUT VFMT 6H Mean F10 3 REAL4 xbar 0 for j 0 j n j S val j xbar var varts s if var 0 0 var var n 1 WRITE OUTPUT LISTIO CSTR Variance REAL4 var 0 return C BIAS TRANSLATION void long ex001 1 var float val xbar var static long j static float s printf ANALYSIS n n n xbar var 0 0 for j 0 j n j xbar val j xbar n printf Mean 10 3f n xbar for j 0 j n j S val j xbar var s s if var 0 0 var n 1 printf Variance 16 6E n var return 4 8 11 13 12 13 11 8 9 13 11 12 13 10 11 In general the objective of FORTRAN bias is to make the C
141. TEGER is defined via the FIs and switches The following shows the same FORTRAN translated with the setting void demo static short i j static int m n k 1 WRITE OUTPUT LISTIO INT2 i INT2 j INT4 k INT4 1 INT4 m INT4 n 0 return Note I and J are short and the rest are int Everything else is the same What has changed is only how the variables are labeled 2 17 Treatment of Internally Generated Constants Ka Ks If PROMULA FORTRAN encounters a constant subprogram argument which must be passed by address it introduces a variable to contain this value The K switch allows you to control the storage type of these variables The K switch has two settings Setting Meaning Ks Declare internal constants as static C bias default Ka Declare internal constants as auto FORTRAN bias default Consider the following simple FORTRAN fragment SUBPROGRAM DEMO CALL ALPHA I 3 0 STOP END Its default translation is as follows void demo extern void alpha static float K1 3 0 static long i 55 PromulaFortran Translator User s Manual ftnini argc argv NULL alpha amp i amp K1 STOP NULL In this translation a static variable K1 is introduced whose value contains the appropriate constant The address of this variable is then passed to the subprogram The Ks switch produces the above Under Ka the translation is as
142. TF GET NEXT FORMAT SPECIFICATION eeeeeceeetennne 196 FIOOPEN OPEN CURRENT FORTRAN 197 FIORALPH READ ALPHABETIC INFORMATION eene e ene 198 FIORBIV FORTRAN READ BINARY 22222 198 FIORCHK CHECK FIXED FORM INPUT FIELD e eene nennen nnn nennen rre renun 199 FIORDB READ FORTRAN BOOLEAN VECTOR eee e e e e ese sa usse nasus ase ananas nana 199 FIORDC READ FORTRAN CHARACTER VECTOR eeeeeeeeeenene ene ene n ener ener nnns nsus nsi ines nsi ese ese hse a ese senes ases u 200 FIORDD READ FORTRAN DOUBLE PRECISION VECTOR seseseserererererererererererererererererererererererererererererererere 200 FIORDF READ FORTRAN FLOATING POINT 8 201 FIORDI READ FORTRAN SHORT INTEGER 201 FIORDL READ FORTRAN LONG INTEGER e en eene nnnnnnn ener 202 FORDS READ FORTRAN STRING ics reete rete ck desc ce tee eR ep eee dete aide ag 202 FIORDT READ FORTRAN TRUTH VALUE 203 FIORDU READ FORTRAN UNSIGNED CHAR VECTOR eese ene emen en ennn nnns nnns nsns nsns ines e isi esa e nsns esas 203 FIOR
143. TIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE AND ALL SUCH WARRANTIES ARE EXPRESSLY AND SPECIFICALLY DISCLAIMED IN NO EVENT SHALL GREAT MIGRATIONS LLC BE RESPONSIBLE FOR ANY INDIRECT OR CONSEQUNTIAL DAMAGES OR LOST PROFITS EVEN IF GREAT MIGRATIONS LLC HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGE Some states do not allow the limitation or exclusion of liability for incidental or consequential damages so the above limitation or exclusion may not apply to you TRADEMARK PROMULA is a registered trademark of Great Migrations LLC DEFINITION OF PURCHASE The definition of your particular purchase is specified in the Great Migrations LLC License Agreement which came with the PROMULA FORTRAN product and which you completed and returned to Great Migrations LLC If you have any questions about your rights or obligations as a PROMULA FORTRAN user or believe that you have not received the complete PROMULA FORTRAN package that you purchased please contact Great Migrations LLC 7453 Katesbridge Ct Dublin Ohio 43017 614 761 9816 PromulaFortran Translator User s Manual Table Of Contents PLEASE READ THIS SECTION P 1 1 INTRODUCTION as sssecnsctisvcastsnnvenscesenusssndsedetesaucecscssvsescestecsnssetecsteedevshstsnesdescecevansbotecvacesencesseed suesesencsvotedesnssssecsbeasessesvese 2 1 1 USER SUPPORT eee er tam idR RU een repens RIP 2 132 WHATISPROMULATORTRANN A ite e e i ten
144. TRAN 66 the value of v is not compared with that of e2 until the bottom of the loop therefore the loop is always executed once In FORTRAN 77 the comparison of v is performed at the top of the loop therefore if 1 exceeds e2 initially the loop is never incremented Officially do loops are to be executed as follows 1 The expressions e1 e2 and e3 are evaluated and then converted to the type of the control variable v if necessary yielding the values m1 m2 and m3 2 The control variable v is assigned the value of m1 3 The iteration count is established as follows ic MAX INT ms mi 4m3 m3 mtc where mtc is the minimum iteration count and equals 0 if FORTRAN 66 conventions are desired and 1 if FORTRAN 77 conventions are desired 4 If icis not zero the loop is executed else execution continues beyond the end of the loop 5 The control variable is incremented by the value m3 ic is decremented by 1 and execution loops back to step 4 To translate DO loops in the official manner then requires introducing two temporary variables for each DO loop the iteration counter and the value of the increment since this increment may be changed within the loop Consider the following FORTRAN subroutine SUBROUTI INTEGE DO 10 WRITE 10 CONTI RETUR END Qe lt R U m In this loop all of the controls ar
145. Translator User s Manual NTEGER 4 num rev um num rev 2 ETURN ND Ed m B H The default translation of this code is as follows void test4 heading P1 char heading int Pl extern void test5 extern char Xmodelhead typedef struct long hd hedsiz Length of header long hd numtab No of tables in file long hd filrev Model file revision date long hd sysprim No of system primitive types j Cmodelhead auto Cmodelhead Tmodelhead Cmodelhead Xmodelhead test5 amp Tmodelhead gt hd_numtab amp Tmodelhead gt hd_filrev Tmodelhead gt hd_hedsiz Tmodelhead hd numtab Tmodelhead hd filrev WRITE 1 LISTIO STRG heading 10 0 WRITE 1 LISTIO INT4 Tmodelhead gt hd_hedsiz INT4 Tmodelhead gt hd_numtab 0 WRITE 1 LISTIO INT4 Tmodelhead gt hd_filrev INT4 Tmodelhead gt hd_sysprim 0 return void test5 num rev long num rev num num rev 2L return char 1 16 3 5 1 Simple Keyword Replacement To begin the detailed discussion of keyword replacement let us assume that the user prefers FORTRAN looking variable type names He has made the following additions to the fortran h file define INTEGER4 long define CHARACTER char and wishes to have the corresponding changes made in his PROMULA output The following demo cnf f
146. VERSIONS CL CS CO C1 C2 14 2 3 1 Arithmetic with Short Integer 17 2 4 DETAILED C OUTPUT FORMAT CF1 CF2 CF4 CF8 CF16 20 2 5 TREATMENT OF CHARACTER VARIABLES CHD CHR CHS CHV esee ener ener 24 2 5 1 Initializing Character Values ie e eI ed ee eee e eee PE ee en UI oi 26 2 2 2 Subprogram Arguments ete eee ree eto eee RR e Pe ee Wes rE Ee a 27 23 SUDSTHRgs de p eU ee ee etii eee Gee diee terere De dier o iere rece eoe ud 3l 2 5 4 Character Concatenatlonss used he eet een e evi vee Baa nedivsatemandalvodtetverescirenider teebsee o eut iue dt 32 2 5 5 Character Treatment Conclusion ccccccccccssseeecececccccssescececcccussseseecccessususeceececessuuseccececeusuuscecessseauaaencececessuuanense 33 2 6 APPEARANCE OF COMMENTS IN C OUTPUT CMO CM2 ccccccccccccecssssscececceeesenssceceeeceesenseaeceeecescessaaseeeess 33 2 7 TREATMENT OF DATA INITIALIZATIONS DA DC DR o cceccececcccccecessssseceeceeceessasececececseneaueaeceeeeceseseaeceeeceesesssaeeeeees 34 2 7 1 Overview of Initialization Problem is ce t e ed e ede e ea ee FC te Pe d Re Ub e dea 35 2 7 2 The Initialization SwitGhes 3 a ete recette be eas mee Dread eek e e evt 37 2 8 TURN ON DEBUGGING MODE DB 55 n
147. a ib 8 8 lt WRITE OUTPUT LISTIO INT4 ic 0 STOP NULL long ialpha ia gval P1 P2 char ia gval int P1 P2 static long ialpha 72 PromulaFortran Translator User s Manual WRITE OUTPUT LISTIO STRG ia P1 STRG gval P2 0 ialpha P1 P2 return ialpha This translation is correct and will probably produce acceptable results There are two disadvantages to the prototype approach First all the function prototypes must be known note that these can be produced by PROMULA FORTRAN Once produced the translation can then be done a second time using them The other problem is that the variables 1A and IB are still being treated as integer variables even though it is clear from the context that they are being used to contain character values The P2048 flag forces variables to have an explicit character type when this appears to be obvious from the context As is true of many other flags and switches this one is a translation aid it is not guaranteed to produce correct results in all cases The translation of the above using P2048 without any additional prototype needed is the following void main argc argv int argc char argv extern long ialpha static char ia 8 tlt tlt t t n Ext DE static char ib 8 P W L O mt lYt tt dgt PE static long ic ftnini argc argv NULL ialpha ia ib 4 4 lt WRITE OUTPUT LISTIO
148. able length record processing logic Variables whose size exceed this threshold are allocated their own storage areas from system memory The default setting of this threshold value is 512 On memory limited machines large programs that do massive data initializations might want to change this value Since this is a tuning value as opposed to an actual allocation value no direct message pertaining to it is generated 2 29 4 QXnumb Size of External Information Storage PROMULA saves the names and types of external symbols that have been declared via EXTERNAL statements and that have been passed to other functions but whose final type is not known See the discussion of the P128 prototyping control flag for detailed information on this topic PROMULA also saves names of those subprograms that have initializations for COMMON blocks See the discussion of link time processing of COMMON data modules the Lm and Ls command line switches for a discussion of this storage The default setting for the QXnumb switch is 2000 If the external symbols storage area is exceeded the fatal error 519 78 PromulaFortran Translator User s Manual The unresolved external symbol storage of nnnn bytes is insufficient use the QXn flag to increase it 18 generated 2 29 5 QHnumb Size of Include File Information Storage When include files are being translated independently of the source codes containing them via the P16384 switch a storage ar
149. able positions within the block are calculated using alignment and wordsize specifications that are supplied via the dialect definition Actual variable references then become references to the COMMON block name plus the calculated position The COMMON variable identifiers within the block disappear The code generated is efficient however if you intend to maintain the code in its C form it is difficult to read This technique is highly recommended for those who are using PROMULA FORTRAN simply as a preprocessor to their C compiler with no real interest in the intermediate C output Gv Defines COMMON variables as independent external symbols This technique should only be used when you are building a function library or when all COMMON blocks are always defined in precisely the same way This setting removes the COMMON blocks from the translation thus it is the inverse of the technique above which removes the variables within the blocks With this technique the variables themselves become external symbols If appropriate this setting produces very readable and clean code however if used in the wrong context it can produce hash Ga This treatment assumes that the user has himself allocated or assigned memory for the common areas thus the COMMON variable itself becomes a structure pointer as opposed to a structure as assumed by all other treatments The COMMON symbol itself is defined as an instantiated pointer to that structure thus giving a v
150. alculated this value during the translation phase 2 The maximum memory space that can be used is 254 times the memory block size which is 1024 bytes in this implementation The current maximum is then 254K which is clearly ample given a 640K overall memory limit on the typical MS DOS PC 3 The maximum file size that can be accommodated is limited by the maximum length of the block status vector Each K of memory space allows for a megabyte of virtual space given the 1024 blocksize being used here So this is no problem A minor problem faced by this algorithm is that the criteria for writing a block to the disk is based on its use pattern consequently a block high in the file might be written before some blocks below it thus creating potential holes in the file Though such holes are acceptable to many operating systems we have decided to avoid them here The virtual algorithm must therefore remember whether a given block has been placed on the disk because it has valid information or simply to fill a hole 5 5 1 The Virtual Memory Management Algorithm The virtual memory management system is controlled via three basic storage elements 2 The block status vector 3 The memory block address vector 4 The virtual memory block The block status vector is the key to this algorithm There must be one entry for each block of memory To keep this vector as memory efficient as possible its entries can be stored as unsigned bytes The particula
151. allocation is slow as compared to auto variables and the allocation itself must be done explicitly The advantage is that the heap is generally quite large Using PROMULA FORTRAN you have complete control over the allocation of variables in the translation 1 14 A Sample Translation to C The output from PROMULA FORTRAN does not look like machine translation To present some of the design features of the translator consider the following example program which computes the mean and variance of a set of values Input Output Notes SUBROUTINE EX001 VAL N XBAR VAR void ex001 float val int n float xbar float var 1 DIMENSION VAL 0 0 auto int j VAR 0 0 auto float s DO 10 J 1 N xbar var 0 0 XBAR XBAR VAL J for j20 j n j xbar 1 2 10 CONTINUE xbar n 3 6 XBAR XBAR N for j20 j lt n j 5 DO 15 J 1 N s val j xbar 3 4 S VAL J XBAR var s s VAR VAR 5 5 5 15 CONTINUE n 1 7 VAR VAR N 1 5 RETURN END Note 1 that the parameter n is not declared as a pointer since it is not changed within the routine PROMULA FORTRAN uses what are called prototypes of subprogram arguments so that it can generate optimal calling sequences These prototypes may be specified by the user or may be determined internally by the translator The above was internally determined by the translator Note 2 that
152. alue A zero if the numbers are the same else a one See also None 140 PromulaFortran Translator User s Manual 6 5 CPXCOS Compute the Short Complex Cosine Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex complex cpxcos a complex a Contains the value to be transformed Description Computes the value of the short complex cosine of a short complex number By definition eli a 4 g i a cos a E oe Return value The short complex result See also float fifsncs Compute single precision sin cosine 6 6 CPXCPX Convert Two Floats to Short Complex Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex complex cpxcpx d1 d2 double di Contains the real value double d2 Contains the imaginary value Description Forms a short complex number whose real and imaginary parts are specified values Return value The short complex result See also None 141 PromulaFortran Translator User s Manual 6 7 CPXDBL Convert Double Precision to Short Complex Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex complex cpxdbl
153. alue Virtual address of values to be written int nval Number of values to be written Description Writes a virtual vector of long truth values to the current output file in accordance with the current format specification Note that in this function each individual value is assumed to have its own corresponding format specification if a formatted write is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowrt Write FORTRAN long truth vector vmsuse Use a virtual value 6 244 VMSWRU Write FORTRAN Virtual Unsigned Character Vector Synopsis include fortran h PROMULA FORTRAN function declarations int vmswru value nval long value Virtual address of characters to be written int nval Number of characters to be written Description Writes a virtual vector of character values to the current output file in accordance with the current format specification Note that in this function each individual character is assumed to have its own corresponding format specification if a formatted write is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes 258 See also fiowru vmsuse 6 245 VMSWVB Synopsis finclude fortran h typedef str
154. ample code under the Gp option note that the common structures are defined as simple arrays of char with the appropriate length at the end of the compilation char Xalpha 428 char Xbeta 200 These declarations are needed to force the linker to allocate sufficient space for each common area Now most but not all linkers will allocate such areas on an appropriate byte boundary to ensure that no alignment errors will occur relative to the first byte For those linkers which do not perform this allocation automatically the modified Gp switch can be used The Gpc switch is the default for Gp which says to declare the external areas as simple arrays of char Gps declares them as arrays of short thus forcing 2 byte alignment Gpl declares them as arrays of long thus forcing 4 byte alignment and Gpd declares them as arrays of double thus forcing 8 byte alignment 2 15 Iname Name of File Containing Inline Functions Various C compilers have conventions which allow you to mark certain functions as having special properties A typical such marker is Inline which when it precedes a function declaration indicates that the code for that function is to be compiled inline There are other such markers as well in other C environments especially the Macintosh and Windows An inline functions file simply contains a list of function names each entered on a separate line Those functions are assigned the special marker Thus as an example consider th
155. and line switch anadat void vanadat float short double double void manadat float short double SUBROUTINE TEST DIMENSION A 10 B 20 C 100 REAL 8 ABAR AVAR BBAR BVAR CBAR CVAR CALL ANADAT A 10 ABAR AVAR CALL ANADAT B 20 BBAR BVAR CALL ANADAT C 100 CBAR CVAR RETURN In this function ANADAT will be translated to use manadat when the vector is in memory and vanadat when the vector is virtual The translation is as follows void test extern void vanadat extern void manadat static double abar avar bbar bvar cbar cvar static float a 10 b 20 static long c 32 131 PromulaFortran Translator User s Manual manadat a 10 amp abar amp avar manadat b 20 amp bbar amp bvar vanadat c 100 amp cbar amp cvar return This translation is exactly as desired 4 5 Global Symbols and Prototypes To interface with external subroutines functions and common data areas PROMULA must know the naming conventions and parameter types for each This information is supplied via function prototypes using either standard ANSI C notation or an extended notation which allows changing the name of the external in the target language These prototypes are entered within separate prototype files which are read at runtime via the Rfilename command line switch The notation and use of these files is described in detail in the earl
156. and or used It is needed because the PDP dialect contains no OPEN statement as such Return value None the function is void See also fiofdata Establish file data fiolun Establish FORTRAN unit number fioshl Shift string left 6 214 PDPCLOSE PDP FORTRAN Subroutine Close Synopsis finclude fortran h PROMULA FORTRAN function declarations void pdpclose lun int lun Logical unit number Description This function closes the file assigned to a logical unit number It is needed because the PDP dialect contains no CLOSE statement as such 243 PromulaFortran Translator User s Manual Return value None the function is void See also fioclose Close the file fiolun Establish FORTRAN unit number 6 215 PDPCVTIM PDP FORTRAN External Function CVTTIM Synopsis include fortran h PROMULA FORTRAN function declarations void pdcvtim midn hrs min sec tic long midn Number of ticks since midnight short hrs Returns hours value short min Returns minutes value short sec Returns seconds value short tic Returns ticks value Description Converts a value which contains the number of clock ticks since midnight for a 60 cycle clock into its corresponding hours minutes seconds and ticks values The input value was computed as follows hour 60 min 60 sec 60 tic Return value None the function is void See also None 6 216 PDPGTIM PDP FORTRAN External Function
157. are dynamically allocated to allow for arbitrary complexity across subprograms and substring expressions are calculated via a temporary in some cases Though CHd conversion does make the relation between FORTRAN source and C output more complex than do the other two switches its result is highly readable and produces correct results in all cases that we have seen The CHd approach is sufficiently robust to deal with all situations and is therefore the default setting The CHv CHaracter vector switch is by far the most complicated approach It has all the robustness of the CHd setting It differs from CHd in that intermediate descriptor variables are created which contain the address and length information The addresses of these intermediate variables are then passed as arguments This approach has the advantage that there is a one to one correspondence between arguments in the FORTRAN and the C The disadvantage is that additional logic is needed to create the descriptor variables In the following subsections a series of examples are presented to show how the four switches CHr CHs CHd and CHv treat FORTRAN character manipulation 2 5 1 Initializing Character Values The first example shows how character variables are initialized It presents a CHARACTER type declaration with associated initializations Though character initialization is performed in the same manner via the four character manipulation switches the technique required brings ho
158. are independent of them The listing below shows the default translation of the above example void demo static float b 5 5 4 0 0 0 0 0 0 0 0 static float x 5 1 0 2 0 3 0 4 3405 4 5 0 5 0 5 0 0 0 2 0 3 0 4 0 5 0 0 0 0 0 0 0 0 0 0 0 4 0 5 0 0 0 0 0 0 0 0 0 5 0 static float 10 205 2000 2 072 O20 7 20 2 005 2 05 2 05 2 0 static float 4 2 static long j 4 static long i static char 1 1x 10f10 1 static char F2 1x i10 WRITE OUTPUT FMT F1 1 REAL4 a DO 5 REAL4 b 0 WRITE OUTPUT FMT F1 1 DO 10 REAL4 c 0 WRITE OUTPUT FMT F2 1 INT4 j 0 for j 1L j lt 5L j WRITE OUTPUT FMT F1 1 MORE for 1 0L i lt 5L i WRITE REAL4 x i j 1 MORE WRITE 0 STOP NULL Note that values not explicitly defined in the DATA statement are given a value of 0 blanks for character strings Note also that the initializations of x and J are correct The following listing shows the same translation using the DR command line switch void demo static float b 5 static float x 5 5 static float c 10 static float a static long j static long i static int ftnsiz 1 1 5 1 1 25 1 1 10 36 PromulaFortran Translator User s Manual static namelist DATAVAR p b 6 ftnsiz x x 6 ftnsizt323 c c o6 ftnsi
159. are to be done using long arithmetic As an example consider the following simple FORTRAN code that computes the square of a weighted mean UBROUTINE DEMO VAL NVAL WEIGHT SQ ENSION VAL EGER 2 NVAL WEIGHT POW 0 2 10 I 1 NVAL AR XBAR VAL I NTINUE R XBAR NVAL WEIGHT XBAR POW URN H X rH 10 D H ADUNUxKXaAXUUKXHOUOUN ZAowowoowg The default C output for this example looks as follows void demo val nval weight sq int nval weight float val sq static int Pow static long i 15 PromulaFortran Translator User s Manual static float xbar xbar 0 0 Pow 2 for i 0 i nval i xbar valti xbar long nval weight sq pow xbar double Pow Notice that there are only two casts shown The long cast on the product between 1 and weight is needed in case this product exceeds the maximum value of a short integer This topic is discussed below in a subsection on arithmetic with short integer variables The other cast is on the variable Pow in the exponentiation Though exponentiation is a binary operator in FORTRAN it is not in C therefore in this cast the double is necessary in order to make the parameter for the pow function have the proper type The xbar parameter does not require a forced cast because C automatically promotes a float to a double when it passes it by value The same o
160. atement is entered into the configuration file It describes the path and filename conventions to be used on the target platform and how these conventions are to be obtained from the source specifications The approach taken is to describe how source pathnames are to be translated into target pathnames Syntax PATHNAMES dirchar REPLACE slt1s2t2 LOWER UPPER PREFIX tname EXCLUDE TERMINATION tc sname 1 tname 1 sname n tname n END Where dirchar is the directory separation character in the source pathname 51615212 sequence of character pairs tname is a target language pathname or pathname prefix sname is a source language pathname or pathname prefix tc is a pathname termination character used in the source pathname The required dirchar specification specifies the character used to separate the pathname components in the original source codes This character is replaced by the equivalent character in the target pathnames For example moving from PRIME FORTRAN to UNIX a character would be replaced by a character Moving from MS DOS to UNIX would replace a character with a character The optional REPLACE parameter specifies additional characters to be replaced in the source names As many pairs of characters as are needed may be included The standard PRIME language description for example contains the following specification for this option
161. by value arguments X Y 2 50 Dollar signs as initial symbols in identifiers Z name 2 51 Location of FORTRAN files to be included 2 2 Specifying your C output bias Bc Bf Bo In the October 1988 issue of Computer Language Mark Davidson reviews PROMULA FORTRAN and declares it the clear winner over its competitors however he says Although the code produced by the translator is not immediately understandable Since that review we have surveyed the users of PROMULA FORTRAN extensively to determine how they feel the output code can be made as understandable as possible and have found no clear consensus However users can be divided into three broad categories 1 Those who want to continue using their present FORTRAN dialect as their programming language For those users the C output is of no importance as such It is merely an intermediate step It should be designed to compile as quickly as possible and to produce an efficient as possible executable 2 Those who are presently FORTRAN programmers but who want to or must become C programmers For them the C output should be as close to the original FORTRAN as possible to ease the transition 3 Those who are C programmers who must now take over a FORTRAN code For them the C output should look as much like a standard C program as possible This general issue is referred to as user bias Those users who want optimized code have the optimized bias those who want FORTRAN like c
162. cation For users who wish to declare variables auto the Yi flag can be used to check for variables that are used before they are initialized It is this class of variables that must be static In essence the Yi flag tells PROMULA to override an auto storage allocation for a variable whose value appears to be used before it is changed The actual algorithm used is simplistic and does not pretend to catch all problem variables As a simple example of the use of this variable consider the following code in which the variable 1 is being used without being explicitly set SUBROUTINE DEMO 5 K T1 d WRITE I J K RETURN END The translation of this code using SS and SAO is as follows void demo auto long j k i J 5 k itj WRITE OUTPUT LISTIO INT4 i INT4 j INT4 k 0 return The variable i is auto In all likelihood codes having structures such as this would not produce the correct result with the above translation Using the Yi flag produces the following output in which i has been forced to static void demo auto long j k static long i J 5 k itj WRITE OUTPUT LISTIO INT4 i INT4 j INT4 k 0 return The Yi flag is not a complete solution to the problem of uninitialized variables and variables that are to retain their values however it can help to avoid problems 2 44 DO Loop Counter Reduction Optimization XI Yl 99 As is also
163. char argv static long i j ftnini argc argv NULL puts ERROR 145 I J K STOP NULL This setting allows you to process all errors at once and to go ahead and form an executable however when you execute it you will get error messages when the offending statements are encountered Finally the ER4 switch specifies that syntax errors cause warnings only This switch is like the ones above except that the following is entered into the C output SYNTAX ERROR errno statement Thus the program above produces the following C code under ER4 void argv int argc char argv static long i j ftnini argc argv NULL SYNTAX ERROR 145 I J STOP NULL Obviously this setting treats errors simply as warnings however the resultant executable will not behave correctly 2 11 FORTRAN Input Format Used Fsnum Ft Ff Fv F9 In the good old days the one thing that was always the same was the basic line format used to enter FORTRAN programs But all good things must end Now there are at least 5 different major variations of the FORTRAN entry format that we know of The F command line switch allows you to specify the entry format that you are using There is an extensive discussion of the different formats in the PROMULA FORTRAN Compiler Manual That discussion will not be repeated here The individual settings associated with this flag are mutually exclusi
164. ci The imaginary part of the number complex typedef struct double dr The real part of the value double di The imaginary part of the number 151 PromulaFortran Translator User s Manual dcomplex dcomplex dpxcpx sng complex sng Contains the value to be transformed Description Forms a double complex number whose real and imaginary parts correspond to a single complex number Return value The double complex result See also None 6 29 DPXDBL Convert Double Precision to Double Complex Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of the number dcomplex dcomplex dpxdbl db1 double dbl Contains the value to be transformed Description Forms a double complex number whose real part is a specified value and whose imaginary part is zero Return value The double complex result See also None 6 30 DPXDIV Double Complex Division Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of the number dcomplex dcomplex dpxdiv a b dcomplex a The numerator dcomplex b The denominator 152 PromulaFortran Translator User s Manual Description Divides two double complex numbers to form a third Return value The double complex result of the division See also None 6 31 DPXDPX C
165. claration form which is compatible with all C compilers rather than the ANSI form which is compatible only with contemporary compilers The default translation for the following FORTRAN code SUBROUTINE DEMO CALL IALPHA Hello World STOP END SUBROUTINE IALPHA IA GVAL CHARACTER IA GVAL PRINT IA GVAL END is as follows void demo ga extern void ialpha ialpha Hello World 5 6 STOP NULL void ialpha ia gval P1 P2 lt char ia gval int P1 P2 WRITE OUTPUT LISTIO STRG ia P1 STRG gval P2 0 Note that the function arguments are listed and then typed separately The P256 flag will produce the following translation 69 PromulaFortran Translator User s Manual void demo void extern void ialpha ialpha Hello World 5 6 STOP NULL void ialpha char ia char gval int Pl int P2 WRITE OUTPUT LISTIO STRG ia P1 STRG gval P2 0 In this form the argument list also contains the typing information 2 27 9 P512 Make Parameters Always Take Explicit Value Type There is a difference between various dialects of FORTRAN as to the treatment of untyped PARAMETERs All contemporary dialects require that PARAMETERS be typed in the same manner as other variable symbols however some assign untyped PARAMETERS the type of the constant which they represent Consider the following FORTRAN code
166. clude fortran h PROMULA FORTRAN function declarations int fiowvs str nstring char str Character string to be written int nstring Length of string Description 226 PromulaFortran Translator User s Manual Writes a single character string to the current output file in accordance with the current format specification This function corresponds to the FORTRAN 77 convention in which values and not just l values can be written Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowbiv Write binary values fiowrs Write string vector 6 183 FIOWVT Write FORTRAN Truth Value Synopsis finclude fortran h PROMULA FORTRAN function declarations int fiowvt val unsigned long val Truth value to be written Description Writes a truth long logical value to the current output file in accordance with the current format specification This function corresponds to the FORTRAN 77 convention in which values and not just l values can be written Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowbiv Write binary values fiowrt Write truth value vector 6 184 FIOWVU Write FORTRAN Character Value Synopsis include fortran h PROMULA FORTRAN function declara
167. code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiordd Read FORTRAN double precision vector vmsdel Change a virtual value 6 227 VMSRDF Read FORTRAN Virtual Floating Point Values Synopsis include fortran h PROMULA FORTRAN function declarations int vmsrdf value nval long value Virtual address of values to be read int nval Number of values to be read Description Reads a virtual vector of single precision floating point values from the current input file in accordance with the current format specification Each value is assumed to have its own corresponding format specification if a formatted read is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiordf Read FORTRAN floating point vector vmsdel Change a virtual value 6 228 VMSRDI Read FORTRAN Virtual Short Integer Vector Synopsis include fortran h PROMULA FORTRAN function declarations int vmsrdi value nval long value Virtual address of values int nval Number of values to be read Description 250 PromulaFortran Translator User s Manual Reads a virtual vector of short fixed point values from the current input file in accordance with the current format specification Each value is assumed to have its own corresponding format spec
168. ctional part is a decimal point followed by zero or more decimal digits The exponent part consists of an E e D or d followed by an optional sign and a sequence of decimal digits The parameters to this function are as follows Name Description of Use str Contains the alphanumeric string to be converted nstr Contains the number of characters in the string Note that the string is not necessarily NULL terminated The following global variables are also used by this function Name Description of Use fioerc If the conversion encounters a character which is not part of the notation then this variable returns the position of that character 212 PromulaFortran Translator User s Manual fiondec Returns the number of decimal places in the fractional part of the number plus 1 Thus a value of zero means there was no decimal point and one means there was a decimal but no fractional digits fioblkn If blanks are normal tie breakers then this variable is zero else if it is 1 blanks are simply ignored and if itis 1 blanks are treated as zero Return value The double precision value of the string as computed See also fiostoi Convert string to integer 6 154 FIOSTOI Convert String to Integer Synopsis include fortran h PROMULA FORTRAN function declarations int fiostoi s char s Pointer to the string pointer Description Converts an alphanumeric string to an integer value Its parameter points the location
169. ctor Synopsis include fortran h PROMULA FORTRAN function declarations int fiowrb bool nval unsigned short bool Points to values being written int nval Number of values to be written Description Writes a vector of Boolean values to the current output file in accordance with the current format specification In this context the term Boolean value refers to a short logical value The output display for a logical value consists of a sequence of fw 1 blanks followed by a T or an F where fw is the field width T is used for nonzero values and F is used for zero values Note that this function also supports the FORTRAN 66 convention under which alphabetic information may be stored in the logical values Return value 217 PromulaFortran Translator User s Manual A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Perform error processing fiofout Process output format specification fiowalph Write alphabetic information fiowtxt Write a text record 6 165 FIOWRC Write FORTRAN Character Vector Synopsis include fortran h PROMULA FORTRAN function declarations int fiowrc c nval char c Points to characters to be written int nval Number of characters to be written Description Writes a vector of character values to the current output file in accordance with the current
170. d alpha flt dbl isht lng int isht double dbl long lng float flt WRITE OUTPUT LISTIO REAL4 f1t REAL8 dbl INT2 isht INT4 lng 0 return Notice that PROTOTYPE definitions effect both the translation of references to subprograms and the definitions of subprograms In this sense PROMULA is very different than C itself Clearly the version driven by the prototype is the better version both from the standpoint of readability and from the standpoint of efficiency 4 3 External Name Clash A problem that pervades C in particular is the external name clash problem One tends to use many different libraries with C It is not at all unusual to have the same names used by different libraries As an example we recently worked with a FORTRAN program which had two generalized I O routines named fread and fwrite which stood for FORTRAN READ and FORTRAN WRITE Since these are the names of the standard C I O functions they needed to be changed The prototype system in PROMULA allows this to be handled very easily Consider the following example translated using the prototype definition shown fread void ftread short long short fwrite void ftwrite short long short SUBROUTINE DEMO INTEGER A B 10 C 20 CALL FREAD 1 A 1 CALL FWRITE 2 A 1 CALL FREAD 1 B 10 CALL FWRITE 2 B 10 CALL FREAD 1 C 20 CALL FWRITE 2 C 20 RETURN
171. dbl double dbl Contains the value to be transformed Description Forms a short complex number whose real part is a specified value and whose imaginary part is zero Return value The short complex result See also None 6 8 CPXDIV Short Complex Division Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex complex cpxdiv a b complex a The numerator complex b The denominator Description Divides two short complex numbers to form a third Return value The short complex result of the division See also None 6 9 CPXDPX Convert Double Complex to Short Complex Synopsis 142 PromulaFortran Translator User s Manual include fortran h PROMULA FORTRAN function declarations typedef struct double dr The real part of the value double di The imaginary part of the number dcomplex float cpxdpx dbl dcomplex dbl Contains the value to be transformed Description Forms a short complex number whose real and imaginary parts correspond to a double complex number Return value The short complex result See also None 6 10 CPXEXP Short Complex Exponential Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex complex cpxexp a complex a The value to be tran
172. de fortran h PROMULA FORTRAN function declarations void ftnini Description This function is called as the first executable statement of any main function produced by the translator It contains the declarations of the variables needed by the FORTRAN I O system The parameters argc and argv are simply the command line arguments passed to the main function and then passed here Via the dialect description file keyword 27 the user has the option to specify the command switches explicitly In this case those switches are contained in the third parameter separated by blanks This function scans the command line strings looking for arguments of the following form Argument Description Cnumber Establishes the runtime conventions code 0 Standard conventions 1 Prime extended character set Inumber Assigns the unit whose number is indicated to standard input Onumber Assigns the unit whose number is indicated to standard output Tnumber Assigns the unit whose number is indicated to be a terminal i e reads from standard input writes to standard output This function does call the BLOCK DATA function which is supplied with this function library as a dummy routine This function references no other functions Though this function must initialize many global variables it avoids making any references to the other runtime function libraries Thus if you are doing no FORTRAN style input output or no virtual memory accesses then
173. ding upon the context of its use See the discussion in the FORTRAN compiler manual on controlling runtime behavior for a discussion of this topic The U command line switches establish unit numbers they do not establish runtime unit number file connections This is done at runtime 2 35 File to Receive Prototype Definitions Wname The W switch allows you to write function prototype files These files are used to define the arguments and types of functions being called with the FORTRAN being translated The format of prototype files conforms to the ANSI C standard The Wname switch will write prototypes either for functions referenced or defined or both Which prototypes are written are controlled by the command line switch If there is no extension supplied with the name of the file then an extension of pro will be assumed 2 36 Miscellaneous Control Flags Y1 Y2 There are various translation options which are difficult to classify under a general topic The Y1 and Y2 switches control these The Y1 switch changes the treatment of entry points while the Y2 switch controls the output form of parameter identifiers Of course as with the other numeric flags the composite switch Y3 performs both operations 89 PromulaFortran Translator User s Manual 2 36 1 The Treatment of Entry Points Y1 The following FORTRAN source code contains a main entry and two entry points The parameters associated with the entries al
174. discussed in the chapter containing the installation instructions for PROMULA FORTRAN and its runtime library 133 PromulaFortran Translator User s Manual 5 2 General FORTRAN Operations This group of functions is named with the prefix ftn They include those operations needed to support the actual FORTRAN language conventions character manipulation 3 way branching generalized DO loops etc Also included in this group are the interface functions used by the C and FORTRAN bias to simplify the translation of FORTRAN I O statements 5 3 Input Output Operations This group of functions is named with the prefix fio As has been discussed in other parts of this manual the FORTRAN input output statements have not been translated into standard C type input output operations Rather they have been translated into function calls much like the calls that would be generated by an actual FORTRAN compiler Each FORTRAN statement is broken into a series of function calls If the FORTRAN operation specifies any error operations then an initial call is made to a function fiostatus which establishes the error processing conventions to be used and a final call is made to ioerror which generates any error branches requested If no error processing is specified by the FORTRAN statement then an error causes the program to exit 5 3 1 Runtime Error Messages The actual I O errors generated along with the name of the function generating them are listed b
175. discussed in the section on array subscript expressions and in the section on the DO statement when the only purpose of a DO loop counter is to subscript arrays within the loop that counter can be reduced to simplify the array subscript expressions The Yl switch turns this optimization on and is the default for all biases The switch turns this optimization off As an example of this optimization consider the following simple FORTRAN code which performs a matrix multiplication SUB DIM DO DO C I DO C I 10 CON 15 CON RET END Notice that in FORTRAN subscripts normally start at 1 therefore the loop variables in the above example also all start at ROUTIN ENSION 15 15 J J 10 K J INUE INUE URN PromulaFortran Translator User s Manual E MATMUL A 10 15 B 15 20 C 10 20 TQ 1 20 0 0 255 C I J A I K B K J 1 Now the default C output for this example with active is shown below void matmul a b c float a 15 10 b 20 15 c 20 10 static int i j k for i 0 i lt 10 it 3 0 j 20 retur In C subscripts always start at 0 therefore the DO loop counters themselves have been changed to start at zero The listing c 3 for n 0 31 7 jtt Kd 0 0 k lt 15 k i c j il a k i1 b j k below shows the C output with XI active void matmul
176. ds produced conform exactly to the FORTRAN standard specification The C output using the Yf flag looks as follows void demo static int i j static float a b c printf 51d 51d 10 2 10 2f 10 2f n 1i j a b c 1214 1214 16 6 16 6F 16 6 E n i j a b c 98 PromulaFortran Translator User s Manual In this translation the FORMAT string has been converted into an equivalent printf style string In the list directed form printf elements have been selected to correspond to each element in the list Note that these elements are in the FORTRAN dialect definition file and may be changed by you The C printf conventions are reasonable for simple formatting operations however they lack the power of the FORTRAN FORMAT string The output produced by the Yf flag is an approximation only of that produced by the runtime library In particular it does not conform to the FORTRAN standard specification Use the Yf flag only if you are moving away from FORTRAN completely and if your application does not require particular formatting conventions 2 43 Initialization Check for Auto Variables Xi Yi Traditional FORTRAN programs assume that local variables maintain their values through calls to a given subprogram As a result of this fact the default treatment of all variables by PROMULA is to make them static See the section on the storage allocation switches for more information on variable allo
177. dth NCnum The NCn switch controls the margin to be associated with inline comments when the CM1 or CM2 switches are active The default value for this switch is 35 This value specifies that the inline comment associated with a particular output statement should be displayed on the same line as the statement beginning at column position 35 If the statement is longer than 35 characters the comment should begin immediately after the end of the statement Other settings have equivalent meanings A switch setting of NCO will force all inline comments to follow their associated statements directly 2 24 Upper and Lower Braces Convention in C 0 NU1 NU2 NL1 NL2 A major area of disagreement about the readability of C has to do with the placement of braces in nested structures If the brace placement convention does not match your desired convention you will probably have trouble feeling comfortable with the output By using the brace placement flags in conjunction with the indentation width switches discussed in the sections on Nnumb and NCnum any reasonable placement convention can be achieved The NUO switch specifies that the upper brace is to be on the same line with the statement causing the nesting increase The NUI switch specifies that the upper brace is on the next line at the indentation point used for the old nesting level The NU2 switch specifies that the upper brace is on the next line at the indentation point used for the new
178. e See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Do requested error processing fiofout Process output format specification fiowalph Write alphabetic information fiowtxt Write a text record 6 172 FIOWRU Write FORTRAN Unsigned Char Vector Synopsis include fortran h PROMULA FORTRAN function declarations 221 PromulaFortran Translator User s Manual int fiowru c nval unsigned char c Points to characters to be written int nval Number of characters to be written Description Writes a vector of unsigned character values to the current output file in accordance with the current format specification In this implementation unsigned char is derived from the nonstandard FORTRAN type LOGICAL 1 Therefore L formatting conventions are assumed Note that in this function each individual character is assumed to have its own corresponding format specification if a formatted write is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Do requested error processing fiowrb Perform short Boolean output 6 173 FIOWRX Write FORTRAN Complex Vector Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary par
179. e fortran h PROMULA FORTRAN function declarations long fifisign mag sgn long mag The magnitude for the result long sgn The sign for the result Description Returns a value after the transfer of a sign The result is if sgn is at least zero else it 15 mag Return value The value as computed above See also None 6 85 FIFMAX0 FORTRAN Intrinsic Function MAX0 Synopsis include fortran h PROMULA FORTRAN function declarations long fifmax0 al a2 long al First value to be compared long a2 Second value to be compared Description Determines which of two long values is the largest Return value The largest long value See also None 6 86 FIFMAX1 FORTRAN Intrinsic Function MAXI Synopsis include fortran h PROMULA FORTRAN function declarations long fifmaxl al a2 float al First value to be compared float a2 Second value to be compared Description 175 PromulaFortran Translator User s Manual Determines which of two double precision values is the largest Return value The largest value converted to a long integer See also None 6 87 FIFMINO FORTRAN Intrinsic Function MINO Synopsis finclude fortran h PROMULA FORTRAN function declarations long fifminO0 al a2 long al First value to be compared long a2 Second value to be compared Description Determines which of two long values is the smallest Return value The smallest long value See also None 6 88 FIFMI
180. e hd_qpis hd_qpsupis firstnoninp hashmax hd_hashval swab flag xlit flag blanks xlit yet blank bytes Initialization entry Initialization calculate some useful things noninpbase hd qpis hd qpsupis firstnoninp noninpbaset1L First non input node Determine practical maximum for no of strings we can hash using PUTSTR entry Knuth s vol 3 says that the avg no collisions 1 2 1 1 R where R is HASH table density it might be nice to stop at 98 full 0 98 since this translates to an average of 25 5 collisions per string entered hashmax hd hashval 2L hd 1 1001 Save local values of SWAB and XLIT flags and convert blank string to match the language EBCDC or ASCII of the model file CREATOR swab flag arg swab flag xlit flag arg xlit flag if xlit flag amp fifi4log blanks xlit yet trlito amp blank bytes amp K1 blanks xlit yet fifi4log long TRUE HF X return modi This form has the same effect as the FORTRAN original however for both the DDL and DML comments the lack of a trailer makes the code hard to read and for the DML comments a header seems appropriate as well COMMENTS 0 ipm xxm mim m xm 1 xn whim xxm mim m xm 1 xj END Using this specification produces a much nicer looking translation which still seems to capture the flavor of the original long modi arg_swab_flag arg_xli
181. e assignments the two assignment statements have been simplified into one Notice that this simplification precedes the braces removal operation Thus the subprogram above processed under Ya and Yb produces the following result void demo static int ival 10 jval 10 i 1 0 i 10 i ival i jval i 10 return Now the two assignments have been collapsed into a single statement and the for loop now contains only that statement Consequently under the Yb flag the braces can be removed The Yb flag is discussed further in another section in this chapter 2 38 Treatment of Single Statement Nesting Brace Xb Yb C allows any conditional statement to form a compound statement with a single statement while FORTRAN allows this only for the IF statement When the YB switch is active this compounding is performed whenever possible while when the XB switch is active a compound statement is formed only if the source was compound YB is the default for the C and optimized biases while XB is the default for the FORTRAN bias The following simple subprogram contains a DO loop a simple IF statement and a block IF statement SUBROUTINE DEMO DIMENSION IVAL 10 JVAL 10 DO 10 I 1 10 IVAL I 10 JVAL I 10 10 CONTINUE IF I EQ 10 WRITE I equals 10 93 PromulaFortran Translator User s Manual IF I EQ 11 THEN WRITE I equals 11 END IF RETURN END
182. e blocks themselves are removed and the variables within them become external variables directly 4 final technique is used when an exact block layout is required The COMMON block is declared as an external pointer to a char The variable positions within the block are calculated with user supplied alignment and size specifications Actual variable references then become references to the COMMON block name plus the calculated position 1 13 Allocation of Local Variables Most FORTRAN compilers allocate a fixed unique storage location to all variables be they COMMON or local The equivalent allocation in C is called static and by default all variables in the translation are declared as static variables C however has two additional storage allocation methods auto and dynamic Auto variables are stored on the program stack automatically when a function is called and are automatically removed when the function is exited The advantage of these variables is that they occupy memory only when needed and their physical allocation is quick and transparent to the C program The disadvantage of auto variables is that the program stack tends to be quite short thus the number and size of auto variables are very limited PromulaFortran Translator User s Manual Dynamic variables are stored on the heap via explicit calls to a function and must also be freed explicitly via a function call The disadvantage of dynamic variables is that their
183. e configuration file is used in contexts where the symbol table must be changed in order to make a program work Typically it is used when a program is being moved between platforms with different word sizes Syntax RESTRUCTURE ident type END Where ident is the identifier of a variable in the program being processed type isthe FORTRAN type to be assigned to that variable To understand the use of the RESTRUCTURE statement consider the following simple FORTRAN program written for a platform with 36 bit words and 6 bit characters PROGRAM TEST3 DIMENSION IALPHA 2 DATA IALPHA 6HHello 6HWorld 1 FORMAT 1X 2A6 PRINT 1 IALPHA STOP END In this program the INTEGER array IALPHA is being used to contain character data This is a standard convention in many FORTRAN dialects still in use today The default translation for this code is a mess void main int argc char argv static long ialpha 2 static int ftnsiz 1 1 2 114 PromulaFortran Translator User s Manual static namelist DATAVAR ialpha ialpha 5 ftnsiz static char DATAVAL SDATAVAR ialpha Hell o SEND static char 1 1x 2a6 ftnini argc argv NULL fiointu char DATAVAL 0 2 fiornl DATAVAR 1 NULL WRITE OUTPUT FMT F1 1 DO 2 INT4 ialpha 0 STOP NULL Since the variable is declared as an INTEGER runti
184. e esas nsns a 234 FTNREW FORTRAN REWIND STATEMENT eee e nsns nsus etere e e e etere e 235 FTNSAC FORTRAN STORE A CHARACTER STRING eeeseeeeenene e ene e ene ener ener nnns nsns ise esiti se ase essa nsns a 236 FTNSALLO FORTRAN STRING ALLOCATION eeeeeenerene e e e ene n nnnm nn nn nnns nnn sese sns is uses s esi ese esa esa ese nsns ases ue 237 FTNSCOMP FORTRAN STRING COMPARISON eee ene e 237 FTNSCOPY FORTRAN STRING C PY nennen ene e Eere En E 237 FTNSLENG FORTRAN STRING 2 etete e e 238 FTNSUBS FORTRAN SUBSTRING EVALUATION eee e e e 238 FTNSTOP FORTRAN STOP STATEMENT 239 FTNWEF FORTRAN END FILE STATEMENT eese e e e ene nnne nnnn nnns nnns nsns nsus ase e ese sese ese nsn asa ese ase nnns a 239 FTNWRIT FORTRAN WRITE STATEMENT e ene 240 FTNXCONS FORTRAN EXACT REPRESENTATION CONSTANT eceeeeeeeenene e e emen n esas 241 P77GETU PRIME FORTRAN 77 FUNCTION F77 GETU 2 242 P77NLENA PRIME FORTRAN 77 SUBROUTINE 242 P77TNOUA PRIME FORTRAN 77 SUBROUTINE TNOUA
185. e following inline function file test inl ialpha when used via the command line switch 1test to process the following FORTRAN program PROGRAM TEST I IALPHA J JALPHA J STOP END FUNCTION IALPHA J IALPHA J 6 RETURN END FUNCTION JALPHA J IALPHA J 6 RETURN END The translation produced is as follows void argv int argc char argv 53 PromulaFortran Translator User s Manual Inline extern long ialpha extern long jalpha static long i j k ftnini argc argv NULL ialpha amp j k jalpha amp j STOP NULL _Inline long ialpha j long j static long ialpha ialpha j 6L return ialpha long jalpha j long j static long jalpha ialpha ialpha j 6L return jalpha Note that the function IALPHA has been assigned the special marker while the function JALPHA has not The actual marker Inline can be easily changed via a configuration file as discussed in the following chapter 2 16 Target C Int Type Is Il An interesting and excellent feature of C is the way in which it defines fixed point integer values as short meaning a relatively narrow range of values but requiring less storage long meaning a relatively wide range of values but requiring more storage int meaning the most efficient fixed point storage technique which might be short or long depending upon the machine FORTRAN
186. e nsigdig specifies the number of significant digits in the value Return value None the function is void The global variable iocrec is updated to contain the new field and the variable fionchar is updated to reflect the new character count in the coded communication record See also fioitos Convert integer to string fiorndv Round floating point value display fioshr Shift display string right 6 160 FIOWEF FORTRAN Write End of file Synopsis include fortran h PROMULA FORTRAN function declarations int fiowef Description This function writes an end of file to a file Return value 215 PromulaFortran Translator User s Manual A zero if all went well else an error code See the general discussion of the FORTRAN I O capabilities for a listing of the possible error codes See also None 6 161 FIOWHEXO Write Hexadecimal or Octal Constant Synopsis include fortran h PROMULA FORTRAN function declarations char fiowhexo base ival vlen int base Base of the exact representation constant char ival Value of exact representation constant int vlen Length of value in bytes Description Displays a numeric value as a hexadecimal or octal character string Return value The function returns a pointer to the exact representation string for the value See also fifibit Insert a bit fifrbit Reverse bits byte order fifxbit Extract a bit 6 162 FIOWLN Write FORTRAN End of Line Sy
187. e of a floating point argument using single precision Return value The single precision value as computed above See also fifsncs Compute sine or cosine 6 50 FIFDATE External Function DATA 161 PromulaFortran Translator User s Manual Synopsis include fortran h PROMULA FORTRAN function declarations void fifdate cl ncl char cl Returns the date in string form int ncl Length of the date return string Description Obtains the current date from the operating system Return value The current date in mm dd yy form See also fioconv General dialect convention flags 6 51 FIFDDIM FORTRAN Intrinsic Function DDIM Synopsis include fortran h PROMULA FORTRAN function declarations double fifddim al a2 double al First value in difference double a2 Second value in difference Description Computes a positive difference Return value If a1 is greater than a2 then the value of a1 a2 is returned else zero is returned See also None 6 52 FIFDINT FORTRAN Intrinsic Function DINT Synopsis finclude fortran h PROMULA FORTRAN function declarations double fifdint a double a Argument containing the value Description Truncates its double precision argument to its integer value In particular fifdint a 0 if 1 the largest integer with the same sign that does not exceed a if a gt 1 Return value 162 PromulaFortran Translator User s Manual The double precision
188. e result of the concatenation is no longer needed its memory must be freed This adds a bit of additional code The CHv translation of this fragment is as follows void demol extern void demo2 static char sum 13 initial 4 static string T1 NULL 0 static string T2 NULL 0 Jj ftnscopy initial 4 ONE 3 NULL ftnscopy sum 13 initial 4 PLUS 5 initial 4 NULL ftnsallo amp T1 C 2 sum 13 NULL OPEN 1 FILEN Tl a Tl n 0 free Tl a ftnsallo amp T1 sum 13 PLUS THREE 10 NULL ftnsallo amp T2 sum 13 PLUS TWO 8 NULL demo2 amp T1 amp T2 free Tl a free T2 a return Notice that the call to demo2 now receives the descriptor pointers directly This is the only context in which the CHv formulation is simpler than the CHd one 2 5 5 Character Treatment Conclusion The default CHd flag does produce slightly longer translations in some complicated cases Alternatively the code it produces is almost always faster It is strongly recommended that the CHr and CHs flags only be used in special cases where a particularly simple conversion is being performed The flag should only be used if external libraries or system functions which assume descriptors must be called In this case the user can modify the string type to reflect the actual required type See the chapter on the configuration file for more information on this topic 2 6 A
189. e use of long in the keyword list therefore the following alternative demo cnf file would produce the same result KEYWORDS long 13 CHARACTER END 3 5 2 Pattern Strings for COMMON blocks In the default translation COMMON blocks generate three different symbols distinguished by the initial first letter The X indicates the global storage area the C indicates the structure definition and the T indicates local pointer The pattern strings used to produce this convention are X C and T Suppose instead that you wish to label the global storage area with the suffix area to label the structure definition with the prefix struct and to have the local pointer simply use the COMMON name The following KEYWORDS statement would make this change KEYWORDS _ os struct TS ow END The following translation is produced void test4 heading P1 char heading int P1 extern void test5 extern char modelhead area typedef struct long hd_hedsiz Length of header long hd_numtab No of tables in file long hd_filrev Model file revision date long hd_sysprim No of system primitive types struct modelheag auto struct modelhead modelhead struct modelhead modelhead area test5 amp modelhead hd numtab amp modelhead hd filrev modelhead hd hedsiz modelhead hd numtab modelhead hd filrev WRIT
190. e variable The translation of this loop using the 77 conventions looks as follows 83 PromulaFortran Translator User s Manual void demo static int v el e2 e3 D1 D2 for v el D1l e3 D2 e2 v D1 D1 D2 gt 0 D2 v D1 WRITE OUTPUT LISTIO INT4 v 0 return Note that there are two temporary variables introduced D1 contains the value of the increment and D2 contains the value of the iteration count Thus the first statement in the C for loop performs steps 1 2 and 3 from above The second statement in the C for loop performs the step 4 logic and the last statement performs the step 5 logic Note that the calculations are arranged in such a manner that none of the original e1 e2 or e3 expressions need to be evaluated more than once PROMULA also performs two other simplifications with for loops First if the body of the loop can be reduced to a single statement then the simple form of the or statement is used Also the statement number is deleted if it is only being used as the terminating point of the loop When 66 conventions are in effect the loop must be forced to execute at least once To implement this the C ternary operator is used The following is the translation of the above with the 1 command line option set which requests 66 style DO loops void demo static int v el e2 e3 D1 D2 for v el Dl e3 D2 D2 e2 v D1 D1 gt 0 D2 1 D2 gt 0 2 1 WRITE OUTPUT LISTIO
191. ea 512 characters large is allocated to control this process See the discussion of the P16384 switch for a detailed discussion of this topic The QHnumb switch can be used to change this value If the include file information storage area storage area is exceeded the fatal error 531 The include file storage of nnnn bytes is insufficient use the QHn flag to increase it 18 generated 2 29 6 QWnumb Word Size of Source Platform The typical word size for contemporary platforms is 32 bits with 8 bits per character This information is internally encoded as bits per word 100 bits per character Thus the default word size specification value is 3208 When translating source codes from platforms with different word sizes the word size value can be changed via the QWnumb switch Note that moving codes between machines of different word sizes is potentially very difficult The C output produced will retain the characteristics of the source platform If explicit word size assumptions were made in the source then those will probably not be well formed on the target platform 2 30 Specify a Configuration File Rname The R switch allows you to read additional configuration files These files contain function prototypes and miscellaneous other configuration information The format and use of configuration files are discussed in Chapter 3 Suffice to say here that the Rname switch will read configuration information from the files named
192. eceive the listing If the named file already exists it is truncated If the specified name under either PA or PN has no extension an extension of 1st is assumed The PH and PW flags specify the desired output page height and width The default setting for PH is 80 lines and that for PW is 132 characters The reports produced are quite large and the minimum setting allowed for PW is 80 characters Note that the symbol table widths are adjusted to fit into the specified page size however the source code listing is truncated if the source line display becomes too wide 2 29 Quantity Control Flags QInumb QEnumb QDnumb QXnumb QHnumb QWnumb Generally PROMULA examines its environment to determine the resources and values that it assigns to its internal processing controls In some cases the user needs to be able to override the default settings for the quantities assigned in particular the following settings may be controlled 77 PromulaFortran Translator User s Manual Flag Quantity controlled QInumb Size of compacted statement storage QEnumb Size of the line number table QDnumb Size of a data block QXnumb Size of external information storage QHnumb Size of include file information storage QWnumb Word size of source platform 2 29 1 QInumb Size of Compacted Statement Storage As PROMULA processes the raw FORTRAN source it strips all blanks and comments and stores the resulting compacted statement for compilation The si
193. ecified long linear address This function takes no responsibility for ensuring that no block boundaries are encountered downstream from the specified byte This function assumes that the calling function intends only to change the information The parameter is as follows Name Description of use ioiad The linear address for which a pointer is desired Return value A pointer to the byte at the requested linear address See also vmsptr Obtain Virtual Byte Pointer 245 PromulaFortran Translator User s Manual 6 219 VMSGLOB Virtual Global Access Synopsis include fortran h PROMULA FORTRAN function declarations void vmsglob iop int iop Description This function loads and saves global variables from a virtual PROMULA database as specified in the globals file This version is a dummy only to satisfy the external references from vmsopn and vmscls Return value None the function is void See also None 6 220 VMSLOAD Load a Virtual Vector Synopsis include fortran h PROMULA FORTRAN function declarations char vmsload nbyte ioiad long nbyte Number of bytes to be allocated long ioiad Virtual address of values Description Returns a pointer to a storage area on the heap which is at least as long as the requested number of bytes If the bytes cannot be obtained then this function exits to the operating system If the bytes can be obtained they are loaded with the information at the specified address o
194. ect and portable FORTRAN processor It runs on multiple platforms and supports both the ANSI FORTRAN 66 and ANSI FORTRAN 77 standard dialects as well as a large number of common extensions such as those found in the following commercial compilers VAX FORTRAN PDP FORTRAN PRIME FORTRAN Data General FORTRAN and Sun FORTRAN Some Fortran 90 extensions are also supported In cases where different versions of FORTRAN have conflicting features or conventions a dialect selection option switch can be used to select the desired set The particular dialect options which the compiler supports are as follows Option Dialect Reference Mvax Vax FORTRAN Programming in VAX FORTRAN AA D034D TE September 1984 Digital Equipment Corporation Mpdp PDP FORTRAN PDP 11 FORTRAN Language Reference Manual AA 1855D TC December 1979 Digital Equipment Corporation Mp77 Prime FORTRAN 77 FORTRAN 77 Reference Guide Fifth edition Release T1 0 21 0 January 1988 Prime Computer Inc Mpiv Prime FORTRAN IV The FORTRAN Reference Guide FDR3057 by Anthony Lewis March 1980 Prime Computer Inc Msun Sun FORTRAN Sun FORTRAN Reference Manual 800 3418 10 March 1990 Dun Microsystems Inc Mufn Unisys FieldData Sperry Univac Series 1100 FORTRAN V Level 4R1 Programmer FORTRAN Reference 1979 Sperry Rand Corporation Mhny Honeywell FORTRAN RCS FORTRAN Reference Manual Order Number DG75 Rev 3 GTEDS December 1981 GTE Data Services Incorporated Mf90 Fortran
195. ee also None 6 109 FIOFEND End Format Processing Synopsis include fortran h PROMULA FORTRAN function declarations void fiofend Description Clears the format control variables to end processing with the current format Return value None the function is void See also None 187 PromulaFortran Translator User s Manual 6 110 FIOFFLD Get Next Free Form Field Synopsis finclude fortran h PROMULA FORTRAN function declarations void fioffld Description This function locates the start of the next unit of information when free form reads are being performed It skips over blanks and commas If an end of record is encountered then it reads the next physical record from the current file The major complexity in this routine has to do with repeat counts and null values Note that when a character is encountered all remaining values are set to null Return value The function returns a 1 if there is a non null value to be read In this case 1oicol points to the first significant character of this value A zero is returned if a null value is to be read From the standpoint of the calling function this means that the corresponding input value is to remain unchanged See also fiospace Skip over white space 6 111 FIOFINI Initialize A FORTRAN Format Synopsis include fortran h PROMULA FORTRAN function declarations int fiofspec 0 Current format specification char fiosadr NU char fiocfmt
196. efore by default PROMULA FORTRAN declares FORTRAN integers as long If on your platform C ints are already long or if in your source FORTRAN dialect FORTRAN integers are short PROMULA FORTRAN may be told to translate INTEGER as int or as short Note 3 that in C all symbols must be defined prior to their use therefore PROMULA FORTRAN declares all parameters and local variables explicitly Even in the FORTRAN bias no attempt is made to maintain a direct correspondence between FORTRAN declarations and C declarations Note 4 that the declaration of the val vector differs between the C and the FORTRAN biases In the FORTRAN bias the C brackets notation is used which though different from standard FORTRAN notation is preferred by many to the traditional C pointer notation used by the C bias This notation may be controlled via command line switches Note 5 that most implementations of FORTRAN treat variables as though they were static i e each variable is assigned its own memory location which remains undisturbed even when the routine containing it is not active The user may specify that variables be declared as auto In general the user of PROMULA FORTRAN has extensive control over the allocation and memory status of variables Note 6 that C and FORTRAN have very different ways of specifying coded write conversions For the FORTRAN bias all of the original FORTRAN machinery is maintained The actual WRITE statement is translated into
197. eing performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Do specified error processing fiordi Read short integer value 6 131 FIORDD Read FORTRAN Double Precision Vector Synopsis include fortran h PROMULA FORTRAN function declarations int fiordd value nval double value Points to values to be read int nval Number of values to be read Description Reads a vector of double precision floating point values from the current input file in accordance with the current format specification Each value is assumed to have its own corresponding format specification if a formatted read is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also 200 PromulaFortran Translator User s Manual fioerror Perform FORTRAN I O error processing fiofinp Get next formatted input specification fiostod Convert string to double fioffld Get next free form input field fioralph Read alphabetic information fiorchk Check current fixed input field 6 132 FIORDF Read FORTRAN Floating Point Values Synopsis include fortran h PROMULA FORTRAN function declarations int fiordf value nval float value Points to values to be read
198. elow The identifier of each error is a mneumonic which specifies the function which generated it and its cause The codes specified are those returned if the I O statement requests an error status value Error Code Function Description ELUN EOF 1 fiolun End of file encountered EOPN EOF 1 fioopen End of file encountered ERBV EOF 1 fiorbiv End of file encountered ERTX EOF 1 fiortxt End of file EWTX EOF 1 fiowtxt Write beyond end of file ELUN RDO 106 fiolun Write to read only file EINT NAF 107 fiointu No active file structure ELUN NAF 108 fiolun No active file structure EINT TMF 109 fiointu Too many files open ELUN TMF 110 fiolun Too many files open ENAM TMF 111 fioname Too many files open ESIO TMF 112 fiostio Too many files open ECLO PCF 113 fioclose Physical close failure EOPN POF 114 fioopen Physical open failed ECLO POF 115 fioclose Physical open failure EBCK FRT 116 fioback At front of file EBCK DIR 117 fioback Direct access file ELUN PWF 118 fiolun Physical write failed ERWV PWF 119 fiorwbv Physical write failure EWBV PWF 120 fiowbiv Physical write failure EWEF PWF 121 fiowef Physical write failure ERDB IFS 122 fiordb Invalid format specification ERDD IFS 123 fiordd Invalid format specification ERDF IFS 124 fiordf Invalid format specification ERDI IFS 125 fiordi Invalid format specification ERDL IFS 126 fiordl Invalid format specification ERDS IFS 127 fiords Invalid format specification ERDT IFS 128 fiordt Invalid format speci
199. en 120 3 5 3 Pattern String for External Functions ic e etes e e ae De e e i a te ER ere Re Hb atn 121 3 5 4 Pattern String for Subprogram Surrogates esee eene neret 122 3 5 5 Pattern Strings for VAX Descriptors isi c dentes e HR e Ye De i E e SEESE ve EE eR e eee Sees 123 3 6 THE CONFIGURATION PRAGMA 123 3 7 THE CONFIGURATION STATEMENT cccccccccssscesssecssececseecsecesssecsseceseccsueceseccssecesssecssecesssecsseceesesensecessseeseeessesenseeees 125 4 THE CONFIGURATION FUNCTION PROTOTYPEG scsssscssssssssscssssssssscsssccssscssssssssscsssscsssssssnsssssssssnessesssees 127 4 1 FUNCTION PROTOTYPE SYNTAX eicesdvcticedstedeneceucncdue sesseenteaseodedescedeateesuvssoeyces 127 4 2 VALUE PARAMETERS RO e ETE REANO TE ERE TE CI PP XR PRECES ERR EGRE 128 4 3 EXTERNAL NAME CLASH essetis re er EE RESP PTes eate e de Fe Ren EE PEE Ea Ee e PETERE 129 2 4 MUETIPEE EORMS Rer acra toed SERO ERO RERO ETHER Te RETE EDI 130 4 5 GLOBAL SYMBOLS AND PROTOTYPES ccssesscosscescessensceueceececsensceuecsececeensnsueeseesceessnneesecesceesenncesseescessenseesseesceseenseeses 132 4 6 RENAMING IDENTIFIERS 132 5 OVERVIEW OF RUNTIME LIBRARY ccssscsssscsssscsssccssscsscccssessssccssnsssssscssnesssscsscnsssssscsssssssssscnsssssssssnessssssees 133 5 1 NAMING AND ORGANIZATION O
200. ents the virtual manager from exhausting all memory before the dynamic manager has a chance to get its share The maximum setting for this parameter is 254 Return value This function returns a code indicating whether or not the open was successful The values are as follows Code Meaning 0 went well the virtual file is ready for use 1 The existing virtual file could not be opened 2 The size of the existing file could not be determined 3 The new virtual file could not be created 4 Sufficient memory for the block status vector could not be allocated See also vmsglob Manages virtual global variables 6 222 VMSPTR Get Virtual Byte Pointer Synopsis include fortran h PROMULA FORTRAN function declarations unsigned char vmsptr fd ioiad delta int fd Handle of virtual file long ioiad Linear address of desired byte int delta Change flag Description 247 PromulaFortran Translator User s Manual Returns a pointer to the byte at a specified long linear address This function takes no responsibility for ensuring that no block boundaries are encountered downstream from the specified byte If the calling function specifies that the byte being addressed is to be changed then this function marks the virtual block structure containing the byte accordingly The parameters for this function are as follows Name Description of use ioiad The linear address for which a pointer is desired delta A flag which if non zero
201. eo ses da RR eei EE PEE 2 1 3 COMPILER ADVANTAGES tet cedet exec ida n eed le ties t iub Es 3 1 4 TRANSLATORS ADVANTAGES redit Cesc irte PEE 3 1 5 HOW PROMULA FORTRAN WORKS icr Ee trei edere ed e che vt eee avene Ta PUn Tene e tree Er uoa x a 5 1 6 RATIONALE FOR DEVELOPING PROMULA FORTRAN ccccccccccecsssssssceeceecsesessececececsenenseaeseeeccesenssaeeeeececsensaasseeeeeenes 5 1 7 DOWNSIZING MAINFRAME CODES FOR USE ON THE PC DOS PLATFORM 6 1 8 DEALING WITH FORTRAN DIALECT PROBLEMS c cccccccscesssssececececcesseaececececsenssaeceeececeesesssesececeeseaaeaesececeenenseaaeeeeceenes 6 1 9 DEALING WITH C TYPES AND FORTRAN TYPES cccccccecsssssececececcesseaeeecececseseaeeecececsesesceeececseseaeaeceeseseseneaaeeeeceeses 7 1 10 DEALING WITH FORTRAN INPUT OUTPUT IN Cu ccccccccccecssssscececececsenseaececececsessaaeceeececsensnaececceceseneaeaeseeeeeeseneaeseeeeeenes 7 Led EO ba 1Y EA D h ET Sd 8 1 12 EATING WITH COMMON BEOGKS tete rie tse E 8 1 13 ALLOCATION OF LOCAT r e a a a E a N SAS 8 114 AsSAMPLE TRANSLATION TO O a a Rea Rae ea 9 2 Se PELO hope 10 2 1 COMMAND LINESYNTAX aen ndr UH epe aH GM i 10 2 2 SPECIFYING YOUR C OUTPUT BIAS BC BF BO ccccccccccccssssssecececcesesscaececececsessnaececececescesuececececeessaaeaeceeeesenensnaeteeecs 12 2 3 ARITHMETIC CON
202. er case 2 5 Treatment of CHARACTER Variables CHd CHr CHs CHv By far the most difficult task to be faced by the processor is that of dealing with FORTRAN character manipulation The problem is that despite the efforts of the various standards committees FORTRAN as a living computer language has no unified approach to character manipulation FORTRAN was originally designed to do formula translation It had little use for characters other than to label the results of calculations FORTRAN 66 has no CHARACTER data type and has no character manipulation statements other than formatted I O operations Character data is simply hidden in whatever variable is convenient Once a programmer stores character data in a numeric variable it is up to him to ensure that he does not use that variable for anything else 24 PromulaFortran Translator User s Manual As the language matured a CHARACTER type was added and some minimal machinery was included to manipulate these types of variables However since there were many FORTRAN 66 programs still in existence the use of the new CHARACTER type did not preclude using the old character management techniques as well Things such as equivalencing numeric variables and character variables were completely valid When the 77 FORTRAN standards were established the committee rightly concluded that the way in which languages like IBM FORTRAN level H had implemented characters was clearly not machine transp
203. er of that block which immediately precedes this block in the least recently used chain A zero in this member means that this block is the least recently used block vbmru number of that record which is immediately more recently used than the current block or zero if this block is the most recently used block vbdat the actual data contained within this block As can be seen from the above structure the virtual memory system keeps track of the order of reference of the blocks When a block must be moved into memory and there are no empty memory blocks the least recently used memory block is flushed 5 5 2 Virtual Memory Global Variables Internally the virtual memory system functions are controlled via a set of statically allocated global variables These are as follows Variable Description of use 137 vmsbadr vmsblock vmsfd vmslru vmsnblk vmsnwr PromulaFortran Translator User s Manual The addresses in memory of virtual blocks Note that there are a maximum of 254 virtual memory blocks therefore this array is statically allocated A pointer to the block status vector as described above The file descriptor for the virtual file Number of the least recently used block The number of blocks on the virtual file The number of bytes currently written to the virtual file This value is needed for the whole avoidance logic 5 6 Single Precision Complex Arithmetic This group of functions is named with the
204. er s Manual DPXCJG COMPUTE THE DOUBLE COMPLEX CONJUGATE eceeeeeete ee ene emen nnnm nn nn nnne nenne senes use e esa ese nsns nsns un 150 DPXCMP DOUBLE COMPLEX COMPARISON eceeeenere e n ene enn etere e e e e ere senes esas ases 150 DPXCOS COMPUTE THE DOUBLE COMPLEX 151 DPXCPX CONVERT SHORT COMPLEX TO DOUBLE COMPLEX eee ee ene ene n ennnnn ener nnns nsns esses 151 DPXDBL CONVERT DOUBLE PRECISION TO DOUBLE COMPLEX eee eee e enenenenennnnn ener nsns nsns nsns esas nsns 152 DPXDIV DOUBLE COMPLEX DIVISION eeeeeeeene e e e 152 DPXDPX CONVERT TWO DOUBLES TO DOUBLE COMPLEX eee e ene emen ener ener nnns nsns nsns ese e ese senes nsns a 153 DPXEXP DOUBLE COMPLEX EXPONENTIAL eeceeeeerene 153 DPXIMA COMPUTE IMAGINARY OF DOUBLE COMPLEX eee ee e e e emen nnnn nnns ntes ese eset nsi isis 154 DPXLOG DOUBLE COMPLEX NATURAL LOGARITHM ecce 154 DPXLOGI10 DOUBLE COMPLEX BASE 10 LOGARITHM eee e 155 DPXLONG CONVERT DOUBLE COMPLEX TO _0 155 DPXMUL DOUBLE COMPLEX MULTIPLICATION eee ern ene enne nnnm nn enhn ener nnns nsns e nsus e usn ise ase esas 155 DPXNEG COMPUTE THE DOUBLE COMPLEX NEGATIVE
205. er to its double precision argument In particular fifnint a fifidint 0 5 ifa fifidint a 0 5 ifa A IV where ifidint is the FORTRAN intrinsic function IDINT Return value The value as computed above 177 PromulaFortran Translator User s Manual See also fifidint FORTRAN intrinsic function IDINT 6 92 FIFRAD50 FORTRAN External Function IRAD50 Synopsis include fortran h PROMULA FORTRAN function declarations int fifrad50 icnt input output int icnt Number of characters to be converted void input Input characters to be converted void output Radix 50 characters return area Description The ifrad50 function converts a specified number of ASCII characters to Radix 50 Conversion stops on the first non Radix 50 character encountered in the input or when the specified number of characters have been converted Radix 50 is a special character representation in which up to three characters can be encoded and packed into 16 bits The Radix 50 characters and their corresponding code values are Character Radix 50 Value space 0 A Z 1 26 d 1 26 27 28 Q 9 30 39 The radix values are stored three characters per word according to the formula cl 40 40 C2 40 c3 where the characters are numbered from left to right If the number of characters is not a multiple of three then it is treated as though it were padded to the right with blanks i e 0 No
206. error return variable 239 PromulaFortran Translator User s Manual Return value A zero if the endfile went well else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Perform I O error processing fiolun Establish unit number fiointu Establish an internal file fiowef Endfile to current file fiostatus Set I O Error Status 6 208 FTNWRIT FORTRAN Write Statement Synopsis finclude fortran h PROMULA FORTRAN function declarations int ftnwrit lun int lun Logical unit number Description Executes the FORTRAN WRITE statement when translated via a non optimized user bias In addition to the logical unit number this function takes a variable number of parameters which specify the actual data to be supplied to the detailed write functions The parameter type codes passed to this function are as follows Code Parameter Description of Use 1 long Points to an error return variable 2 Indicates that list directed I O is being performed 3 char int Points to a full FORMAT specification 4 char Points to FORMAT string 5 int Specifies a record number 9 Specifies that more operations are to be performed 10 int Specifies that a series of operations are desired 11 Introduces a short integer value 12 Introduces a double precision value 13 Introduces a short logical value 14 Introduces a char value 15 Introduces a long value
207. error will be entered into the C output in the form SYNTAX ERROR errno statement Thus the program above produces the following C code under 1 void main argc argv int argc char argv static long i j ftnini argc argv NULL SYNTAX ERROR 145 I J K STOP NULL Any attempt to compile the output will cause errors at that time This setting allows you to see all your errors at one time but blocks you from actually using the results The ER2 switch specifies that syntax errors block at link time It behaves exactly like 1 above except that the following is entered into the C output SYNTAXERROR errno Statement Thus the program above produces the following C code under ER2 void main argc argv int argc char argv 47 PromulaFortran Translator User s Manual static long i j ftnini argc argv NULL SYNTAXERROR 145 I J K STOP NULL This setting allows you to process all errors at once and to go ahead with a provisional compilation however the unsatisfied external SYNTAXERROR will block any attempt to link the program The ER3 switch specifies that syntax errors cause a runtime error to be issued This switch behaves exactly like 1 and ER2 above except that the following is entered into the C output puts ERROR errno statement Thus the program above produces the following C code under ER3 void main argc argv int argc
208. erted to upper or lowercase respectively Since some system s pathnames are case sensitive while others are not it is important to specify one of these options For example though most PRIME pathnames are shown in uppercase most transfer programs create lowercase names when transferring files to UNIX therefore the standard PRIME language description contains a specification of LOWER for this option For initial testing and use of PROMULA FORTRAN for particular small projects the simplest approach is simply to move all source files including the INCLUDE files into the user s local directory To do this PROMULA FORTRAN must be told to ignore all directory information in the source pathnames Under this alternative all characters up to and including the last occurrence of the directory component separations character are stripped from the source pathname This is achieved via the EXCLUDE option 208 PromulaFortran Translator User s Manual A possible alternative structure for the INCLUDE files for a UNIX implementation might be to copy all of these files into some subdirectory where they would retain the same relative structure as they had on the PRIME The PREFIX pname option allows a directory specification to be added to the front of all source pathnames Another alternative might be to copy all include files into a single subdirectory with no additional structure This effect can be achieved by using PREFIX in conjunction with the
209. ery clean looking translation Examples of the use of this flag are presented here The following FORTRAN subroutine contains two COMMON blocks SUBROUTINE DEMO CHARACTER C1 15 C2 10 COMMON ALPHA A 10 10 B 5 C D COMMON BETA C1 10 C2 5 A l1 1 g B 5 D C1 1 3 12 C2 4 RETURN END Below is the default translation for this example which uses the Gp setting void demo extern char Xalpha Xbeta l typedef struct float a 10 10 b 5 c 4d Calpha auto Calpha Talpha Calpha Xalpha typedef struct char GLlil50 c21501 Cbeta auto Cbeta Tbeta Cbeta Xbeta Talpha gt a 0 0 Talpha gt c 50 PromulaFortran Translator User s Manual Talpha gt b 4 Talpha gt d ftnscopy Tbeta c1 42 10 Tbeta gt c2 30 10 NULL return char Xalpha 428 char Xbeta 200 In the function itself COMMON blocks ALPHA and BETA are defined simply as arrays of char note NOT pointers to char Next there are typedefs whose names are constructed by appending a C to the block name which defines the local structure of the block Next there is an auto pointer defined name constructed by adding a T to the block name which points to the start of the char storage area This double definition with a typedef and then a variable of that type is needed because the assignment of the address of the common storage area requires a cast to avoid a compiler error or warni
210. es of this character in some operating environments e g the Korn shell in UNIX the alternative character may be used to mark the project file name The processing of the project file will produce the following three kinds of files 1 The translated FORTRAN files named fileI c 2 A global prototypes file project name h 3 A global variables file project name c To avoid conflicts make sure that the name of the project is different than the names of all the project member files 106 PromulaFortran Translator User s Manual 3 CONFIGURATION FILE PROMULA translates various dialects of FORTRAN source code to a variety of possible C source code outputs It has a powerful dialect processing component for controlling both the syntax of the FORTRAN dialect it accepts as input and the form of the C source code it produces as output This is accomplished via an external configuration file which contains detailed statements for customizing the translation process These statements include the following SWITCHES specifies command line switches COMMENTS controls format of comments PATHNAMES controls location and naming of include filenames RESTRUCTURE restructures variables in the source code KEYWORDS changes output keywords and pattern strings PRAGMA configures special comments into C pragmas controls treatment of the dollar sign Function prototypes control the translation of functions and their arg
211. ess FORTRAN codes and later when they compile the C outputs In these cases the simple extern declaration of rfunc can interfere with the full ANSI declaration The P32 flag simply removes the extern declarations for functions defined via user supplied prototypes The same translation as above with the additional use of P32 produces the following translation void alpha ia gval long ia float gval extern long ifunc lt static float r static long i rfunc ia i ifunc gval WRITE OUTPUT LISTIO INTA i REAL4 r 0 STOP NULL float rfunc i long i static float rfunc rfunc fifiabs i return rfunc In this translation note that func still has an extern declaration but that rfunc no longer has 2 27 6 P64 Write PFC Style Prototypes not C Type As is discussed extensively in the section on the treatment of CHARACTER variables PROMULA FORTRAN must add a character length argument to the subprograms which use FORTRAN CHARACTER variables As a result the argument list for such functions differs in their C form from their original FORTRAN form Consider the following FORTRAN subroutine with two CHARACTER 10 arguments SUBROUTINE ALPHA IA GVAL CHARACTER 10 IA GVAL PRINT IA GVAL STOP END The default prototype written for this function under the default CHd character treatment switch is as follows 67 PromulaFortran Translator User s Manual vo
212. ested logical or integer value See also fiofinqu Generic file information function 192 PromulaFortran Translator User s Manual 6 117 FIOFWSP Skip Format White Space Synopsis finclude fortran h PROMULA FORTRAN function declarations void fiofwsp Description This function is a utility used by the FORMAT processing functions to skip over white space within the specification string White space consists of blanks and boundaries between the independent lines of the specification Return value None the function is void The effect of its processing is reflected in the various global format control variables See also None 6 118 FIOINTU Establish FORTRAN Internal Unit Synopsis finclude fortran h PROMULA FORTRAN function declarations int fiointu intu rsize action char intu Pointer to internal storage int rsize Record size int action Specifies read 1 or write 0 Description When some action is to be performed on an internal unit typically a character storage area this function initializes a file structure to point to this internal unit and makes this file structure the current one Note that if the record size is zero the internal unit is a set of pointers to C style strings and not FORTRAN style strings The form of the creation of the INTERNAL depends upon the type of the action to be performed This action code is as follows Code Action to be performed 0 An internal file is to be opened wi
213. expressions require pointer arithmetic at run time and therefore produce inefficient code especially on the PC when large or huge memory models are in effect Gd This is a variation of the Gp option except that it may produce better code for some environments Under Gp local auto variables are introduced as structure pointers while under Gd the same syntactic effect is achieved via a defined symbol 49 PromulaFortran Translator User s Manual Gs Defines COMMON blocks as static structures with varying internal composition In particular this technique defines each occurrence of the COMMON block as an external structure whose members are defined in the same manner as they are defined in the subprogram being translated This technique produces readable code and generates relatively efficient code The problem is that many compilers consider redefining an external with a detailed structure repeatedly in different functions to be a typing error Some give warnings and others consider this a fatal error If this technique works with your compiler and if it does not offend your own view of symbol typing then this technique is highly recommended Note that it can be used freely with the Om flag discussed elsewhere Gr Defines COMMON blocks as raw static vectors of char This technique is used when an exact block layout is required or when the efficiency of the Gs technique is desired in an environment where it cannot be used Here the vari
214. f the following is true 1 The increment is a variable expression 2 The maximum is a complex expression or is changed inside the loop 3 The control variable is changed inside the loop Most DO loops do not meet the above criteria therefore a much simpler form of the loop can be used Consider the following which shows the same example as before except the increment is fixed constant SUBROUTINE DEMO INTEGER V E1 E2 DO 10 V El E2 3 WRITE V 10 CONTINUE RETUR END Now there is no need to introduce any temporary variables since the simplistic view of the DO loop gives the same result void demo static int v el e2 for v el v lt e2 vt 3 WRITE OUTPUT LISTIO INT4 v 0 return The final point to be made about DO loops is their close relation to array subscripts In many instances the DO loop control variables can be reduced if the control variable is used only as an array subscript The following example shows this SUBROUTINE DEMO INTEGER V E2 DIMENSION A 10 10 DO 10 V 1 E2 A V V 10 0 10 CONTINUE RETURN END In this example the source of the array subscripts are DO loop counters and those counters are used only as subscripts therefore the counters can be reduced thus simplifying the subscript expression and simplifying the translation of the DO loop This particular optimizat
215. f the number of significant digits and sign required to represent the value is less than the specified width the unused leftmost portion of the field is filled with blanks If it is greater than the width asterisks are entered instead of numeric digits If a minimum digit count is specified the output field consists of at least that many digits and is zero filled as necessary If the minimum digit count is zero and the value is Zero then the field is simply blank filled regardless of any sign control in effect The parameter value contains the value to be converted The conversion control parameters are specified via global variables Return value None the function is void The global variable iocrec is updated to contain the new field and the variable fionchar is updated to reflect the new character count of the coded communication record See also fioitos Convert integer to string fioshr Shift display string right 6 122 FIOLUN Establish FORTRAN Unit Number Synopsis finclude fortran h PROMULA FORTRAN function declarations int fiolun lun action int lun Logical unit number of file int action Action code for subsequent use Description Before any action can be performed on a FORTRAN file the logical unit number must be associated with an existing FORTRAN file structure If there is no already existing structure for the unit number then this function will attempt to create one The form of this creation depends up
216. fd int option char str int ns Description e is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing e error codes Pointer to current FORTRAN file Converts an integer to a string Specifies presence of error checking Code for actual error encountered Returns an error code or zero IOFDATA FORTRAN File Data fortran h PROMULA FORTRAN function declarations ata option str ns Specifies which data is being specified String information String length or integer information This function is used to specify the various file data options associated with the current FORTRAN file structure The particular data being specified is defined by the opt ion parameter as follows Option Description of data 1 Specifies the file name to be assigned to the file This name may have no more than 39 characters If the name is NULL or if it is all blank then no name is assigned If the file is opened with no name assigned then the open function will request a name from stdinp via stdout Specifies the status of the file Only the first character of the string parameter is used as follows Char Meaning Old the file already exists do not create it Nn New the file does not exist create it even if this means destroying an existing file with the same name 5 5 Scratch the same as new except that the file is removed when it is closed Unknown if the file e
217. fication EWRB IFS 129 fiowrb Bad format specification EWRS IFS 130 fiowrs Bad format specification EWRT IFS 131 fiowrt Bad format specification EWVL IFS 132 fiowval Bad format specification 134 PromulaFortran Translator User s Manual ENXF EFS 133 fionxtf End of format string ENXF BTF 134 fionxtf Bad T format ENXF BUS 135 fionxtf Bad B business format string ENXF BBF 136 fionxtf Bad BN Z format ENXF_WID 137 fionxtf Missing width specification ENXF_DEL 138 fionxtf Missing terminating delimiter ENXF_HOL 139 fionxtf Bad Hollerith string ERCK_BUF 140 fiorchk Internal buffer exceeded ERDX_MLP 141 fiordx Missing left parenthesis ERDX_COM 142 fiordx Missing comma ERDX MRP 143 fiordx Missing right parenthesis Error Code Function Description ERDZ MLP 144 fiordx Missing left parenthesis ERDZ COM 145 fiordx Missing comma ERDZ MRP 146 fiordx Missing right parenthesis ERNL MNI 147 fiornl Missing namelist identifier ERNL MVI 148 fiornl Missing variable identifier ERNL UVI 149 fiornl Undefined variable identifier ERNL SSV 150 fiornl Subscripted scalar variable ERNL NNS 151 fiornl Nonnumeric subscripts ERNL TMS 152 fiornl Too many subscripts ERNL EQL 153 fiornl Missing equals sign ERNL BSI 154 fiornl Bad string input ERNL MLP 155 fiornl Complex missing left pren ERNL COM 156 fiornl Complex missing comma ERNL MRP 157 fiornl Complex missing right pren ESTD NNC 158 fiostod Nonnumeric character in field 5 4 Noncomplex Intrinsic Func
218. fill character Description Shifts a character string right a specified number of places The spaces thus created are set equal to the specified fill character This function is typically used during detailed editing of displays during various numeric conversions Return value 210 PromulaFortran Translator User s Manual None the function is void See also None 6 150 FIOSPACE Skip White Space in Record Synopsis finclude fortran h PROMULA FORTRAN function declarations int fiospace Description This function is a utility used by the list directed input processing functions to skip over white space within the input record White space consists of blanks and newlines Return value If the function reads a new record during its processing it returns a 1 else it returns a zero See also fiortxt Read next text record 6 151 FIOSTATUS Set FORTRAN I O Error Status Synopsis include fortran h PROMULA FORTRAN function declarations void fiostatus iostat error long iostat Address of error status variable int error Error testing switch Description If the FORTRAN I O runtime system encounters an error it sets an error code and calls function fioerr The behavior of that function depends upon how the code using the I O system is doing error processing This function establishes the error code return variable and the error checking level Return value None the function is void See also None 6
219. finclude fortran h PROMULA FORTRAN function declarations void ftnblkd Description This function represents a dummy entry point for those FORTRAN programs which have no BLOCK DATA subprograms It is called by the FORTRAN runtime initialization function Return value None the function is void See also None 6 191 FTNCLOSE FORTRAN Close Statement Synopsis 230 PromulaFortran Translator User s Manual include fortran h PROMULA FORTRAN function declarations int ftnclose int lun Description Executes the FORTRAN CLOSE statement when translated via a non optimized user bias In addition to the logical unit number this function takes a variable number of parameters which specify the actual data to be supplied to the close operation The parameter type codes passed to this function are as follows Code Parameter Description of Use Ends the list of parameters 1 long Points to an error return variable Return value A zero if the close went well else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiointu Establish an internal file fioclose close current file fioerror Perform I O error processing fiolun Establish unit number fiostatus Set I O Error Status 6 192 FTNCMS FORTRAN Compare Strings Synopsis include fortran h PROMULA FORTRAN function declarations int ftncems sl1 nl1 s2 n2 char 5
220. follows void demo extern void alpha auto float K1 3 0 static long i ftnini argc argv NULL alpha amp i amp K1 STOP NULL In this translation the variable K1 is declared as being auto Under this declaration its value will be refreshed upon each invocation of the subroutine In some cases the use of Ka and Ks can produce different results as the following discussion explains A constant is a fixed value Note that in most implementations of FORTRAN including this one the value of a constant can be changed by passing it to a subprogram as a parameter As an example the following code will produce a sequence of values from 2 to 11 PROGRAM DEMO DO 10 I 2 10 f 1 10 CONTINUE Gi t W O Gq H 2 E ALPHA 1 UBROUTINE BETA I I The default translation of this example is as follows void main argc argv int argc char argv extern void alpha beta static long K1 1 static long i ftnini argc argv NULL for i 2L i 10L i alpha amp K1 beta amp K1 STOP NULL void alpha i long i 56 PromulaFortran Translator User s Manual i i 1L return void beta i long i WRITE OUTPUT LISTIO INT4 i 0 return The reason for the sequence of values is that the value of the constant is stored in a constants table or in the C translation
221. for runtime data 25 namelist structure name for namelist 26 declaration which precedes includes 27 ftnini argc argv NULL runtime initialization function call 28 vmsopn argc argv runtime initialization call for virtual memory 29 void main argc argv main program declaration 30 int argc declaration for main argc argument 31 char argv declaration for main argv argument 32 define define keyword 33 undef undef keyword 116 Keyword if switch case default break goto for return DATAVAR DATAVAL FIRST IENTRY RETNUMB ELP 0 0 0 0 oO OL 0 0 OL 0 0 0 dpxzero cpxzero int NULL fiointu char fiornl union sizeof INT2 REAL8 LOG2 BYTE INT4 REAL4 LOG4 LOGI DCMPLX STRG DO CSTR 7 7 25 25 15 2 2c 3 3d 12 121d 16 16 6E 2 2c 3 3d 25 25 15LE PromulaFortran Translator User s Manual Description of use in C if keyword switch keyword case keyword default keyword break keyword goto keyword for keyword return keyword namelist structure for runtime data data for runtime data marks end of runtime data variable which record first entry into function entry point number variable alternate return variable end local processing statement label zero of type of INTEGER 2 zero of type of REAL 8 zero of type LOGICAL 2 zero of type INTEGER 1 zero of type INTEGER 4 zero of type REAL 4 zero of type LOGICAL 4 zero of type LOGICAL 1
222. format specification Note that in this function each individual value is assumed to have its own corresponding format specification if a formatted write is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowrl Write FORTRAN long integer vector vmsuse Use a virtual value 6 242 VMSWRS Write FORTRAN Virtual Vector of Strings Synopsis include fortran h PROMULA FORTRAN function declarations int vmswrs value nstring nval long value Virtual address start of strings int nstring Length of each string int nval Number of strings to be written Description Writes a virtual sequence of fixed length strings stored one after another to the current output file in accordance with the current format specification Each string is assumed to have its own corresponding format specification if a formatted write is being performed Return value 257 PromulaFortran Translator User s Manual A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowrs Write FORTRAN string vector vmsuse Use a virtual value 6 243 VMSWRT Write a Virtual Long Truth Value Vector Synopsis include fortran h PROMULA FORTRAN function declarations int vmswrt value nval long v
223. format specification Note that in this function each individual character is assumed to have its own corresponding format specification if a formatted write is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Do requested error processing fiowrih Perform short integer output 6 166 FIOWRD Write FORTRAN Double Precision Vector Synopsis include fortran h PROMULA FORTRAN function declarations int fiowrd value nval double value Points to values being written int nval Number of values to be written Description Writes a vector of double precision values to the current output file in accordance with the current format specification Note that in this function each individual value is assumed to have its own corresponding format specification if a formatted write is being performed Return value 218 PromulaFortran Translator User s Manual A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Do requested error processing fiofout Process output format specification fiowalph Write alphabetic information fiowval Write value 6 167 FIOWRF Write FORTRAN Single Precision Vector Synopsis include fortran h PROMULA FORTRAN func
224. g num rev num num rev 2L return char Xmodelhead 16 3 5 4 Pattern String for Subprogram Surrogates As is discussed under the Y4 command line switch a typical FORTRAN optimization involves making local surrogates for subprogram parameters Using Y4 with the example used above produces the following translation only last subprogram shown void test5 Pnum Prev long Pnum Prev auto long num Pnum auto long rev Prev num num rev 2L goto ELP ELP Pnum num The manner in which the local surrogates are formed can easily be changed Consider the following configuration file KEYWORDS ps Local copy END which will change the look of the translation considerably void test5 Local num copy Local rev copy long Local num copy Local copy auto long num Local num copy auto long rev Local rev copy num num rev 2L goto ELP ELP Local_num_copy num 122 PromulaFortran Translator User s Manual 3 5 5 Pattern Strings for VAX Descriptors Users of VAX FORTRAN often want to treat characters as descriptors The CHv flag gives this basic treatment Consider the following simple code UBROUTINE DEMO I J HARACTER 12 NAME HARACTER 4 INITIAL INDEX NAME INITIAL JINDEX NAME INITIAL ETURN E ADUHAAN z The output of the code above using CHv command
225. g both a starting position and length are needed The starting position is simply min 1 while the length can be computed as max min 1 In the simple case where the substring range is constant PROMULA FORTRAN performs the computation at compile time When the range is not constant the computation must be inserted in the code This can be seen in the second and third examples The third example shows the problem Since the value of min is needed twice in the computations the min expression is evaluated twice This double evaluation is inefficient Under the CHd and CHv approach the descriptor is used This looks as follows void demol static char name 12 initial 4 static int i j static string Tl ftnscopy initial 4 name 4 4 NULL ftnscopy initial 4 name i 1 j i 1 NULL 31 PromulaFortran Translator User s Manual Tl n fifindex name 12 initial 4 1 Tl a name T1 n Tl n j Tl n ftnscopy initial 4 T1 a T1 n NULL return Under this approach the two members of the T1 are computed using the min only once In the following copy request the members of T1 are then used rather than the source Notice that CHd and CHv also translate character assignment differently This issue is discussed in the next subsection under character concatenation 2 5 4 Character Concatenations It is from its treatment of string concatenation that CHd CHaracter dynamic gets its name Both CHr and CHs allow for static conca
226. g is reflected in the various global format control variables See also fiortxt Read next text record ftnxcons Process Exact representation constant fiorchk Check fixed form input field 6 127 FIORBIV FORTRAN Read Binary Values Synopsis finclude fortran h PROMULA FORTRAN function declarations int fiorbiv value nvalue void value Points to values being read int nvalue Number of bytes to be read Description 198 PromulaFortran Translator User s Manual This function reads binary values from a file Return value A zero if all went well else an error code See the general discussion of the FORTRAN I O capabilities for a listing of the possible error codes See also None 6 128 FIORCHK Check Fixed Form Input Field Synopsis include fortran h PROMULA FORTRAN function declarations void fiorchk nfield int nfield Description This function controls the physical reading of text records If the logical end of record has been reached the next physical record is read If a physical end of record has been reached prior to the logical end of record the physical end of record is extended by padding the record with blanks Return value None the function is void The effect of its processing is reflected in the various global format control variables See also None 6 129 FIORDB Read FORTRAN Boolean Vector Synopsis include fortran h PROMULA FORTRAN function declarations int fiordb bool n
227. g to each C function written to a separate file with the same name as the function The O switch gives you this capability Its mutually exclusive settings are as follows Setting Meaning Om Writes each function to a separate file with the same name as the function and with the extension C Os Writes all output to the same file This is the default Because of the fact that many C constructs such as define and typedef have module scope the Om flag can also be used to avoid a problem with the treatment of COMMON blocks as discussed in the section on the global variables convention flag The cleanest treatment can be achieved via the Gs switch which defines COMMON blocks as static structures with varying internal composition In particular this technique defines each occurrence of the COMMON block as an external structure whose members are defined in the same manner as they are defined in the subprogram being translated This technique produces readable code and generates relatively efficient code The problem is that many compilers consider redefining an external with a detailed structure repeatedly in different functions to be a typing error If you are using such a compiler then the Om switch can be used to overcome it 2 27 Miscellaneous Prototyping Control Flags Pnumb P numb The P command line switch introduces 15 flags for use in dealing with the prototypes and other output issues faced by PROMULA FORTRAN What brings these flags to
228. gether is that all are primarily focused on the differences in semantics between FORTRAN and C The flags and their values are as follows 1 Include definitions of int functions when listing external function references 2 Include ANSI prototypes with declarations of functions passed as arguments 4 Exclude referenced function prototypes from the prototype output file 8 Exclude defined function prototypes from the prototype output file 16 Ignore prototypes for definitions 32 Treat user prototypes as system functions 64 Write PROMULA FORTRAN style prototypes not C type 128 Write all function decls to header file 256 Use ANSI C function declarations 512 Make parameters always take explicit value type 1024 Exclude undefs from the translation 2048 Force variables to have explicit character type 4096 Define equivalences via a define 62 PromulaFortran Translator User s Manual 8192 Use parameter identifiers in equivalences 16384 Display include files separately Any combination of these flags can be obtained by summing the desired values and then associating that sum with the P switch Alternatively individual flags may be turned on by using the P numb form of the switch 2 27 1 P1 Include Definitions of int Functions The first flag forces the definitions of int functions when listing the external functions referenced within a given prototype This flag is on by default If a referenced function is not defined in a
229. han 6 characters is nonstandard 184 utest for N864 Declarative statements following executable statements is nonstandard 223 utest for N851 The use of inline comments is nonstandard 295 utest for N872 Equivalencing of type character 80 with IBUF of type INTEGER 2 is nonstandard 343 utest for N853 The standard delimiter for a character constant is the single quote 431 utest for N858 The identifier PRANDOM_INDX with more than 6 characters is nonstandard 474 utest for N806 Omitting the comma after the I FORMAT specification is nonstandard 545 utest for N823 COMMON block DIR1 has character 1 variable VFORMS and an unspecified variable OPTEV 722 utest for N801 The INCLUDE statement is nonstandard 1 STRUC6 INC N801 The STRUCTURE statement is nonstandard 800 utest for N833 The non parenthetical form of the PARAMETER statement is nonstandard 41 PromulaFortran Translator User s Manual As can be seen the EL3 level generates many messages and is primarily intended for those who are trying to write pure standard conforming code 2 9 2 Annotated Listing of Source Code The following listing was produced via the ES option for a simple subprogram referencing a single include file PROMULA FORTRAN Compiler V4 00 Date 08 11 92 Time 09 35 Page 1 File utest for
230. har s Main string int ns Length of main string char c Substring int nc Length of substring Description Returns the location of a substring within a string Note that the arguments are two FORTRAN style strings which means that the length of each string must also be sent Return value If the substring occurs within the main string the result is an integer indicating the starting position relative to 1 of the first occurrence of the substring within the main string If the substring does not occur within the main string a zero is returned See also None 6 82 FIFIPOW FORTRAN Intrinsic Function IPOW Synopsis include fortran h PROMULA FORTRAN function declarations long fifipow a b long a Value to be raised to a power long b Power to be used Description Raises a long integer value to a long integer power Return value The long integer result See also None 6 83 FIFISHF FORTRAN Intrinsic Function ISHFT Synopsis include fortran h PROMULA FORTRAN function declarations long fifishf long a Value to be shifted int n Number of positions to be shifted Description Shifts the long argument left or right the specified number of bit positions depending upon whether the position count is positive or negative Return value 174 PromulaFortran Translator User s Manual The long integer result See also None 6 84 FIFISIGN FORTRAN Intrinsic Function ISIGN Synopsis includ
231. he C language form PromulaFortran Translator User s Manual 2 COMMAND LINE To use PROMULA FORTRAN in its simplest mode type its name pfc followed by the name of the FORTRAN file that you wish to process and then press the Enter or Return key This operation assumes that the following files are stored either in the default path or more typically in the path specified via the operating system pfc exe The PROMULA FORTRAN executable pfc pak The description of the FORTRAN and C dialects to be used To compile codes you need to use whatever conventions are appropriate for the C compiler that you are using Typically this means that the file fortran h should be copied into the same directory or disk which contains the other include files for the compiler and the appropriate PROMULA FORTRAN runtime library file pfcmsc lib or libpfc a Or pfcvms olb should be copied into the same directory which contains the other library files for the linker Note that details on the installation and placement of the PROMULA FORTRAN files are included with the installation instructions which accompanied the PROMULA FORTRAN product There are three ways in which to modify the behavior of PROMULA FORTRAN The easiest way is via command line switches This chapter discusses the use of PROMULA FORTRAN via its command line switches In addition you may supply a configuration file which gives you detailed control over particular aspects of the trans
232. he character arguments Again notice that the index system function is unaffected by these games The CHd result below shows the best approach void demol i Jj pnt sir Jy extern int jindex static char name 12 initial 4 j fifindex name 12 initial 4 xi initial 12 4 return void demo2 i j int Lp ty extern int jindex static int name 3 initial j fifindex name 12 amp initial 4 jindex name amp initial return void demo3 i j inte extern int jindex static char name 12 static int initial j fifindex name 12 amp initial 4 i jindex name amp initial 12 return With the lengths all moved to the back the calls to jindex at least always begin with the two pointers Remember that we are not dealing only with well written FORTRAN codes Things such as the above often occur in real FORTRAN programs Finally the listing below shows the translation using the CHv switch void demol i j int i extern int jindex static char name 12 initial 4 static string Tl NULL 0 static string 2 NULL 0 xj fifindex name 12 initial 4 Tl a name 1 12 T2 a initial T2 n 4 xi jindex amp T1 amp T2 return void demo2 i j Int i extern int jindex static int name 3 initial j fifindex name 12 amp initial 4 i jindex name amp initial
233. ia float gval extern long ifunc extern float rfunc static float r static long i r rfunc ia i ifunc gval WRITE OUTPUT LISTIO INT4 i REAL4 r 0 STOP NULL float rfunc a float a g static float rfunc rfunc fifabs a lt return rfunc 2 27 5 P32 Treat User Prototypes as System Functions When PROMULA FORTRAN translates a reference to an external subprogram which is not a system function it always includes an extern declaration for that function This is true even when user prototypes are being read by PROMULA FORTRAN Consider the following FORTRAN code UBROUTINE ALPHA GVAL RFUNC IFUNC GVAL RINT I R OP z UNCTION RFUNC I FUNC ABS I ETURN BE ug hnmdog mnzuou Z when translated using the following prototype file float rfunc long produces the following C translation void alpha ia gval long ia float gval extern long ifunc 66 PromulaFortran Translator User s Manual extern float rfunc lt static float r static long i r rfunc ia i ifunc gval WRITE OUTPUT LISTIO INT4 i REAL4 r 0 STOP NULL float rfunc i long i static float rfunc rfunc fifiabs i return rfunc Note that the function rfunc is explicitly given an extern declaration Many users of PROMULA FORTRAN use the same prototype files when they proc
234. ialization of local variables This is the most efficient method and works for most cases The alternative is to use a combination of FORTRAN READ NAMELIST and internal file capabilities to set the data initialization values at runtime This method is less efficient because it requires the construction of tables in the code and because it must be executed This method is always used for initializations to COMMON and when you use the DR data runtime switch on the command line 2 7 1 Overview of Initialization Problem The following example shows a set of data initializations which highlight several important points about DATA statements in FORTRAN SUBROUTINE DEMO DIMENSION B 5 X 5 5 C 10 DATA A 4 2 B 1 5 4 DATA X J 1I I 1 J9 J 1 5 1 2 2 3 3 4 4 5 5 DATA J 4 C J J 1 10 10 2 0 1 FORMAT 1X 10F10 1 2 FORMAT 1X 110 WRITE 1 A B WRITE 1 C WRITE 2 J 35 PromulaFortran Translator User s Manual DO 10 J 1 5 WRITE 1 X J I I 1 5 10 CONTINUE STOP END Note first in the initialization of x that the value of the dummy subscript J is used as a terminal point for the subscript 1 This is allowed in DATA statements Note next that in the following DATA statement a program variable J is given the value of 4 and then a dummy variable J in the DATA statement is allowed to run from 1 to 10 Dummy variables in DATA statements are not program variables and
235. ic memory This is the default memory for variables It is equivalent to the normal FORTRAN memory allocation It is wasteful in that local variables all have unique memory locations but is simple to implement Also static storage has memory between calls This means that local variables can retain their values between calls to their code This is a trick that is often used by older FORTRAN source codes Second there is auto storage This storage is allocated to the stack each time a function is called When the function exits the storage is returned to the stack for use by other functions The advantages of auto storage are that it is fast easy to use and economizes on total storage used The disadvantages are that it has no memory and that it is very limited on PC platforms where program stacks are typically quite short and always less than 64K on the PC Third there is dynamic storage Dynamic storage is typically taken from that storage available on the machine which is not already allocated to the program This is generally called the heap In C it is accessed via the malloc function and is returned via the ree function The advantages of dynamic storage is that it is most likely to be abundant and economizes total storage used The disadvantages are that it must be explicitly allocated each time a function is entered and freed each time a function is exited PROMULA automatically inserts the code to do this when dynamic storage is used Fo
236. id alpha char char int int Alternatively this prototype becomes void alpha char int char int when the CHs switch is used Depending upon the context either of these prototypes would be appropriate for use with a later C compilation however if the intent is to use the prototypes with PROMULA FORTRAN then the following is needed void alpha string string This prototype does not have the character treatment assumption already decomposed This prototype is produced when the P64 switch is used 2 27 7 P128 Write All Function decls to Header File The following sample FORTRAN code shows a potentially serious problem for PROMULA FORTRAN PROGRAM DEMO CALL IALPHA Hello World STOP END FUNCTION IALPHA GVAL CHARACTER IA GVAL PRINT IA GVAL IALPHA LEN IA LEN END This code shows a function which writes the concatenation of two strings and then returns the length of that concatenation The only reference to this function is via a CALL statement i e the fact that IALPHA also returns a value is being ignored The default translation for this code is shown below void main argc argv int argc char argv extern void ialpha lt ftnini argc argv NULL ialpha Hello World 5 6 STOP NULL long ialpha ia gval P1 P2 lt char ia gval int P1 P2 static long ialpha WRITE OUTPUT LISTIO STRG ia P1 STRG gval
237. ier sections of this chapter An additional problem that comes up is that some of the runtime libraries have been implemented via FORTRAN and others via C Some FORTRANS and or some Cs append additional characters to each external symbol be it a function or a subroutine This requires that groups of global symbols but not all use a modified naming convention These modifications are achieved via GLOBAL strings which may be entered into prototype files The GLOBALS string consists of two characters only the function or subroutine prefix and suffix characters As an example consider the following piece of a prototype file for a FORTRAN based system in which the FORTRAN compiler appends an underscore character GLOBALS void actday long long long long void valdt long long long short void fixdt long long long long long long long short void weeknd long long long long long long long short The actual public names are actdat_ valdt fixdt_ and weeknd The GLOBALS string preceding these tells PROMULA to make this change in the target C 4 6 Renaming Identifiers Only A final capability of the configuration file is to allow the user to enter simple identifier renaming requests The following notation oldname newname in a prototype input file will rename all occurrences of oldname with newname 132 PromulaFortran Translator User s Manual 5 OVERVIEW OF RUNTIME LIBRARY The PROMULA FO
238. ification if a formatted read is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiordi Read FORTRAN short integer vector vmsdel Change a virtual value 6 229 VMSRDL Read FORTRAN Virtual Long Integer Vector Synopsis include fortran h PROMULA FORTRAN function declarations int vmsrdl value nval long value Virtual address of values int nval Number of values to be read Description Reads a virtual vector of long fixed point values from the current input file in accordance with the current format specification Each value is assumed to have its own corresponding format specification if a formatted read is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiordl Read FORTRAN long integer vector vmsdel Change a virtual value 6 230 VMSRDS Read FORTRAN Virtual String Synopsis include fortran h PROMULA FORTRAN function declarations int vmsrds value nstring nval long value Virtual address of start of strings int nstring Length of each string int nval Number of strings to be read Description 251 PromulaFortran Translator User s Manual Reads a virtual sequence of fixed length strings stored one af
239. ilation errors c During execution by the debugger of the C compiler For UNIX based systems the dbx debugger may be used and it can reference either the translated C code or the lines of the original FORTRAN source code 1 4 Translator Advantages PromulaFortran Translator User s Manual As a FORTRAN to C translator PROMULA FORTRAN offers a number of advantages over other FORTRAN to C translators Multi Dialect Translate the standard FORTRAN 66 and FORTRAN 77 dialects as well as various other Processing extended dialects such as VAX IBM VS PDP PRIME Honeywell and Data General FORTRAN A particular FORTRAN dialect is selected at the time of purchase Additional FORTRAN dialects may be purchased later Cross Platform Translate your code on one platform for compilation linking and running on a second platform Availability Completeness The syntax processor handles almost all existing standard and extended FORTRAN dialects It also supports a number of Fortran 90 features such as structures and pointer variables On the execution side its comprehensive runtime library covers the full functionality of all of the above FORTRAN dialects full FORTRAN I O complex arithmetic NAMELIST DECODE ENCODE DEFINE FILE B FORMAT VAL DESCR STRUCTURE RECORD embedded and inline comments etc Correctness Since PROMULA FORTRAN compiles to C i e performs both a syntactic and a semantic transformation to C the transla
240. ile will do this KEYWORDS 6 INTEGER4 13 CHARACTER END When test4 is translated using the Rdemo command line switch the following translation will be produced void test4 heading P1 CHARACTER heading int P1 extern void test5 extern char Xmodelhead typedef struct INTEGER4 hd_hedsiz Length of header INTEGER4 hd_numtab No of tables in file INTEGER4 hd_filrev Model file revision date INTEGER4 hd_sysprim No of system primitive types 119 PromulaFortran Translator User s Manual j Cmodelhead auto Cmodelhead Tmodelhead Cmodelhead Xmodelhead test5 amp Tmodelhead gt hd_numtab amp Tmodelhead gt hd_filrev Tmodelhead hd hedsiz Tmodelhead hd numtab Tmodelhead hd filrev WRITE 1 LISTIO STRG heading 10 0 WRITE 1 LISTIO INT4 Tmodelhead hd hedsiz INT4 Tmodelhead hd numtab 0 WRITE 1 LISTIO INT4 Tmodelhead gt hd_filrev INT4 Tmodelhead gt hd_sysprim 0 r void test5 num rev INTEGER4 num rev num num rev 2L return char Xmodelhead 16 Notice that there are several different char occurrences within the output Some pertain directly to the CHARACTER type while others are used for various memory allocation specifications Keyword 13 is the one for the CHARACTER type Alternatively there is only on
241. iles referenced in the current subprogram then this section is omitted The philosophy behind the design of the symbols referenced table is that the user will use this table when he wants information about a particular symbol whose status he may not be familiar with but whose identifier he knows or has seen somewhere in the listing The report consists of a single alphabetical list of each symbol referenced in the subprogram For each symbol its object type its storage status its binary type and a comment are provided The object type is straightforward There are thirteen possible entries constant parameter variable subroutine function intrinsic namelist entry statefunc structure record pointer and common These names correspond directly to the possible FORTRAN object types It should be mentioned that members within structure definitions are treated simply as variables or records This convention is compatible with the approach of using a single alphabetized list of all symbols The storage status of a symbol can be one of four different things For subprogram arguments it is specified as argument For variables in common blocks it is the name of the common block containing the variable For members of structures it is the name of structure containing the member For simple variables it is one of the following static auto dynamic Or virtual See the Sa and Ss switches for a description of static versus auto storage Dynamic and vi
242. ility If a program were written in pure FORTRAN 77 then any one of the existing compilers could probably process it with no problem Unfortunately few programs are written in pure 77 FORTRAN and as time passes the possible hosts for these programs disappear Our treatment of CHARACTER data attempts to deal with all of the dialects with which we are familiar If a program hides its characters in numeric variables no problem If a program mixes character and non character variables in COMMON blocks or across subroutine parameters no problem If a program assumes that a CHARACTER 20 in one place is a CHARACTER 1 20 in another place no problem If it worked in the original then it will work in the C version A character variable is treated simply as a sequence of chars It has no other structure A CHARACTER 4 is identical to a CHARACTER 2 2 a CHARACTER 1 4 or a sequence of characters hidden in an INTEGER 4 The elements of a CHARACTER array are stored one after another with no intervening NULL characters or character counts or intermediate character pointers This storage of character data is simple and allows the easy simulation of the various tricks allowed by the various dialects If FORTRAN had no subprograms there would be no problems Unfortunately FORTRAN subprograms allow for variable length characters and the various system operations allow for the total mixture of character strings of any length To deal with a variable character
243. ill be promoted to double when passed by value The Yr flag on the other hand treats single precision real arithmetic completely separately from double No automatic promotion is assumed In addition single precision real constants have an F appended 2 47 Subscript Pointer Notation Xs Ys Ysv Ysf There is a difficult translation issue associated with array subscript expressions In FORTRAN an array is a set of elements identified by a single name The dimension of the array merely specifies how many elements there are and how they are organized The elements in an array are always stored contiguously in memory in row major order That is the leftmost subscript varies the fastest as elements are accessed in storage order When a FORTRAN array is defined its minimum and 101 PromulaFortran Translator User s Manual maximum subscript bounds may be any value positive or negative just so long as the minimum bound is less than the maximum bound If no minimum bound is specified a minimum of one is assumed In C there is a strong relationship between pointers and arrays Though the array notation and pointer notation may be mixed fairly freely the pointer version will in general be more efficient Multidimensional arrays are stored contiguously in memory in column major order That is the rightmost subscript varies the fastest as elements are accessed in storage order When defining an array the user may only specify its size the
244. ing of the possible error codes See also fioltos Convert long to string fioerror Do requested error processing fiofout Process output format specification fiowalph Write alphabetic information fiowtxt Write a text record 6 169 FIOWRL Write FORTRAN Long Integer Vector Synopsis include fortran h PROMULA FORTRAN function declarations int fiowrl value nval long value Points to values to be written int nval Number of values to be written Description Writes a vector of long integer values to the current output file in accordance with the current format specification Note that in this function each individual value is assumed to have its own corresponding format specification if a formatted write is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioltos Convert long to string fioerror Do requested error processing fiofout Process output format specification fiowalph Write alphabetic information fiowtxt Write a text record 6 170 FIOWRS Write FORTRAN Vector of Strings Synopsis finclude fortran h PROMULA FORTRAN function declarations int fiowrs str nstring nval char str Points to start of strings int nstring Length of each string int nval Number of strings to be written Description 220 PromulaFortran Translator User
245. ingle dollar sign will cause a to be printed in the corresponding position in the output field fixed dollar Multiple dollar signs will be replaced by printing characters in the number and a single will be printed in the position immediately to the left of the leftmost printing character of the number floating dollar Asterisk Asterisks may be preceded only by an optional fixed sign and or a fixed dollar Asterisks in positions used by digits of the number will be replaced by those digits The remainder will be printed as asterisks field filling Zed Z If the digit corresponding to a Z in the output number is a leading zero a space blank will be printed in that position otherwise the digit in the number will be printed leading zero suppression Number sign The number sign indicates a digit position not subject to leading zero suppression the digit in the number will be printed in its corresponding portion whether zero or not zero non suppression Decimal point A decimal point indicates the position of the decimal point in the output number Only the sign and either trailing signs or credit CR may follow the decimal point Comma Commas may be placed after any leading character but before the decimal points If a significant character of the number not a sign or dollar precedes the comma a comma will be printed in that position If not preceded by a significant character a space will be p
246. int fioname strg ns char strg Name of file int ns Number of characters in file name Description The FORTRAN INQUIRE statement allows the user to inquire about file status either via its logical unit number or via its name If the name reference is being used then this function is called If there is no already existing structure for a unit with this name then this function will attempt to create one Return value A zero if all went well else an error code See the general discussion of the FORTRAN I O capabilities for a listing of the possible error codes See also fifindex Find one substring in another fioerror Performs standard error processing 6 124 FIONXTF Get Next Format Specification Synopsis 196 PromulaFortran Translator User s Manual include fortran h PROMULA FORTRAN function declarations void fionxtf Description Gets the next format specification from the format list and sets the external variables fiofspec fioiwd and fiondec to indicate what it is The above variables are set as follows Specification Type fioiwd fiondec free form 0 0 0 nH 1 n 0 el cn 1 n 1 1 n c cn 1 n nX 2 n TRn 2 n Tn 3 n TLn 4 n SS 5 SP 6 BN 7 BZ 8 9 10 n 11 1 2 Aw 12 w Lw 13 w Iw 14 w Fw d 15 w d Dw d 16 w d Ew d 17 w d Gw d 18 Ww d B ssss 19 Return value None the funct
247. int nval Number of values to be read Description Reads a vector of single precision floating point values from the current input file in accordance with the current format specification Each value is assumed to have its own corresponding format specification if a formatted read is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Perform FORTRAN I O error processing fiofinp Get next formatted input specification fiostod Convert string to double fioffld Get next free form input field fioralph Read alphabetic information fiorchk Check current fixed input field 6 133 FIORDI Read FORTRAN Short Integer Vector Synopsis include fortran h PROMULA FORTRAN function declarations int fiordi value nval short value Vector of values int nval Number of values to be read Description Reads a vector of short fixed point values from the current input file in accordance with the current format specification Each value is assumed to have its own corresponding format specification if a formatted read is being performed Return value 201 PromulaFortran Translator User s Manual A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror
248. ioffld Get next free form input field fioralph Read alphabetic information fiorchk Check current fixed input field 6 137 FIORDU Read FORTRAN Unsigned Char Vector Synopsis include fortran h PROMULA FORTRAN function declarations int fiordu c nval unsigned char c Points to characters to be read int nval Number of characters to read 203 PromulaFortran Translator User s Manual Description Reads a vector of character values from the current input file in accordance with the current format specification In this implementation unsigned char is derived from the nonstandard FORTRAN type LOGICAL 1 Therefore L formatting conventions are assumed Note that in this function each individual character is assumed to have its own corresponding format specification if a formatted read is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also int fiordb Read short Boolean value int fioerror Perform FORTRAN I O error processing 6 138 FIORDX Read FORTRAN Complex Values Synopsis include fortran h PROMULA FORTRAN function declarations int fiordx value nval complex value Points to values to be read int nval Number of values to be read Description Reads a vector of single precision complex values from the current input file in accordance with the current format
249. ion derives from a particular pattern in FORTRAN programs It is unlikely that any generalized optimizer found in a C compiler would detect or perform this optimization void demo static int v e2 static float a 10 10 for v20 v e2 a v v 10 0 85 PromulaFortran Translator User s Manual return PROMULA supports both FORTRAN 66 and FORTRAN 77 dialects In most cases this merely means ignoring various restrictions placed on the language in the 77 standard and then adding the 77 constructs There is no real conflict between the two dialects except for one thing DO statements DO is probably the most used statement in the language except for assignment What was in the minds of the committee It must have been a very long night Anyway with FORTRAN 66 the following code fragment will compute a value of 1 while in 77 it will produce a value of zero IVAL 0 N 0 DO 10 I 1 N IVAL IVAL 1 10 CONTINUE This is because in 77 a DO loop is not executed if at the initial entry the minimum exceeds the maximum while in 66 all loops are always executed once The above is only half of the nightmare As we have discussed earlier when the 77 standard was announced vendors had to find a way to still support the older programs Without fail all FORTRAN compilers vended since the 77 standard have supported a control switch which tells the compiler what to do about the DO statement As will be discussed
250. ion is void See also fioerror Do requested error processing fiofmtv Get constant format value fiofwsp Skip over white space 6 125 FIOOPEN Open Current FORTRAN File Synopsis include fortran h PROMULA FORTRAN function declarations int fioopen 197 PromulaFortran Translator User s Manual Description This function opens the file associated with the current FORTRAN file as specified in the global variable iocurf Return value A zero if the open went well else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioitos Converts a short integer to a string fioerror Performs standard error processing 6 126 FIORALPH Read Alphabetic Information Synopsis include fortran h PROMULA FORTRAN function declarations void fioralph alpha nalpha nfield char alpha Character string int nalpha Length of string int nfield Width of field Description This function blank fills a character string and then reads characters from the current input record into the string Read characters are left justified in the string Any characters beyond the end of the string are discarded If input is free form and if the characters read begin with a single quote then the material within the quoted list is entered into the character string with reducing to a single quote Return value None the function is void The effect of its processin
251. ior to translation if that program is suspected of playing any games with internal representations It should be pointed out that the storage of prototype information read from a configuration file has been carefully optimized which means that there are minimal performance penalties for using large prototype files 4 1 Function Prototype Syntax In C a function prototype declaration defines the name return type and storage class of a function In addition it can define the types of some or all of the arguments for that function The prototype declaration has the same format as the function declaration except that it is terminated by a semicolon and the argument identifiers are optional If used argument identifiers have scope only within the prototype declaration and serve only as place holders i e int a b c is the same as int int int The syntax used for prototypes by PROMULA is similar to that used by C however it is unfortunately not the same The reason is that additional information must be supplied to the translator since additional problems are introduced by the weak typing conventions of FORTRAN The syntax of the PROMULA prototype definition is as follows fname type name type c type clI type name Dj Where fname isthe name of the subprogram used in the FORTRAN code type isone ofthe following C binary type specifiers void 127 PromulaFortran Translat
252. is not a multiple of three it is treated as though it were padded to the right with blanks i e 0 160 PromulaFortran Translator User s Manual Note they are called radix 50 because 40 is 50 when written in octal Return value None the function is void See also None 6 48 FIFCHAR FORTRAN Intrinsic Function CHAR Synopsis finclude fortran h PROMULA FORTRAN function declarations char fifchar iv int iv Value to be converted Description Converts a numeric display code or lexical value or collating weight into its character code The point of this function is that character values on the host processor are not necessarily the same as those on the machine for which a given FORTRAN program was written All numeric display code references in a source FORTRAN program are passed through this function either by the translator directly or by the other runtime functions included in this library Note that if you wish this function to return some value other than the host processor values you must modify it Typically this modification would take the form of a lookup table reference Return value A pointer to a 1 character string containing the character corresponding to the display code See also None 6 49 FIFCOS FORTRAN Intrinsic Function COS Synopsis finclude fortran h PROMULA FORTRAN function declarations float fifcos x float x Argument containing the value Description Computes the cosin
253. ithin in each module is given a unique name and the message alerts the user that he must provide an ftnblkd function which calls the individual module initialization control functions Note that the actual convention used to form the initialization names can be easily changed via a configuration file as discussed in the following chapter 2 20 Inclusion of Line Numbers for Debugging Ln L0 For platforms that support the UNIX dbx debugger or its equivalent FORTRAN codes can be debugged using the original source FORTRAN lines To activate this capability the Ln switch is used This switch tells PROMULA FORTRAN to include line number and source file information in its intermediate C output so that it can later be used by the debugger For best effects the following three switches should also be used Bo LO Qe32000 See the relevant sections in this chapter for detailed descriptions of these switches Bo selects the optimized bias This is needed to give PROMULA FORTRAN maximum flexibility in maintaining a clear correspondence between the source FORTRAN and the target C LO specifies that no newlines should be inserted in the listing This is needed to maintain the line numbering system Qe32000 allocates room for the emissions table which is needed to generate the line number information 59 PromulaFortran Translator User s Manual 2 21 FORTRAN Dialect Selection Flags Mdialect The PROMULA FORTRAN processor is a general purpose multi dial
254. ival n i ival 0 n 0 for i 1 i lt n i ival ival 1 86 PromulaFortran Translator User s Manual return This translation conforms with the 77 standard because C checks the conditional part of the for statement prior to the first execution The following shows the same thing using the 1 setting void demo static int ival n i ival 0 n 0 for i 1 i 1 i lt n i ival ival 1 return To achieve the desired result an extra condition has been placed into the for statement to force it through the code at least once Finally the following shows the same thing translated via the T2 switch void demo static int ival n i D2 ival 0 n 0 for i 1 D2 n itl D2 gt 0 D2 i 1 ival ival 1 return This switch forces the computation of the trip count variable as specified by the FORTRAN 77 standard 2 33 Treatment of Internally Generated Temporaries Ta Ts During the processing of FORTRAN codes PROMULA must occasionally introduce temporary variables The contexts which generate these variables are as follows 1 When a computed value must be passed by reference to a subprogram 2 When a DO statement must compute a trip count value to control the number of times it executes 3 When a variable DIMENSION contains a computed value All of these contexts are discussed elsewhere in this chapter This section discusses the sto
255. l differ In addition the logic of the construct assumes that the value of B set via the call to Is remembered in the calls to IDOUBLE and to MULTIPLY A Q D ISET_BVALUE 10 E IDOUBLE H H 5 MULTIPLY A 2 4 Zw tJ zi EGER A R B OUBLE A B ETURN J DHHHES z HPE 2 ll Jj DK 1 G Il We A B a lt H AWE ZEE EGER FUNCTION IDOUBL ira D IPLY_A A R BVALUE B E will be ET BVALU We recently processed a large body of VAX FORTRAN code which made extensive use of this ability to successively set argument values though independent entry points The default translation produces the following C output void main argc argv int argc char argv extern void iset bvalue extern long idouble multiply a static long K1 10 static long K2 5 static long K3 2 static long K4 4 iset bvalue amp K1 ftnini argc argv NULL WRITE OUTPUT LISTIO INT4 idouble amp K2 INT4 multiply_a amp K3 amp K4 0 static long E0000 int IENTRY long a r b static long idouble switch IENTRY ENTRY a r b 0 goto IDOUBLE case 1 goto MULTIPLY A case 2 goto IS IDOUBLE idouble a return idouble MULTIPLY A a r idouble a
256. l shown as i3 2 The same fragment with the CF1 switch is as follows 21 PromulaFortran Translator User s Manual void test define i3 3 define i5 i3 2 extern char Xalpha typedef struct int ival jval kval Calpha auto Calpha Talpha Calpha Xalpha Talpha gt ival Talpha gt jval Talpha gt kval i5 switch Talpha gt ival case 1 goto S10 case 2 goto S20 case 3 goto S30 default break S10 WRITE OUTPUT LISTIO STRG IVAL 1 8 0 S20 return S30 STOP NULL undef 13 undef 15 Now the expressions Talpha gt jval Talpha gt kval 15 and 13 2 have been widened but the remainder of the C output is as before The same fragment with the CF2 switch is as follows void test define i3 3 define i5 13 2 extern char Xalpha typedef struct int ival jval kval Calpha auto Calpha Talpha Calpha Xalpha Talpha gt ival Talpha jval Talpha kval i5 switch Talpha ival case 1 goto S10 case 2 goto S20 case 3 goto S30 default break S10 WRITE OUTPUT LISTIO STRG IVAL 1 8 0 S20 return 530 STOP NULL undef i3 undef i5 In this version blanks are not inserted in expressions but the switch statement is now vertical Note that the appearance of the braces and indentation is controlled via the NU and NL switches which are described in another section of this chapter The CF4 switch produces the following
257. lation process and specifies how individual parameters associated with particular subroutines and functions are to be processed The use of configuration files is discussed in Chapter 3 Finally you may replace the entire dialect description file pfc pak that is controlling the PROMULA FORTRAN translations Special dialect definition files may be obtained from Great Migrations LLC upon request 2 1 Command Line Syntax The syntax for using PROMULA FORTRAN from the operating system level is as follows PFC filename optl opt2 opt3 Where filename is the name of the FORTRAN file to be converted optionally preceded by a drive and path specification If no extension is supplied an extension of for is assumed 10 opti PromulaFortran Translator User s Manual is a single character or group of characters in upper or lower case indicating a particular switch followed by any particular information to be associated with that switch The switch letter s plus its associated information may contain no embedded blanks The various options are separated by blanks The following is an alphabetic listing of the command line switches the number of the section in which each is discussed and a brief description The remaining sections of this chapter will discuss these switches in detail Switch B C OAT OOO a tee ae 2 2 an 2 C P Options c f o 1 5 0 1 2 num d r s v 0 1 2 0 1 2 3 0 1 2 3
258. ler in that in order to translate the source code it must also validate it i e check it for any syntax errors Syntax errors may arise for a variety of reasons 1 An actual syntax error 2 Use of a dialectal feature not supported by the currently active language definition file 3 Animproperly used command line switch 4 A problem with a subprogram parameter The basic assumption in the design of PROMULA FORTRAN is that existing valid FORTRAN programs are being processed therefore a syntax error means that a mismatch of some sort exists between PROMULA FORTRAN and the dialect being processed By default therefore PROMULA FORTRAN stops processing when a syntax error occurs The switch ERO specifies this behavior The switches 1 through 4 specify that processing should continue despite the syntax error With PROMULA FORTRAN if processing is to continue despite an error then PROMULA FORTRAN must be told how to produce the C output given a statement in the source program which it cannot translate The through switches differ in how the error statement appears in the C output Consider the following simple program which contains an illegal character in its second statement PROGRAM TEST I JQ Kk STOP END The 1 switch specifies that syntax errors block at C compilation time In other words when a syntax error is encountered a message is issued but processing continues The statement producing the
259. lex negative of a short complex value Return value The complex result See also None 6 17 CPXPOL Short Complex Conversion to Polar Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex complex cpxpol a complex a The value to be converted Description Converts a complex number into its polar form By definition r i in Cartesian form is r exp 1 0 Return value The short complex result of the conversion See also None 146 PromulaFortran Translator User s Manual 6 18 CPXPOW Raise Short Complex to a Power Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex complex cpxpow a b complex a The value to be raised to a power complex b The value of the power Description Computes the value of a short complex raised to a short complex power Return value The complex result of the calculation See also cpxexp Computes a short complex exponential complex cpxlog Computes a short complex logarithm complex cpxmul Multiplies two short complex values 6 19 CPXREAL Compute Real Part of Short Complex Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary
260. llows void demo 3 int i j 27 PromulaFortran Translator User s Manual extern int jindex static char name 12 initial 4 i fifindex name 12 initial 4 j jindex name initial return In this version notice first that the FORTRAN intrinsic function index is converted to fifindex which is the PROMULA FORTRAN runtime library function equivalent In addition the lengths of the two character strings are passed as well since ifindex needs this information The treatment of FORTRAN system functions is not affected by CH flags Under CHr the call to the jindex subprogram is quite different Here the subprogram is simply passed pointers to the start of the two character arrays No length information is given If jindex does not need this information then the above is the cleanest treatment however in most cases it is insufficient The treatment under the CHs flag looks as follows void demo 3 rnt Say extern int jindex static char name 12 initial 4 1 fifindex name 12 initial 4 j jindex name 12 initial 4 return Under this treatment the jindex subprogram is treated in exactly the same manner as the index function The lengths of the character strings follows directly after the pointers to their starting points This treatment works in most cases but not in the case where character and non character arguments are both passed to the subprogram The CHd result is shown be
261. long i auto float a b c d a float ftnalloc 40L b float ftnalloc 400L c float ftnalloc 4000L 81 ELP PromulaFortran Translator User s Manual d float ftnalloc 40000L i long ftnalloc 4L it 1 for i 0L i ic i 101 br icr i at i goto ELP ftnfree char ftnfree char ftnfree char ftnfree char ftnfree char 101 10L L 10L pb i 20 1 d c b a dt it it i 30 10L 10L cC it i 10L This probably went too far We really did not want the simple variable i to be dynamic Try again with SS to turn off static SAO to use auto for small variables and SD10 to make everything else dynamic The following is the result void demo auto long i auto float a b c d ELP a float ftna b float ftna c float ftna d float ftna for i 0L i 10 c b ati i i 10 goto ELP char char char char ftnfree ftnfree ftnfree ftnfree m lloc lloc lloc lloc i L L bid d c b a it iti 10L 10L 401 400001 4001 40001 L EFL t20 dis i i i 10L i 30 10L 10L As a final example do the same thing as above except replacing dynamic storage with virtual storage To do this enter SS
262. low void demo 3 rnt sae eee extern int jindex static char name 12 initial 4 1 fifindex name 12 initial 4 xj initial 12 4 return This treatment differs from the one under CHr in that character lengths are passed to the subprogram and it differs from CHs in that the lengths are not directly associated with their pointers rather they are collected and placed at the end of the call Note again that the call to the system function index is unaffected Finally the treatment under the CHv flag is as follows void demo 3 IE Ju extern int jindex static char name 12 initial 4 static string T1 NULL 0 static string T2 NULL 0 1 fifindex name 12 initial 4 Tl a name 1 12 T2 a initial T2 n 4 j jindex amp T1 T2 return 28 PromulaFortran Translator User s Manual In this version the temporary variables T1 and T2 are introduced to contain the addresses and lengths Note that the string variable type can easily be replaced by a particular descriptor type as required by particular platforms This topic is discussed extensively in the chapter on the configuration file under the topic of keyword replacement More complicated treatments of the character arguments than the CHs treatment are needed because of examples such as the following
263. ly Prime FORTRAN 77 the normal internal character representation is a seven bit code with the 8th bit always set on while the extended characters have the 8th bit off The normal external character representations use the opposite conventions This function converts string constants to the extended representation to make them compatible with other internal characters Return value A pointer to the result string See also None 6 102 FIOBACK Backspace a FORTRAN File Synopsis include fortran h PROMULA FORTRAN function declarations int fioback Description This function backspaces the file associated with the current FORTRAN file as specified in the global variable fiocurf For files with fixed record lengths this operation involves moving backwards in the file by the length of one record For text files this means moving backwards in the file to the position immediately following the second carriage return line feed character pair which immediately precedes the current record For other types of files no backspace is possible Note that if the file is currently positioned at its beginning no operation is performed Return value A zero if the backspace was successful else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Performs standard error processing 6 103 FIOBFOUT Business Format Output Synopsis include fortran h P
264. maginary part of the number complex complex cpxsub a b complex a The left hand value complex b The right hand value Description Subtracts two short complex numbers to form a third Return value The short complex result of the subtraction See also None 6 23 DPXABS Compute the Double Complex Absolute Value Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of the number dcomplex double dpxabs a dcomplex a Contains the value to be transformed Description Computes the value of the double complex absolute value The result is double precision Return value The double precision result See also None 6 24 DPXADD Double Complex Addition Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value 149 PromulaFortran Translator User s Manual double ci The imaginary part of the number dcomplex dcomplex a b dcomplex a The left hand value dcomplex b The right hand value Description Adds two double complex numbers to form a third Return value The double complex result of the addition See also None 6 25 DPXCJG Compute the Double Complex Conjugate Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of
265. me initialization is needed In addition since PROMULA assumes no more that 4 characters per word the phrase world is thrown away Using the QW3606 switch improves the situation slightly however the result will not run on a 32 bit platform void main argc argv int argc char argv static long ialpha 2 static int ftnsiz 1 1 2 static namelist DATAVAR ialpha ialpha 5 ftnsiz static char DATAVAL SDATAVAR ialpha Hello World SEND static char F1 1x 2a6 ftnini argc argv NULL fiointu char DATAVAL 0 2 fiornl DATAVAR 1 NULL WRITE OUTPUT FMT F1 1 DO 2 INT4 ialpha 0 STOP NULL because the long variable ialpha is now expected to contain 6 characters We need to tell the translator that this variable is really CHARACTER 6 The following configuration file does this SWITCHES QW3606 RESTRUCTURE IALPHA CHARACTER 6 END Using this configuration file the following correct translation is produced void main argc argv int argc char argv static char ialpha 12 115 PromulaFortran Translator User s Manual B ter static char F1 1x 2a6 ftnini argc argv NULL WRITE OUTPUT FMT F1 1 D0 2 STRG ialpha 6 0 STOP NULL 3 5 The Configuration KEYWORDS Statement Most of the actual text output of PROMULA is formed by combining sequences
266. me the difference between the approaches to character strings between FORTRAN and C The following FORTRAN code simply initializes a character array NAME with the names Charles Frederick Andrew Mary and Mary again Note that the notation used is straightforward and that a simplified notation can be used to give the last two names the same value SUBROUTINE DEMO CHARACTER 10 NAME 4 8 Charles Frederick Andrew 2 Mary WRITE NAME END 26 PromulaFortran Translator User s Manual The default C representation for this example is shown below void demo static char 50 eg x Rx reti pat OIR YAY mn ud Lu at ty xy yt NR t Et P Pa bet t ES TR i ms us P RS Y H WRITE OUTPUT LISTIO DO 5 STRG name 10 0 Note first that name is simply declared as an array of char containing 50 entries structure has been reduced to a simple vector The compiler itself however remembers the structure and uses it when necessary Note second that the initialization of the name array shows up one of the quirks of C There is no way to specify a sequence of characters without using cumbersome notation because the notation Charles would generate a null terminated string and the notation Charles is not accepted by many compilers eve
267. multiline comment The alternative using CM2 produces the following void test i j k long i j k Compute the value of I using the following relation les J K I 5 j jt k 5L In this output each FORTRAN comment line becomes an equivalent C comment line Note that the NCn switch described in another section of this chapter describes the conventions used for inline comments 2 7 Treatment of DATA Initializations Da Dc Dr The FORTRAN language has an excellent notation for describing the initial values to be assigned to FORTRAN variables The following summarizes the FORTRAN syntax DATA nlist clist nlist clist Where nlist is a list of names to be initially defined Each name in the list can take one of the forms var is a variable name array is an array name element is an array element name i e subscripted array name substring is a substring of a character variable or array element dolist is an implied DO list of the form dlist i init term incr 34 PromulaFortran Translator User s Manual Where dlist is a list of array element names and implied DO lists Subscript expressions must consist of integer constants and active control variables from DO list i is an integer variable called the implied DO variable init is an integer constant symbolic constant or expression specifying the initial value as for DO loops term is an integer constant symbolic cons
268. n Computes the value of the double complex sine of a double complex number By definition e i a g i sin a 2 i Return value The double complex result See also None 6 43 DPXSROOT Compute Double Complex Square Root Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of the number dcomplex dcomplex dpxsroot a dcomplex a Contains the value to be transformed 158 PromulaFortran Translator User s Manual Description Computes the square root of a double complex value Return value The double complex result See also None 6 44 DPXSUB Double Complex Subtraction Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of the number dcomplex dcomplex dpxsub a b dcomplex a The left hand value dcomplex b The right hand value Description Subtracts two double complex numbers to form a third Return value The double complex result of the subtraction See also None 6 45 FIFAMAX0 FORTRAN Intrinsic Function AMAX0 Synopsis finclude fortran h PROMULA FORTRAN function declarations float fifamax0O al a2 long al First value to be compared long a2 Second value to be compared Description Determines which of two long integer values is the largest Return value The largest long
269. n be controlled 1 theuse of whitespace surrounding operators 2 the amount of whitespace to be used in the switch statement 4 the case of identifiers 8 the amount of whitespace in structure and common definitions 16 the display of constant parameters as constants The values of the CF switches may also be combined to obtain composite effects thus any CF switch value between 0 and 31 is valid The default setting is CFO Consider for example the following FORTRAN code fragment SUBROUTINE TEST COMMON ALPHA IVAL JVAL KVAL PARAMETER 13 3 15 13 2 IVAL JVAL KVAL 15 GOTO 10 20 30 IVAL 10 WRITE IVAL 1 20 RETURN 30 STOP END The default translation for this fragment is as follows void test define i3 3 define i5 13 2 extern char Xalpha typedef struct int ival jval kval Calpha auto Calpha Talpha Calpha Xalpha Talpha gt ival Talpha gt jval Talpha gt kval i5 switch Talpha gt ival case 1 goto S10 case 2 goto S20 case 3 goto S30 default break S10 WRITE OUTPUT LISTIO STRG IVAL 1 8 0 S20 return S30 STOP NULL undef 13 undef 15 There is no whitespace within the expression Talpha jval Talpha kval i5 the typedef and the switch statement are horizontally written to reduce whitespace the identifiers in C have all been reduced to lower case and the definition of i5 is stil
270. n places Tn 3 Move to position n TLn 4 Skip left n places SS 5 Set the plus sign to a space SP 6 Set the plus sign to a BN 7 Set blanks to null no operation for output BZ 8 Set blanks to zero no operation for output 9 Physically write the current record nP 10 Set the floating scale factor to n 11 End of Format write current record Return value None the function is void however the global variable ioier may be set to an error code if a problem is encountered See also fionxtf Get next format specification fiowtxt Write text record 6 116 FIOFVINQ Inquire About File Value Synopsis include fortran h PROMULA FORTRAN function declarations long fiofvinq option int option Specifies data being inquired about Description This function is used to inquire about the various file data options associated with the current FORTRAN file structure which return an integer or logical value The particular value being inquired about is defined by the option parameter as follows Option Description of data 2 Inquire about existence of file 3 Inquire as to connection status of file 4 Inquire as to file s external unit number 5 Inquire whether file has a name 13 Inquire about file record length 14 Inquire about file current record number Note that this function is needed as distinct from the generic file information function because type conversions are required on the returned value Return value The requ
271. n the virtual database Return value A pointer to the start of the storage area See also ftnalloc Allocate dynamic memory vmsvect Virtual vector input or output 6 221 VMSOPN Open A Virtual Memory File Synopsis include fortran h PROMULA FORTRAN function declarations int vmsopn argc argv 246 PromulaFortran Translator User s Manual int argc Number of command line strings char argv An array of pointers to the strings Description Opens a virtual file for processing This may either be an existing file or a new file This function is called to initialize the virtual file system by any program requiring the use of the virtual file system This function is passed to the command line strings which were passed to the main function by the operating system This function scans those command line strings looking for two particular arguments as follows Argument Description Vname Specifies the name of a file which is to be used as the virtual disk file This may be an existing PROMULA array datafile or it may be a file to be created during the execution of this program Svalue Specifies that the name specified by the V parameter is to be created during the execution of this program The parameter value specifies the size of that file in bytes Zvalue Specifies the maximum number of sheets to be allocated for virtual memory This parameter is needed when both dynamic and virtual allocations are being made It prev
272. n though it seems completely clear and unambiguous Note third in the initializations another point about which more will be said in other sections The definition of Mary must be repeated twice since C has no equivalent of n value notation used in FORTRAN Note also that FORTRAN character strings are always padded with blanks and not nulls as is the convention in C It is this fact that makes C and FORTRAN character manipulation completely incompatible For C then FORTRAN CHARACTER variable is treated simply as an array of char and not as a C string The discussion in this chapter of the Dr Data runtime flag presents an alternative approach to the above in which a small amount of runtime efficiency can be sacrificed to improve the readability of the above 2 5 2 Subprogram Arguments The subsection above highlighted that aspect of character manipulation which is common to all four approaches to character manipulation This subsection describes the major way in which the three approaches differ Consider the following simple FORTRAN code fragment which calls the FORTRAN system function index and a user written function jindex Both these functions take two character string arguments SUBROUTINE DEMO I J CHARACTER 12 NAME CHARACTER 4 INITIAL INDEX NAME INITIAL J JINDEX NAME INITIAL RETURN END The simplest treatment of this code is obtained via the CHr flag It is as fo
273. nal compilation It is very common for FORTRAN preprocessors to have such a capability as well Consider the following simple FORTRAN code which uses a conditional compilation notation SUBROUTINE TEST6 CALL RENAME modxx lio model lio IF UNIX CALL RENAME modxx lio model lio ENDIF CALL RENAME modxx lio model lio STOP END This particular notation uses three particular comment strings 4IF 1 ENDIF introduces a conditional expression introduces a statement within a conditional block closes a conditional block Though the notation varies from site to site the above is typical Since these conventions are based on C we need to have some mechanism for describing them to PROMULA To see the problem the following is the default translation of the above void test 6 extern void rename f The translator rename f modxx lio model 1io 9 9 IF UNIX CALL RENAME modxx lio model lio ENDIF rename f modxx lio model 1io 9 9 STOP NULL completely ignores the true meaning of the condition block The whole thing is treated as a comment The fact that the should be ignored when going to C is easily established using the YD command line switch This switch is described in the previous chapter It gives the following translation void test6 extern void rename f
274. ng As was said above the big problem with this translation technique is that it must use the gt operator This operator does pointer arithmetic which on the PC can be relatively inefficient with some memory models Finally the storage areas for the blocks themselves are reserved as global storage areas with the maximum size defined for each block The translation below is produced via the Gd setting void demo extern char Xalpha Xbeta l typedef struct float a 10 10 b 5 c 4d Calpha define Talpha Calpha Xalpha typedef struct char cLILI50 e21501 Cbeta define Tbeta Cbeta Xbeta Talpha gt a 0 0 Talpha gt c Talpha gt b 4 Talpha gt d ftnscopy Tbeta c1 42 10 Tbeta gt c2 30 10 NULL return undef Talpha fundef char Xalpha 428 char Xbeta 200 This translation is very similar to the one above except that the definition of the symbol Tbeta is accomplished via a define as opposed to a variable The translation below which uses the Gs setting is our favorite if you and your compiler can accept it void demo extern struct float a 10 10 b 5 c d Xalpha extern struct char c1 150 c2 50 Xbeta Xalpha a 0 0 Xalpha c Xalpha b 4 Xalpha d ftnscopy Xbeta c1 2 10 Xbeta c2 30 10 NULL return char Xalpha 428 char Xbeta 200 51 PromulaFortran Translator User s Manual This translation comes
275. ng constant is limited C has a very neat notation for defining a constant set of pointers to constant strings Thus a long format string can be written very conveniently as a sequence of individual lines as follows static char fmtO1 T2 20A4 T2 MASS M UNITS 2A4 T2 LENGTH L UNITS 2A4 T2 TIME T UNITS 2A4 H Using this notation indefinitely long FORMAT specification strings can be written as a sequence of blocked lines Note that the FORMAT routines ignore the breaks between the lines so they may be broken in any way Simple FORMAT specification strings may of course still be written as standard string constants such as the following static char fmt02 F10 0 The FORMAT specification string itself has the same content as a FORMAT would have in a standard FORTRAN environment It must begin with a left parenthesis and end with a right parenthesis See the general discussion of FORTRAN I O capabilities for a detailed description of the format specifications There are two parameters to be supplied for this function as follows Name Description of use fmt Points to the format control string As discussed above it may either be a pointer to a series of strings or it may be a simple string pointer It may also be a NULL meaning either that free format is to be used or that a binary operation is to be performed nfmt Provides additional information about the above
276. nopsis include fortran h PROMULA FORTRAN function declarations int fiowln Description Completes the current read operation by flushing the current format statement and writing out the current record Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiofout Process output format specification fiowtxt Write text record fiofend End current format fioerror Do requested error processing 6 163 FIOWNL Process FORTRAN WRITE DATALIST Statement 216 PromulaFortran Translator User s Manual Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct char nmname char nmvalu int nmtype int nmadr namelist int fiownl name nname nlident namelist name NAMELIST group int nname Number of variables in group char nlident Identifier of NAMELIST group Description Writes the information in a namelist data group Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowdbl Write double precision values fioerror Perform FORTRAN I O error processing fiowtxt Write current text record fiodtos Convert floating point number fioitos Convert integer to string 6 164 FIOWRB Write FORTRAN Boolean Ve
277. ns void fiortxt Description Reads the next physical text record into the coded communications record for detailed processing Both internal and external files are processed Return value None the function is void The effect of its processing is reflected in the various global format control variables See also fioerror Perform error processing 209 PromulaFortran Translator User s Manual 6 147 FIORWBV FORTRAN Rewrite Binary Values Synopsis include fortran h PROMULA FORTRAN function declarations int fiorwbv value nvalue void value Points to values being written int nvalue Number of bytes to be written Description This function rewrites binary values to a file Return value A zero if all went well else an error code See the general discussion of the FORTRAN I O capabilities for a listing of the possible error codes See also None 6 148 FIOSHL Shift String Left Synopsis include fortran h PROMULA FORTRAN function declarations void fioshl s n char s String to be shifted int n Number of places to be shifted Description Shifts a character string left a specified number of places The spaces removed are lost Return value None the function is void See also None 6 149 FIOSHR Shift String Right Synopsis include fortran h PROMULA FORTRAN function declarations void fioshr s n fill char s String to be shifted int n Number of places to shift char fill The
278. ntation might be to copy all of these files into some subdirectory where they would retain the same relative structure as they had on the source system The PREFIX pname option allows a directory specification to be added to the front of all source pathnames Another alternative might be to copy all include files into a single subdirectory with no additional structure This effect can be achieved by using PREFIX in conjunction with the EXCLUDE option All source structure would be excluded and then would be replaced by the desired target subdirectory name In some cases no generic translation scheme will work Certain names might have to be changed on an individual basis The final list of sname t name pairs achieves this end Each pathname is first translated using the generic specifications on the PATHNAME statement itself The resultant pathnames are compared with the snames in the list If the first n characters of a pathname match the n characters of an sname then those n characters are stripped and the associated tname is added to the front of the name As can be seen from the above it will be necessary to organize the INCLUDE files in the new environment Once that organization has been completed the PATHNAMES component of the configuration file can be used to describe that structure No changes need be made in the FORTRAN source code INCLUDE and INSERT statements 3 4 The Configuration RESTRUCTURE Statement The RESTRUCTURE statement in th
279. nts are concerned the translation of the DML is straightforward since the ordering of the DML in the target language is typically the same as that in the source language For the DDL however the ordering of symbol definitions is often quite different in the target language therefore the comments must also often be rearranged The algorithms for associating comments with symbols are complex and by no means perfect The COMMENTS statement itself does not deal with the ordering of comments but rather deals with the source and target form of the individual comment lines themselves Syntax COMMENTS DDL specification DML specification END Where DDL specification says how comment lines within the DDL are to be interpreted and translated DML specification says how comment lines within the DML are to be interpreted and translated Each specification consists of three sections Within each section the fields are as follows Section Content Description 1 0 Do not precede comment blocks with a standard header 1 str Precede comment blocks with the indicated header 2 61 A series of individual comment introducers and their 827 replacements in C 3 0 Do not follow comment blocks with a standard trailer 1 str Follow comment blocks with the indicated trailer As an example consider the following FORTRAN code FUNCTION MODI ARG SWAB FLAG ARG XLIT FLAG PURPOSE General purpose initialization for MODI
280. ode are replaced by z The default translation of the following code UNCTION HELLOS 15 75 IMENSION WORLDS 10 20 ELLOS WORD I J ETURN ND Ei 50 m UO is as follows float helloZ iZ jZ long iZ 3jZ extern float word static float helloZ helloZ word iZ j2 return helloZ 125 PromulaFortran Translator User s Manual Using a configuration statement of S would give the following float helloS iS jS long iS 7S extern float word static float helloS helloS 15 35 return helloS Alternatively if 515 to be left alone use float 110 15 35 long i j extern float word static float hello hello word i j return hello 126 PromulaFortran Translator User s Manual 4 THE CONFIGURATION FUNCTION PROTOTYPES As has been discussed in several other places FORTRAN passes all function arguments by name i e it passes the address of a parameter to a subprogram as opposed to its value C on the other hand allows argument values to be passed directly as well as allowing the passing of argument addresses In instances where the value of the argument is scalar and where its value is not being changed by the subprogram passing the value of that argument is far more efficient than passing its address Whenever possible the translation from FORTRAN to C should use call by value The problem is that it is not always pos
281. ode have the FORTRAN bias and those who want C like code have the C bias When using PROMULA FORTRAN this output bias can be selected via a single command line switch The default bias is the FORTRAN bias which can be selected via the Bf switch The C and optimized biases are however fully supported by PROMULA FORTRAN and may be activated by the Bc and Bo command line switches When one of the bias switches is used it should always go first on the command line since it controls the selection of various other switch values Mixed bias can be produced by following the bias switch itself with other switches 12 PromulaFortran Translator User s Manual Regardless of the user bias the output of PROMULA FORTRAN does not look like machine translation To present some of the design features of the different biases consider the following example program which computes the mean and variance of a set of values The figure below shows the original FORTRAN source code along with the default translation produced by PROMULA FORTRAN for the FORTRAN and C biases FORTRAN SOURCI 10 1S FORTRAN SOURCI 10 I9 07 SUBROUTINE EXO001 XBAR VAR DIMENSION VAL N WRITE 1x a ANALYSIS XBAR 0 VAR 0 0 10 J 1 N XBAR XBAR VAL J CONTINUE XBAR XBAR N WRITE 6H Mean F10 3 XBAR DO 15 J 1 N S VAL J XBAR VAR VAR 5 S CONTINUE IF VAR NE 0 0 VAR VAR N 1 WRI
282. of conventions as to what actual units to associate with these statement forms The U command line switch allows you to control this number Note that the PROMULA FORTRAN runtime library associates no significance to any particular unit number value A unit number may be any integer value The URnum command line switch tells PROMULA to use unit number num with READ or ACCEPT statements for which no unit number is explicitly shown The default setting of UR specifies that READ statements for which no unit is explicitly shown should be directed to standard input The UWnum command line switch tells PROMULA to use unit number num with WRITE or PRINT statements for which no unit number is explicitly shown The default setting UW specifies the WRITE or PRINT statements for which no unit number is explicitly shown should be directed to standard output The UPnum command line switch tells PROMULA to use unit number num with PUNCH statements for which no unit number is explicitly shown The default setting UP specifies that PUNCH statements for which no unit number is explicitly shown should be directed to standard output Note that many FORTRAN runtime systems attach special significance to certain unit number values i e many assume the unit 5 is standard input while unit 6 is standard output Others assume that unit 0 is standard input Especially interesting is the Prime dialect which assumes that unit 1 is both standard input or standard output depen
283. ommand line switches available and the use of a configuration file Virtually every aspect of the look of the output can easily be controlled by you If you are still not satisfied please contact our user support staff We feel very strongly that PFC can support any reasonable objective or bias and would like the opportunity to prove it PromulaFortran Translator User s Manual 1 INTRODUCTION The PROMULA FORTRAN to C Translator generates C code which can be compiled by a standard C compiler to produce executable code It is operational on a wide variety of platforms More than just a translator PROMULA FORTRAN adds value to the code during the translation virtual memory logic dynamic memory logic references to external databases and application management systems and integration with other operating environments Finally PROMULA FORTRAN has a dialect management component which allows the customization of the package in terms of both the FORTRAN dialect to be accepted by the translator and the form of the C translation output The source code processing component of PROMULA FORTRAN is completely compatible with the one used by the PROMULA FORTRAN Compiler The manual for the PROMULA FORTRAN Compiler is included with this manual That manual contains a description of the FORTRAN language supported controlling runtime behavior the PROMULA interface and error messages That discussion applies directly to PROMULA FORTRAN as well and will not be
284. on be performed This function performs that conversion using whatever facilities are available with a particular platform This function converts the floating point number val to a character string It stores precisely ndig digits in dspdig followed by a null byte If the number of digits in val exceeds ndig the last digit is rounded if the number of digits is less than ndig then dspdig is padded with zeros The parameter pdecpt points to an integer value giving the position of the decimal point with respect to the beginning of the string and psign returns zero if val is positive else one Return value None the function is void See also None 6 107 FIOERROR Perform FORTRAN I O Error Processing Synopsis include fortran h PROMULA FORTRAN function declarations int fioerror clear int clear Should error control be cleared 185 Description PromulaFortran Translator User s Manual If the FORTRAN I O runtime system encounters an error it sets an error code and calls this function This function either sets an error return value or exits to the operating system with an error message In the case where an error code is returned to the calling function the parameter clear specifies whether or not the error processing control variables should be cleared prior to the return Return value A zero if ther of the possibl See also fiocurf fioitos fioerch fioier fiostat 6 108 F Synopsis include void fio
285. on the command line There may be as many SWITCHES statements as desired in a given configuration file however the set of SWITCHES statements must always be the first set of statements in the configuration file If multiple configuration files are being processed a SWITCHES statement may only appear in the first such file Note that the Rname switch may appear within a SWITCHES statement The following is an example of a SWITCHES statement 107 PromulaFortran Translator User s Manual SWITCHES P 4096 P 8192 CF11 Il Mvax CHv Ga Yd QH1000 Note that the SWITCHES statement often interacts with other statements following it in the configuration file Many examples of this interaction appear in the following discussion of the other statements in the configuration file 3 2 The Configuration COMMENTS Statement Many FORTRAN programs have relatively complicated conventions for marking comments of various types These conventions often do not work well when transformed directly to C As has been mentioned elsewhere in this manual the hardest single aspect of FORTRAN codes to deal with in translation is the commenting conventions In general the statements of a FORTRAN program can be divided into two groups 1 those that define the symbols to be used referred to as data definition language or DDL 2 those that specify the actual operations to be performed referred to as data management language or DML Insofar as comme
286. on the type of the action to be performed This action code is as follows 195 PromulaFortran Translator User s Manual Code Action to be performed 0 A file is to be opened with this logical unit number if there is an open structure currently associated with this number close the file 1 A structure for this logical unit number has been created If it cannot be found then an error has occurred A coded read is to be performed If there is no structure defined then create one and open the file A binary read is to be performed If there is no structure defined then create one and open the file A coded write is to be performed If there is no structure defined then create one and create the file A binary write is to be performed If there is no structure defined then create one and create the file A miscellaneous operation is to be performed If there is no structure defined then create one and open the file 7 A simple inquiry is being made ON tA amp Return value A zero if all went well else an error code See the general discussion of the FORTRAN I O capabilities for a listing of the possible error codes See also fioclose Closes the current FORTRAN file fioerror Performs standard error processing fioopen Opens the current FORTRAN file fiortxt Read next text record 6 123 FIONAME Establish FORTRAN Unit by Name Synopsis include fortran h PROMULA FORTRAN function declarations
287. on to this occurs when processing numeric constants A constant with a decimal point or exponent specified may be used as either a float or double value parameter and a numeric constant without a decimal indication with a value in the range 32767 to 32767 may be used in either a long or short environment In a long context an L is appended The following simple example shows this void alpha float double short long UBROUTINE DEMO S CALL ALPHA 1 234 10E5 32 32 RETURN END SUBROUTINE ALPHA FLT DBL ISHT LNG REAL 8 DBL INTEGER 2 ISHT WRITE FLT DBL ISHT LNG RETURN z The prototype above used in conjunction with the FORTRAN code above produces the following output void demo extern void alpha 128 PromulaFortran Translator User s Manual alpha 1 24 1000000 0 32 32L return void alpha flt dbl isht lng int isht double dbl long 1ng float flt WRITE OUTPUT LISTIO REAL4 flt 8 1 INT2 isht INT4 1ng 0 return In this output the 32 being passed via the long parameter is shown as a long constant The following is the same translation as the above but without the prototype supplied void demo extern void alpha static float K1 1 234 static long K3 32 static double T2 T2 1000000 0 alpha amp K1 amp T2 amp K3 amp K3 return voi
288. onvert Two Doubles to Double Complex Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of the number dcomplex dcomplex dpxdpx d1 d2 double d1 Contains the real value double 42 Contains the imaginary value Description Forms a double complex number whose real and imaginary parts are specified values Return value The double complex result See also None 6 32 DPXEXP Double Complex Exponential Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of the number dcomplex dcomplex dpxexp a dcomplex a The value to be transformed Description Computes the double complex exponential of a double complex value Return value 153 PromulaFortran Translator User s Manual The double complex exponential See also None 6 33 DPXIMA Compute Imaginary of Double Complex Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of the number dcomplex double dpxima a dcomplex a Contains the value to be transformed Description Computes the imaginary part of a short complex value The result is double precision Return value The double precision result See also None 6 34 DPXLOG Double Complex
289. or User s Manual short double unsigned short char long float unsigned long dcomplex complex string name is any valid identifier to be used in the C output c is one of the following special characters jndicates a memory pointer to the indicated type indicates a virtual pointer to the indicated type indicates that a value conversion should be made to the indicated type Only a single prototype definition may occur on each record and the notation below in PROMULA type name means that the function has no arguments It does not mean that the function has some unspecified number of unspecified arguments The type specifiers have their traditional C interpretation The complex and dcomplex types refer to single precision complex and double precision complex respectively The string type refers to the FORTRAN style character string which was discussed extensively in the section on the CHARACTER type Remember that a string argument actually consists of a pointer and a length specification Depending upon the character translation scheme being used these two may not even appear together in the actual C function declaration 4 2 Value Parameters If no c specifier follows the type specification a value parameter is being defined When a reference to this argument is processed the value of the argument is passed and not its address If the actual argument does not have the proper type then an error occurs The only excepti
290. or User s Manual void demo static int i j k 1 static float a 10 b 42 c 60 d 120 aref a 3 0 aref b 2 6 4 1 aref c 3 3 4 4 5 2 aref d 1 2 1 3 1 4 1 3 aref a i 4 aref b 2 6 3 5 aref c i 3 2 4 k 6 aref d i 2 2 3 k 4 1 7 Notice that the subscript values are now exactly as entered in the original FORTRAN code Again as with Ysv it is up to the user to provide an appropriate implementation of the _aref function or macro 2 48 Unformatted Write Optimization Xu Yu It is a rather strange fact about FORTRAN that the only file type that requires special internal formatting above that used by the normal I O system is the unformatted file The problem is that unformatted files receive variable length records It is up to the runtime library to ensure that records are not overflowed and to ensure that each new read begins at the start of a record To make this possible the runtime library when it writes unformatted records must put a length value at the front and back of each record This convention creates a file structure which is efficient to read backspace and in general use To write the file however requires repositioning the file twice for each record written To avoid this repositioning overhead a flag is available that tells PROMULA to compute the lengths of records before they are written The following is a simple FORTRAN code which performs a series of unformatted wri
291. ords and explicitly ignores all blanks in its source statements many perfectly readable FORTRAN programs do not meet the above requirements PROMULA FORTRAN translates FORTRAN 66 as well as FORTRAN 77 programs The user can control those aspects of the two languages which conflict by selecting options in the translation process The translator accepts split tokens token sequences without separators blank lines comment lines within continuation sequences and other potential translation ambiguities You do not have to clean up the code to use PROMULA PromulaFortran Translator User s Manual FORTRAN So if you have an old 66 program punched on a 026 keypunch with a drum card on as few cards as possible and you can get it onto a disk PROMULA FORTRAN can process it for you Also PROMULA FORTRAN knows the difference between DO10I 1 5 and DO10I 1 5 Or DO 10 I 1 5 and DO 10 I 1 5 and translates them all correctly PROMULA FORTRAN deals with all of the violations of standard character manipulation including hiding character values in arithmetic variables mixing character and noncharacter variables in COMMON blocks equivalencing character variables with arithmetic variables and assuming that CHARACTER N and CHARACTER 1 N have the same memory representations Finally in cases where different versions of FORTRAN have conflicting features or conventions a dialect selection option switch can be used to select the desired set The pa
292. ortable and was not acceptable Therefore they made all FORTRAN 66 techniques of hiding character data in numeric variables illegal In addition they placed several restrictions on the placement and handling of character variables all intended to improve the portability of the language After the FORTRAN 77 standard came out however no compiler vendor felt that he could afford to lose his customers by telling them that they had to convert all of their FORTRAN programs into the new standard therefore all compiler vendors put features into their compilers which allowed existing programs to still operate The standards committee however had totally ignored the problem of what to do with existing programs aside here is that we first attempted to write a processor which would take older dialects of FORTRAN to the 77 standard This effort proved impossible The languages are simply incompatible FORTRAN 77 does not have the openness of C Since there was no standard mixed 66 77 FORTRAN all the vendors created their own Still today as new compilers are created even for the PC they differ in how they deal with this problem of 66 versus 77 FORTRAN The problem is now compounded in that particular programs are being written not for either standard but rather for the incompatible hybrid combinations In designing PROMULA FORTRAN we decided that the major barrier to using existing FORTRAN programs in a new machine environment was this incompatib
293. oth functions Using the default Ych switch the following is produced void demo long iv static char c 1 static long i i gs 56 Alternatively the Xch switch produces the following translation void demo iv long iv static char c 1 static long i i fifichar 8 c fifchar 56 This version would only be needed if some special character conversion logic were needed 2 41 Treatment of FORTRAN D debugging statements In the default FORTRAN dialect supported any FORTRAN statement with a non whitespace non numeric character in column is treated as a comment The Yd command line switch modifies this rule slightly When Yd is active statements with a D in column 1 are treated as though the D were a blank i e they are not comments This facility is used for inserting debugging statements comments into a FORTRAN program The default Xd flag treats D statements as simple Consider the following simple FORTRAN subprogram which contains a D type comment SUBROUTINE DEMO A B C DIMENSION A 10 B 6 7 C 3 4 5 D WRITE Demo successfully entered 96 PromulaFortran Translator User s Manual A 5 5 0 B 2 3 6 0 C I 2 3 7 0 RETURN END The default intermediate C output for this subprogram looks as follows void demo a b c float a b 7 6 c 5 41 3 WRITE Demo successfully entered a 4
294. other Finally the CF16 switch produces the following result void test define i3 3 define i5 5 extern char Xalpha typedef struct int ival jval kval Calpha auto Calpha Talpha Calpha Xalpha 23 PromulaFortran Translator User s Manual Talpha gt ival Talpha gt jval Talpha gt kval i5 switch Talpha gt ival case 1 goto S10 case 2 goto S20 case 3 goto S30 default break S10 WRITE OUTPUT LISTIO STRG IVAL 1 8 0 S20 return S30 STOP NULL undef i3 undef 15 In this version the definition of i5 has been simplified to its simple value of 5 The effects of the switches may be combined by adding their values For example CF11 1 2 8 could be used to maximize whitespace It produces the following result void test define i3 3 define i5 i3 2 extern char Xalpha typedef struct int ival int jval int kval Calpha auto Calpha Talpha Calpha Xalpha Talpha gt ival Talpha gt jval Talpha gt kval 15 switch Talpha gt ival case 1 goto S10 case 2 goto S20 case 3 goto S30 default break S10 WRITE OUTPUT LISTIO STRG IVAL 1 8 0 S20 return S30 STOP NULL undef 13 undef 15 With this setting the expressions are widened with whitespace and the switch and typedef statements have a vertical form however parameter values are still shown in their original form and identifiers are in low
295. parameters have the same case conventions that they had in the FORTRAN source 2 37 Treatment of Multiple Assignments Xa Ya C allows multiple assignments to the same value to be written together When the Ya switch is active the processor looks for such assignments and combines them whenever possible The Xa switch excludes this optimization The Ya flag is active for the C and optimized biases The FORTRAN bias does not by default perform these combinations since they destroy the correspondence between source and output statements Therefore for the default FORTRAN bias Xa is active The following code shows a FORTRAN subprogram which contains multiple assignments i e the ival and 1val vectors are both being set equal to 10 SUBROUTINE DEMO DIMENSION IVAL 10 JVAL 10 DO 10 I 1 10 IVAL I 10 92 PromulaFortran Translator User s Manual JVAL I 10 10 CONTINUE RETURN END The default intermediate C output for this subprogram is shown below void demo static int ival 10 jval 10 i for i20 i 10 i ival i 10 jval i 10 return The C output maintains a one to one correspondence with the FORTRAN original The multiple assignments are not simplified The same fragment under the Ya flag produces the following void demo static int ival 10 jval 10 i for i 0 i 10 i ival i jval i 10 return Since Ya allows multipl
296. part of the number complex float cpxreal a complex a Contains the value to be transformed Description Computes the real part of a short complex value The result is double precision Return value The double precision result See also None 6 20 CPXSIN Compute the Short Complex Sine 147 PromulaFortran Translator User s Manual Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex complex cpxsin complex a Contains the value to be transformed Description Computes the value of the short complex sine of a short complex number By definition eli a _ sin a 2i Return value The short complex result See also None 6 21 CPXSROOT Compute Short Complex Square Root Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex complex cpxsroot a complex a Contains the value to be transformed Description Computes the square root of a short complex value Return value The short complex result See also None 6 22 CPXSUB Short Complex Subtraction Synopsis 148 PromulaFortran Translator User s Manual include fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The i
297. performs the initial call to test 58 PromulaFortran Translator User s Manual The ftnblkd function is the default BLOCK DATA subprogram which is called each time a program is started Consequently the COMMON initializations within test are now treated as an extension of the BLOCK DATA system and behave in the same way as BLOCK DATA initializations Unfortunately this approach only works if all COMMON initializations are performed in the same module In cases where multiple modules are being combined perhaps via libraries to form a single executable there will be multiple ftnb1kd functions created This multiple creation will then cause a linker error The Om switch is used to deal with this problem The following translation is produced for the above using Om void test extern char Xblock typedef struct long i j Cblock auto Cblock Tblock Cblock Xblock static int FIRST 1 static namelist DATAVAR i Xblock 5 NULL j Xblock 4 5 NULL static char DATAVAL SDATAVAR ios j 6 SEND if FIRST FIRST 0 fiointu char DATAVAL 0 2 fiornl DATAVAR 2 NULL return WRITE OUTPUT LISTIO INT4 1 gt INT4 Toblock gt Jj 0 STOP NULL void BD_test lt 3 test In addition the following message is set to standard output Creating COMMON data module BD_test With this approach the initialization call w
298. ppearance of COMMENTS in C Output CM0 CM1 CM2 PROMULA FORTRAN translates both embedded and inline comments Embedded comments are FORTRAN statements with a C or an in column 1 Inline comments usually are placed on the same line as a FORTRAN statement starting to the right of an VAX FORTRAN convention or a PRIME FORTRAN convention which indicates where the statement ends and the comment begins 33 PromulaFortran Translator User s Manual The comment control switches CMO CM1 and CM2 simply specify whether comments from the source program should be included in the translation and what form to use if they are included The CMO switch excludes comments the CM1 switch includes them and blocks comments which follow each other while CM2 includes comments but does not block them Normally for the optimized bias is active and for the other biases including the default is active To see the difference between CM1 which blocks comments and CM2 which does not consider the following code SUBROUTINE TEST I J K C Compute the value of I using the following relation C J K I 5 I J K 5 END The default C output which uses CM1 produces the following output void test i j k long i j k Compute the value of I using the following relation J K I 5 xi jt k 5L In this output the comments associated with the statement have been grouped and combined into a single C
299. prefix cpx Single precision complex arithmetic is handled entirely via function references since C contains no complex arithmetic of its own The functions are straightforward as can be seen in the following function descriptions The basic structure used to implement the complex data type is as follows typedef struct float cr float ci complex The real part of the value The imaginary part of the number 5 7 Double Precision Complex Arithmetic Double precision complex arithmetic is handled entirely via function references since C contains no complex arithmetic of its own The functions all begin with the prefix dpx and are straightforward as can be seen in the following function descriptions The basic structure used to implement the double complex data type is as follows typedef struct double cr The real part of the value double ci The imaginary part of the number dcomplex 138 PromulaFortran Translator User s Manual 6 RUNTIME LIBRARY FUNCTION DESCRIPTIONS The following sections contain the descriptions for the functions in the PROMULA FORTRAN Runtime Library 6 1 CPXABS Compute the Short Complex Absolute Value Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex float cpxabs a complex a Contains the value to be transformed Description Compute
300. psis finclude fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of the number dcomplex int fiowvz val dcomplex val Double precision value to be written Description 228 PromulaFortran Translator User s Manual Writes a double precision complex value to the current output file in accordance with the current format specification This function corresponds to the FORTRAN 77 convention in which values and not just l values can be written Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowbiv Write binary values fiowrd Write double precision vector 6 187 FTNADS FORTRAN Add Strings Synopsis finclude fortran h PROMULA FORTRAN function declarations char 51 1 52 2 char 51 Pointer to first string int nl Length of first string char s2 Pointer to second string int n2 Length of second string Description Concatenates two FORTRAN style strings into a scratch storage array and returns a pointer to that array Calls to this function may be nested The maximum length of the scratch storage array is set at 256 Return value A pointer to the result of the concatenation See also None 6 188 FTNALLOC Allocate Dynamic Memory Synopsis include fortran h PROMULA FO
301. q static int Pow static long i static float xbar xbar 0 0 Pow 2 for i 1L i long nval i xbar val short i 1 xbar double long nval long weight sq pow xbar double Pow As specified the float double casts are no longer forced but all fixed point casts are forced At the C2 casting level the following is the result void demo val nval weight sq int nval weight float val sq static int Pow static long i static float xbar xbar 0 0 Pow 2 for i 0 i nval i xbar valti xbar double long nval weight sq pow xbar double Pow Now only the mixed mode cast is forced Integer arithmetic is again free so the loop reduction is also allowed 2 3 1 Arithmetic with Short Integer Variables 17 PromulaFortran Translator User s Manual In moving from one environment to another one must always be concerned with the accuracy of floating point arithmetic however there is also a real problem with fixed point arithmetic even when dealing with identical word sizes This section concerns itself with short integer arithmetic in non short integer environments There is a real semantics issue here the same program compiled in different environments behaves differently even though these environments have the same word size Consider the following FORTRAN program which does short integer additions multiplications and divisions in a variety of contexts
302. qual to num but less than any other threshold setting should be stored as a static variable The default setting for this storage type is 550 SD Specifies that no variables are to be stored as dynamic This is the default setting for this storage type SDnum Where num is greater than or equal to zero specifies that any variable whose size in bytes is greater than or equal to num but less than any other threshold setting should be stored as a dynamic variable SZ Specifies that no variables are to be stored as dynamic but taken from the virtual file manager SZnum Where num is greater than or equal to zero specifies that any variable whose size in bytes is greater than or equal to num but less than any other threshold setting should be stored as a dynamic variable but should have its values initialized via the virtual memory manager and should have its values returned to that manager when freed SV Specifies that no variables are to be stored as virtual This is the default setting for this storage type SVnum Where num is greater than or equal to zero specifies that any variable whose size in bytes is greater than or equal to num but less than any other threshold setting should be stored as a virtual variable 80 PromulaFortran Translator User s Manual The effect of these settings is to allocate memory into size ranges Variables are then allocated to a given storage range The example below gives a summary of the storage switch use
303. r User s Manual Reads a sequence of fixed length strings stored one after another from the current input file in accordance with the current format specification Each string is assumed to have its own corresponding format specification if a formatted read is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Perform FORTRAN I O error processing fiofinp Get next formatted input specification fioffld Get next free form input field fioralph Read alphabetic information 6 136 FIORDT Read FORTRAN Truth Value Vector Synopsis include fortran h PROMULA FORTRAN function declarations int fiordt bool nval unsigned long bool The values to be read int nval The number of values to be read Description Reads a vector of truth values long logical values from the current input file in accordance with the current format specification Note that in this function each individual truth value is assumed to have its own corresponding format specification if a formatted read is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Perform FORTRAN I O error processing fiofinp Get next formatted input specification f
304. r of the include are concerned since the include could well be included multiple times within a given file Each file defines its own internal variable to keep symbols such as defines from occurring multiple times This convention which is needed to overcome the scope differences between FORTRAN and C presents an excellent opportunity to do away with the undefs discussed EST1 h EST2 h int argc char argv include TEST1 include TEST2 ftnini argc NULL READ INPUT LISTIO DO int na INT4 Talpha gt ia 0 WRITE OU char Xalpha 1200 ifndef ICF_TEST1 Define the needed sizes KY define na 100L define nb 200L define ICF_TEST1 fendif ICF_TEST1 ifndef ICF_TEST2 Define the common data endif ICF TEST2 extern char Xalpha typedef struct long ia na Calpha auto Calpha Tal ib nb lpha Calpha ifndef ICF TES 2 define ICF TES 2 endif ICF TEST2 PUT LISTIO DO int nb INTA4 Talpha ib 0 Xalpha earlier at least insofar as the PARAMETERs within include files are concerned To give some idea of how complicated the actual production of the above files is the listing file produced is shown below If Line N1 Translation 1 define SPROTOTYPI IH 76 PromulaFortran Translator User s Man
305. r values in this vector are as follows Code Meaning 136 PromulaFortran Translator User s Manual 0 This block is not in memory and has no valid data on disk See discussion above on holes 1 This block is not now in memory but has data on disk 2 255 This block is in memory and is stored in block i 2 Given the block status vector the current status of any memory block can be determined via a single memory access no searching for the block is required Note that the block status vector is dynamically allocated when the virtual memory system is initialized Assuming first that the desired block is specified as being in memory by the block status vector then the memory block address vector contains a pointer to the actual virtual memory block Since there is an absolute maximum of 254 memory blocks the maximum number of entries in the memory block address vector is 254 therefore this vector is statically allocated The virtual memory blocks are dynamically allocated until there are either 254 of them or until available dynamic memory is exhausted Each block is defined via the vmsbt yp structure defined as follows define VMSBSIZ 1024 typedef struct int vbdel int vblru int vbmru long vbadr unsigned char vbdat VBLKSIZ vmsbtyp The members of the structure are defined as follows Name Description of use vbdel A flag indicating whether or not the data in this block has been changed vblru The numb
306. rage status of the temporary variables generated The following is a simple code that generates all three types of temporary variables SUBROUTINE DEMO A N M INC DIMENSION A N 1 M 1 INTEGER V DO 10 V 1 Mtl INC CALL TEST A V V 10 0 10 CONTINUE RETUR END 87 PromulaFortran Translator User s Manual The default C intermediate output for this example looks as follows void demo a n m inc int n m inc float a extern void test static int T1 T2 v D4 D5 static float T3 Tl n 1 T2 m 1 for v 1 D4 inc D5 m 1 v D4 D4 D5 gt 0 D5 v D4 T3 1 1 1 10 0 test amp T3 return The temporaries T1 and T2 are introduced to contain the values of the variable dimensions for the array a The temporaries D4 and D5 are introduced to contain the increment values and trip count value for the DO statement Finally the temporary T3 is introduced to contain the temporary value whose address must be passed to subroutine test Notice that all of these temporaries are assigned to static storage because the default storage for all variables is static see the section on the storage control flags If the above example were compiled using a different storage convention say SS SAO then the following output would be produced void demo a n m inc int n m inc float a extern void test auto int T1 T2 v D
307. ranslating it with the following prototype produces the memory version shown below anadat void manadat float short double double void manadat val n xbar var int n double xbar var float val static long j static float s xbar 0 0 var 0 0 for j 0L j n j xbar 1 3 xbar xbar double n 130 PromulaFortran Translator User s Manual for j OL j n j S val j xbar varts s var var double n 1 return Notice that the name of the function has been changed to agree with the prototype name Now translating the identical FORTRAN code with the following prototype produces the virtual code shown anadat void vanadat float short double double void manadat val n xbar var BACON double xbar var long val static long j static float s xbar 0 0 var 0 0 for j OL j n j xbar float vmsuse val j 4 xbar xbar double n for j OL j n j S float vmsuse val j 4 xbar var varts s var var double n 1 return This code treats val as a long virtual address and not as an array All references to val are made through the vmsuse function Again the name of the function has been changed via the prototype Now the following prototype and FORTRAN code can be translated In the translation we will tell PROMULA that C is to be virtual via an Sv100 comm
308. rc ee ie C cobs e IE ehe ees e e exte it e bee rine eR 40 2 9 ECHO CONTROL OPTIONS ES ET EX EZ EP EL 40 2 9 T Warnings Notes and CommeHhls d eee e o eq te Pe TRY Re Poe ER eb dea dee i ace eve e 41 2 9 2 Annotated Listing of Source Code eee eme ote Pee pe fe dee ete RE Gives ee 42 2 9 3 Symbol Listing and Cross Reference Table nene iia nete enne 43 2 0 4 Tntermediate Compiler Tables ea p rH ERU EHE pe 45 2 9 5 Annotated Listing of C Output i ERE EROR en ee eee HE 46 2 10 TREATMENT OF SYNTAX ERRORS ERO ERI ER2 4 46 2 11 FORTRAN INPUT FORMAT USED FSNUM FT FF FV F9 nennen eet haare 48 2 12 SOURCE FORTRAN INTEGER TYPE FIS FIL ener ener nennen nete nnn n essen neta tasas 49 2 13 GNAME NAME OF FILE CONTAINING GLOBAL SYMBOLS cccccccecsessssececececesensaececececsensaececececeeseaanaeceesesesenneaeeeeecs 49 2 14 COMMON VARIABLES CONVENTION GA GD GP GS GR 49 2 14 1 Overall Alignment Control with Gp Gpc Gps Gpl Gpd sse 53 2 15 INAME NAME OF FILE CONTAINING INLINE FUNCTIONS esee 53 2 16 TARGET CINT TYPE IS IE iecit to eee OH EE GE EEUU RR Eten ads EE eee UI TER EGRE 54 2 17 TREATMENT OF INTERNALLY GENERATED
309. reated on the fly in a calling parameter list or I O statement and then copies the specified strings into it Return value None the function is void however the parameter str is updated to contain a pointer to the start of the created string and its length Note that if this function is unable to obtain the needed scratch memory it exits with a message See also None 6 202 FTNSCOMP FORTRAN String Comparison Synopsis include fortran h PROMULA FORTRAN function declarations int ftnscomp sl 1 char sl int nsl Description This function performs string comparisons in which at least one of the two operands represents a concatenation of smaller strings Once any required concatenations have been performed the comparison proceeds character by character either until the end of both strings is reached or until a specific character difference is found If the strings are of unequal length the shorter string is treated as though it were padded with blanks to the length of the longer string Return value if no character difference is found n ifa character in the first string is less than the corresponding character in the second string n ifa character in the first string is greater than the corresponding character in the second string See also ftncms Compare two FORTRAN style strings 6 203 FTNSCOPY FORTRAN String Copy 237 PromulaFortran Translator User s Manual Synopsis include fortran h
310. repeated in this manual 1 1 User Support If you are a licensed and registered user of PROMULA FORTRAN you are entitled to user support from Great Migrations LLC If you encounter a problem that you cannot resolve on your own by referring to this User s Manual you may call or write us Great Migrations LLC PFC Support 7453 Katesbridge Ct Dublin Ohio 43017 614 761 9816 Your comments and suggestions about the product are always welcome If possible we will provide help over the telephone However if the problem involves an apparent translation problem or a runtime library problem we will probably need a copy of your source FORTRAN We will protect the full confidentiality of any sample codes that you send to us If the problem does uncover a problem either with the translator or with the runtime library you will be supplied with a corrected copy of PROMULA FORTRAN and or the Runtime Library as soon as we have made those corrections All purchasers of PROMULA FORTRAN will be notified of any revised versions and will be given the option to purchase them at a nominal update cost At your request Great Migrations LLC staff will also provide technical consulting services to assist you in software conversion projects 1 2 What is PROMULA FORTRAN PROMULA FORTRAN is a compiler which will process FORTRAN codes of almost any dialect on almost any platform that supports a standard C compiler It is also a comprehensive FORTRAN to
311. result void TEST define I3 3 define I5 13 2 extern char XALPHA 22 typedef struct int IVAL JVAL KVA CALPHA auto CALPHA TALPHA TALPHA gt IVAL TA PromulaFortran Translator User s Manual L CALPHA XALPHA PHA gt JVAL TALPHA gt KVAL I5 switch TALP HA gt IVAL case 1 goto S10 case 2 goto S20 case 3 goto S30 default break S10 WRI S20 return S30 STOP NULL undef I3 undef I5 Here all user supplied identifiers are contemporary C style E OUTPUT LISTIO STRG IVAL 1 8 0 shown in upper case the traditional FORTRAN style as opposed to lower case the Using the CF8 switch gives the following C result void test define i3 3 define i5 13 2 extern char Xalpha typedef struct int ival int jval int kval Calpha auto Calpha Talpha Talpha gt ival default break S10 Calpha Xalpha Talpha gt jval Talpha gt kval i5 switch Talpha gt ival case 1 goto S10 case 2 goto S20 case 3 goto S30 WRIT S20 return S30 STOP NULL undef i3 undef i5 In this version the typedef statement that the appearance of the braces and section of this chapter E OUTPUT LISTIO STRG IVAL 1 8 0 used to define the structure of the COMMON block is shown vertically Note again indentation is controlled via the NU and NL switches which are described in an
312. rf Write single precision vector 6 180 FIOWVI Write FORTRAN Short Integer Value Synopsis include fortran h PROMULA FORTRAN function declarations int fiowvi val 225 PromulaFortran Translator User s Manual short val Short value to be written Description Writes a short integer value to the current output file in accordance with the current format specification This function corresponds to the FORTRAN 77 convention in which values and not just l values can be written Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowbiv Write binary values fiowri Write short integer vector 6 181 FIOWVL Write FORTRAN Long Integer Value Synopsis include fortran h PROMULA FORTRAN function declarations int fiowvl val long val Long value to be written Description Writes a long integer value to the current output file in accordance with the current format specification This function corresponds to the FORTRAN 77 convention in which values and not just l values can be written Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowbiv Write binary values fiowrl Write long integer vector 6 182 FIOWVS Write FORTRAN String Value Synopsis in
313. rinted in this position unless the comma is in an asterisk field In that case an will be printed in that position Credit CR 183 PromulaFortran Translator User s Manual The characters CR may only be used as the last two rightmost characters of the string If the number is positive two spaces will be printed following it If negative the letters CR will be printed Return value None the function is void The global variable iocrec is updated to contain the new field and the variable fionchar is updated to reflect the new character count in the coded communications record See also None 6 104 FIOCLOSE Close Current FORTRAN File Synopsis include fortran h PROMULA FORTRAN function declarations int fioclose Description This function closes the file associated with the current FORTRAN file as specified in the global variable iocurf Return value A zero if the close was successful else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Performs standard error processing 6 105 FIOCPATH Convert Pathname Synopsis finclude fortran h PROMULA FORTRAN function declarations void fiocpath pathname cinfo char pathname Pathname to be converted unsigned char cinfo Conversion information table Description Converts a pathname as it appears in an OPEN or open related statement into a form compatible with
314. routines PARAMETERS LOGICAL ARG SWAB FLAG R 108 PromulaFortran Translator User s Manual jOGICAL ARG XLIT FLAG R RETURN VALUE one SIDE EFFECTS one Incorporates features to support portable LIO file handling 8 Apr 1986 IMPLICIT INTEGER A Z Initialization entry Initialization calculate some useful things NONINPBASE HD OPIS HD OPSUPIS FIRSTNONINP NONINPBASE 1 First non input node Determine practical maximum for no of strings we can hash using PUTSTR entry Knuth s vol 3 says that the avg no collisions 1 2 1 1 R where is HASH table density it might be nice to stop at 98 full R 0 98 since this translates to an average of 25 5 collisions per string entered HASHMAX HD HASHVAL 2 HD HASHVAL 100 Save local values of SWAB and XLIT flags and convert blank string to match the language EBCDC or ASCII of the model file CREATOR SWAB FLAG ARG SWAB FLAG XLIT FLAG ARG XLIT FLAG IF XLIT FLAG AND NOT BLANKS XLIT YET THEN CALL TRLITO BLANK BYTES 4 BLANKS XLIT YET TRUE NDIF ETURN ND Ej UU on Within this code the is being used to mark the comments and multiple occurrences of this symbol are used for emphasis The default translation looks as follows long modi arg swab flag arg xlit flag
315. rticular dialects which the compiler supports are as follows 1 9 Dealing with C Types and FORTRAN Types PROMULA FORTRAN gives you access to all of the following standard C types signed char unsigned char short signed int short unsigned int long signed int long unsigned int float double and to float complex and double complex structured types Though PROMULA FORTRAN has the usual set of default FORTRAN types you can specify which FORTRAN type should connect to which C type If your dialect of FORTRAN includes nonstandard types such as BOOLEAN or BYTE or INTEGER 3 then you can include these In addition FORTRAN compilers differ as to whether logical types are bit sequences or simply TRUE FALSE values You may also select this If bit sequences are assumed the logical operators become bit manipulation operators 1 10 Dealing with FORTRAN Input Output in C The approach taken to FORTRAN input output statements by PROMULA FORTRAN is straightforward the C programs should behave identically to their FORTRAN counterparts All standard I O statements are accepted including long FORMAT statements These tend to be rejected by C compilers because their initialization strings exceed the compiler limit on the length of individual static strings C input output is managed via calls to functions in the C library C has no input output statements as such FORTRAN input output statements are translated into calls to functions in the PROMULA FORTRAN
316. rtran Translator User s Manual SUBROUTINE TEST IFUNC RFUNC R RFUNC 99 I IFUNC 56 7 PRINT I R STOP END When processed using the Yv switch which allows call by value arguments the following C translation is produced void test ifunc rfunc int ifunc float rfunc static float r static int i rfunc 99 i ifunc 56 7 WRITE OUTPUT LISTIO INT4 i REAL4 4r 0 STOP NULL Notice that the types of the arguments are not specified in the declarations for ifunc and rfunc Using the P2 flag produces the following result void test ifunc rfunc int ifunc float float rfunc int static float r static int i r rfunc 99 i ifunc 56 7 WRITE OUTPUT LISTIO INT4 i REAL4 r 0 STOP NULL Now the function declarations use the full ANSI C form for the function declarations 2 27 3 P4 P8 Exclude Referenced or Defined Prototypes When processing a FORTRAN code PROMULA FORTRAN builds an internal table containing the prototypes for all subprograms encountered either by reference or by definition If desired these prototypes can be written to a file via the W switch described elsewhere in this chapter The P4 and P8 flags associated with the P switch block either those functions only referenced via the code or those functions only defined in the code The following FORTRAN code UBROUTINE ALPHA IA GVAL RFUNC
317. rtual variables can be created via the PROMULA interface described in the PROMULA FORTRAN Compiler manual The type of a variable is simply its type specification For records it is the structure type of the record The comment associated with the symbol is a function of its object type For constant integer parameters i e those that may be used in other declaration statements the comment specifies the value of the parameter as specified or computed For variables or records the dimensionality is given and for arrays the total size in bytes For subprograms the number of arguments is given along with the assumed type of each argument Note that in the C tradition PROMULA FORTRAN makes extensive use of subprogram prototypes and always makes certain that argument types are consistent If they are not it issues a warning The symbol reference by line number table is simply that Along with the line number a use type code is also specified D means defined M means modified U means used and P means passed to a subprogram If a symbol has multiple references in a single statement then only one reference is reported Note that if include files are involved then the line number is followed by the include file sequence number If include files are not involved then a simple sequence number is used PROMULA FORTRAN Compiler V4 00 Date 08 11 92 Time 09 35 Page 2 File utest for Include files used in unit Seq Filename
318. runtime library These functions perform input output the way FORTRAN does Though the general form of input output statements is quite standard throughout the FORTRAN dialect community no two FORTRAN compilers have the identical set of options This is the major area in which dialects differ With the PROMULA FORTRAN dialect manager you may describe the individual options associated with your particular dialect In addition PromulaFortran Translator User s Manual you may specify the form of the function to be referenced in the translation Then you can simply write the needed function and include it in your version of the runtime library 1 11 Runtime Library The PROMULA FORTRAN runtime library is a set of approximately 250 functions designed for use with the PROMULA FORTRAN translator It may also be used by those FORTRAN programmers who wish to program in C but who do not wish to give up the input output conventions formatting controls and intrinsic functions which they have grown used to Initially C codes using this library can be produced by translating FORTRAN programs into C using the PROMULA FORTRAN translator Once in C the programs may then be maintained by using these functions If there is one certainty it is that no two FORTRANS behave in the same way especially with regard to their runtime libraries Thus if your conventions differ from the ones used here you may alter the library code Alternatively your version of F
319. s Manual include fortran h PROMULA FORTRAN function declarations char fifstrgv str nc char str String to be converted int nc Number of characters in string Description Copies a character string into a local buffer and then pads that buffer to 8 characters with blanks so that the string can be compared to a character string which has been hidden in a noncharacter variable Note that such hiding always pads the numeric with blanks out to its natural size This function is needed for situations like F K EQ where is a noncharacter variable The translation of this would be if k typeof k fifstrv Y 1 Return value A pointer to a buffer containing the string value See also None 6 97 FIFSYSTM FORTRAN External Function SYSTEM Synopsis finclude fortran h PROMULA FORTRAN function declarations int fifsystm cl ncl char cl String containing command to be executed int ncl Length of the execution string Description Executes the command contained in the character string and then resumes execution of the current program The problem addressed by this function is that system requires a nul terminated string while the string received is blank padded with no terminator Return value A system dependent integer status from the command In UNIX systems the status return is the value returned by exit See also None 6 98 FIFTAN FORTRAN Intrinsic Function TAN Synopsis finclude fortran h
320. s Manual Writes a sequence of fixed length strings stored one after another to the current output file in accordance with the current format specification Each string is assumed to have its own corresponding format specification if a formatted write is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Do requested error processing fiofout Process output format specification fiowalph Write alphabetic information fiowtxt Write a text record 6 171 FIOWRT Write FORTRAN Truth Value Vector Synopsis include fortran h PROMULA FORTRAN function declarations int fiowrt bool nval unsigned long bool Points to values to be written int nval Number of values to be written Description Writes a vector of truth values to the current output file in accordance with the current format specification In this context the term truth value refers to a long logical value The output display for a logical value consists of a sequence of fw 1 blanks followed by a T or an F where fw is the field width T is used for nonzero values and F is used for zero values Note that this function also supports the FORTRAN 66 convention under which alphabetic information may be stored in the logical values Return value A zero if there is no error flag set else an error cod
321. s characters to be read int nval Number of characters to read Description 252 PromulaFortran Translator User s Manual Reads a virtual vector of character values from the current input file in accordance with the current format specification Note that in this function each individual character is assumed to have its own corresponding format specification if a formatted read is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiordu Read FORTRAN character vector vmsdel Change a virtual value 6 233 VMSSAVE Save a Virtual Vector Synopsis include fortran h PROMULA FORTRAN function declarations void vmssave svec nbyte ioiad char svec Starting address of the information long nbyte Number of bytes to be saved long ioiad Virtual address of values Description Saves a vector of values on the virtual database and then releases the memory being used for that vector Return value None the function is void See also ftnfree Free dynamic memory vmsvect Virtual vector input or output 6 2324 VMSUSE Use Virtual Information Synopsis finclude fortran h PROMULA FORTRAN function declarations unsigned char vmsuse ioiad long ioiad Offset of information to be used Description Returns a pointer to the byte at a specified long linear address This func
322. s present otherwise fills evalue with blanks Return value None the function is void See also None 166 PromulaFortran Translator User s Manual 6 63 FIFHBIT FORTRAN High Bit Management Synopsis include fortran h PROMULA FORTRAN function declarations void fifhbit chrs nbyte unsigned char chrs Vector of characters int nbyte Number of bytes in vector Description In several dialects of FORTRAN especially Prime FORTRAN 77 the normal internal character representation is a seven bit code with the 8th bit always set on while the extended characters have the 8th bit off The normal external character representations use the opposite conventions This function converts from one representation to the other by toggling the high order bit in each character of the input vector Return value None the function is void See also None 6 64 FIFIZABS FORTRAN Intrinsic Function I2ABS Synopsis finclude fortran h PROMULA FORTRAN function declarations short fifi2abs a short a Value whose absolute value is needed Description Computes the absolute value of a short integer argument Return value The short integer result See also None 6 65 FIFIZDAT FORTRAN External Function IDDDATE Synopsis include fortran h PROMULA FORTRAN function declarations void fifi2date mm dd yy short mm Returns month value short dd Returns day value short yy Returns year value Description 167
323. s the second to it Under either case demo2 will not produce the right result The CHd approach is general String concatenations themselves are simpler and dynamic allocations are allocated memory at runtime The following is the default CHd C output for the above void demol extern void demo2 static char sum 13 initial 4 static string T1 T2 ftnscopy initial 4 ONE 3 NULL ftnscopy sum 13 initial 4 PLUS 5 initial 4 NULL ftnsallo amp T1 C 2 sum 13 NULL OPEN 1 FILEN Tl a Tl n 0 free T1 a ftnsallo amp T1 sum 13 PLUS THREE 10 NULL 32 PromulaFortran Translator User s Manual ftnsallo amp T2 sum 13 PLUS TWO 8 NULL demo2 T1 a T2 a T1 n T2 n free Tl a free T2 a return In the case of assignment the ftnscopy function replaces the combined effect of ftnads and ftnsac as presented earlier The ftnscopy function takes a variable number of arguments terminated by a NULL It concatenates the second and beyond strings directly into the first In the case of dynamic allocations the string type introduced in the previous subsection is combined with a function ftnsallo This function computes the length needed for the concatenation dynamically allocates memory for it and then does the copies The results are stored in the st ring temporary passed as the first argument This approach is completely general The only problem is that once th
324. s the value of the short complex absolute value The result is single precision Return value The single precision result See also None 6 2 CPXADD Short Complex Addition Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex float cpxadd a b complex a The left hand value complex b The right hand value Description Adds two short complex numbers to form a third Return value The complex result of the addition 139 PromulaFortran Translator User s Manual See also None 6 3 CPXCJG Compute the Short Complex Conjugate Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex complex cpxcjg a complex a Contains the value to be transformed Description Computes the complex conjugate of a short complex value By definition conj a a cr i a ci Return value The complex result See also None 6 4 CPXCMP Short Complex Comparison Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex int cpxcmp a b complex a The left hand value complex b The right hand value Description Compares two short complex numbers Return v
325. sa oe esa nne s un 229 FTNBACK FORTRAN BACKSPACE STATEMENT ececeeeetene e e e ene nne ener nnns ns ase e ese sese sata sese setas 230 FINBLKD FORTRAN BLOCK RDA A etre ette cete tete trece tele cere de ehe ee ete caido dd 230 FTNCLOSE FORTRAN CLOSE STATEMENT eee 230 FTNCMS FORTRAN COMPARE STRINGS eese 231 FINFREE FREE DYNAMIC MEMORY 232 FTNINI INITIALIZE FORTRAN PROCESSING eene e ene esas ese ase ese esa ise 232 FTNLUN ESTABLISH FILE FOR LOGICAL UNIT NUMBER eere ee emen nnnm ener nnns nsns ese usus ns e nsa uses s nsns nsu 233 viii 6 196 6 197 6 198 6 199 6 200 6 201 6 202 6 203 6 204 6 205 6 206 6 207 6 208 6 209 6 210 6 211 6 212 6 213 6 214 6 215 6 216 6 217 6 218 6 219 6 220 6 221 6 222 6 223 6 224 6 225 6 226 6 227 6 228 6 229 6 230 6 231 6 232 6 233 6 234 6 235 6 236 6 237 6 238 6 239 6 240 6 241 6 242 6 243 6 244 6 245 PromulaFortran Translator User s Manual FTNOPEN FORTRAN OPEN e enne nnnm enhn e etere ere 233 FTNPAUSE FORTRAN PAUSE STATEMENT eeeeeeenere e e e ene e nnne enhn nn nse senis uses e nsu ese ese e ese ases esas ese ese nnns ue 234 FTNREAD FORTRAN READ STATEMENT eceeeeetene ee e e emen nnnm nn nn e uses e ese ese ise ese es
326. sformed Description Computes the short complex exponential of a short complex value Return value The short complex exponential See also None 6 11 CPXIMA Compute the Imaginary Part of a Short Complex Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex 143 PromulaFortran Translator User s Manual float cpxima a complex a Contains the value to be transformed Description Computes the imaginary part of a short complex value The result is single precision Return value The single precision result See also None 6 12 CPXLOG Short Complex Natural Logarithm Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex complex cpxlog a complex a The value to be transformed Description Computes the short complex natural logarithm of a short complex value Return value The complex natural logarithm See also None 6 13 CPXLOG10 Short Complex Base 10 Logarithm Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex complex 10910 a complex a The value to be transformed Description 144 PromulaFortran Translator User
327. sible tell whether call by value is possible simply from the source code Some other device is needed to tell the translator which arguments can be passed by value Another problem with FORTRAN is that it is weak typing By this is meant that it is sometimes valid to pass data of differing binary types via the same subprogram argument The translator needs to know when this is valid and if it is not valid it needs to know which binary type is the expected one Many perfectly valid FORTRAN programs fail either at compilation time or execution time because of differing typing conventions and differing internal representations There is no general solution to the translation of weak typing FORTRAN programs The translator must be told what to do Some device is needed to describe subprogram arguments to the translator A similar problem has faced C programmers as well Consequently the new ANSI C has introduced the notion of a function prototype which describes the arguments of functions in terms of their binary type and in terms of their pointer status The conventions developed there are exactly those needed by the translator although they need to be extended slightly to deal with virtual variables multiple forms and external name clash The C prototype system and its extensions to PROMULA are discussed in this section It is strongly recommended that you take the time to develop a set of prototypes for the subprograms within a FORTRAN program pr
328. sis finclude fortran h char fioitos numb int numb Description Performs standard error processing Read text record Convert Integer to String PROMULA FORTRAN function declarations Value to be converted Converts an integer number to a character string Return value A pointer to the string result See also None 6 120 FIOLREC Synopsis finclude fortran h int fiolrec irec long irec Description Position a FORTRAN File on a Record PROMULA FORTRAN function declarations Record number desired This function positions the file associated with the current FORTRAN file as specified in the global variable fiocurf at the beginning of a specified record 194 PromulaFortran Translator User s Manual Return value A zero if the positioning was successful else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also None 6 121 FIOLTOS Convert Long Integer to String Synopsis finclude fortran h PROMULA FORTRAN function declarations void fioltos value long value Value to be converted Description Converts a long integer value to display form and stores it at the current position in the coded communications record right justified in a fixed length field In particular the output field consists of blanks if necessary followed by a minus sign if the internal value is negative or an optional plus sign otherwise I
329. slate all errors into non linkable FORTRAN ERROR function calls d Translate all errors into executable print error message statements e Translate all errors into warning comments PromulaFortran Translator User s Manual Conversion As part of our standard maintenance agreement if after translating working FORTRAN Support application with PROMULA FORTRAN you find that it does not yield the same results in C as it does in FORTRAN we will gladly revise or extend our translator and or its runtime library until you achieve complete reproducibility of results This guarantee applies only to pure FORTRAN applications hybrid applications that contain external possibly non FORTRAN components are outside the scope of our standard maintenance agreement and are handled as part of our system conversion services on a consulting basis upon request Documentation According to a Computer Language review October 1988 the documentation for PROMULA FORTRAN is excellent It meticulously details the translation process all the runtime library routines and more In addition the language reference component of the documentation describes the actual FORTRAN dialect that PROMULA FORTRAN supports 1 5 How PROMULA FORTRAN Works The PROMULA compiler is based on the proven FORTRAN to C translator which was released to the PC market almost six years ago as Version 1 0 of PROMULA FORTRAN and has shown in a number of FORTRAN code migration projects
330. so None 6 40 DPXPOW Raise Double Complex to a Power Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of the number dcomplex dcomplex dpxpow a b dcomplex a The value to be raised to a power dcomplex b The value of the power Description Computes the value of a short complex raised to a short complex power Return value The complex result of the calculation See also dpxexp Computes a double complex exponential dpxlog Computes a double complex logarithm dpxmul Multiplies two double complex values 6 41 DPXREAL Compute Real Part of Double Complex Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of the number dcomplex 157 PromulaFortran Translator User s Manual double dpxreal a dcomplex a Contains the value to be transformed Description Computes the real part of a short complex value The result is double precision Return value The double precision result See also None 6 42 DPXSIN Compute the Double Complex Sine Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct double cr The real part of the value double ci The imaginary part of the number dcomplex dcomplex dpxsin a dcomplex a Contains the value to be transformed Descriptio
331. specification Each value is assumed to have its own corresponding format specification if a formatted read is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Perform FORTRAN I O error processing fioffld Get next freeform input field fiordf Read floating point values fiospace Skip white space in record fiostod Convert string to double 6 139 FIORDZ Read FORTRAN Double Complex Values Synopsis include fortran h PROMULA FORTRAN function declarations int fiordz value nval 204 PromulaFortran Translator User s Manual dcomplex value Points to values to be read int nval Number of values to be read Description Reads a vector of double precision complex values from the current input file in accordance with the current format specification Each value is assumed to have its own corresponding format specification if a formatted read is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Perform FORTRAN I O error processing fioffld Get next freeform input field fiordd Read floating point values fiospace Skip white space in record fiostod Convert string to double 6 140 FIOREC
332. ssed through this function either by the translator directly or by the other runtime functions included in this library Note that if you wish this function to return some value other than the host processor values then you must modify it Typically this modification would take the form of a lookup table reference Return value 172 PromulaFortran Translator User s Manual The character value directly See also None 6 79 FIFIDIM FORTRAN Intrinsic Function IDIM Synopsis include fortran h PROMULA FORTRAN function declarations long fifidim al a2 long al First value in difference long a2 Second value in difference Description Computes a positive difference Return value If a1 is greater than a2 then the value of al a2 is returned else zero is returned See also None 6 80 FIFIDINT FORTRAN Intrinsic Function IDINT Synopsis include fortran h PROMULA FORTRAN function declarations long fifidint a double a Argument containing the value Description Truncates its double precision argument to a long integer value In particular fifidint a 0 if lt 1 the largest integer with the same sign that does not exceed aif gt 1 Return value The value as computed above See also None 6 81 FIFINDEX FORTRAN Intrinsic Function INDEX Synopsis include fortran h PROMULA FORTRAN function declarations 173 PromulaFortran Translator User s Manual int fifindex s ns c nc c
333. standard or two available If you have a relatively small FORTRAN 77 program and wish to use it on a new platform in a C environment then you have a variety of compilers and or translators available to you but we think PROMULA FORTRAN will give you the best results However if you have a serious FORTRAN program and you do not wish to do any changes by hand then we know that PROMULA FORTRAN is your only current alternative We originally developed PROMULA FORTRAN for use in our own consulting business because we were unable to find a FORTRAN compiler which would effectively and accurately process the typical FORTRAN programs which we wished to migrate to the PC or other workstations Having once developed the translator we discovered that far more than just translation could be achieved during the conversion Many other problems could also be solved by translating to C 1 7 Downsizing Mainframe Codes for Use on the PC DOS Platform Many FORTRAN programs are written with the assumption that a very large memory is available This is because most machines above the PC class have virtual operating systems A standard error free translation of a statement like the following DIMENSION A 10000 20 B 10000 20 would be static float a 20 10000 5 20 10000 The fact that this is syntactically error free is of little significance since no C compiler presently available for the PC will accept it Faced with this problem what do you do In
334. t forced on 1 by the CL setting in fact causes a long cast on weight This is how this convention forces short arithmetic to be done as long arithmetic It might be instructive to see this example again not with CO but with Cs which will not force these long promotions 16 PromulaFortran Translator User s Manual void demo val nval weight sq int nval weight float val sq static int Pow static long i static float xbar xbar 0 0 Pow 2 for i 1L i long nval i xbar float double xbar double val short i 1 xbar float double xbar double nval weight sq float pow double xbar double The other effect of forcing the casts is that this process blocks the DO loop reduction algorithm Thus in the initial version the loop was reduced to start at zero but now it starts at one We decided to do this since forcing the casting level probably means that you are very concerned about the arithmetic being performed Since loop reduction performs additional integer arithmetic we turn it off As one might fear the cast to double is too late The division is done at the integer 2 level and the possibly overflowed result is promoted to double Thus using CO to fix the short arithmetic problem only shows us what the problem is Moving up to C1 the following example shows the C output using the C1 casting level void demo val nval weight sq int nval weight float val s
335. t of parameters 1 long Points to an error return variable Return value A zero if the rewind went well else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Perform FORTRAN I O Error Processing fiointu Establish an internal file fiolun Establish unit number fiorew Rewind current file fiostatus Set I O Error Status 6 200 FTNSAC FORTRAN Store a Character String Synopsis include fortran h PROMULA FORTRAN function declarations void ftnsac sl nl s2 n2 char 51 Pointer to receiving string int nl Length of receiving string char s2 Pointer to sending string int n2 Length of sending string Description Stores a FORTRAN style character string into another string and then pads that string with blanks Note that some care must be taken with overlaying copies Since not all platforms support an overlayed copy this is done via an explicit loop Return value None the function is void See also None 236 PromulaFortran Translator User s Manual 6 201 FTNSALLO FORTRAN String Allocation Synopsis include fortran h PROMULA FORTRAN function declarations typedef struct char a Pointer to the character storage int Length of the string string void ftnsallo str string str String to be allocated Description This function allocates scratch storage space needed to store a string concatenation c
336. t of the number complex int fiowrx value nval complex value Points to values being written int nval Number of values to be written Description Writes a vector of single precision complex values to the current output file in accordance with the current format specification Note that in this function each individual component of the complex value is assumed to have its own corresponding floating point format specification if a formatted write is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowrf Write vector of floats 6 174 FIOWTXT Write Text Record 222 PromulaFortran Translator User s Manual Synopsis finclude fortran h PROMULA FORTRAN function declarations void fiowtxt Description Writes the current text record to the current file followed by a new line Then it sets the length of the current text record to Zero Return value None the function is void See also fioerror Perform error processing 6 175 FIOWVAL Write Floating Point Value Synopsis include fortran h PROMULA FORTRAN function declarations void fiowval value double value Value to be converted Description This utility function controls the conversion of floating point values to string form under the control of a FORMAT specification The actual form of the outp
337. t_flag PURPOSE General purpose initialization for MODI routines PARAMETERS OGICAL SWAB FLAG R OGICAL XLIT FLAG R FE RETURN VALUE None KK SIDE EFFECTS None Incorporates features to support portable LIO file handling 8 Apr 1986 111 PromulaFortran Translator User s Manual long arg swab flag arg xlit flag extern void trlito static long K1 4 static long modi noninpbase hd qpsupis firstnoninp hashmax hd hashval swab flag xlit flag blanks xlit yet blank bytes Initialization entry Initialization calculate some useful things m noninpbase hd qpis hd qpsupis firstnoninp noninpbaset1L First non input node Determine practical maximum for no of strings we can hash using PUTSTR entry Knuth s vol 3 says that the avg no collisions 1 2 1 1 R where R is HASH table density it might be nice to stop at 98 full 0 98 since this translates to an average of 25 5 collisions per string entered b 0X FF F 0X F X hashmax hd hashval 2L hd 1 1001 Save local values of SWAB and XLIT flags and convert blank string to match the language EBCDC or ASCII of the model file CREATOR swab flag arg swab flag xlit flag arg xlit flag if xlit flag amp fifi4log blanks xlit yet
338. tant or expression specifying the terminal value as for DO loops incr is an integer constant symbolic constant or expression specifying the increment as for DO loops clist is a list of constants or symbolic constants specifying the initial values Each item in the list can take the form Gy EFC Elele y Where is a constant or symbolic constant is a repeat count that is an unsigned nonzero integer constant or the symbolic name of such a constant The repeat count can repeat the value of a single constant or can repeat the values of a list of constants enclosed in parentheses To specify repetition of a complex constant another set of parentheses must be used In addition to DATA statements themselves the declaration of any variable be it in a DIMENSION statement or a type statement or a COMMON statement may be followed by a set of initialization values enclosed in slashes The biggest problem in the translation of initializations from FORTRAN to C relates to a limitation in C syntax Simply stated there is no equivalent of the n value notation in FORTRAN Thus if you need to initialize a large array with a single value you must write that value over and over again The other problem has to do with initializing a numeric variable with a character constant which C does not allow in its initializations The default setting for PROMULA FORTRAN uses the C static variable initialization notation for FORTRAN DATA init
339. te they are called radix 50 because 40 is 50 when written in octal Return value The number of characters actually converted See also None 6 95 FIFRBIT FORTRAN Intrinsic Function Synopsis include fortran h PROMULA FORTRAN function declarations void fifrbit bits nbyte unsigned char bits Bit vector to be reversed int nbyte Number of bytes in bit vector 178 PromulaFortran Translator User s Manual Description Reverses the order of the bytes in a bit string Return value None the function is void See also None 6 94 FIFSIN FORTRAN Intrinsic Function SIN Synopsis include fortran h PROMULA FORTRAN function declarations float fifsin x float x Argument containing the value Description Computes the sine of a floating point argument using single precision Return value The value as computed above See also fifsncs Compute sine or cosine 6 95 FIFSNCS FORTRAN Single Precision Sine Cosine Synopsis include fortran h PROMULA FORTRAN function declarations float fifsncs x cosflag float x Argument containing the value int cosflag Perform cosine flag Description Computes the sine or cosine of a single precision value which has been promoted to double via the C promotion rules Return value The double precision value as computed above See also None 6 96 FIFSTRGV FORTRAN String Value Conversion Synopsis 179 PromulaFortran Translator User
340. ted C code is both compilable and correct that is it yields reproducible results See Validation above Language Translate FORTRAN to maintainable C Migrate to C for source code maintenance and further Migration development The C code generator actually offers three output options or biases a The FORTRAN bias generates C output which is as close to the original FORTRAN as possible and is aimed at easing the transition of those users who are presently FORTRAN programmers but wish to or must become C programmers b The C bias generates C output which looks much like a standard C program and is aimed at those users who are C programmers but must now take over the maintenance of a FORTRAN code c The optimized bias generates C output which is designed to compile as quickly as possible and to produce an efficient as possible executable module This output is not very readable and is aimed at those users who wish to continue to program in FORTRAN For these users the C output is of no importance as such It is merely an intermediate step and serves as input to the C compiler Error When a FORTRAN error occurs an error message is issued and the user has the option to exit or Processing continue processing command line switch allows the user to select from five different error handling options a Stop translation when the first error is encountered b Translate all errors into non compilable FORTRAN ERROR statements c Tran
341. tenations but deal only with simple examples involving concatenations done on the fly Consider the following FORTRAN fragment SUBROUTINE DEMO1 CHARACTER SUM 13 INITIAL 4 INITIAL ONE SUM INITIAL PLUS INITIAL OPEN 1 FILE C SUM CALL DEMO2 SUM PLUS THREE SUM PLUS TWO RETURN END This fragment shows not only a concatenation associated with an assignment but also concatenations as passed to the operation and as passed as subprogram arguments The C for this fragment under CHs would be as follows Er z void demol extern void demo2 static char sum 13 initial 4 ftnsac initial 4 ONE 3 ftnsac sum 13 ftnads ftnads initial 4 PLUS 5 9 initial 4 13 OPEN 1 FILEN ftnads C 2 sum 13 15 0 demo2 ftnads sum 13 PLUS THREE 10 23 ftnads sum 13 PLUS TWO 8 21 return In this code a simple function tnads which returns a pointer to a char is used to perform all the concatenations Each call to ftnads adds another string to a static array In the case of the assignment statement the temporary result in tnads is copied into the result variable immediately after it is computed For the OPEN and the argument cases however the result is left in ftnads For the OPEN this does not matter but in the case of the arguments the second call to tnads either destroys the first argument or concatenate
342. ter 2 9 Echo Control Options ES ET EX EZ EP EL Except for its initial banner PROMULA FORTRAN is normally silent unless a fatal error is encountered Messages are sent to standard output and may be redirected or may be sent to a listing file see section on the PN and PA switches Note that error messages themselves are discussed in the PROMULA FORTRAN Compiler manual The echo control options may be used to send additional information to the listing output This information includes the following Option Information sent to listing output Warnings about potentially serious inconsistencies usages in FORTRAN source code EL2 Comments about possibly non portable code or about code that may be incorrect and the warnings from above 40 PromulaFortran Translator User s Manual EL3 Notes about standard FORTRAN violations and other miscellaneous observations and comments and warnings from above EP A listing of the intermediate pseudo code produced by the compiler ES An annotated listing of the source code ET An annotated listing of the C output code EX An alphabetical listing of the symbols used in the source along with a summary description of each and a cross reference listing of symbol references by line number EZ A listing of the intermediate symbol information produced by the compiler 2 9 1 Warnings Notes and Comments The form and meaning of the warnings notes and comments are discussed in the
343. ter another from the current input file in accordance with the current format specification Each string is assumed to have its own corresponding format specification if a formatted read is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiords Read FORTRAN character strings vmsdel Change a virtual value 6 231 VMSRDT Read FORTRAN Virtual Truth Value Vector Synopsis include fortran h PROMULA FORTRAN function declarations int vmsrdt value nval long value Virtual address of values to be read int nval The number of values to be read Description Reads a virtual vector of truth values long logical values from the current input file in accordance with the current format specification Note that in this function each individual truth value is assumed to have its own corresponding format specification if a formatted read is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiordt Read FORTRAN truth value vector vmsdel Change a virtual value 6 232 VMSRDU Read FORTRAN Virtual Unsigned Character Vector Synopsis include fortran h PROMULA FORTRAN function declarations int vmsrdu value nval long value Virtual addres
344. ter vector vmsuse Use a virtual value 6 238 VMSWRD Write FORTRAN Virtual Double Precision Vector Synopsis include fortran h PROMULA FORTRAN function declarations int vmswrd value nval long value Virtual address values being written int nval Number of values to be written Description Writes a virtual vector of double precision values to the current output file in accordance with the current format specification Note that in this function each individual value is assumed to have its own corresponding format specification if a formatted write is being performed 255 PromulaFortran Translator User s Manual Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowrd Write FORTRAN double precision vector vmsuse Use a virtual value 6 239 VMSWRF Write FORTRAN Virtual Single Precision Vector Synopsis include fortran h PROMULA FORTRAN function declarations int vmswrf value nval long value Virtual address of values being written int nval Number of values to be written Description Writes a virtual vector of single precision values to the current output file in accordance with the current format specification Note that in this function each individual value is assumed to have its own corresponding format specification if a formatted write is being performed Return value
345. tes to unit 1 UBROUTINE DEMO A B C NR NC RITE 1 A RITE 1 A I I 1 6 RITE 1 B C ETURN Euzzzou z The default translation under Xu is as follows void demo int nr nc float a b c static int i WRITE 1 REAL4 a nr 0 WRITE 1 MORE for i 0 i lt 6 i WRITE REAL4 amp a i 1 MORE WRITE 0 WRITE 1 REAL4 b nc REAL4 c nr nc 0 return The listing is easy to read and corresponds to the original FORTRAN quite closely When executed it will be reasonably efficient but each write will require that the file be repositioned twice Using the Yu flag the following C output is obtained void demo int nr nc IMENSION A NR B NC C NR NC 104 PromulaFortran Translator User s Manual float a bI c static int T1 i T2 T3 Tl nr 4 fiouwl amp T1 WRITE 1 4 0 2 24 fiouwl amp T2 WRITE 1 MORE for i 0 i 6 i WRITE REAL4 amp a i 1 MORE WRITE 0 nc 4 nr nc 4 fiouwl amp T3 WRITE 1 REAL4 b nc REAL4 c nr nc 0 return In this output temporary variables are computed which contain the lengths of the records to be written These values are then passed to a runtime function
346. text the term Boolean value refers to a short logical value The output display for a logical value consists of a sequence of fw 1 blanks followed by a T or an F where fw is the field width T is used for nonzero values and F is used for zero values 254 PromulaFortran Translator User s Manual Note that this function also supports the FORTRAN 66 convention under which alphabetic information may be stored in the logical values Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowrb Write FORTRAN Boolean vector vmsuse Use a virtual value 6 237 VMSWRC Write FORTRAN Virtual Character Vector Synopsis include fortran h PROMULA FORTRAN function declarations int vmswrc value nval long value Virtual address of characters to be written int nval Number of characters to be written Description Writes a virtual vector of character values to the current output file in accordance with the current format specification Note that in this function each individual character is assumed to have its own corresponding format specification if a formatted write is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowrc Write FORTRAN charac
347. th the constant reductions optimizations turned off void demo static int i j k 1 static float a 10 b 42 c 60 d 120 ac3 l 2 0 2 1 4 1 6 1 3 1 4 1 5 1 4 3 2j d 1 14 1 14 1 1 1 1 4 3 2 3 ati 1 4 b 2 1 j 1 6 5 ct ti 1 2 14 k 1 4 3 6 d i 1 2 14 k 1 1 1 4 3 2 7 Two additional subscripting options are provided for those who want to write their own approach to subscripting the Ysv and Ysf flags The listing below shows the C intermediate output with the Ysv switch active void demo static int i j k 1 static float a 10 b 42 c 60 d 120 aref a 2 0 aref b 1 6 3 1 aref c 2 3 3 4 4 2 aref d 0 2 0 3 0 4 0 3 aref a i 1 4 aref b 1 6 3 1 5 aref c i 1 3 1 4 k 1 6 aref d i 1 2 1 3 k 1 4 1 1 7 In this output a generalized symbol _aref is provided followed by a pointer to the start of the array followed by a sequence of ordered pairs containing the subscript value and the size of the dimension associated with that value The user can then either write a macro in fortran h or a function to perform the actual subscript calculations The Ysf flag uses the same approach as Ysf except that the subscript values are not offset to make them conform to the C convention The following shows the output for Ysf 103 PromulaFortran Translat
348. th the specified storage area as its starting address The size parameter specifies the overall size of the area If there is an open structure currently using this starting address then it is simply rewound 1 An internal file with this starting address has been created If it cannot be found then error has occurred 2 A coded read is to be performed If there is no structure defined then create one which will be removed at the end of the operation this is the standard behavior 3 A binary read is to be performed If there is no structure defined then create one which will be removed at the end of the operation 193 PromulaFortran Translator User s Manual 4 A coded write is to be performed If there is no structure defined then create one which will be removed at the end of the operation this is the standard behavior 5 A binary write is to be performed If there is no structure defined then create one which will be removed at the end of the operation 6 A miscellaneous operation is to be performed If there is no structure defined then an error has occurred 7 A simple inquiry is being made If there is no structure defined then create one which will be removed at the end of the operation Return value A zero if all went well else an error code See the general discussion of the FORTRAN I O capabilities for a listing of the possible error codes See also fioerror fiortxt 6 119 FIOITOS Synop
349. the brute force approach you would probably go through the code maybe all 10000 lines of it and change the references to arrays A and B to some sort of disk reference You would probably also analyze how these arrays are referenced so that you could make the disk accesses as efficient as possible In processing large programs on the limited memory model of the PC DOS platform PROMULA FORTRAN takes a virtual memory approach In PROMULA FORTRAN you can tell the translator that all variables larger than a certain number of bytes should be treated as virtual memory disk variables The runtime library has a very efficient set of virtual memory management routines and the translator replaces references to the specified variables with function calls to the virtual memory functions The subscript calculations are replaced by virtual memory address calculations Thus PROMULA FORTRAN allows you to bring an entire class of FORTRAN programs to the PC which cannot now be processed easily by any other product available 1 8 Dealing with FORTRAN Dialect Problems Other translators deal primarily with FORTRAN 77 require that tokens contain no blanks and be separated and treat FORTRAN statement names as reserved words FORTRAN codes not meeting these specifications must be changed by the user manually Our experience has been that many FORTRAN programs are written in FORTRAN 66 or in some mixed dialect of FORTRAN 66 and FORTRAN 77 Since FORTRAN has no keyw
350. the code associated with those components is not loaded with your program at run time 232 PromulaFortran Translator User s Manual Return value None the function is void See also ftnblkd BLOCK DATA subprogram 6 195 FTNLUN Establish File for Logical Unit Number Synopsis include fortran h PROMULA FORTRAN function declarations define txtfile FILE The type of a text file txtfile ftnlun lun int lun Logical unit number of file Description Before a standard C I O function can access a FORTRAN file it must have access to the stream pointer associated with the logical unit number The logical unit number must be associated with an existing FORTRAN file structure If there is no already existing structure for the unit number then this function will attempt to create one Return value The stream pointer associated with the logical unit number See also fiolun Establish FORTRAN unit number 6 196 FTNOPEN FORTRAN Open Statement Synopsis include fortran h PROMULA FORTRAN function declarations int ftnopen lun int lun Logical unit number Description Executes the FORTRAN OPEN statement when translated via a non optimized user bias In addition to the logical unit number this function takes a variable number of parameters which specify the actual data to be supplied to the open function The parameter type codes passed to this function are as follows Code _ _ Parameter Description of Use
351. the target platform The actual conversion information is stored in a conversion information table which has the following structure Byte Description of content 0 Directory separation character 1 Exclude directories from pathname flag 2 Case conversion code 0 none toupper 2 tolower 3 Other conversions string A Prefix to be added to name length characters 5 Conversions list two Pascal style strings 184 PromulaFortran Translator User s Manual 6 Nul termination byte The conversion information table itself is constructed via function fiorpath Return value None the function is void however the content of the pathname parameter is altered to reflect the conversion See also fiorpath Reads the path conversion information used here fioshl Shift string left fioshr Shift string right 6 106 FIODTOS Convert Double Value to String Synopsis include fortran h PROMULA FORTRAN function declarations void fiodtos val ndig pdecpt psign dspdig double val Value to be converted int ndig Number of digits to produce int pdecpt Returns position of decimal point int psign Returns the sign of the value char dspdig Returns the string produced Description ANSI C expects that all conversions of floating point values to string be performed via the sprintf function Though this can be done most generalized applications prefer to perform their own editing operations and require only a raw conversi
352. this chapter In essence the Yp flag ensures that passed subprogram argument types are consistent As a result of this flag prototypes are formed when subprograms are first referenced or defined When the first reference to a subprogram disagrees with the later definition of that subprogram the Yp switch must enforce the prototype This can be seen in the following FORTRAN code UBROUTINE ALPHA IA RFUNC IA IFUNC GVAL RINT I R OP z UNCTION RFUNC A FUNC ABS A ETURN BEjug 7unmo mnuco in which the function RFUNC is called with an INTEGER argument however when the RFUNC function is later defined that argument is a FLOAT The translation of this code under the Yp switch is shown below void alpha ia gval long ia float gval extern long ifunc extern float rfunc static float r static long i r rfunc ia i ifunc gval WRITE OUTPUT LISTIO INT4 i REAL4 r 0 STOP NULL 65 PromulaFortran Translator User s Manual float rfunc a long a lt static float rfunc rfunc fifiabs a lt return rfunc Note that the argument a is shown and is used as a long to agree with the earlier reference If this behavior is not desired the P16 flag will block the Yp switch from changing defined function argument types The translation of the test code using both Yp and P16 becomes the following void alpha ia gval long
353. tic long ia 2 static long ib 2 static long ic 71 PromulaFortran Translator User s Manual static int ftnsiz 1 1 2 1 1 2 static namelist DATAVAR Mia la b5 ftnsiz ib ib b ftnsizt3 static char DATAVAL SDATAVAR ia hell o ib 2 Wor ld SEND ftnini argc argv NULL fiointu char DATAVAL 0 2 fiornl DATAVAR 2 NULL ic ialpha ia ib lt WRITE OUTPUT LISTIO INT4 ic 0 STOP NULL long ialpha ia gval P1 P2 char ia gval int P1 P2 static long ialpha WRITE OUTPUT LISTIO STRG ia P1 STRG gval P2 0 ialpha P1 P2 return ialpha When the IALPHA function is called PROMULA FORTRAN does not know that that function will expect the lengths of the character vectors One way to solve this problem is to provide an explicit prototype as shown below long ialpha string string Using this prototype to augment the translation produces the following translation void main argc argv int argc har argv extern long ialpha static long ia 2 static long ib 2 static long ic static aint ftnsiz l1 1 2 1 1 2 static namelist DATAVAR Ve la 10 o gt static char DATAVAL SDATAVAR ia hell o ib Wor ld SEND ftnini argc argv NULL fiointu char DATAVAL 0 2 fiornl DATAVAR 2 NULL ic ialpha i
354. tion can be done via the fortran h file which is included with each C output from PROMULA FORTRAN This file is discussed in this chapter as well 5 1 Naming and Organization of Functions The PROMULA FORTRAN runtime library functions are divided into six general groups Associated with each group is a three letter prefix which is part of each function name The groups are as follows Prefix General Description Performs single precision complex arithmetic dpx Performs double precision complex arithmetic fif Noncomplex intrinsic function fio Performs an input output operation ftn Performs a general FORTRAN operation p77 PRIME FORTRAN functions pdp PDP FORTRAN functions vms virtual memory management functions The descriptions of the individual functions are given in alphabetic order in the following chapter The remainder of this chapter gives a summary discussion of the six functions a listing of the runtime error produced a description of the pfclib h header file which is included with every runtime function and finally a description of the fortran h header file which is included by pfclib h and which is also included by each C output file produced by PROMULA FORTRAN The final file needed by every runtime function is the plat form h file This file is used to achieve transportability of the C source code over a wide variety of platforms This header file is also included by the shrouded source code of PROMULA FORTRAN itself It is
355. tion declarations int fiowrf value nval float value Points to values being written int nval Number of values to be written Description Writes a vector of single precision values to the current output file in accordance with the current format specification Note that in this function each individual value is assumed to have its own corresponding format specification if a formatted write is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Do requested error processing fiofout Process output format specification fiowalph Write alphabetic information fiowval Write value 6 168 FIOWRI Write FORTRAN Short Integer Vector Synopsis include fortran h PROMULA FORTRAN function declarations int fiowri value nval short value Points to values to be written int nval Number of values to be written Description Writes a vector of short integer values to the current output file in accordance with the current format specification Note that in this function each individual value is assumed to have its own corresponding format specification if a formatted write is being performed 219 PromulaFortran Translator User s Manual Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a list
356. tion declarations long fifi2pow a b short a Value to be raised to a power short b Power to be used Description Raises a short integer value to a short integer power Return value The short integer result See also None 6 73 FIFIZSHF FORTRAN Intrinsic Function I2ZSHFT Synopsis include fortran h PROMULA FORTRAN function declarations short fifi2shf a n short a Value to be shifted int n Number of positions to be shifted 170 PromulaFortran Translator User s Manual Description Shifts the short argument left or right the specified number of bit positions depending upon whether the position count is positive or negative Return value The short integer result See also None 6 74 FIFIZ2SIGN FORTRAN Intrinsic Function IZ2SIGN Synopsis include fortran h PROMULA FORTRAN function declarations short fifiZ2sign mag sgn short mag The magnitude for the result short sgn The sign for the result Description Returns a value after the transfer of a sign The result is if sgn is at least zero else it is mag Return value The value as computed above See also None 6 75 FIFIABS FORTRAN Intrinsic Function IABS Synopsis include fortran h PROMULA FORTRAN function declarations long fifiabs long a Value whose absolute value is needed Description Computes the absolute value of a long integer argument Return value The long integer result See also None 6 76
357. tion takes no responsibility for ensuring that no block boundaries are encountered downstream from the specified byte This function assumes that the calling function intends only to reference the information and not to change it The parameter is as follows Name Description of use 253 PromulaFortran Translator User s Manual ioiad The linear address for which a pointer is desired Return value A pointer to the byte at the requested linear address See also vmsptr Obtain Virtual Byte Pointer 6 235 VMSVECT Virtual Vector Input Output Synopsis include fortran h PROMULA FORTRAN function declarations void vmsvect fd vec nbyte ioiad iop int fd Handle of virtual file char vec Vector being read or written long nbyte Number of bytes to be processed long ioiad Virtual address of values int iop Operation code 1 write 0 read Description Reads or rewrites a vector of bytes for a specified virtual file beginning at a specified address Return value None the function is void See also vmsptr Get virtual byte pointer 6 236 VMSWRB Write FORTRAN Virtual Boolean Vector Synopsis include fortran h PROMULA FORTRAN function declarations int vmswrb value nval long value Virtual address values being written int nval Number of values to be written Description Writes a virtual vector of Boolean values to the current output file in accordance with the current format specification In this con
358. tions This group of functions is named with the prefix fif Many of the noncomplex intrinsic functions are part of the C standard C libraries Insofar as the implementation of the noncomplex intrinsic functions is concerned the following functions are assumed to be part of the library supplied with your C compiler Name Prototype Description of computation acos double acos double arccosine asin double asin double arcsin atan double atan double arctangent atan2 double atan2 double double arctangent of y x cos double cos double cosine cosh double cosh double hyperbolic cosine exp double exp double exponential fabs double fabs double absolute value floor double floor double largest integer less than log double log double natural logarithm log10 double log10 double base 10 logarithm sin double sin double sine sinh double sinh double hyperbolic sine sqrt double sqrt double square root tan double tan double tangent tanh double tan double hyperbolic tangent 135 PromulaFortran Translator User s Manual The functions above are referenced directly by PROMULA FORTRAN in its translations of the intrinsic functions In addition the following functions are used internally in the implementation of the remaining functions Name Prototype Description of computation modf double modf double double Gets components of a value pow double pow double double Raises value to a power 5 5 Virtual Memory System This group of
359. tions int fiowvu c unsigned char c Character value to be written Description Writes a single unsigned character value to the current output file in accordance with the current format specification This function corresponds to the FORTRAN 77 convention in which values and not just l values can be written Return value 227 PromulaFortran Translator User s Manual A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowbiv Write binary values fiowru Write character vector 6 185 FIOWVX Write FORTRAN Complex Value Synopsis finclude fortran h PROMULA FORTRAN function declarations typedef struct float cr The real part of the value float ci The imaginary part of the number complex int fiowvx val complex val Single precision value to be written Description Writes a single precision complex value to the current output file in accordance with the current format specification This function corresponds to the FORTRAN 77 convention in which values and not just l values can be written Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiowbiv Write binary values fiowrf Write single precision vector 6 186 FIOWVZ Write FORTRAN Double Complex Value Syno
360. try Points eene tentent teen tenete 90 2 36 2 Output Form of Parameter Identifiers Y2 esses eene trennen rentre eerte tenerent enne 92 2 37 TREATMENT OF MULTIPLE ASSIGNMENTS XA YA 92 2 38 TREATMENT OF SINGLE STATEMENT NESTING BRACE XB YB cccsessssccececsesenececececeeseuscecececeesenseaeceeececseneaeeeeecs 93 2 39 CONSTANT REDUCTION OPTIMIZATION 94 2 40 CHARACTER OPTIMIZATION SWITCHES 96 2 41 TREATMENT OF FORTRAN D DEBUGGING STATEMENTS esee 96 2 41 1 Treatment of Other Debugging Statements Ydstring eese eren net entrent 97 2 42 USE OF PRINTF STYLE FORMATTING XF YF cccecsccccccecssssaeceeeccesensnaececececsessaaecesececeecesasseseceeseseanaaeseeececsenseseeeees 98 2 43 INITIALIZATION CHECK FOR AUTO VARIABLES Y1 cccsssccccccecesssssecececcesesssececececeesessaecececeeceseaeseeeeeceennsaeeeeees 99 2 44 DO LOOP COUNTER REDUCTION OPTIMIZATION XL 99 2 45 SUBPROGRAM ARGUMENT TYPE CHECKING XP ntes enne ten rena annees 101 2 46 SINGLE PRECISION REAL ARITHMETIC XR 101 2 47 SUBSCRIPT POINTER NOTATION XS YS YSV
361. ual 2 include fortran h 3 void main argc argv 4 int argc 5 char argv 6 7 include TEST1 h 2 8 ifndef TESTI 1 9 1 1 10 Define the needed sizes 1 11 12 include TEST2 h 2 13 ifndef ICF_TEST2 2 14 2 15 Define the common data 2 16 Ef 1 17 define na 1001 1 18 define nb 200L 2 19 endif ICF_TEST2 2 20 extern char 1 2 21 typedef struct 2 22 long ia na ib nb 2 23 Calpha 24 auto Calpha Talpha Calpha Xalpha 25 ftnini argc argv NULL 26 READ INPUT LISTIO DO int na INT4 Talpha gt ia 0 27 WRITE OUTPUT LISTIO DO int nb INT4 Talpha gt ib 0 28 1 29 define ICF_TEST1 1 30 fendif ICF_TEST1 2 31 ifndef ICF_TEST2 2 32 define ICF_TEST2 2 33 fendif ICF TEST2 34 char Xalpha 1200 Note that the writing of the three files is completely intertwined 2 28 Listing File Control PAname PHnumb PNname PWnumb As described earlier the E flags produce various types of listing and reports about the code being translated its content compiled form and its C form The PA PH PN and PW switches determine the characteristics of the file to receive the listing PAname specifies the name of a file to which the listing currently being produced is to be appended If no file with the specified name exists then one is created PNname on the other hand specifies the name of a new file to r
362. uct int vbdel int vblru int vbmru int vbblk int vbhan char vbdat 1024 vmsbtyp void vmswvb vb vmsbtyp vb Description PromulaFortran Translator User s Manual Write FORTRAN character vector Use a virtual value Write a Virtual Block PROMULA FORTRAN function declarations Virtual File System parameters Physically writes the data block associated with a virtual memory block back onto its virtual file Return value None the function is void See also None 259
363. uild a library and wish to simplify the externals structure then this setting gives a very clean and efficient translation We use it often but be careful The final translation of this example is produced by the Ga setting and is shown below void demo extern struct Xalpha float a 10 10 b 5 c 4d Calpha define Talpha Calpha extern struct Xbeta char c1 150 c2 50 Cbeta define Theta Cbeta Talpha a 0 0 Talpha c Talpha b 4 Talpha d ftnscopy Tbeta c1 2 10 Tbeta c2 30 10 NULL return undef Talpha undef 52 PromulaFortran Translator User s Manual Note that we now have a simple pointer to a structure and no actual structures allocated This technique is most typically used when C code is already present in the application or when special techniques are being used to reference the COMMON blocks A final comment on the treatment of common blocks has to do with the need to form multiple symbols from the original COMMON identifier By default PROMULA FORTRAN uses T C and X prefixes These conventions can be easily changed via a configuration file as discussed in the following chapter Remember in establishing these conventions however that many versions of FORTRAN allow local symbols or even COMMON members to have the same identifier as a COMMON block 2 14 1 Overall Alignment Control with Gp Gpc Gps Gpl Gpd In examining the translation of the s
364. uments The configuration file is used with the R option on the pfc command line as follows pfc file name for Rconf cnf The rules for writing directives in configuration files are described in the context of the following sections Suffice to say here that the Rname switch will read configuration information from the files named The Rname switch may occur multiple times to allow the reading of multiple files If there is no extension supplied with the first file an extension of cnf will be assumed If there is no extension supplied for additional files an extension of pro is assumed The reason for this is that typically when multiple files are read the first contains the configuration information while the additional files contain function prototype information It should be noted that configuration files may contain command line switch information thus a common use of a configuration file is to supply commonly used switches independently of the command line 3 1 The Configuration SWITCHES Statement The PROMULA translation process is controlled via many command line switches These switches give the user easy access to the translation process however there are often many needed To simplify the organization of complex groups of switches the configuration file contains a SWITCHES statement Syntax SWITCHES Si SH Where Sl sn are any valid sequence of command line switches entered exactly as they would be entered
365. urth there is virtual storage Virtual storage is actually maintained in a disk file As it is accessed it is paged into memory The advantages of virtual storage are that it is practically unlimited and like static storage has memory Also virtual memory is unique in the fact that it remains viable after the program has completed execution This topic is discussed in the chapter on the PROMULA interface in the FORTRAN Compiler User s Manual The disadvantages of virtual memory are that it is slower than the other memory types and that the code using it tends to be difficult to read Also there are problems associated with utility functions which use both virtual and non virtual pointer arguments Such functions require two versions This topic is discussed in the chapter on configuration prototypes which allow you to deal with this problem The S switch allows you to specify storage threshold values for five different memory types Its settings are as follows SA Specifies that no variables are to be stored as auto This is the default setting for this storage type SAnum Where num is greater than or equal to zero specifies that any variable whose size in bytes is greater than or equal to num but less than any other threshold setting should be stored as an auto variable SS Specifies that no variables are to be stored as static SSnum Where num is greater than or equal to zero specifies that any variable whose size in bytes is greater than or e
366. ut display depends upon the particular specification For all formats the output field consists of blanks if necessary followed by a minus sign if the value is negative or an optional plus sign otherwise For F format this is followed by a string of digits that contains a decimal point representing the magnitude of the value Leading zeros are not provided except for an optional zero immediately to the left of the decimal point if the magnitude of the value is less than one The leading zero also appears if there would otherwise be no digits in the output field For E format this is followed by a zero a decimal point the number of significant digits specified and an exponent of a specified width For G format the display type depends upon whether or not all significant digits can be displayed in F format If they can F format is used if they cannot E format is used For B format business formatting conventions are used See function fiobfout for details For all display types if the number of digits required to represent the value is less than the specified width the unused leftmost portion of the field is filled with blanks If it is greater than the width asterisks are entered instead of the representation Return value None the function is void The global variable iocrec is updated to contain the new field and the variable fionchar is updated to reflect the new character count in the coded communications record See also
367. utput using the Cs switch is as follows void demo val nval weight sq int nval weight float val sq static int Pow static long i static float xbar xbar 0 0 Pow 2 for i 0 i lt nval i xbar valti xbar nval weight sq pow xbar double Pow It is identical with the above except that the cast on the short integer conversion is not shown Be careful this version might produce an incorrect result See the subsection below for a detailed discussion of arithmetic with short integer variables Let us now go to extremes The following is the same output using the CO cast which requests that all casts be explicitly shown void demo val nval weight sq int nval weight float val sq static int Pow static long i static float xbar xbar 0 0 Pow 2 for i 1L i lt long nval i xbar float double xbar double val short i 1 xbar float double xbar double long nval long weight sq float pow double xbar double Pow In all likelihood you would never want to PROMULA FORTRAN in this mode unless you are interested in seeing all of the promoting that actually goes on Notice for example that all floating point calculations are promoted to double and then reduced back to float Note that constants also get promoted thus in the for statement the value of 1 assigned to i is now shown as 1L Finally note that the long cas
368. vailable for several platforms IBM PC Apple Macintosh VAX VMS VAX ULTRIX SUN UNIX IBM AIX 386 UNIX and other UNIX workstations as well as IBM mainframes PROMULA FORTRAN is a portable C program and can be installed via shrouded source code on platforms not listed above provided they support a standard C compiler Run Time PROMULA FORTRAN comes with an extensive runtime library which reproduces in C the full Library functionality of FORTRAN FORTRAN I O complex arithmetic etc The library is available in C source code and can be recompiled with any standard C compiler This means that you can use PROMULA FORTRAN as a cross compiler i e translate your FORTRAN code to C in one platform using PROMULA FORTRAN then compile link and run the translated C code on a second platform using a C compiler Validation The compiler has been tested on several platforms with Version 2 0 of the FORTRAN compiler Validation System from the Federal Software Testing Center and passed the test at the full validation level on all platforms Integration Upgrade your FORTRAN applications by integrating them naturally with GUI libraries and other C based software on new platforms Debugging Extensive error checking is done at three steps of the process a During compilation of the FORTRAN to C by PROMULA FORTRAN b During compilation of the translated C code by the C compiler In principle after a successful Step a above this step yields no comp
369. val unsigned short bool The values to be read int nval The number of values to be read Description Reads a vector of Boolean short logical values from the current input file in accordance with the current format specification Note that in this function each individual Boolean value is assumed to have its own corresponding format specification if a formatted read is being performed Return value A zero if there is no error flag set else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fioerror Do requested error processing fioffld Get next free form field 199 PromulaFortran Translator User s Manual fiofinp Process input format specifications fioralph Read alphabetic information fiorchk Check fixed read field 6 130 FIORDC Read FORTRAN Character Vector Synopsis include fortran h PROMULA FORTRAN function declarations int fiordc c nval char c Points to characters to be read int nval Number of characters to read Description Reads a vector of character values from the current input file in accordance with the current format specification In this implementation signed char is derived from the nonstandard FORTRAN types BYTE or INTEGER 1 Therefore formatting conventions are assumed Note that in this function each individual character is assumed to have its own corresponding format specification if a formatted read is b
370. value as computed above See also None 6 53 FIFDMAX1 FORTRAN Intrinsic Function DMAXI Synopsis include fortran h PROMULA FORTRAN function declarations double fifdmaxl al a2 double al First value to be compared double a2 Second value to be compared Description Determines which of two double precision values is the largest Return value The largest double precision value See also None 6 54 FIFDMIN1 FORTRAN Intrinsic Function DMIN1 Synopsis include fortran h PROMULA FORTRAN function declarations double fifdminl al a2 double al First value to be compared double a2 Second value to be compared Description Determines which of two double precision values is the smallest Return value The smallest double precision value See also None 6 55 FIFDMOD FORTRAN Intrinsic Function Synopsis include fortran h PROMULA FORTRAN function declarations double fifdmod num dem double num The numerator for the calculation double dem The denominator for the calculation Description 163 PromulaFortran Translator User s Manual Computes the value of the remainder of num divided by dem If dem is zero the result is zero For nonzero values the result is calculated as follows _ num num floor dem dem where floor is the standard C floor function Return value The value as computed above See also None 6 56 FIFDNINT FORTRAN Intrinsic Function DNINT Synopsis include fortran h
371. ve and are as follows Fsnum Selects the standard fixed format with an ending column of n The default setting is Fs72 which is that good old format referred to above 48 PromulaFortran Translator User s Manual Ft Selects tab format which comes from the VAX FORTRANS Ff Selects the free form format which is relatively typical of those FORTRANS that accepted terminal input Fv Selects the VS FORTRAN free form format F9 Selects the Fortran 90 format 2 12 Source FORTRAN Integer Type FIs There is variation between FORTRAN compilers as to whether the default type of the INTEGER specification should be INTEGER 2 or INTEGER 4 In fact an interesting aspect of many modern FORTRAN compilers is that the user may specify whether the default integer type is to be a short 16 bit representation or a long 32 bit representation on the command line The FIs and command line switches in PROMULA FORTRAN allow for this specification The FIs specification says that the default integer type is short INTEGER 2 while specifies that it is long INTEGER 4 The default setting for this switch is for the standard FORTRAN dialect 2 13 Gname Name of File Containing Global Symbols The PROMULA application management system and PROMULA FORTRAN can be used in tandem to upgrade existing codes The interface between these two systems truly adds value to existing FORTRAN programs The global symbols file whose default extension is
372. when any conversions need to be performed Normally however it only shows these conversions in the C output if they would not be performed automatically by the C compiler Alternatively the convention that operands of the same type are not promoted can cause trouble for short integer arithmetic since most FORTRAN compilers promote all short integer calculations to long Explicit conversions in C can be forced in any expression with a cast which looks as follows type expression This cast converts the expression to the indicated type The C casting level option allows one to control which casts are forced in the C Both aspects of promotion are controlled via this switch The switch itself may occur more than once on the command line Its individual settings are as follows Setting Meaning CO Specifies that all promotions between operands of different types be forced in the C output Specifies that all promotions between different operands involving any integer types be forced but that conversions between float and double not be forced C2 Specifies that only those conversions between fixed point and floating point be forced but that other conversions not be forced C3 Is the default and specifies that only those casts needed to maintain the integrity of a calculation be maintained Cs Specifies that short integer calculations are to be done using short arithmetic CL Is the default and specifies that short integer calculations
373. when moving from FORTRAN to C This reordering is greatly complicated by the need to retain the content of each include separately Consider the following FORTRAN source along with two include files TEST FOR PROGRAM DEMO INCLUDE TEST1 INC INCLUDE TEST2 INC READ IA PRINT IB END TEST1 INC C Define the needed sizes C PARAMETER NA 100 PARAMETER NB 200 TEST2 INC C Define the common data COMMON ALPHA IA NA IB NB The default translation is shown below void main argc argv int argc char argv Define the needed sizes Define the common data xf define na 1001 define nb 2001 Li Li extern char Xal lpha 75 PromulaFortran Translator User s Manual typedef struct long ia na ib nb Calpha auto Calpha Talpha Calpha ftnini argc argv NULL READ INPUT LISTIO DO int na INT4 Talpha ia 0 undef na undef nb char Xalpha 1200 Xalpha WRITE OUTPUT LISTIO DO int nb INT4 Talpha gt ib 0 In this translation the information from the include files has been incorporated into the translation Using the P16384 flag three output files are created test c TEST1 h and TEST2 h These are shown below test c void main argc argv H The test c file corresponds precisely with the original FORTRAN insofar as the placement and orde
374. with the Dr switch set void main argc argv int argc char argv extern void demo1 static float a 50 static char name 16 static int ftnsiz 1 1 50 1 1 4 static namelist DATAVAR a a 6 ftnsiz name name 16 ftnsiz t3 static char DATAVAL SDATAVAR a 10 0 0 40 1 0 name Fred 2 Mary Joe END ftnini argc argv NULL fiointu char DATAVAL 0 2 fiornl DATAVAR 2 NULL demol STOP NULL void demol a float al static float val static long i static namelist DATAVAR val amp val 6 NULL static char DATAVAL SDATAVAR val 55 6 S END static FIRST 1 if FIRST FIRST 0 fiointu char DATAVAL 0 2 fiornl DATAVAR 1 NULL for i OL i lt 50L i a i a i val return In this translation PROMULA FORTRAN has constructed logic and simple static initializations which allow the variable values to be read into the actual variables at runtime Using this technique the value strings can be kept simple Also any problems having to do with weak typing or hiding characters are easily solved This runtime initialization gives us all the flexibility we need but it does require that extra code be linked with the program and that the value be stored twice Finally the listing below shows the C output when using the Da switch void main
375. xists it is opened if it does not exist then it is created Specifies the access type of the file Only the first character of the st ring parameter is used as follows Char Meaning 5 5 Sequential the file is opened for sequential access D d Direct the file is opened for direct access 186 PromulaFortran Translator User s Manual 4 Specifies the form of the file Only the first character of the st ring parameter is used as follows Char Meaning Ef the records are formatted the records are unformatted 5 Specifies the record size for the file as contained in the parameter ns The parameter string is ignored 6 Specifies the convention for treating blanks in numeric input fields Only the first character of the string parameter is used as follows Char Meaning N NULL blank characters in numeric formatted input fields are ignored except that a field of all blanks is zero Z ZERO Blanks are treated as zeros 7 Specifies that the file is readonly 8 Specifies that the file is to be open for shared access 9 Specifies the record type 10 Specifies the carriage control conventions 11 Specifies a pointer to the associated variable 12 Specifies the maximum number of records on the file 13 Specifies a pointer to receive the number of the unit assigned to the file 14 Specifies the associated buffer address along with its size 15 Specifies a buffer block size Return value None this function is void S
376. y be turned off via a command line switch Note 9 that C has and operators which take advantage of the fact that most computers have increment and decrement operators The translator uses these operators whenever possible for all biases Note 10 that C allows any conditional statement to form a compound statement with a single statement while FORTRAN allows this only for the IF statement In the C bias this compounding is performed whenever possible while in the FORTRAN bias a compound statement is formed only if the source was compound Note 11 that C has operators like z Z etc The use of these operators ensures that the address of the left hand side of the assignment will not be computed more often than is necessary The C bias uses the operators the FORTRAN bias does not Note 12 that the statement labels and CONTINUE statements ending the two DO loops are not needed for any other purpose Unneeded statement labels are always removed 2 3 Arithmetic Conversions CL Cs C0 C1 C2 C3 14 PromulaFortran Translator User s Manual In C if a binary operator like or has operands of different types the lower type is promoted to the higher type automatically by the compiler Alternatively if operands of the same type are combined then the result is of that type As PROMULA FORTRAN processes source code it keeps track of the types of all operands involved in expressions and determines
377. ying values ina FORTRAN DATA statement Return value A zero if the read was successful else an error code See the general discussion of FORTRAN I O capabilities for a listing of the possible error codes See also fiostod Convert string to double fioerror Perform FORTRAN I O error processing fiofend End formatted processing fioralph Read alphabetic information fiospace Skip white space in record fiostoi Convert string to integer fioffld Get next free form field 6 145 FIORPATH Read Pathname Conversion Information Synopsis include fortran h PROMULA FORTRAN function declarations int fiorpath fname pak char fname Name of file containing conversion information unsigned char pak Conversion information table Description Reads a file containing pathname conversion information which establishes a particular pathname translation scheme The actual specification is contained on the file whose name is specified by the fname parameter This specification describes the path and file name conventions to be used on the target platform and how these conventions are to be obtained from the 207 PromulaFortran Translator User s Manual source pathname specifications The approach taken is to describe how source pathnames are to be translated into target pathnames Syntax PATHNAMES dirchar REPLACE slt1s2t2 LOWER UPPER PREFIX tname EXCLUDE
378. ze of the storage area for this compacted statement is 5120 characters This is sufficient for a statement with over 70 continuation lines If extremely long statements are being processed or if memory is at a premium the size of this area may be changed via the QInumb command line switch Note that if the size of the compacted statement storage area is exceeded the fatal error 522 Statement contains more than nnnnn characters use the QIn flag to increase this value 18 generated 2 29 2 QEnumb Size of the Line Number Table When line number information is being generated in the C output via the Ln or DB command line switches this switch is used to indicate the amount of storage to be allocated for this information When Ln is used this switch is required when DB is used the default value established for this switch is 32000 To change this default for DB the QEnumb switch must follow the DB switch of the command line Note that if the size of the line number information table is exceeded the fatal error 532 The line number storage of nnnnn bytes is insufficient use the QEn flag to increase it 18 generated 2 29 3 QDnumb Size of a Data Block When PROMULA encounters data initializations it must allocate memory equal to the size of the variable to store these initializations The values for all variables less than a data block threshold value are allocated into fixed length blocks using PROMULA s standard internal vari
379. zero of type REAL 16 zero of type DOUBLE COMPLEX zero of type COMPLEX int keyword NULL keyword initializes runtime data processing performs runtime data processing union keyword sizeof keyword i o list designator for INTEGER 2 i o list designator for REAL 8 i o list designator for LOGICAL 2 i o list designator for INTEGER 1 i o list designator for INTEGER 4 i o list designator for REAL 4 i o list designator for LOGICAL 4 i o list designator for LOGICAL 1 i o list designator for DOUBLE COMPLEX i o list designator for CHARACTER o list designator for implied DO i o list designator for constant string printf for free form INTEGER 2 printf for free form REAL 8 printf for free form LOGICAL 2 printf for free form INTEGER 1 printf for free form INTEGER 4 printf for free form REAL 4 printf for free form LOGICAL 4 printf for free form LOGICAL 1 printf for free form REAL 16 53 25 15E 25 15E 35 16 6E 16 6E 0 CMPLX Inline include fortran h ftnblkd REAL8 printf for free form DOUBLE COMPLEX printf for free form COMPLEX printf for free form CHARACTER i o list designator for COMPLEX inline function designator include request for fortran h name of BLOCK DATA subprogram i o list designator for REAL 8 117 Idn 164 165 166 169 170 171 211 212 213 214 259 260 261 270 271 272 273 Keyword BD VOID define LPROTOTYPE volatile NULL 0 7 ftnstruc ftnfdata
380. zt6 6 NULL j amp j 5 NULL static char DATAVAL SDATAVAR b 5 4 4 0 0 x 1 0 2 0 3 0 4 0 5 0 0 0 2 0 3 0 4 0 5 0 2 0 0 3 0 4 0 5 0 3 0 0 4 0 5 0 4 0 0 5 0 c 10 2 0 a 4 2 5 4 SEND static FIRST 1 static char F1 1x 10f10 1 static char F2 Ix 110 if FIRST FIRST 0 fiointu char DATAVAL 0 2 fiornl DATAVAR 5 NULL WRITE OUTPUT F F1 1 REAL4 a D0O 5 REAL4 b 0 WRITE F1 1 D0 10 REAL4 c 0 WRITE OUTPUT F F2 1 INT4 35 0 for j 1L j lt 5L j WRITE OUTPUT FMT F1 1 MORE for i OL i lt 5L i WRITE REAL4 x i j 1 MORE WRITE 0 STOP NULL See the discussion of NAMELIST in the FORTRAN Compiler User s Manual for more information on namelist In essence to do the data initialization at runtime PROMULA FORTRAN invents a NAMELIST which contains all of the variables involved in data initializations This is done in the initialization of ftnsiz which contains the array bound specifications and in the initialization of the structure DATAVAR which contains the namelist specification Next the data quantity lists and values are converted into namelist read input format and stored in an internal character file called DATAVAL Finally a variable FIRST is introduced which triggers the calls to the runtime processing

PROMULA FORTRAN to C Translator User`s

Contents

Download Pdf Manuals

Related Search

Related Contents