Trailing-Edge
-
PDP-10 Archives
-
decuslib20-03
-
decus/20-0078/maint/tdii.rno
There is 1 other file named tdii.rno in the archive. Click here to see a list.
.sp 1;.nonum;.lm 9;.ps 50,72
.ts 1,9,17,25,33,41,49,57,65
.pg;.nm ch 1
.ch SIMULA FOR DEC SYSTEM 10#############TD, COMPILER
.nm 1
.st 741021####780302######6###########################Stefan Arnborg
.pg;.lm 9;.p -8,0,10;II######THE COMPILER
.br;------------
.s;II.1####GENERAL COMPILER DESCRIPTION
.S;The compiler consists of three passes, each consisting of a
reentrant high segment and a layout of the low segment. The passes
are referred to as pass 1, pass 2 and pass 3. Their high segment
names are SIMULA.EXE, SIMP2.EXE and SIMP3.EXE. When a program is
compiled, it is processed first by pass 1, then by pass 2 and
finally by pass 3. If several programs or external modules are
compiled, the process is repeated for each source module. The passes
communicate by means of data in the low segment (II.3) and
intermediate files (II.4).
.s;The first pass reads the source, assembles characters to basic symbols
(special symbols, constants, keywords and identifiers), removes comments,
transforms the program to reverse Polish and processes declarations.
Declaration processing includes retrieval of declarations from .ATR files
for external modules and allocation and sorting of declarations.
The second pass binds identifiers to declarations, checks operand/operator
correspondence and generates code.
.s;The third pass edits the code generated to a format acceptable to LINK-10,
edits listings and types error messages.
.nm ch 2
.ch SIMULA FOR DEC SYSTEM 10#############TD, COMPILER
.st 741111####780302######6###########################Olof Bjorner
.nm 1.pg;.lm 9;.i -8;II.2####SYSTEM DECOMPOSITION
.nm ch 2
.S;The decomposition of each pass is described in II.A.2
(pass 1), II.B.2 (pass 2) and II.C.2 (pass 3).
.sp 1;.nonum;.lm 9;.ps 50,72
.ts 1,9,17,25,33,41,49,57,65
.pg;.nm ch 3
.ch SIMULA FOR DEC SYSTEM 10#############TD, COMPILER
.st 741106####780302######6###########################Claes Wihlborg
.nm 1;.pg;.lm 9;.i -8;II.3####LOW SEGMENT ORGANIZATION
.S;The layout of the compiler low segment is:
.nf;.s 2
_^ _^ _^
: : :
:--------: : :--------:
! ! : ! !
! +--------+ !
! PASS 1 ! PASS 2 ! PASS 3 !
! DATA ! DATA ! !
!--------+--------+--------! ca 7100 octal
! !
! GLOBAL DATA !
!--------------------------! 140 octal
! JOBDATA AREA !
+--------------------------+ 20 octal
.s;Pass 1 data is documented in II.A.3
Pass 2 data is documented in II.B.3
Pass 3 data is documented in II.C.3
.f;.s;The global part contains 3 large areas and some smaller areas
(mostly single words). The sizes are (octal radix and production version):
.nf;.lm 17;.p 0,1,6
I/O buffers ca 5100
Stack (YSTK) 1200
Diagnostic messages buffer (ZDM) 230
Miscellaneous, in all ca 120
.f;.lm 9;.p 0,1,4;The I/O buffers, YSTK and ZDM are documented here,
the other areas in the file LOWSEG.MAC.
.p -8,2,8;II.3.1##I/O buffer organisation
.br;-----------------------
.s;There is an I/O buffering scheme designed to minimise the number of
intermediate files written on disk. The scheme works as follows.
.br;The command file and the REL file (and debug output in debug version)
have fixed buffer space allocated and they will not be further considered
here.
.br;The other files have a common pool of 16 buffers which may be
used as buffer rings for I/O or as storage for the entire file contents.
To simplify the description, the I/O activities during a compilation
have been divided into 8 phases:
.nf;.p 0,1,15;
I/O phase Input ! Output
+-- !
! 1 Source ! LS1, XRF, IC1
Pass 1 < 2 Attributes ! DF1
! 3 ! ZSE
+-- !
Pass 2 < 4 IC1, DF1 ! IC2
! ! ATR
+-- !
! 6 ZSE, ATR !
Pass 3 < 7 IC2, Attributes ! ATR
! 8 Source, LS1, XRF! LST
+-- !
.f;.p 0,1,5;The initial setup (for phase 1) is: The LS1 file is given
3 buffers, the XRF file 2 buffers, the IC1 file 5 and the source 2
buffers leaving 2 free buffers.
.lm 9;.p 0,1,5;.nf;
Buffer
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
+----------------------------------------------+
! LS1 ! XRF ! IC1 ! Source ! Free!
+----------------------------------------------+
.f;.s;The files LS1, XRF and IC1 are initially kept in core in their
allocated space. Whenever one of these files exceeds its qouta, it
gets the free buffer space as a buffer ring and is output on disk
from then on. Its initial space is thus freed and will be used as
free space for the next file to overflow, if any. This is governed
by the YBRBUF pointer:
.tp 4;.nf
+---------------------------------------------+
YBRBUF: ! number of free buffers ! first free buffer !
+---------------------------------------------+
.f;At the end of phase 1, all files still kept in core are packed
together to avoid memory fragmentation. The pointers YBRSRC and YBRZSE
are set.
.nf;.p 0,1,10
# 1 2 3 4 5
+-------------------
! LS1 ! XRF ! IC1 !
+---------------------
# _^ _^
YBRSRC -----+ !
YBRZSE -----------+
.f;.s;During phase 2 the file DF1 is written. The space from
(YBRZSE) up to the last 4 buffers (the last 2 buffers if IC1 is still
in core) is used for file storage. If this proves sufficient, YBRZSE
is changed, otherwise the last two buffers (15 and 16) are set up
as a buffer ring and the DF1 file is output to disk.
As there may not exist any free buffers when DF1 is being written,
and external ATR files may be input simultaneously, the ATR files
get buffer space in the pass 1 YDPD structure which is no longer in
use.
.s;During phase 3 the file ZSE is written. The space from (YBRZSE)
up to the last 6 buffers (last 4 if one of IC1 or DF1 still in core
and last 2 if both in core) is computed. As the size of the ZSE file
is known before the file is written, it is also known if it fits into
the free space. If so, the file is put there, otherwise the last two
buffers are used as a buffer ring. Finally YBRZSE and YBRBUF are set
(note that from now on YBRBUF has another meaning).
The situation at the end of phase 3 is:
.nf;.p 0,1,10
# 1 2 3 4 5 6 7 8
+----------------------------------
! LS1 ! XRF ! IC1 ! DF1 ! ZSE !
+---------------------------------------
# _^ _^
YBRSRC -----+ !
YBRBUF, YBRZSE ---------------+
.f;.s;During phase 4, the files IC1 and DF1 are read and the
file IC2 is written. There is no attempt to keep IC2 in core. The
files which reside on disk get a buffer ring of two buffers each.
This is always possible due to the maximum sizes imposed on the
files in the earlier phases. During the allocation of the buffer
rings, YBRBUF acts as a free area pointer. When phase 4 is terminated,
the situation is like this:
.nf;.p 0,1,10
+----------------------------------
! LS1 ! XRF ! Garbage ! ZSE !
+---------------------------------------
# _^ _^
YBRSRC -----+ YBRZSE --------+
.f;.p 0,1,5
During phase 5 the ATR file is written. The space from (YBRZSE) up
to the last 2 buffers is used as file storage. If this is insufficient
the last 2 buffers are used as a buffer ring and the ATR file is output
on disk.
.p 0,1,5;During pase 6 the files ZSE and ATR are read into local pass
3 areas. If they are on disk, they are input in unbuffered dump mode.
Thus no buffers are allocated in this phase.
At the end of phase 6, the situation is:
.nf;.p 0,1,6
+---------------------
! LS1 ! XRF !
+-----------------------
# _^
YBRSRC -----+
.f;.p 0,1,6;During the last two phases each file which is used and
not kept in core will have a buffer ring of two buffers allocated,
except for the ATR file which is written in unbuffered mode.
YBRBUF is used to keep track of file buffer usage and is reset
from YBRSRC at the start of each phase.
.p -8,2,6;II.3.2##The push-down stack (YSTK)
.br;--------------------------
.s;The push-down stack is initialised to detect overflow by hardware
(push-down pointer counter negative) and underflow is trapped by
making a few words at the stack bottom point to an error routine
(different routine in different passes).
.p -8,2,5;II.3.3##The diagnostic messages buffer (ZDM)
.br;------------------------------------
.s;This buffer has space for 50 messages. Each message uses 3 words.
.nf;.p 0,1,12;
0 17 18 35
+------------+------+------+-------------+
! error number ! first line to print!
!------------+------+------+-------------!
! semicol. of line ! last line to print!
+------------+------+------+-------------!
! I1 ! I2 ! I3 !
+------------+-------------+-------------+
#0 11 12 23 24 35
.f;.p 0,1,5;.lm 13;
Explanation:
.p -4,0,3;Word 2: Bits 0-17 Number of semicolons before faulty statement
on first line to be printed.
.p -4,0,6;Word 3: The contents depend on the message, either:
.p -4,0,3;1) ASCII text, 5 characters or less left justified.
.p -4,0,3;2) I1, I2 and I3 are 12-bit numbers which are interpreted as:
.p -4,0,3;2a) A positive integer to be output in decimal ASCII.
.p -4,0,3;2b) Value < 1024: edited as a symbol defined via the SYMBOL macro
in SIMMAC.MAC.
.p -4,0,3;2c) Value >= 1024: Edit as identifier from dictionary.
.lm 9
.nm ch 4
.ch SIMULA FOR DEC SYSTEM 10#############TD, COMPILER
.st 740814####780302######6###########################Reidar Karlsson
.nm 1;.pg;.nf;II.4 File descriptions
.s;II.4.1 SIM command file
II.4.2 Source code file
II.4.3 Attribute file
II.4.4 LS1 List control file
II.4.5 IC1 Intermediate code from pass 1
II.4.6 DF1 Declaration file
II.4.7 ZSE Symbol table
II.4.8 Relocatable binary object code
II.4.9 ATR Intermediate attribute file
II.4.10 XRF Cross reference data file
II.4.11 IC2 Intermediate code from pass 2
II.4.12 Source listing file
II.4.13 SIMERR Error message tables
II.4.14 DEB Debug files
.nf;.p 0,1,25;.ps 55,75
+------------------------------------------------------------------------+
! FILE NAME ! FILE CONTENTS ! W ! R ! CHANNEL! REMARKS !
!------------------------------------------------------------------------!
! nnnSIM.TMP ! Command file ! - ! 1 ! QCHCOM ! only via CCL entry !
! infile.SIM ! Source file ! - !1,3! QCHSRC ! !
! extern.ATR ! Attribute file ! 3 !1,3! QCHEXT ! !
! nnnLS1.TMP ! List control file ! 1 ! 3 ! QCHLS1 ! !
! nnnIC1.TMP ! Intermediate code 1 ! 1 ! 2 ! QCHIC1 ! !
! nnnDF1.TMP ! Declaration file ! 1 ! 2 ! QCHDF1 ! !
! nnnZSE.TMP ! Symbol table ! 1 ! 3 ! QCHZSE ! !
! relfil.ext ! Binary object code !ALL! - ! QCHREL ! name from command !
! nnnATR.TMP ! Intermediate attr. ! 2 ! 3 ! QCHATR ! !
! nnnXRF.TMP ! Cross ref. data ! 1 ! 3 ! QCHXRF ! !
! nnnIC2.TMP ! Intermediate code 2 ! 2 ! 3 ! QCHIC2 ! !
! lstfil.ext ! Source listing ! 3 ! - ! QCHLS3 ! name from command !
! SIMERR.ERR ! Error messages ! - ! 3 ! QCHERR ! created by SUTERR !
! nnnDEB.LST ! Debug output pass 1 ! 1 ! - ! QCHDEB ! only debug version !
! nnnDEB.TMP ! Debug output pass 2 !1,2! - ! QCHDEB ! only debug version !
+------------------------------------------------------------------------+
.s;Table II.4.1 Compiler files.
################nnn = job number. W written in pass. R = read in pass.
.f;.ps 50,72;.pg;.lm 9;.p -8,0,10;II.4####FILE DESCRIPTIONS
.s;All temporary compiler files follow the TOPS-10 naming
conventions. Their names have the following form:
.br;####nnnXXX.TMP
.br;where nnn is the decimal job number with leading zeroes, XXX is
a mnemonic for the use of the file.
.br;Two new file name extensions (types) are introduced with the SIMULA
system: SIM for SIMULA language source ASCII files and ATR for binary
attribute files created and read by the SIMULA compiler.
.p 0,2,10;All files used within the compiler are listed in table II.4.1.
The temporary files LS1, IC1, DF1, ZSE, ATR and XRF may either be kept in
core or output on disk. Space for 16 disk buffers of standard size
is allocated in the low segment for these files. See low segment organization
(II.3). The files LS1, IC1 and XRF which are created at the same time
during pass 1 processing, will stay in core unless they exceed 3,
5 and 2 buffers respectively. The files DF1, ZSE and ATR are created
consecutively in that order, and if they fit in in the free buffer
area, they will be kept in core.
The other files listed in table II.4.1 will always be on disk.
If an I/O error occurs in pass 1, two things can happen.
A LOOKUP error on either the source file or the attribute file
will cause an error message on the TTY stating that the file was not
found. A new command to the compiler is expected.
Any other I/O error in pass 1 and any I/O error in pass 2 will generate
a terminating error message and pass 3 will be called immediately.
If a terminating error occurs in pass 3, an attempt is made to edit
the error messages found so far. If more than one terminating error
is detected (e.g_. failure to read the error message tables from SIMERR.ERR),
the message "?UNRECOVERABLE ERRORS" will be typed on the TTY and the
next command is processed or the compilation is terminated by an EXIT
UUO if the compiler commands came from COMPIL.
.p -8,2,15;II.4.1##SIM command file
.br;----------------
.lm 17;.p -8,1,1;FILE:###nnnSIM.TMP
.p -8,1,1;PURPOSE:When the SIMULA compiler is called via the CCL entry,
an attempt is made to read the command string from a TMPCOR file named
SIM. If that fails, the file DSK:nnnSIM.TMP is read if found.
.s;If the SIMULA compiler is invoked directly by R SIMULA, commands
are read from the TTY. No command file is involved in that case.
If the TMPCOR file SIM is present, it is copied and deleted by I1CM.
.p -8,1,4;OPEN:###OPEN and LOOKUP on channel QCHCOM in buffered ASCII
line mode (.IOASL). Done in I1CM.
.p -8,1,1;INPUT:##The file is read in I1MOVE.
.p -8,1,3;DELETE:#The file is finally deleted in T3 when all commands
have been processed.
.p -8,1,10;RECORD LAYOUT: The data in the file is separated into lines,
each line taken as a command to the SIMULA compiler. The first null
character is taken as end of file. The command format is similar to
that accepted by other compilers used under TOPS-10:
.br;##RELFILE [,LISTFILE] = SOURCE-1 [,SOURCE-2]...
.br;The general form of each file specification is:
.br;##DEV:FILE.EXT[PATH]
.br;Compiler switches may appear after LISTFILE as /SWITCH
and after SOURCE-files as (SSSS...), where S is the short one letter form
(-S also allowed).
.s;Default for DEV is DSK, defaults for EXT are REL, LST and SIM resp.
PATH is the users default path (usually the logged in UFD) unless
explicitly given.
.p 0,1,5;The last line in a command file may be of the form
.br;##filespec!
.br;where filespec should be the name of a program (default device
is SYS:). The program will be RUN with run offset = 1 (CCL entry).
.p 0,1,6;A few examples follow:
.s;##.COMPILE prog.sim
.br;will give the command file:
.s;##PROG=PROG.SIM
.s;##.LOAD prog1+prog2/CREF
.BR;=>PROG2,PROG2/C=PROG1.SIM,PROG2.SIM
.br;##LINK!
.s;##.EXECUTE/LIST prog1,prog2/NOLIST(-W)
.br;=>PROG1,LPT:PROG1=PROG1.SIM
.BR;##PROG2=PROG2.SIM(-W)
.BR;##LINK!
.P 0,1,10
It is assumed in the examples that prog.SIM, prog1.SIM and prog2.SIM
exist on the user's default area on the disk structure. If another
file with the same name but with a file extension different from
SIM exists, another compiler may be invoked if the extension is not
explicitly given in the COMPILE (LOAD, EXECUTE) command.
.pg;.lm 17;.p -16,0,1;II.4.2##Source code file
.br;.i -8;----------------
.p -8,1,5;FILE:###Input file specification of the form DEV:FILE.EXT[PATH].
Defaults: DEV=DSK, EXT=SIM, PATH=default directory, usually the logged in
directory (UFD).
.p -8,1,3;PURPOSE:The SIMULA program to be compiled must exist as one or
more disk files.
.p -8,1,4;OPEN:###The file is opened on channel QCHSRC in buffered binary mode
(.IOBIN) in the LC (line control) module of pass 1. The file is looked up in
O1SCOP which is called from LC.
.p -8,1,3;INPUT:##Read in pass 1 by LC via O1SC.
.p -8,1,2;CLOSE:##Closed in O1SCCL which is called from T1.
.p -8,1,4;OPEN:###Opened again in pass 3 on the same channel, but
in buffered ASCII mode (.IOASC), in I3S, which is called from E3LICS.
.p -8,1,4;INPUT:##Normally read in pass 3 via O3RSC called from E3IN.
If an illegal end of program is found, the rest of the source file
will be input at O3SCLS.
.p -8,1,1;CLOSE:##File closed, channel released in T3L, called from
E3.
.p -8,1,10;RECORD LAYOUT: A DECsystem-10/20 SIMULA program is a sequence
of symbols from the DECsystem-10 character set (ASCII 7-bit code).
The data in the file is separated into lines. A line is delimited
by a line feed (LF), vertical tab (VT) or form feed (FF) character.
A line may contain at most 135 characters. For further information,
see DECsystem-10 SIMULA Language Handbook.
.st 740814####780302######6###########################R Karlsson/L Enderin
.lm 17;.p -16,2,15;II.4.3##Attribute file
.p -8,0,0;--------------
.p -8,1,5;FILE:###<E-name>.ATR
.p 0,0,0;Normally, <E-name> == <I-name>,
the name of the external class or procedure
(first 6 characters, possibly with trailing blanks).
.p -8,1,4;PURPOSE:The attribute file contains declarative information about
a separately compiled class or procedure.
.p -8,2,5;PASS1 PROCESSING
.nf;.lm -8;.p 0,1,5;Definitions: The general external declaration has the form:
##<E-header> <E-item> [,<E-item>]...
##where
# <E-header> ::= EXTERNAL ([<type>] PROCEDURE ! CLASS)
# <E-item> ::= <I-name> _!
# <I-name> = [<dev>:] <E-name> [<ppn>]
.j.f;If <E-item> has no '=' sign, <E-name> = <I-name>.
.lm 17;.p -8,2,5;OPEN:###For each external item (<E-item>) found in the source file,
the file <E-name>.ATR will be looked up on channel QCHEXT in synchronous
buffered binary mode (.IOBIN, with IO.SYN bit set).
The lookup is done by O1EXLU, called from O1EXOP, after O1EX.O has been called
to initialize channel QCHEXT.
Each ATR file which is found is processed as shown below.
If the file is not found directly, a further search can normally be done, unless
the "=" convention was used to specify an explicit file.
.s;If all ATR files have not been found by direct lookup, they may still
be found in library files.
The next file to be looked up is defined by the /SEARCH switch(es) given
to the compiler via OPTIONS statements or via the command string.
See [LH2], Section 7.1.1.
Several files can be defined, and they are looked up in an order which is the
reverse of the order of definition.
.br;If no file is found in this way, an error message is issued.
.br;Otherwise, it is checked whether the file found is a library file or a
separate attribute file. The first word of a library file must contain
XWD#14,177, signifying an index block. The index block contains a table of
entry names. Each name in that table is checked against the list of
external items not yet resolved.The SIMULA name, <I-name>, is used for
the comparison.
The index blocks form a list within the file. Each index block is searched
in this way until no more index blocks exist or the list of external
items is exhausted.
.br;If more items remain, the next file on the search list is searched,
etc.
Once a match occurs,the first word of the ATR module corresponding to
<I-name> is found via the table and read in from the buffer.
.s;The ATR file is now ready to be processed.
If the file is NOT a library file,
it is assumed to correspond exactly to the external item.
The start of the ATR module depends on the version of SIMULA which produced it.
The old version could not be included in a library because no entry or name
block was included. The current ATR files are made to look like REL files in
order to allow manipulation by FUDGE2 or MAKLIB.
.br;Before returning to DPEXT, O1EXOP scans the file
up to the first word of the actual
ATR information.
.p -8,1,5;INPUT:##Actual input of blocks from the file is done by O1EXT,
called from DPEXT.
.p -8,2,2;CLOSE:##The file is closed by O1EXCL.
.p -8,3,10;PASS 3 PROCESSING
.s;.lm 9;When compiling a procedure or class separately, pass 3
looks up any old ATR file with the same SIMULA name.
Pass 1 informs pass 3 of the whereabouts of this file via global
data: YATRDEV:YATRFN[YATRPPN], YATROFS(=word offset in library file).
(In TOPS-20 SIMULA, YATRJFN is passed on as well as the size YATRSZ).
.br;If the file is found, its contents are compared to the new information
in the intermediate ATR file generated in Pass 2.
.lm 17;.p -8,2,5;OPEN:###The old ATR file is looked up on channel QCHEXT in
buffered binary synchronous mode (.IOBIN).
I3E does the OPEN and calls I3ELO to look up the file.
If LIBSIM.ATR is looked up, I3FI finds the submodule within
the ATR library just like O1EXFI does in Pass 1.
Once the (sub)module is found, it is scanned up to the first word of ATR info
and is ready for processing.
If no old ATR file is found in this way, a new file is to be created.
.p -8,1,5;INPUT:##By O3RA via M3ATR.
.p -8,1,5;CLOSE:##If the attribute file is similar to the new ATR information
given by the intermediate ATR file, it is closed and the channel is released.
If differences are found, the old file is deleted and a new file is written in
its place.
.p -8,1,2;OUTPUT:#O3WATR writes the new ATR file if any. The file is
written in dump mode (.IODPR=16).
.p -8,3,10;RECORD LAYOUT:
.nf;.s;The ATR file consists of 3 parts:
.p 0,1,6;1) Entry and name blocks as in the REL file:
.p 0,1,5;1a) ENTRY block (type 4):
.s;###XWD 4,1
.br;###EXP 0 ;relocation bits
.br;###RADIX50 0,/<I-name>/
.f;.s;<I-name> is the SIMULA name of the class or
procedure.
This information, excluding the zero relocation word,
is placed in the index block of a library file and is used
when the submodule is looked up.
.p 0,1,8;1b) NAME block (type 6):
.s;###XWD 6,1
.br;###EXP 0
.br;###RADIX50 0,/<E-name>/
.s;<E-name> is determined from the REL file name (<E-name>.REL) given
in the command string to the compiler.
Assuming default device, extensions, and path, the command string is
.br;##<E-name>,<L-name>=<S-name>
.br;i.e_. the file <S-name>.SIM is compiled to give the REL file <E-name>.REL,
the ATR file <E-name>.ATR, the list file <L-name>.LST.
.p 0,1,10;2) ATR information block:
.p 0,1,0;This is disguised as a REL file comment block (type 0):
.s;###XWD 0,M
.br;####<N words of ATR information>
.s;The word count, M, is computed from the actual count N by regarding every
18th word as a relocation word which is not included in M.
This is because LINK-10, LOADER, FUDGE2 and MAKLIB handle REL files in this
fashion.
A dummy zero word may sometimes have to be added to the end of the block
to make the count come out correctly.
.br;The ATR information is:
.p 0,1,5;2a) Heading:
.s;One word, zero for a SIMULA procedure, the entry name in
RADIX50 for MACRO-10 or FORTRAN procedures, and a unique identifier with
a '%' as third character for a class.
.p 0,1,5;2b) Attributes:
.s;<A-list>::=ZQU[ZHB[<A-list>]...<zeroword>]
.s;The attributes form an A-list. See II.6 for the definition of ZQU and ZHB.
.p 0,1,5;2c) Information for checking:
.s;A list of ZHE(QQUACH) records (see II.6)
is created, where each record corresponds to an external item referenced by this
module. The list ends with a zero word.
.p 0,2,10;3) END block (type 5):
.s;###XWD 5,2
.br;###EXP 0,0,0
.p 0,1,0;This delimits the ATR module within a library file.
.p 0,1,5;In the old record layout, only part 2 was present.
The format was changed to allow creation of library files by FUDGE2 or MAKLIB,
which treat only REL files.
A SIMULA ATR file thus looks like a REL file with an ENTRY block, a NAME block,
a COMMENT block and an END block.
.p -8,3,20;RECORD LAYOUT for library ATR files:
.p 0,2,0;An ATR file with the layout described above may be made part of
an ATR library file. FUDGE2 or MAKLIB can be used to create a library out of
separate ATR files or other libraries. The format is that of an indexed
REL file library:
.p 0,1,10;1) The first block is an index block (type 14), with the structure:
.p 4,1,0;XWD 14,177
.p 4,0,0;<index table>
.p 4,0,0;XWD -1,next
.p 0,1,0;next is the number of the next index block in the file or -1 if no more
index blocks exist.
.p 0,1,10;<index table> consists of a series of <moduleitem>'s, where
<moduleitem> is:
.p 4,1,6;XWD 4,1
.p 4,0,0;RADIX50 0,/<I-name>/
.p 4,0,0;XWD W,B
.s;W is the word offset within block B of the file where the submodule
corresponding to <I-name> starts.
.p 0,2,10;The second block and following blocks (except for inserted index
blocks) contain ATR file information copied into contiguous blocks of words.
If the first block has a pointer to another index block, that index block
is placed before the first submodule referred by an entry name in it.
The second and following index blocks are similar. The chain of index blocks
is finished by a block whose last non-zero word contains -1.
.p -8,2,10;INPUT of a library ATR file:
.p 0,1,0;Initially, 3 buffers are allocated. When the first block is recognized
as an index block, it is disconnected from the buffer ring and 2 buffers are
left to form a new ring. Following the lookup of an entry in the
index block, a number of disk blocks are skipped to find the start of the
module or the next index block if lookup was not successful.
Only one index block is retained, placed in the first buffer of the
initial 3 buffers. Since the file is read in synchronous mode,
stopping after each bufferful, USETI can be used to locate the
correct block before issuing the IN UUO.
.p 0,2,1;Otherwise, an ATR library is handled like a separate ATR file.
.lm 17;.p -16,2,15;II.4.4##LS1 List control file
.i -8;.br;---------------------
.p -8,1,1;FILE:##nnnLS1.TMP
.p -8,1,4;PURPOSE:The LS1 file is written by pass 1 and read by pass 3. It
contains information needed by pass 3 to edit the source file listing.
.p -8,1,1;OPEN:
.p -8,0,1;OUTPUT:#If the file size does not exceed 3 disk buffers,
it will stay in core between passes. When the limit is exceeded, O1LS1
is called and opens a disk file on channel QCHLS1 in buffered binary
mode (.IOBIN) at O1OPLS.
.p -8,1,1;CLOSE:##Closed in O1LSCL called from T1.
.p -8,1,4;REOPEN:#If the file is not core resident, it is looked up
in I3L (pass 3) and then input by calls to O3RS during editing by E3.
At the end of E3, the file will be deleted by a call to T3L.
.p -8,1,10;RECORD LAYOUT: The LS1 file contains two types of records,
control records and text records. Control records are a full word
or a halfword and are distinguished from text records by a 1 in the
rightmost bit (bit 17 or 35).
.nf;.br;The following control records exist:
.s;Type 1 Not used
Type 2 BEGIN record
Type 3 END record
Type 4 Line record
Type 5 Not used
Type 6 PAGE record
Type 7 LIST record
Type 8 NOERR record
Type 9 Source identification record
.f;.s;Records of types 4 and 9 are created by the LC module and all other types
by I1.
The LCLS1 subroutine edits control records. Text records are edited
by the LCLS1T subroutine.
.p 0,1,7;Control record formats:
.s;The first seven bits of a control record contain a letter in ASCII
code identifying the record type, except for types 6, 7 and 8 (see
below).
.nf;.p 0,1,10;Type 2 BEGIN record:
.s
0 6 17 34 35
+-------+-----------+-----------------+-+
! 'B' ! ! !1!
+-------+-----------+-----------------+-+
.s;Bits 0-6 = ASCII 'B'.
Created by the LC module.
.nf;.p 0,1,10;Type 3 END record:
.s
0 6 17 34 35
+-------+-----------+-----------------+-+
! 'E' ! ! !1!
+-------+-----------+-----------------+-+
.s;Bits 0-6 = ASCII 'E'.
Created by the LS module.
.nf;.p 0,1,10;Type 4 Line record:
.s
0 6 7 8 9 16 17
+-------+-+-+--------+-+
! 'I' ! ! ! !1!
+-------+-+-+--------+-+
.s;.f;This record is a halfword and is defined as ZLEREC in
the SIMMAC parameter file.
.nf;.p 0,1,20
Bits Symbol Meaning
---- ------ -------
0- 6 ZLEID Always ASCII 'I'
# 7 ZLESRC = 1 if this line has a standard line
# number
# 8 ZLEOK = 1 if ZLESRC = 1 and line number is in
# sequence
9-16 ZLEIND Code for break character combination for
# line
# 17 ZLEBIT Always = 1
.p 0,1,10;The break character code has the following meaning:
.s;
Value Symbol Break characters
----- ------ ----------------
# 0 QLDEL LF (line feed)
# 1 QVDEL VT (vertical tab)
# 2 QFDEL FF (form feed)
# 3 QLVDEL LF+VT
# 4 QLFDEL LF+FF
# 5 QEFDEL End of file
.f;.p 0,1,5;Code QEFDEL is used only when the last line of
a source file is not terminated by a break character.
Type 4 records are packed into full words whenever possible.
If a type 4 record is followed by a full word control record
it is padded to the right if necessary with a halfword containing
a 1 in the last bit.
.nf;.p 0,1,7;Type 6 PAGE record
.s
0 6 7 13 20 35
+-------+--------+--------+---------------+-+
! n ! n ! 'P' ! !1!
+-------+--------+--------+---------------+-+
.p 0,1,7;or
0 6 7 13 35
+-------+--------+------------------------+-+
! n ! 'P' ! !1!
+-------+--------+------------------------+-+
.p 0,1,7;or
0 6 35
+-------+---------------------------------+-+
! 'P' ! !1!
+-------+---------------------------------+-+
.f;.p 0,1,7;This record is created when a PAGE switch is found
either in the command string or in an OPTIONS statement.
"nn" is the number of lines per page, which may precede 'P'
in the switch.
If the switch is followed by :"text string", this string will follow
in one or more text records immediately after the page record.
The text string is delimited by the next control record.
.nf;.p 0,1,7;Type 7 LIST record
.s
0 6 7 13 35
+-------+--------+------------------------+-+
! - ! 'L' ! !1!
+-------+--------+------------------------+-+
.p 0,1,7;or
0 6 35
+-------+---------------------------------+-+
! 'L' ! !1!
+-------+---------------------------------+-+
.f;.p 0,1,6;This record is created when a -LIST or LIST switch is found
in a command string or i an OPTIONS statement. A LIST record is also
created if a COMPILE-class command generates LPT: for the list file or
if the command string contains a CREF switch.
.nf;.p 0,1,7;Type 8 NOERR record
.s
0 6 7 13 35
+-------+--------+------------------------+-+
! - ! 'N' ! !1!
+-------+--------+------------------------+-+
.p 0,1,7;or
0 6 35
+-------+---------------------------------+-+
! 'N' ! !1!
+-------+---------------------------------+-+
.f;.p 0,1,6;This record is created when a -NOERR or NOERR switch is found
in a command string or an OPTIONS statement.
.nf;.p 0,1,7;Type 9 Source identification record
.s
0 6 35
+-------+---------------------------------+-+
! ! ! !
+-------+---------------------------------+-+
.f;.p 0,1,4;This record is created when a new source file is being scanned,
i.e_. when a "false" EOF word is found in the buffer (see LC module).
It is followed by four or more text records:
.nf;.p 0,1,25
0 35
+-------------------------------------------+
! device name !
+-------------------------------------------+
! file name !
+--------------------+----------------------+
! extension ! !
+--------------------+----------------------+
.s;If no path was given or just a ppn, the 4th word is:
+--------------------+----------------------+
! project no or 0 ! programmer no or 0 !
+--------------------+----------------------+
.s;If an SFD path was specified, the 4th word is:
+--------------------+----------------------+
! -n-1 ! 0 !
+-------------------------------------------+
.s;followed by n+1 words:
+--------------------+----------------------+
! project no ! programmer no !
+--------------------+----------------------+
! SFD name 1 !
+-------------------------------------------+
: :
: :
+-------------------------------------------+
! SFD name n !
+-------------------------------------------+
.f;.p 0,1,3;This information is used by pass 3 to look up the source
file(s).
.p 0,1,10;The following sequence of records will always start the
LS1 file:
.nf;- L
- N
S
<first source identif. rec>
-
-
-
.p 0,1,7;Records belonging to a line will appear in the order:
.nf;
+-----------------+----------------------+------------+
! ! ... ! ! ! !
+-----------------+----------------------+------------+
0 or more records ! type 2 and/or type 3 ! type 4
type 5-8 ! if BEGIN and or END !
.f;.lm 17;.p -16,3,15;II.4.5##IC1##Intermediate code from pass 1.
.i -8;.br;----------------------------------
.p -8,1,2;FILE:###nnnIC1.TMP
.p -8,1,3;PURPOSE:The intermediate code file IC1 contains information
about executable constructions in the source text.
.p -8,1,5;OPEN:
.p -8,1,4;OUTPUT:#The file will stay in core until it exceeds 5 disk
buffers in size. When this limit is reached, O1IC1
is called and a file is opened on channel QCHIC1
in buffered binary mode (.IOBIN) at O1OPIC.
.p -8,1,2;CLOSE:##The file is closed in O1ICCL called from T1.
.p -8,1,3;REOPEN:#A LOOKUP on the file is performed in O2OP if the
file does not already reside in core.
.p -8,1,3;INPUT:##O2IS is called to read a symbol and O2IV to read a
constant value.
.p -8,1,1;DELETE:#The file is deleted in O2EX.
.p -8,1,9;RECORD LAYOUT: IC1 is, with a few exceptions, a reverse
Polish string in which the elements are symbols represented
in the syntax description below by upper case letters.
For example PLUS is the symbol for the binary operator + .
Some symbols are followed by parameters (e.g. CONR(v1)(v2)
represents a real constant whose value is the actual
parameters that follow in the next two halfwords). Each
symbol and parameter occupies one halfword.
.s;In the description the special symbol ID is used as a
generic symbol representing an identifier. Each
differently spelt identifier is allocated a number by the
Lexical scanner which is equivalent to its index in the
symbol table. This lexical identifier number is used in
the intermediate files to specify the identifiers. Strictly
there should be a symbol ID(identifier) but in practice
there is no single ID symbol. The identifier appears
alone acting as both symbol and parameter.
This is possible because the lexical identifier numbers
occupy a range disjoint from the range of all other
symbol values (i.e. 1025-4095).
.lm 0;.p 0,2,15;SYNTAX DESCRIPTION:
.s;.lm 9;The first characters of lines have the following
meaning:
.nf;; comment line
I syntax production for IC1
D syntax for corresponding construct in DF1
.lm 17;
.f;.i-8;F#######follows a line tagged with I and shows fixup
locations for a construction
.br;The second character of a line may be a + sign.
.br;.i-8;#+######Then the line is a continuation of the last line
with the same first character and this line should
be concluded with a + sign.
.nf;.lm 0;.p 0,2,10;Program:
.s
I<program>::= BPROG <class declaration> EPROG |
D ZHE
I BPROG <procedure declaration> EPROG |
D ZHE
I BPROG <statements> EPROG
D ZHE
.p 0,1,15;;Declarations:
.s
I<declare>::= |
I <declaration> <declare>
.s
I<declaration>::= <array declaration> |
I <procedure declaration> |
I <class declaration> |
I <switch declaration> |
I <line> |
I ERROR |
I PURGE
I DEBUG (cp)
.p 0,1,10;
; ERROR may occur anywhere in a declaration or statement where
; a symbol may otherwise occur, and does not affect the syntax
; of remaining symbols.
; PURGE may occur anywhere where ERROR might otherwise occur
; but this symbol terminates the currently innermost statement
; (i.e. symbols that could otherwise have followed are
; suppressed).
; DEBUG is used to control compiler testing and is allowed
; always where a symbol can occur. The (cp) halfword following
; is the component name (two sixbit characters) and an action
; code in the remaining bits.
.nf;.p 0,1,10;
I<class declaration>::= ID BEGCL <declare> EDCL <statements> +
F f+2 f+3
.s
I+ <classend>
.s
; The fixup number is obtained from ZHEFIX in the ZHB record
; for the class.
.p 0,1,5;.nf
I<classend>::= INNER <statements> ENDCL |
F f+4 f+5
I IENDC
F f+4 f+5
.p 0,1,5
I<procedure declaration>::=ID BEGPR <declare> <statements> ENDPR
F f+2 f+3
D ZHE
; The fixup number above is obtained from the ZQU entry of the
; procedure in pass 2
.p 0,1,5;
I<array declaration>::= <bounds> ADEC |
I ID <array declaration>
.p 0,1,5;
I<bounds>::= <expression> <expression> BOUND |
I <bounds> <bounds>
.p 0,1,5;
I<switch declaration>::=ID SWITCH <swlist>SWEND
F f+1
.p 0,1,5;
I<swlist>::= |
I <expression> SWEL <swlist>
.p 0,1,5;
I<line>::= LINE (sc) (li)
; LINE is a generic symbol consisting of the current line
; number and with the high order bit set. The literals
; following are a semicolon count and the last line number.
; Last line number <= current line number.
.p 0,1,15;
;Statements:
.s;
I<statements>::= |
I <statement> <statements>
.p 0,1,5;
I<statement>::= <line> |
I ERROR |
I PURGE |
I DEBUG (cp)
I <conditional> |
I <assignment statement> |
I <denotes statement> |
I <goto statement> |
I <if-goto> |
I <activation> |
I <for statement> |
I <while statement> |
I <connection> |
I <prefixed block> |
I <block> |
I <label definition> |
I <procedure call>
.p 0,1,5;
I<conditional>::= <expression> IFST (f) <statements> FIX (f) |
F f
I <expression> IFST (f) <statements> JUMP (g) +
.p 0,1,5;
I+ FIX (f) <statements> FIX (g)
F f g
.p 0,1,5;
I<assignment statement>::=<expression> <expression> BECOM |
I <expression> <assignment statement>
.p 0,1,5;
I<denotes statement>::= <expression> <expression> DENOT |
I <expression> <denotes statement>
.p 0,1,5;
I<goto statement>::= <expression> GOTO
I<if-goto>::= <expression> <expression> IFTRU |
I <expression> <expression> IFTRE (g) +
.p 0,1,5;
I+ <statements> FIX (g)
.p 0,1,5;
I<activation>::= <expression> <expression> ACTIV (code) |
I <expression> ACTIV (code)
; The code following ACTIV is a bit mask describing the
; activation clause:
; bit meaning
; 17 reactivate
; 16 before
; 15 after
; 14 at
; 13 delay
; 12 prior
.p 0,1,5;
I<for statement>::= <controlled variable> <for list> FORDO +
F f +
.p 0,1,5;
I+ <statement> ENDFO
F+ f+1
.p 0,1,5;
; The fixup number is obtained from the ZHE entry of the for statement
.p 0,1,5;
I<controlled variable>::=ID CVBE |
D ZHE
I ID CVDE
D ZHE
.p 0,1,5;
I<for list>::= <for list element> |
I <for list> <for list element>
.p 0,1,5;
I<for list element>::= <expression> FORSI |
I <expression> <expression> FORWH |
I <expression> <expression> <expression> FORST
.p 0,1,5;
I<while statement>::= <expression> WHILE (f)<statements> JUMP (f) +
F f +
.p 0,1,5;
I+ FIX (f+1)
F+ f+1
.p 0,1,5;
I<connection>::= <expression> INSPE (f) <dolist> <otherw>
.p 0,1,5;
I<dolist>::= DO <statements> ENDDO |
D ZHB
I <whendo>
.p 0,1,5;
I<whendo>::= |
I <whendo> ID WHEDO <statements> ENDDO
F f+2
D ZHB
.p 0,1,5;
I<otherw>::= OTHER <statement> FIX (f+3) |
F f f+3
I NOTHR
F f f+3
.p 0,1,5;
I<prefixed block>::= <reference> BEGPB <declare> EDPB +
F f+2 f+3 +
D ZHB
.p 0,1,5;
I+ <statements> PBEND
F+ f+4
; The fixups for prefixed blocks are obtained
; from the ZHB entry of the block
.p 0,1,5;
I<block>::= BBLK <declare> <statements> EBLK
D ZHE
.p 0,1,5;
I<label definition>::= FIX (f)
F f
.p 0,1,5;
I<procedure call>::= <expression> <line>
.p 0,1,5;
.p 0,1,5;
.p 0,1,5;
;Expressions:
.p 0,1,5;
I<expression>::= <simple> |
I <expression> <unop> |
I <expression> <expression> <binop> |
I <expression> <expression> <expression> IFEX
.p 0,1,5;
I<simple>::= <constant> |
I <reference>
.p 0,1,5;
I<unop>::= PAREN |
I UMIN |
I UPLUS |
I NOT |
I THIS |
I NEW |
I DELOP
; DELOP is used for error recovery
.p 0,1,15;
I<binop>::= PLUS |
I MINUS |
I DIV |
I MULT |
I IDIV |
I POW
I LESS |
I NLESS
I GRT |
I NGRT |
I EQ |
I NEQ |
I DEQ |
I NDEQ |
I IS |
I IN |
I QUA |
I AND |
I OR |
I IMP |
I EQV
.p 0,1,5;
I<reference>::= <id1> |
I <id1> <alist>
.p 0,1,5;
I<id1>::= ID |
I <expression> ID DOT
.p 0,1,5;
I<alist>::= LB <tail> |
I LP <tail>
.p 0,1,5;
I<tail>::= RP |
I <expression> <tail>
.p 0,1,15;
I<constant>::= TRUE |
I FALSE |
I NONE |
I NOTEXT |
I CONC (v1) |
I CONI (v1) (v2) |
I CONR (v1) (v2) |
I CONLR (v1) (v2) (v3) (v4) |
I CONT (ra) (le) |
I UDEF
; The text constant is of length le and starts at ra from the
; program break at run time. UDEF is used for error recovery.
.sp 1;.nonum;.lm 9;.ps 50,72
.ts 1,9,17,25,33,41,49,57,65
.pg;.nm ch 4
.ch SIMULA FOR DEC SYSTEM 10#############TD, COMPILER
.nm 26
.st 740814####780302######6###########################Reidar Karlsson
.pg;.lm 9;.p -8,0,10;II.4.6##DF1 Declaration file
.br;--------------------
.s 2;FILE:###nnnDF1.TMP
.lm 17;.p -8,1,5;PURPOSE: The declaration file contains declarative
information from the source text in a special form arranged for input
to pass 2.
.p -8,1,10;OUTPUT:#The file output is controlled by DP code. DP is called from T1.
.br;DF1 will stay in core as long as it fits in the free buffer area.
When all buffers available to DF1 are filled, nnnDF1.TMP will be
entered on channel QCHDF1, which is opened for device DSK in binary
mode (.IOBIN) by O1OPDF. The buffers are output by O1DF1. O1DFCL
closes the file.
.p -8,1,5;INPUT:##The file is looked up by O2OP if it was written on DSK in pass
1. Buffers from the file are read by O2D1. O2EX deletes the file.
.p -8,1,10;RECORD LAYOUT: DF1 contains 4 record types: ZHE, ZHB,
ZQU and ZQQ. Declared or specified quantities are represented by ZQU
records. Blocks are represented by ZHE records (FOR statements, procedure
bodies, and reduced subblocks) or by ZHB records (procedures, classes,
prefixed blocks and connection statements). ZHB records contain all
the fields of ZHE records and some additional information.
At the end of the DF1 file is a list of ZQQ records, one for each
external declaration. All declarations of a block will be at the start
of that block in the order: parameters, non-data quantities and value
quantities, texts and finally ref variables and arrays.
Then follow the sublists for declared classes and procedures, containing
class attributes and procedure parameters. These sublists may contain
further sublists which will follow next, etc.
.p 0,1,3;A detailed description of the four record types is found
in II.6 Declaration structures.
.p -16,2,15;II.4.7##ZSE Symbol table
.i -8;-----------------
.p -8,1,2;FILE:###nnnZSE.TMP
.p -8,1,8;PURPOSE: The symbol table is a copy of the dictionary created
in pass 1. It contains the names of all differently spelled identifiers
appearing in the compiled program. The names are in SIXBIT code and
only the first 12 characters are saved for long identifiers.
The ZSE file is used in pass 3 to build the symbol table used by SIMDDT
and to edit error messages and cross reference listing, if any.
.p -8,1,8;OUTPUT:#The O1ZS procedure, called from T1, decides whether the file should
stay in core or be output to disk. If the symbol table is too big
to fit in the core buffer area alloted, the disk file is entered on
channel QCHZSE by O1ZSOP in buffered binary mode (.IOBIN), written and then
closed by O1ZSCL.
.p -8,1,5;INPUT:##If the file is not core resident, it is opened in unbuffered dump
mode (.IODMP) on channel QCHZSE by pass 3 (procedure O3RZ, called
from I3), read into core and deleted.
.p -8,1,5;RECORD LAYOUT: Each identifier in the source code has
a two word entry in the symbol table, referenced by the identifier
index.
.nf;.p 0,1,8
+-----------------------+
! I ! D ! E ! N ! T ! I !
(INDEX-1024)*2 -> !-----------------------!
! F ! I ! E ! R ! ! !
+-----------------------+
0 35
.f;.p 0,1,4;The two words contain the first 12 characters of the identifier
in SIXBIT code, left justified. The unused positions are zero (SIXBIT
for space).
.p -16,2,15;II.4.8##REL Relocatable binary object code
.i -8;-----------------------------------
.p -8,1,5;FILE:###Name as specified in command line, normally infile.REL,
where infile.SIM is the input source file.
.p -8,1,6;PURPOSE:Code generated by the SIMULA compiler is output
in relocatable binary format on a disk file. LINK-10 accepts this
file as input and produces an executable program from this file and
some other relocatable object code files.
.p -8,1,6;OPEN:###The file is opened on channel QCHREL in buffered
binary mode (.IOBIN) at I1END, and the file is entered in O1RLOP called
from O1SCOP after the source file has been opened.
.p -8,1,3;OUTPUT:#The output routines in the 3 passes are named
O1RL, O2REL and O3WIB respectively.
.p -8,1,8;CLOSE:##The file is closed in T3I called at the end of M3.
If the file is just a header for an external FORTRAN or MACRO procedure,
T3R, called at the end of M3ATR,
closes it with an option that causes the file to be discarded
(creation inhibited).
.lm 9;.p 0,1,10;RECORD LAYOUT: REL file records are LINK-10 items of the
types: 0, 1, 2, 3, 4, 5, 6, 7, 10, 17 and ASCIZ (type code > 3777,
first 7-bit byte a printable ASCII character).
These record types are accepted also by the LOADER, except for type
17 and ASCIZ. Thus LOADER cannot be used when external classes or
procedures are referred to in the SIMULA program.
The formats of these items are described in detail in LINK-10 Programmer's
Reference Manual, Appendix A.
.br;The first item in the REL file is type 4 (ENTRY) defining .MAIN
as the entry point name if a main program is compiled, the external name
for a separately compiled class or procedure.
The processor code for the SIMULA compiler (octal 15) is output in
the next item, which is type 6 (NAME).
LINK-10 recognizes this processor code.
The entry and name records are output directly after opening the file
in O1RLOP. The third item is a code item (type 1), which starts with
a copy of the lookup information for the first source file, followed
by any literals (text literals from pass 1, other literals from
pass 2).
.br;The rel file has the following general layout:
.nf;.p 0,1,25;
+----------------+
! ENTRY (type 4) !
!----------------!
! NAME (type 6) !
!----------------!
! 1 or more CODE ! Source code identification, text literals,
! (type 1) ! and absolute constants output in pass 2.
!----------------!
! mixed type 1, 2!
! (SYMBOLS) and !
! type 10 (INTER-! IC2 code converted by pass 3 (M3).
! NAL REQUEST) !
!----------------!
! START ADDRESS ! Only for main programs. Output by M3.
! (type 7) !
!----------------! Definition of global symbols and external
! 1 or more type ! requests for entry points in external
! 2 records ! classes and procedures.
!----------------!
! END (type 5) ! Output in M3.
+----------------+
.lm 17;.f;.p -16,2,15;II.4.9##ATR Intermediate attribute file
.i -8;--------------------------------
.p -8,1,2;FILE:###nnnATR.TMP
.p -8,1,5;PURPOSE: The attribute file contains declarative information
about a separately compiled external class or procedure.
.p -8,2,8;PASS 2 PROCESSING:
.P 0,1,0;When an external class or procedure is compiled, the EXAT
routine, called from O2EX, will create a skeleton attribute file.
.p -8,1,10;OUTPUT:#The file will stay in core if it fits in its alloted buffer area,
otherwise it will be output in unbuffered record dump mode
(.IODPR) on channel
QCHATR via calls to O2ATR, where the first call will open the channel
and enter the file.
.p -8,1,3;CLOSE:##The last buffer is output and the file is closed
when EXAT returns to O2EX.
.p -8,2,10;PASS 3 PROCESSING:
.P -8,1,5;INPUT:##If not core resident, the file is read into core in unbuffered
dump mode (.IODMP) on channel QCHATR and closed, all i/o performed
in O3ATR called from M3ATR.
.p -8,1,10;UPDATE:#The file created in pass 2 contains identifier
indices instead of sixbit names, and unique identifier fields are
zero.
.br;M3ATR translates identifier indices to sixbit names via the symbol
table (ZSE).
.p -8,1,10;OUTPUT:#If an old attribute file exists and shows some essential
attribute difference, or if an old attribute file does not exist,
new unique numbers (RADIX50 identifiers) are generated. Any old attribute
file x.ATR is renamed to x.QTR (after deleting x.QTR if present already),
and a new x.ATR file is written. Here x is the relfile name (normally
the same name as the source file). A new ATR file is also generated
if the old ATR module was found in a library ATR file, even if the
attributes are not different, in that case using the old unique numbers.
The ATR file copy is created so that the ATR file is always in the
same area as the corresponding REL file. The QTR file is saved to
provide a backup in case the compilation was not useful (REL file
error for example).
O3WATR writes the ATR file (binary unbuffered dump mode .IODMP).
O3WARE handles QTR files.
.p -8,1,3;RECORD LAYOUT: See II.4.3 Attribute file.
.p -16,2,15;II.4.10#XRF Cross reference data file
.i -8;-----------------------------
.p -8,1,2;FILE:###nnnXRF.TMP
.p -8,1,8;PURPOSE:If the C switch is specified to the compiler, a
cross reference data file is created in pass 1. Pass 3 will process
this file and create a cross reference listing at the end of the source
program listing. Note that the system program CREF is n#o#t involved
in creating a SIMULA cross reference listing.
.p -8,1,6;OUTPUT:#The file will stay in core until it exceeds
2 standard disk buffers. O1XR will then be called to open a file on
channel QCHXRF in buffered binary mode (.IOBIN) via O1OPXR.
.p -8,1,2;CLOSE:##The file is closed in O1XRCL which is called from
T1.
.p -8,1,8;INPUT:##If a cross reference listing is demanded, the SIMXRF procedure
is called from E3. SIMXRF will call I3X to perform a lookup if necessary
(not core resident file). X3READ reads blocks from the file, and T3X
will finally delete the file.
.p -8,1,10;RECORD LAYOUT: Each word in the file has the following format:
.nf;.s;
+----------------------------------------+
!D!E! ! I ! L !
+----------------------------------------+
0 1 6 17 18 35
.f;.s;The D bit is set if the identifier with index I was declared
on line L.
.p 0,0,2;The E bit is set if the identifier is declared as external.
.p 0,0,2;The I field is the identifier index (1024 <= I <= 4095).
.p 0,0,2;The L field is the line number.
.p 0,1,3;A word with this layout is output by pass 1 every time an
identifier is found.
.p -16,2,15;II.4.11#IC2 Intermediate code from pass 2
.i -8;----------------------------------
.p -8,1,2;FILE:###nnnIC2.TMP
.p -8,1,10;PURPOSE: Code generated in pass 2 is output in 5 different
streams. Since LINK-10 could not handle more than one low segment
relocation counter, four of those streams are output on an intermediate
code file. The fifth code stream, containing absolute literals (constants),
is output to the REL file directly. The four streams on the IC2 file
consist of code stream and prototype stream (item type 1013), symbol
table (type 1014), and line number table (type 1015).
.br;Note that our definition of record types 1013, 1014, 1015 may be
at variance with existing or future definitions in LINK-10 documentation.
This does not matter, however, since these records are only used internally
in the compiler and will be replaced with other record types before
output to the rel file.
.br;References to external symbols (e.g_. entry points in the run
time system and external procedures and classes) are output in IC2
as global requests (item type 2). Internal requests (type 10) are also output
in order to have fixup references resolved.
.p -8,1,4;OPEN:###Opened and entered on channel QCHIC2 in binary
mode (.IOBIN) by O2OP called from I2.
.p -8,1,3;OUTPUT:#Written by O2IC2 called from O2GEN and O2DF.
.p -8,1,3;CLOSE:##Closed in O2EX just before entering pass 3.
.p -8,1,2;REOPEN:#Looked up by I3I called from M3.
.p -8,1,2;INPUT:##Read by O3RI called from M3.
.p -8,1,3;DELETE:#Deleted by T3I called at end of M3.
.lm 9;.p 0,1,14;RECORD LAYOUT: The header word is common to all IC2 records.
It contains the type code in the left half and a counter in the right
half which shows the length of the record except for the header word.
.nf;.p 0,1,6;Header word:
.s; +--------------------------------------+
! TYPE CODE ! WORD COUNT !
+--------------------------------------+
.f;.p 0,1,4;The maximum length of an IC2 record is adapted to the
size of the disk buffer data area (i.e_. 128 words including header word).
.p 0,1,15;.nf;Global requests type 2
----------------------
.s;
0 4 12 18 35
+----+------+-----+------------------+
! 2 ! COUNT !
+----+------+-----+------------------+
! !RCODE! ADDRESS ! Data is grouped
+----+------+-----+------------------+
! 0 ! INDEX ! in pairs
+----+------+-----+------------------+
! !RCODE! ADDRESS !
+----+------+-----+------------------+
! 60 ! UNRSYMBOL !
+----+------+-----+------------------+
! !
.f;.p 0,1,7;ADDRESS, relocated according to RCODE, points to the first
word in a global request chain. The second word in each pair is either
an INDEX (left half = 0) to a global run time symbol (RTSYMBOLS) or
a RADIX50 symbol (UNRSYMBOL) generated as a unique number for an external
procedure or class.
.p 0,1,3;This record is converted in pass 3 to LINK-10 item type 2.
.nf;.p 0,1,15;Internal requests type 10
-------------------------
.s;
0 12 18 35
+-----------+-----+------------------+
! 10 ! COUNT !
+-----------+-----+------------------+
! ! NON-ZERO COUNT !
+-----------+-----+------------------+
! !RCODE! ADDRESS ! Data is grouped
+-----------+-----+------------------+
! !RCODE! VALUE ! in pairs
+-----------+-----+------------------+
! !
.f;.p 0,1,5;Dummy zero word pairs may appear in this type of record.
NON-ZERO COUNT shows number of non-zero words.
.p 0,0,5;ADDRESS, relocated according to RCODE, points to the first
word of a chain for an internal request (i.e_. a fixup reference).
The relocated VALUE defines this request. This record is converted
in pass 3 to a LINK-10 item type 10.
.nf;.p 0,1,17;Generated code and prototypes type 1013
---------------------------------------
.s;
0 18 21 35
+------------------+---+---------------+
! 1013 ! COUNT !
+------------------+---+---------------+
! RELOC.WORD 18 2-bit bytes !
+------------------+---+---------------+
! ! RC! START ADDRESS !
+------------------+---+---------------+
! !
! 17 DATA WORDS !
! !
+------------------+---+---------------+
! RELOC.WORD !
+------------------+---+---------------+
! !
! 18 DATA WORDS !
! !
+------------------+---+---------------+
! !
.f;.p 0,1,10;RELOC.WORD controls the relocation of the the following
18 DATA WORDS:
.nf;.p 0,1,7;
Reloc. byte Data word relocation
------------------------------------
0 No relocation
1 Relocate right half
2 Relocate left half
3 Relocate both halves
.f;.p 0,1,10;A relocated halfword contains an internal relocation
code in its 3 leftmost bits:
.nf;.s;
0 3 17
18 21 35
+-----+-----------------------+
!RCODE! Value to be relocated !
+-----+-----------------------+
.f;.p 0,1,5;The internal relocation will be performed in pass 3 by
adding a relocation constant according to RCODE. The result will then
be finally relocated by LINK-10. RCODE is one of the values:
.s;########QRELCN, QRELCD, QRELPT, QRELLT, QRELST.
.p 0,1,5;The first data word of each record yields the start address
of the following block of DATA WORDS. The relocation code RC is either
QRELCD for generated code or QRELPT for a prototype record.
.br;This record is converted to a LINK-10 item type 1 by pass 3.
.nf;.p 0,1,17;Symbol table type 1014
----------------------
.s;
0 18 21 35
+------------------+---+---------------+
! 1014 ! COUNT ! Relocation groups
+------------------+---+---------------+ as for type 1013
! RELOC. WORD 18 2-bit bytes !
+------------------+---+---------------+
! ! RC! START ADDRESS ! RC = QRELST
+------------------+---+---------------+
! !
! 17 DATA WORDS !
! !
+------------------+---+---------------+
! 0 !ID ! INDEX !
+------------------+---+---------------+
! 0 !I2 ! INDEX !
+------------------+---+---------------+
! !
! 18 DATA WORDS !
! !
+------------------+---+---------------+
! !
.f;.p 0,1,8;The only difference from type 1013 is that two special
relocation codes for right halves of data words may appear, i.e_.
QRELID and QRELI2 which instruct pass 3 to replace the whole word
with the first (QRELID) and second (QRELI2) half of an identifier
in SIXBIT code. The actual identifier is defined by its INDEX in the
ZSE symbol table.
.nf;.p 0,1,17;Line number tables type 1015
----------------------------
.s;
0 18 21 35
+------------------+---+---------------+
! 1015 ! COUNT !
+------------------+---+---------------+
! RELOC. WORD 18 2-bit bytes ! Relocated like 1013
+------------------+---+---------------+
! ! RC! START ADDRESS ! RC = QRELLT
+------------------+---+---------------+
! !
! 17 DATA WORDS !
! !
+------------------+---+---------------+
! !
.f;.p 0,1,5;Type 1015 is introduced so that pass 3 may easily recognise
line number information. The description of data for type 1013 also
applies to type 1015.
.p 0,2,8;Comment block type 0
.br;--------------------
.p 0,1,4;A comment block is output to fill the last words of the disk
buffer after outputting partly filled local code stream buffers. The
padding is done before outputting type 10 blocks.
.lm 17;.p -16,3,15;II.4.12#LS3 List file
.i -8;-------------
.p -8,1,8;FILE:###The name is taken from the command line:
.br;*relfil.REL,lstfil.LST=infile.SIM or
.br;*infile/L (lstfil = infile in this case)
.br;The name is thus whatever stands in the place of lstfil.
Default extension is LST. If the /LIST switch was given in a COMPILE
(EXECUTE, LOAD, DEBUG) command, and LPT is spooled on disk, LPT: will
be given as device. The actual disk file will be named xxxxxx.LPT
where xxxxxx is a created name and the spooled name (lstfil) is taken from
the command.
.p -8,1,7;PURPOSE: LS3 is the list file produced by the compiler.
It contains:
.br;* Listing of source code with line numbering (added or already
part of source file), consecutive numbering of BEGIN keywords in the
left margin and also numbering of END keywords with the number of
the matching BEGIN. Error messages are inserted at the place of error.
.br;* If the Y switch was set,
a table relating line number to code offset (octal).
.br;* If the C switch was given, a cross reference listing.
.br;The pages of the listing are numbered (file page number,
if necessary followed by sub-page number (e.g_. PAGE 3-5). The file
pages are delimited by form feed characters. The heading of each page
shows the compiler version, e.g_. DECsystem-10 SIMULA %4(300).
The date and time of compilation,
the file specification of the source file and its creation time and
date are also shown. Any text given via the PAGE switch is placed
in the heading.
.p -8,1,5;OPEN:###The pass 3 module I3 contains code to open the file
and enter it on channel QCHLS3 (I3LS3, called from E3). The file is
written in buffered ASCII mode (.IOASC).
.p -8,1,5;OUTPUT:#Via O3WS, O3LS3 in the pass 3 module O3. O3WS is
called by E3, T3 and SIMXRF.
.p -8,1,3;CLOSE:##By T3L in the pass 3 module T3.
.p -8,1,3;LIST FILE FORMAT: See any compilation listing. See E3.MAC
assembly listing for further details.
.p -16,1,15;II.4.13#SIMERR Error message tables
.i -8;---------------------------
.p -8,1,3;FILE:###SIMERR.ERR
.p -8,1,6;PURPOSE: SIMERR contains error message tables used in pass
3 (module E3) to create error messages on the listing
and on the terminal.
The file is created by the maintenance utility program SUTERR using
as input
a RUNOFF source file containing appendix C of the DECsystem-10 SIMULA
Language Handbook Part II, which contains all compiler error messages.
.p -8,1,6;INPUT:##E3 calls O3ERR, which opens channel QCHERR in unbuffered dump
mode (.IODMP) and looks up the file SIMERR.ERR (usually on the system
device), then reads it into core and closes the file.
.p -8,1,5;REFERENCES: A detailed description of this file is found
in TD II.5 Diagnostic handling and V.6 Utility program SUTERR.
.p -16,1,15;II.4.14#DEB Debug files
.i -8;----------------
.p -8,1,5;.nf;FILES:##nnnDEB.LST
nnnDEB.TMP
DBLIST.LST
.p -8,1,3;REFERENCES: [1] I.10 Test standard.
####[2] V.4 SUTEDB.
.p -8,1,6;.f;PURPOSE: Special debug files are created in the debug
version of the compiler for the purpose of obtaining symbolic dumps
and traces of the compiler actions.
Pass 1 outputs its debug information in directly readable form as
nnnDEB.LST. Pass 2 debug information is put on nnnDEB.TMP in a special
format, which can later be edited by the utility program SUTEDB to
yield a readable file DBLIST.LST which may also include a copy of nnnDEB.LST.
.p -8,1,6;DEB.LST: In the debug version of the SIMULA compiler, I1
opens channel QCHDEB in buffered binary mode (.IOBIN). The SP module
of pass 1 calls P1INIT of the P1DUMP module to enter the file nnnDEB.TMP
(calling OUTOP). The information which is output depends on the special
debug symbols edited into the source code. See [1].
The end part of nnnDEB.LST may be a readable dump of DF1 and IC1.
These files are read if not core resident on channel QCHIC2 in binary
buffered mode (.IOBIN) via DFDUMP, ICDUMP called from P1DUMP, called
from T1. P1DUMP closes the file nnnDEB.LST via OUTCL (no output if
nothing written).
.p -8,1,8;DEB.TMP: After closing nnnDEB.LST, P1DUMP will
enter nnnDEB.TMP on the same channel (QCHDEB) via O1DBOP.
Finally, pass 1 outputs the symbol table to this file, enabling output
from pass 2 to be edited into symbolic form. [1] shows the format
of records in the DEB.TMP file and how to create a record in the file.
.p 0,1,6;The utility program SUTEDB edits the debug output from pass
2 into a readable symbolic format. See [2].
SUTEDB starts by asking for source file and job number (nnn) of the
nnnDEB.TMP file. If the source file is given, it will be copied as
the first part of DBLIST.LST. Then SUTEDB asks if nnnDEB.LST is present
and copies also that file to DBLIST.LST if the answer is affirmative
and the file actually exists. Finally SUTEDB reads nnnDEB.TMP and
edits the information into the output file DBLIST.LST.
See [1] and [2].
.sp 1;.nonum;.lm 9;.ps 50,72
.ts 1,9,17,25,33,41,49,57,65
.pg;.nm ch 5
.ch SIMULA FOR DEC SYSTEM 10#############TD, COMPILER
.nm 1
.st 740718####780302######6###########################Elisabeth Engstrom
.pg;.lm 9;.p -8,0,10;II.5####DIAGNOSTIC HANDLING
.P -8,2,10;II.5.1##Introduction
.br;------------
.p 0,1,10;All error messages are documented in DECsystem-10 SIMULA
Language Handbook, part II, appendix C, which also contains information
on the behaviour of the compiler and suggested user actions when errors
occur.
The RUNOFF source file for appendix C is used as input to the SUTERR
program which creates the error message file SIMERR.ERR.
That file and a table (ZDM) produced during compilation are used by
the routine E3EM of pass 3 to create the final messages. See VI.3.
.s;The RUNOFF file must be updated when compiler error messages are changed,
replaced or added.
.p -8,3,15;II.5.2##Output of diagnostics
.br;---------------------
.p 0,1,10;Each diagnostic message is output on the user terminal together
with the source code line(s) to which the message refers. The message
is also edited into the list file in the corresponding place (after
the last line of the offending statement).
.br;A maximum of 50 messages will be output, the first 50 encountered.
Warning messages (except for the first 10 warnings)
may however be replaced by error messages that are
occur later, when 50 messages have already been issued.
Error statistics (number of messages of each type) are output at the
end of the compiler listing and on the terminal. If the compiler was
called via the CCL entry (COMPILE etc), no information is output for
programs without diagnostics.
.p 0,1,12;The error statistics look like this:
.nf;.s;NO ERRORS DETECTED
or
ERRORS DETECTED:
####nn TYPE W MESSAGES
####nn TYPE E MESSAGES
? TERMINATION ERROR
.s;.f;nn is the total number of errors and may exceed 50.
.p -8,3,19;II.5.3##Error message format
.br;--------------------
.s;The general format is:
.nf;.s;SIMddd t LINE nnnnnn:mm text
.f;.lm 17;.p -8,1,10;ddd is the message number in octal format
.i -8;t###is the message type:
.br;T termination error
.br;E non-terminating error
.br;W warning message
.i -8;nnnnnn line number
.i -8;mm######number of semicolons before faulty statement on line nnnnnn
.br;:mm is omitted if mm = 0
.lm 9;.f;.p -8,3,15;II.5.4##Number of and types of messages
.br;-------------------------------
.p 0,1,7;The number is in octal format and belongs to certain intervals
depending on error type and pass number of detecting pass.
.br;The intervals are defined by assembly constants in SIMMAC.MAC.
.br;Message number = mn
.nf;.p 0,1,15;
Q1.WAR <= mn < Q1.ERR Warning message pass 1
Q1.ERR <= mn < Q1.TER Error message pass 1
Q1.TER <= mn < Q2.WAR Terminating message pass 1
Q2.WAR <= mn < Q2.ERR Warning message pass 2
Q2.ERR <= mn < Q2.TER Error message pass 2
Q2.TER <= mn < Q3.WAR Terminating message pass 2
Q3.WAR <= mn < Q3.ERR Warning message pass 3
Q3.ERR <= mn < Q3.TER Error message pass 3
Q3.TER <= mn < Q.WAR Terminating message pass 3
Q.TER <= mn < 577 Error messages, identical in all
three passes. The pass number is
often part of the text.
.f;.p -8,4,15;II.5.5##Insertion of an error message in the compiler
.br;---------------------------------------------
.p 0,1,7;When an error is detected in the course of a compilation,
current error information is saved in the ZDM table, which is used
by pass 3 to create the text of the error message. The error information
is reported via a set of UUO's which are trapped by the DM module
present in each pass.
.p 0,1,2;Format: UUOcode t,n
.lm 17;.p -8,1,4;UUOcode:There are 6 different UUO codes depending
on message type:
.br;ERR, ERRI1, ERRI2, ERRI3, ERRT, ERRLI
.p -8,1,3;ERR: This message is fixed, i.e_. no variable information
is edited into the message by pass 3.
.p -8,1,6;ERRIx: x = 1, 2 or 3. These UUO codes are used when
the message should contain the name of an identifier, a symbol specified
by the SIMMAC macro SYMBOL, or a number. The UUO name indicates how
many insertions should be made.
The first argument is taken from AC1 (X1), the second from X2, the
third from X3. The arguments are:
.br;* identifier number >= 1024
.br;* value of current symbol as defined in SIMMAC
.br;* binary number <= 4095 (last 12 bits if greater)
.p -8,1,7;ERRT: This message contains text or a file name to be edited
into the message.
.br;The lookup/enter block address is placed in X1 if the message
contains a file name. The ASCII text can be at most 5 characters left
justified in X1.
.p -8,1,4;ERRLI: This UUO must be used to give the last line
number of the erroneous statement, unless it was given at once.
.lm 9;.p 0,1,3;Apart from the information given via the UUO and the
3 ac's, some global information is used:
.lm 25;.p -8,1,7;YELIN1 number of first line of faulty statement
.i -8;YELIN2 number of last faulty line or 0 if line number
not known
.i -8;YESEM number of semicolons preceding statement on first
line.
.p -8,1,6;t type of message:
.br;# QE error
.br;# QT termination error
.br;# QW warning
.p -8,1,10;n error number in octal radix
.p -16,1,4;Example:
.lm 17;.br;YELIN1 = 7, YELIN2 = 7, YESEM = 0
.br;L X1, idno(A)
.br;ERRI1 QE,362
.lm 9;.p 0,1,3;would give the message after line 7:
.br;SIM362 E LINE 7 IDENTIFIER A IS NOT DECLARED
.lm 9;.p -8,4,15;II.5.6##Error messages in RUNOFF source file
.br;------------------------------------
.p 0,1,10;The SIMULA Handbook part II is created from one or more RUNOFF source
files. The distributed version is SIMLH2.RNM which gives SIMLH2.MAN
when processed by RUNOFF. The essential part in this context is appendix
C, which exists at QZ as SLH2C.RNO. SUTERR assumes that its input
file is called SIMLH2.RNO, thus a copy has to be made with that name
or one of the files has to be renamed temporarily. SUTERR could also
be changed to expect another file, but existing documentation assumes
SIMLH2.RNO throughout.
SUTERR ignores RUNOFF control information (first character of a line
is a dot '.'). The interesting part of the file is delimited by two
headers:
.p 5,1,3;COMPILER SOURCE CODE DIAGNOSTICS:
.P 5,0,1;END OF COMPILER SOURCE CODE DIAGNOSTICS
.P 0,1,5;These lines must not be changed.
SUTERR expects a certain file format and complains if violations are
found. In practice, the format accepted is less definite. The essential
information is found in the error messages, which must contain the
characters SIM at the very start of a line.
Except for these lines, SUTERR expects horisontal tab
(HT, Char(9)) or dot at the start of each line.
Complaints about faulty line formats may however be disregarded.
.p -8,3,15;II.5.7##Format of error message lines
.br;-----------------------------
.s;.nf;.lm 9;SIMddd t ... text
.s;ddd is an octal message number.
t is the message type.
text is the message.
.f;.s;SIMddd followed by one space and the type character must be in
the first 8 characters of a line. Error messages must appear in increasing
order by error number and be confined to the intervals documented
in II.5.4.
.lm 9;.p 0,1,5;If there are no numbers available in the applicable
interval, one of the interval limits must be changed. This change
may mean that messages in another interval must be renumbered, etc,
but enough free numbers are left so that this should not happen.
.br;If the error message extends over more than one line, continuation
lines must start with HT (Char(9)). The text must not contain more
than 15 words and no word may be longer than 15 characters.
.br;Certain code words in the message will be replaced with information
pertaining to the current occurrence of the error. The possible code
words are: XXXX, AAAA and NNNN. There must be no more than three code
words XXXX or NNNN or at most one instance of AAAA.
Code word interpretation:
.lm 17;.p -3,1,3;1. XXXX - name of identifier.
.br;Ex. SIM362 E IDENTIFIER XXXX IS NOT DECLARED
.P -3,1,5;2. XXXX - Special SYMBOL defined in SIMMAC.MAC.
.br;Ex. SIM130 E MISSING XXXX
.BR;where XXXX is ) or ] .
.p -3,1,6;3. AAAA - ASCII text.
.br;Ex. SIM007 W NOT RECOGNIZABLE SWITCH AAAA, SWITCH
.br;<HT>IGNORED
.br;The text must be 5 characters or less.
.p -3,1,3;4. NNNN - binary number.
.br;A small integer (<4096) may be edited into the message.
.lm 9;.p 0,2,4;SUTERR processes only the lines starting with SIM and
continuations of these lines. Example of a segment from the input
file:
.p 0,1,10;
.BR;_.p -9,2,5
.BR;_SIM010 W SWITCH: AAAA MISSPELLED.
.BR;_.skip 1
.BR;<HT>EXPLANATION: AAAA is a switch name.
.BR;_.skip 1
.BR;<HT>COMPILER ACTION: Switch is processed according to first letter.
.BR;_.p -9,2,5
.BR;SIM011 W SWITCH AAAA NOT VALID AFTER PROGRAM START, SWITCH IGNORED
.BR;.skip 1
.BR;<HT>EXPLANATION: AAAA is a switch name. See chapter 3.5.
.BR;_.p -9,2,5
.lm 8;.p -8,3,15;II.5.8##Creation of SIMERR.ERR
.br;----------------------
.s;A new version of SIMERR.ERR is created by the SUTERR program from
an updated RUNOFF file containing the error messages. The input file
must be named (renamed) SIMLH2.RNO.
.br;If SUTERR finds any error message of an incorrect format, this
will be reported on the terminal and the message will be rejected.
.br;If SUTERR reports that either of the tables YE3D or YE3M must
be extended, the following actions should be taken:
.br;* Change QE3D and/or QE3M in SIMMAC.MAC and reassemble SIMMAC.MAC.
.br;* Recompile LOWSEG.MAC.
.br;* Recompile debug version of O3.MAC if used.
.br;* Change the constants QE3M and QE3D and the sizes of the tables
Y3ED and Y3EM in SUTERR.CBL.
.p 0,1,8;SUTERR exits after typing END OF SUTERR on the terminal.
.br;For further information see V.6.
.p 0,1,6;SIMERR.ERR consists of 4 tables preceded by 4 counters.
YE3M and YE3MI both start at a buffer and disk block boundary, any
leftover space in the preceding disk block being padded by zeroes.
.nf;.p 0,2,25;#SIMERR.ERR
+--------------+
! ! Number of words in YE3DL
+--------------+
! ! Number of words in YE3D
+--------------+
! ! Number of words in YE3M
+--------------+
! ! Number of words in YE3MI
+--------------+
! ! YE3DL
! !
+--------------+ The tables are described in II.C.3
! !
! ! YE3D
! !
+--------------+
! !
! ! YE3M
! !
+--------------+
! !
! ! YE3MI
! !
+--------------+
.f;.p 0,2,20;Summary
.br;-------
.p 0,1,3;When inserting a new message in the compiler, the following
must be done:
.s;* An error UUO must be issued by the compiler. See II.5.5.
.br;* Insert the error message text in appendix C of SIMLH2 - II.5.6.
.br;* Create SIMERR.ERR by running the utility program SUTERR - II.5.8.
.br;* Correct message if necessary and rerun SUTERR.
.sp 1;.nonum;.lm 9;.ps 50,72
.ts 1,9,17,25,33,41,49,57,65
.pg;.nm ch 6
.ch SIMULA FOR DEC SYSTEM 10#############TD, COMPILER
.nm 1
.st 741116####780302######6###########################Claes Wihlborg
.pg;II.6 DECLARATION STRUCTURE
.s;.lm 9;.f;The declaration structure consists of 4 record types,
ZQU, ZHE, ZHB and ZQQ. 6 states are defined:
.s;.lm 17;.nf;
1. System component list (SYS1) (II.A.2)
2. Pass 1 intermediate list (II.A.3)
3. DF1 intermediate file (II.4.6)
4. Pass 2 declaration stack (II.B.3)
5. ATR intermediate file (II.4.9)
6. External attribute file (II.4.3)
.s;.lm 9;.nf;
The modules which perform the transformations are:
.s;
# (1)
# ! +--------------------------------------+
# v v !
+----+ +----+ +----+ +----+ +----+ !
! SD !->(2)->! DP !->!CARL!->(4)->! EX !->(5)->! M3 !->(6)
+----+ +----+ +----+ +----+ +----+
.s;
All fields named here are declared in the universal SIMMAC.
.pg;.i -8;II.6.1##ZQU record
----------
.f;.s;The ZQU record holds information about declared quantities.
The size is:
.lm 11;.nf;
4 words in states 3 and 4
3 words in state 2
4 or 6 words in state 1
6 words in states 5 and 6
.s;.lm 9;
FIELD LOCATION SIZE STATE CONTENTS
.s;
ZDETYP 0 <0:2> 3 ALL ZQU%V
ZQUGLO 0 <3> 1 234 Global CLASS/PROCEDURE
.tp 2
ZQUEXT 0 <4> 1 23456 External CLASS or
PROCEDURE
.tp 2
ZQULEN 0 <5> 1 2 ZQUIND contains length
of object
.tp 2
ZQUIS 0 <5> 1 4 Inspected class
attribute
.tp 2
ZQUIB 0 <6> 1 4 Function whose body is
being compiled
.tp 2
ZQUTPT 0 <6> 1 256 Temporary storing of
ZQUPTD
ZQUSYS 0 <7> 1 ALL System component
ZQUTYP 0 <8:11> 4 ALL TYPE code
ZQUHID 0 <12:13> 2 23456 HIDDEN specification
.tp 3
ZQUNOT 0 <14> 1 23456 The HIDDEN PROTECTED
specification has the
NOT form
ZQUMOD 0 <12:14> 3 ALL MODE code
ZQUKND 0 <15:17> 3 ALL KIND code
ZQUTMK 0 <8:17> 10 4 TYPE+MODE+KIND
ZQUSTMK 0 <7:17> 11 4 SYS+TYPE+MODE+KIND
ZQUZHE 0 <18:35> 18 4 Enclosing block header
.tp 4
ZQUTEM 0 <18:35> 18 12 Temporary storing of
3<0:17> if system
element, otherwise of
3<18:35>
.tp 2
ZQUFIX 1 <0:17> 18 1 Link next ZQU at same
level
ZQUIVA 1 <0> 1 23456 Invalid access
.tp 2
ZQUNSB 1 <1:5> 5 23456 Number of subscripts or
virtual match
.tp 2
ZQULID 1 <6:17> 12 234 Lexical identifier
number
.tp 5
ZQUIND 1 <18:35> 18 2 Length
ALL Offset or virtual index
1234 Fixup or HISEG address
5 Code address of
prototype or label
.tp 2
ZQUUSE 2 <0> 1 4 Used when resolving
HIDDEN specifications
.tp 2
ZQUZQU 2 <0:17> 18 4 Pointer to qualifying
class
.tp 2
ZQUQID 2 <0:17> 18 23 Identifier number of
qualifying class
.tp 2
ZQUZB 2 <18:35> 18 4 Parameter and attribute
list header
.tp 2
ZDELNK 2 <18:35> 18 2 Link in pass 1
structure
.p 0,1,3
ZQUNAC 3 <0:2> 3 34 Number of accumulator
to be reserved
ZQUPR 3 <3> 1 34 Proper procedure
.tp 2
ZQUGB 3 <4> 1 34 Can cause garbage
collection
ZQUIO 3 <5> 1 34 I/O editing
.tp 2
ZQULO 3 <6> 1 34 Identifier more than 6
characters
ZQUQIK 3 <7> 1 4 QUICK procedure
ZQUPTD 3 <8> 1 34 Protected attribute
ZQUDD 3 <9> 1 4 Doubly declared
ZQUSNR 3 <10:17> 8 34 System routine number
.tp 2
ZQULNE 3 <18:35> 18 34 Line number of
declaration
ZQU1ID 2:3 72 16 ZQULID in SIXBIT
.tp 2
ZQUUNR 4 36 6 Unique number if label
attribute
ZQU2ID 4:5 72 16 ZQUQID in SIXBIT
.p -8,4,15;II.6.2 ZHE record
----------
.s;.f;The ZHE record holds information about declaration
scopes of blocks and for-statements. The size is:
.nf;.lm 11;3 words in state 2
2 words in states 3 and 4
4 words in states 5 and 6
.s;.lm 9;
FIELD LOCATION SIZE STATE CONTENTS
.p 0,1,3
ZDETYP 0 <0:2> 3 ALL ZHE%V (ZHB%V if used as
prefix)
ZHETYP 0 <3:5> 3 ALL Type code
.s;If ZHETYP is not QQUACH:
.s;
ZHENOI 0 <11> 1 4 No INNER in this class
.tp 2
ZHESOL 0 <12:17> 6 ALL Source level, block
stack index
.tp 3
ZHEDLV 0 <18:35> 18 23456 Offset to access
display element for
this block
.s;
ZHEEBL 1 <0:4> 5 ALL Effective block level
ZHELEN 1 <5:14> 10 ALL Size of block instance
ZHEBNM 1 <15:23> 9 23456 No of reduced subblocks
.tp 2
ZHEFIX 1 <24:35> 12 1234 Base fixup for this
block
.p 0,1,3
ZDELNK 2 <18:35> 18 2 Link in pass 1
structure
.s;If ZHETYP = QQUACH:
.s;
ZHEOID 0 <6:17> 12 23 Own identifier number
.tp 3
ZHELID 0 <18:35> 18 23 Lexical identifier
number of referenced
external
.tp 2
ZHEUNR 1 36 2356 Unique number of
referenced external
.tp 2
ZDELNK 2 <18:35> 18 2 Link in pass 1
structure
ZHEID 2:3 72 6 ZHELID in SIXBIT
.p -8,4,15;II.6.3 ZHB record
----------
.s;.f;The ZHB record holds information of declaration scopes of
procedures, classes, prefixed blocks and connection blocks.
It is a subrecord to the ZHE record. The size is:
.nf;.lm 11;3 words in state 1
4 or 5 words in state 2
5 words in states 3, 4, 5 and 6
.s;.lm 9;
FIELD LOCATION SIZE STATE CONTENTS
.s;
# 0:1 Same as ZHE
.s;
ZHBZQU 2 <0:17> 18 34 Link to corresp. ZQU
ZHBXID 2 <0:17> 18 2 External identifier
ZHBZHB 2 <18:35> 18 4 Prefix link
.tp 2
ZHBZE 2 <18:35> 18 4 Link to local var. list
from param. list header
.tp 2
ZDELNK 2 <18:35> 18 2 Link in pass 1
structure
.s;
ZHBDEV 3 <0:11> 12 2 Device of ATR file
ZHBNRP 3 <0:7> 8 ALL Number of parameters
ZHBVRT 3 <8:15> 8 ALL Number of virtuals
.tp 3
ZHBSBL 3 <16:20> 5 234 EBL of statically
enclosing complete
block
ZHBSTD 3 <21:25> 5 23456 Display level
ZHBSZD 3 <26:30> 5 456 Max value of STD
.tp 2
ZHBUPF 3 <31> 1 ALL Class used as block
prefix
.tp 12
ZHBMFO 3 <31:33> 3 23456 Non-zero for MACRO-10
or FORTRAN procedure.
Type codes:
QEXMAC (1):
MACRO-10/E:CODE
QEXMQI (2):
MACRO-10/E:QUICK
QEXFOR (4):
FORTRAN/E:FORTRAN
QEXF40 (5):
FORTRAN/E:F40 (old)
ZHBLOC 3 <33> 1 ALL Local object (if class)
.tp 3
ZHBNCK 3 <34> 1 23456 Separately compiled
MACRO-10 procedure
with NOCHECK
.tp 2
ZHBKDP 3 <34> 1 ALL Class with class
attributes
.tp 2
ZHBEXT 3 <35> 1 23456 Separately compiled
class or procedure
.s;
ZHBPPN 4 36 2 PPN of ATR file
.tp 2
ZHBUNR 4 36 2346 Unique number if
external
.p -8,4,15;II.6.4 ZQQ record
----------
.s;.lm 9;.f;The ZQQ record holds information about entry points
of separately compiled modules. ZQQ records are output last
in the intermediate declaration file. The size is:
.i 2;.br;2 words in states 2 and 3
.s;.nf;
FIELD LOCATION SIZE STATE CONTENT
.s;
ZQQLNK 0 <0:17> 18 2 Link next ZQQ
ZQQFIX 0 <18:35> 18 23 Fixup
ZQQUNR 1 36 23 Unique number
.sp 1;.nonum;.lm 9;.ps 50,72
.ts 1,9,17,25,33,41,49,57,65
.pg;.nm appendix 1
.ax SIMULA FOR DEC SYSTEM 10#############TD, COMPILER PASS 1
.nm 1
.st 741025####780302######6###########################Claes Wihlborg/LE
.pg;.lm 9;.p -8,0,10;II.A####PASS 1
.p -8,2;II.A.1##Pass description
.br;----------------
.s;The first pass of the SIMULA compiler executes in three phases.
.br;In phase 1, the low segment is initialised and the file
specifications are read.
.br;In phase 2, the source file is read and lexical and syntactical
analysis are performed.
.br;In phase 3, the declarations found in phase 2 are merged with
external attributes (if any) and sorted to be suitable for reading
in pass 2.
.p -8,2,10;II.A.1.1#Important algorithms used
.s;MODULE SH (Symbol Hash)
.br;-----------------------
.br;The symbol table is organised to be searched by a combination of
hash-coding and binary search.
The hashing function maps all identifiers on 128 binary trees by
multiplication of the first six characters with a constant and
taking the middle 7 bits as a hash code.
.p 0,1,10;MODULE LS (Lexical Scanner)
.br;---------------------------
.br;The lexical scanner inputs characters one by one and recognises the basic
entities of the language. Each basic entity is output as one or more
symbols to the caller (the syntax recogniser). The following functions
are performed by the scanner:
.lm 12;.i -3;*##Skips all comments, blanks and special control characters.
.i -3;*##Outputs text constants to the rel file.
.i -3;*##Converts arithmetic constants to internal format.
.i -3;;*##Converts delimiters and reserved words to internal representation.
.i -3;*##Converts identifiers to integer code numbers.
.lm 9;.p 0,1,10;MODULE SR (Syntax Recognition)
.br;------------------------------
.br;The syntax recogniser controls the process of recognising and checking
the structure of the source text. A top-down method for syntax recognition
(recursive descent)
is employed. Each syntactic definition (production rule) is recognised
by its own recognition routine, which calls other recognition routines
recursively, until the analysed sentence is reduced to terminal symbols.
A recognition routine attempts to find an instance of the syntactic
class defined by the production. The right hand side is transcribed
into a sequence of calls on the corresponding recognition routines
for syntactic classes, interspersed with direct comparisons for terminal
symbols. Comparisons are followed by reading the next symbol.
.br;*##Error recovery:
.br;If a recognition routine discovers an error in the source text
it will normally be able to go on by recovering from the error, usually
by ignoring some of the following symbols from the source text.
The general principle is that if a recognition routine can identify
an error, it will recover from that error.
.br;*##Rearrangement of syntactic definitions:
.br;The syntax as published in the Common Base document is not directly
suitable as basis for recognition routines.
The definitions are therefore rearranged, which serves two purposes.
.br;1) Some changes are essential to make the analyser function correctly
in a top-down fashion, e.g_. left recursion must be removed.
.br;2) Other changes make the recognition process more efficient.
.br;Some modifications are:
Removal of redundant definitions, reduction of self-recursion
to iteration, factorisation of definitions and alternatives to avoid
back-tracking. To that end a one symbol look-ahead is also employed.
.br;*##Implementation language:
.br;To ensure standardisation and simplify development the recognition
routines are written in a high level macro language.
.br;Among the powerful facilities available we may mention:
.br;Optimised operand evaaluation - the macro system will automatically
choose the correct instruction depending on the operand being a
constant, in an accumulator or in a memory cell.
.br;Optimised predicate evaluation - the macro system will choose
the instructions needed to evaluate a Boolean expression.
.p 0,1,10;MODULE SD (Syntax Dispatch)
.br;---------------------------
.br;This module is called from SR and takes care of the declarations,
creating the DC1 list. By using an advanced list structure, SD
makes it possible for the DP module to read the DC1 list in a
straightforward manner. A stack with 64 levels is employed in the
list-manipulation code. Each stack level keeps track of 5 linear lists
of declarations.
.p 0,1,10;MODULE DP (Declaration Processing)
.br;----------------------------------
.br;This module transforms the DC1 list and merges it with any external
attributes creating the DF1 file. Only those attributes which are
used (i.e_. mentioned in the compiled code, not only in the external
attribute file), are merged in, thus minimising the file.
.br;*** Problem:
.br;Unfortunately, this can lead to a virtual procedure definition
being "lost" in certain rare cases. A modification of
this algorithm would thus seem to be necessary, but not urgent.
.st 740805####780302######6###########################Lars Enderin
.p 0,2,10;MODULE O1 (INPUT/OUTPUT)
.br;------------------------
.f;.br;This module performs the I/O operations necessary in Pass 1.
The operations are normally straightforward, except for attribute file
handling, which is done by O1EX.O, O1EXBU, O1EXFM, O1EXLU, O1EXNP, O1EXRM,
O1EXRQ, O1EXSO, O1EXT, O1EXTB.
.p 0,1,5;* Chain of ZQU records for externals:
.br;When an EXTERNAL declaration is encountered,
SDEXT makes skeleton ZQU, ZHB and ZHE(QQUACH) records and puts the
ZQU record on the chain starting at YEXZQU. In older versions of SIMULA,
the ATR file would be read at that stage. Reading is now deferred
until the DP module has merged system components with declarations
from the source program.
.s;* Lookup of ATR files:
.br;DPEXFA (find all ATR modules) is called to
scan the YEXZQU chain and input all the necessary information form
ATR files. Looking for ATR files proceeds in two phases:
.i -3;.s;1) Looking up separate ATR files:
.br;Each ATR file is first looked up as a separate file on device
DSK: (assuming the simplest form of external declaration). O1EX.O
opens channel QCHEXT if necessary.
A lookup operation tries to find the relevant ATR file (O1EXLU).
.br;* Reading the first block of the file (O1EXTB):
.br;If the file is found, the ZQU is removed from the chain and the first
block of the ATR info is read by O1EXTB (which calls O1EXBU to allocate
buffers).
.br;* Updating the request block chain:
.br;O1EXRQ places a request block on the YEXZRQ chain (see SEARCH
list handling below).
.br;* Processing the ATR file:
.br;DPEXT is then called and will call O1EXT to read
info from the ATR module.
.br;* O1EXCL will close the ATR file.
.s;Variations on this scheme exist:
.br;* Compiling an external module:
.br;If a global module is being compiled and the module found is an
earlier version of the global module, O1EXNP takes note of where the
ATR module was found (file spec, word within file) for the benefit
of pass 3 which will later read the file. O1EXRQ and DPEXT are not
called.
.s;* Specific ATR file requested:
.br;If the external declaration requested a specific file (device
and/or path explicitly given, or '=' used in declaration),
O1EXFM (findmodules) is called after a successful lookup to check
if the file is a library (started by a LOADER index block) or a separate
ATR file. If the file is not a library,
the file is read directly, otherwise
the relevant module is hopefully found in the library.
In both cases, O1EXRM reads the header info up to the actual ATR
info, then calls DPEXT to process the atr info, using O1EXT to read
words from the file.
.s;At the end of phase 1, all ZQU records have been scanned once.
If some external references have not been found then, phase 2 starts.
.s;.i -3;2) Looking for ATR modules which were not found as separate files:
.br;If any ZQU records remain unresolved on the chain after the first
phase, the SEARCH list of ATR libraries is tried next.
.s;* Definition of the SEARCH list:
.br;The SEARCH list for ATR files is defined as follows:
.br;When compilation starts, SP sets up a DEFAULT search list by calling
I1SRCH.
I1SRCH uses the table YI1SRC to set up an initial chain of ZRQ records,
containing:
.p 8,1,5;ZRQDEV - Device name (sixbit)
.i 8;ZRQFIL - File name (sixbit)
.i 8;ZRQPPN - ppn (or full SFD path pointer)
.p 0,1,5;The chain is handled by O1EXRQ. YRQHEAD points to the LAST inserted
ZRQ record, which points to its predecessor via its ZRQZRQ field, etc,
until ZRQZRQ = 0 is found.
New ZRQ records are put on the chain if an identical file spec
(YRQDEV, YRQFIL, YRQPPN compared to ZRQ....) cannot be found. The ZRQSRC switch
shows if a file belongs to the search list. See request blocks below.
.p 0,1,5;The search list is modified by /SEARCH switches.
.s;/SEARCH:"spec" puts a ZRQ block corresponding to "spec" at the front end
of the chain (pointed to by YRQHEAD). No duplicate entries are created, however,
which means that a ZRQ record stays in the same place in the chain once it is
put there.
.p 0,1,5;If the /SEARCH switch has the form
.br;/-SEARCH:"spec", however, it means that the corresponding ZRQ record should
be removed from the chain. This is not really done by taking the ZRQ record
away, but by resetting the ZRQSRC flag (one bit).
.p 0,1,5;If the /SEARCH switch is given without a file spec, the DEFAULT search list
is reinstated, while any additional files remain.
.p 0,1,5;A negated switch without file spec (/-SEARCH) removes ALL files from the search
list by turning off the ZRQSRC bits.
.p 0,1,2;See also TD II-4.3.
.p 0,1,10;* Request blocks in REL file:
.br;The ZRQ chain described above is also used to define
the request blocks which are written to the REL file. Request blocks are
loader blocks of type 17 containing the information ZRQFIL, ZRQPPN, ZRQDEV
copied from YRQFIL, YRQPPN, YRQDEV when the ZRQ block was defined
(by O1EXRQ). The linking loader (LINK-10) will scan the corresponding file
(extension REL assumed) in library search mode, before scanning SIMLIB.REL.
When O1EX determines that a file - dev:x.ATR[p,pn] - is the correct ATR
file to use for resolving an external reference to a class or procedure,
the file spec (YRQDEV, YRQFIL, YRQPPN) is checked against all records on the
ZRQ chain, starting with YRQHEAD. O1EXRQ does this. If the file is not on
the chain, a new ZRQ record is created (via SDALLOC) and chained via
YRQHEAD and ZRQZRQ. When returning to O1EXLU,
the ZRQOUT bit is set for the record (also
in YRQHEAD to show that there is information available for at least one
request block).
.p 0,1,5;* Output of request blocks or command strings in REL file:
.br;At the end of PASS 1 processing, O1RLRQ is called, provided
YRQHEAD is non-zero, to output any request blocks defined by ZRQ records
with the ZRQOUT bit set.
Only device, ppn and file name without extension are specified. The
extension REL is assumed.
Thus no SFD can be defined. For SFD's, ASCIZ commands are instead put into the
REL file by O1RLOA called from O1RLRQ: "dev:atrfil[path]/SEARCH".
Only ASCIZ strings are output in the TOPS-20 version.
.p 0,1,5;* Scanning the search list:
.br;The ATR files on the search list are scanned in turn. In order
to reduce the necessary file positionings, however, the remaining
ATR files are matched against the index blocks by scanning the index
blocks in order and checking the list against each module name in
the index block before going on to the next index block and eventually
to the next ATR file if necessary. As modules are found in the index
block currently active, the ATR module is read from one of the following
blocks avoiding unnecessary i/o operations by taking advantage of
the fact that each module is placed in the file in the same order
as the index block entries and following the index block.
.br;DPEXFA contains the main loop over the ZQU list and over the search
list, O1EXFM finds all modules matching the ZQU list in the current
library. Each matched ZQU is taken off the list by DPEXRM and processed
by DPEXT which uses O1EXT for I/O. O1EXSO skips overhead information
(entry and name blocks), O1EXTB allocates buffers and reads the first
block off each new file.
At the end of the search list, if there are still unmatched ZQU's,
a compilation error will be issued.
.p 0,1,10;* Possible future modification:
.br;The algorithm described above will have to be modified if INDIRECT
external references are to be included in the program without explicit
declaration. It may possibly be advantageous to read the directories
(UFD, SFD files) corresponding to DSK: saving the ATR file names
in order to avoid unnecessary lookup operations, since most ATR modules
tend to be placed in the libraries anyway. It will perhaps also be necessary
to keep all index blocks in core to handle indirect external references,
if excessive i/o operations should be avoided.
.pg;.i -8;II.A.2##Pass decomposition
.br;-------------------
.nf;.s;.i -8;II.A.2.1 Components.
.s;Pass 1 consists of the following modules:
.s;D1 Miscellaneous tables
DM Diagnostic messages handling
DP Create intermediate declaration file
I1 Set up low segment and read commands
LC Input one line of source code
LS Lexical scanner
O1 Input/Output routines
SD Semantic processing of declarations
SH Symbol table handling
SM Miscellaneous semantic routines used by SR
SP Perform first pass main scan
SR Syntax recognition routines
SYS1 Table of SIMULA system components
T1 Termination routines
.f;.s;The following modules are only present in the debug version
of the compiler:
.s;.nf;P1DUMP Miscellaneous debug routines
TRACE Tracing code (traces each instruction)
TRACEI Tracing code initialisation
.lm 1;.p 0,2,26;II.A.2.2 Main flow.
.s;.nf;
+----------------+ +----------------+ +----------------+
! I1 !----->! SP !----->! T1 !
! Initialisation ! ! Syntax parser ! ! Termination !
+----------------+ +----------------+ +----------------+
! !
v v
+----------------+ +----------------+
! SR ! ! DP !
+-------! Syntax recogn. !-+ ! Decl. process. !
! +----------------+ ! +----------------+
! ! !
! ! !
+----------------+ +----------------+ ! +----------------+
! SM ! ! LS ! +----! SD !
! Syntax, misc. ! ! Lexical scan ! ! Syntax dispatch!
+----------------+ +----------------+ +----------------+
!
+----------------+
! LC !
! Line control !
+----------------+
.pg;.lm 9;.p -8,0,6;II.A.2.3 Module contents
.s;.f;The following list should serve as an index into the compiler
listings, which must be consulted for any detailed study of pass 1
functions.
.nf;.p 0,2,4
MODULE D1
---------
.s;
Tables, no executable code.
.p 0,2,6
MODULE DP
---------
.f;.lm 17;
.p -8,1,4;DP Output DC1 list to DF1 after modifications.
Call DPEXFA to find all external attribute files needed.
Check for conflicting unique id's in external modules.
.p -8,0,4;DPEXT Read an attribute module (separate ATR file or part of
a library). If compiling a main program, generate code for prototypes
and symbol tables of external MACRO-10 and FORTRAN procedures.
.p -8,0,3;DPEXTC Append external attributes to the DC1 list.
Calls itself recursively.
.p -8,0,3;DPEXCR Create external symbol request loader block
for REL file.
.p -8,0,2;DPEXDF Define external name corresponding to a fixup number.
.p -8,0,2;DPSYS Append system elements to the DC1 list.
.p -8,0,4;DPEXFA (findemall). Finds ATR files corresponding
to ZQU records for external declarations.
Tries separate files first, then libraries on search list.
.p -8,0,3;DPEXER Issue error message for external file error.
.p -8,0,3;DPEXGD Get device name from current ZHB (id or 'DSK').
.p -8,0,3;DPEXFS File spec for ATR: YRQFIL:=name, YRQPPN:=ppn.
.p -8,0,3;DPEXRM Remove ZQU record from YEXZQU chain.
.p -8,2,8;MODULE I1
.p -8,0,1;---------
.p -8,1,2;I1CHAR Get next command character.
.p -8,0,2;I1CM Read a command line (TMPCOR, TMP file or TTY).
.p -8,0,2;I1TTYC Read a command line from the TTY.
.p -8,0,2;I1DIR Parse a directory specification.
.p -8,0,2;I1FN Parse a file specification.
.p -8,0,2;I1MOVE Move a line from command buffer to command
line buffer.
.p -8,0,2;I1PC Put a character in command line buffer (PUTCHAR).
.p -8,0,2;I1NAME Scan a name (identifier) from command.
.p -8,0,2;I1PPN Parse and convert an octal number from command file.
.p -8,0,2;I1REN Enter Pass 3 after REENTER command.
.p -8,0,2;I1RX50 Conversion SIXBIT -> RADIX50.
.p -8,0,2;I1SC Create OPEN and LOOKUP blocks for next source file.
.p -8,0,2;I1SRBP Check compiler switches that must appear before
the start of the program text proper.
.p -8,0,8;I1SRTx A set of routines, each of which handles a
compiler switch, e.g_. I1SRTA handles the ARRAY switch. The initial
letter of the switch name is substituted for x to get the name of
the routine. See SIMCOM.HLP or the SIMULA Language Handbook part II,
Sect 3.5.1, for available switches.
.p -8,0,2;I1SW Recognise compiler switches.
.p -8,0,2;I1SWIT Process compiler switches.
.p -8,0,2;I1RECO Recover from switch error.
.p -8,0,2;I1SWER Switch error handler.
.p -8,0,2;I1SRCH Set up standard search list.
.p -8,0,2;SIMULA Primary entry point to compiler (pass 1).
.p -8,0,2;I1OVFL Stack overflow handler.
.p -8,0,2;I1UNFL Stack underflow handler.
.p -8,2,8;MODULE LC
.p -8,0,1;---------
.p -8,1,2;LC Input next source line.
.p -8,0,2;LCEOF End of file in source.
.p -8,0,2;LCLINE Update source line number.
.p -8,0,2;LCLS1 Output record to LS1.
.p -8,0,2;LCMORE
.p -8,0,2;LCNEXT Move source line to line buffer from file buffer
.p -8,0,1;LCPLIN Create LS1 records.
.p -8,2,8;MODULE LS
.p -8,0,8;---------
.p -8,1,2;LS Get next symbol from source line.
.p -8,0,2;LSC Scan character constant.
.p -8,0,2;LSCHBE
.p -8,0,2;LSCHBL Scan past blank characters.
.p -8,0,2;LSCHGR Handle line control characters.
.p -8,0,2;LSCHIG Skip ignorable characters.
.p -8,0,2;LSCHN1
.p -8,0,2;LSCHNP Error messages for non-printable characters.
.p -8,0,2;LSEND Scan end comment.
.p -8,0,2;LSI Handle identifiers and reserved words.
.p -8,0,2;LSIOSC Scan characters from OPTIONS statement.
.p -8,0,2;LSIPAG Scan PAGE switch.
.p -8,0,2;LSIS Scan identifiers and reserved words.
.p -8,0,2;LSLNSM Prepare line number for error message.
.p -8,0,2;LSN Scan numeric constant.
.p -8,0,2;LSNCO Convert INTEGER -> LONG REAL.
.p -8,0,2;LSNDL Check decimal point not last char of number.
.p -8,0,2;LSNEXP Scan exponent part of number.
.p -8,0,2;LSNKEX Look-ahead for keyword.
.p -8,0,4;LSNRDD
.p -8,0,1;LSNRDN
.p -8,0,1;LSNRDS Scan number in non-decimal radix.
.p -8,0,2;LSNSCE Scale number.
.p -8,0,2;LSSEM Update semicolon count.
.p -8,0,2;LST Handle debug symbols in source code.
.p -8,0,2;LSX Handle TEXT constant.
.p -8,0,2;LSXTXT Scan TEXT constant.
.p -8,2,8;MODULE O1
.p -8,0,1;---------
.p -8,1,2;O1DBOP Open file nnnDEB.TMP (only in debug mode).
.p -8,0,2;O1DB6 Output 3 words to nnnDEB.TMP (debug only).
.p -8,0,2;O1DFOP Allocate space for file DF1.
.p -8,0,2;O1OPDF Open file nnnDF1.TMP and output the data which is in core.
.p -8,0,2;O1DF1 Open nnnDF1.TMP unless already open. Output a buffer.
.p -8,0,2;O1DFCL Close nnnDF1.TMP if open.
.p -8,0,2;O1EXBU Set up buffers for ATR file.
.p -8,0,2;O1EXFM Find ATR modules in ATR library file.
.p -8,0,2;O1EXFW Find first word of an index block (ATR library).
.p -8,0,2;O1EXMP (TOPS-20 only) Make sure wanted page of ATR file is mapped
in core.
.p -8,0,2;O1EXMA (TOPS-20) Map ATR file pages on core area.
.p -8,0,2;O1EXNP Note position of the old ATR module (file and word in file).
.p -8,0,2;O1EXRM Prepare for DPEXT to read ATR module (position to first word).
Call DPEXT for processing.
.p -8,0,2;O1EXSO Skip ATR module overhead (position to first word of attr.
list).
.p -8,0,2;O1EXLU LOOKUP ATR file.
.p -8,0,2;O1JFNI (TOPS-20) Get input JFN for file defined by TOPS-10 lookup
block. Open the file for input (O1OJFI).
.p -8,0,2;O1OJFI (TOPS-20) Perform OPENF JSYS. Records file size (SIZEF JSYS).
.p -8,0,2;O1JFN (TOPS-10) Get a JFN for input or output.
Use TOPS-10 lookup block for file spec.
.p -8,0,2;O16TO7 Translation ASCII -> SIXBIT.
.p -8,0,2;O1EXRQ Define request block information.
.p -8,0,2;O1SFDC Copy SFD record. Used e.g_. by O1EXRQ when SFD was specified.
.p -8,0,2;O1EX.O Perform OPEN UUO for channel QCHEXT. (TOPS-20 version just
records device name).
.p -8,0,2;O1EXTB Allocate ATR file buffers, then fall into O1EXT code.
.p -8,0,2;O1EXT Read a buffer from the current ATR file.
.p -8,0,2;O1EXCL Close ATR file. (TOPS-20 version first unmaps file).
.p -8,0,2;O1ICOP Allocate space for IC1 file.
.p -8,0,2;O1OPIC Open file nnnIC1.TMP and transfer pending data from core to
file.
.p -8,0,2;O1IC1 Output a buffer to nnnIC1.TMP, if necessary after opening
the file.
.p -8,0,2;O1ICCL Close nnnIC1.TMP if open.
.p -8,0,2;O1LSOP Allocate space for LS1 file.
.p -8,0,2;O1OPLS Open file nnnLS1.TMP and transfer pending data from core to
file.
.p -8,0,2;O1LS1 Output a buffer to nnnLS1.TMP, if necessary after opening
the file.
.p -8,0,2;O1LSCL Close nnnLS1.TMP if open.
.p -8,0,2;O1RLOP Open rel file and output entry block and name block, also
source file identification (lookup info).
.p -8,0,2;O1RL Output an unrelocated word to REL file.
.p -8,0,2;O1RLR Output a relocated word to REL file.
.p -8,0,2;O1RLS Output a symbol item to REL file (part of type 2 loader block).
.p -8,0,2;O1RLC Output a relocation word to REL file.
.p -8,0,2;O1RLD Close current code item.
.p -8,0,2;O1RLNB Start new rel code buffer.
.p -8,0,2;O1RLIC Start new code item (type 1 loader block).
.p -8,0,2;O1REL Write a buffer to REL file.
.p -8,0,2;O1RLRQ Output request info to REL file for all REL files to be searched by
the linking loader for matching external references corresponding
to ATR files which were used. The info is either type 17 blocks or
ASCIZ strings.
.p -8,0,2;O1RLOA Output ASCIZ string to REL file: "dev:file[path]/SEARCH".
.p -8,0,2;O16BIT Convert SIXBIT -> ASCII calling O1RLOA for each character.
.p -8,0,2;O1OCT As O16BIT, but converts binary integer to octal ASCII characters.
.p -8,0,2;O1RLUN Output comment record (type 0 block) containing all external
references, also any type 17 block corresponding to ATR files.
.p -8,0,2;O1SCOP Look up source file.
.p -8,0,2;O1SC Read a buffer from source file.
.p -8,0,2;O1SCCL Close source file.
.p -8,0,2;O1XROP Allocate space for XRF file in core.
.p -8,0,2;O1OPXR Open file nnnXRF.TMP. Output currents contents of core file area.
.p -8,0,2;O1XRF Output a buffer to nnnXRF.TMP, having opened the file if not
open before.
.p -8,0,2;O1XRCL Close nnnXRF.TMP if open.
.p -8,0,2;O1ZSOP Open file nnnZSE.TMP.
.p -8,0,2;O1ZSE Output buffer to nnnZSE.TMP.
.p -8,0,2;O1ZSCL Close nnnZSE.TMP.
.p -8,0,2;O1ZS Write symbol table.
.p -8,0,2;O1ERR Error message output: ? FILE xxx NOT FOUND
.p -8,0,2;O1PACK Pack core files in core file area.
.p -8,0,2;O1SETB Set up buffer rings.
.p -8,2,8;MODULE SD
.p -8,0,1;---------
.p -8,1,2;SDBEG Start of block, procedure, inspection etc.
.p -8,0,2;SDEND End of block, procedure, inspection etc.
.p -8,0,2;SDZQU Declared quantity.
.p -8,0,2;SDEXT External declaration.
.p -8,0,2;SDSPEC Parameter specification.
.p -8,0,2;SDESPE End of parameter specifications.
.p -8,0,2;SDABEG Start of array segment.
.p -8,0,2;SDAEND End of array segment.
.p -8,0,2;SDPEND Program end.
.p -8,0,2;SDPPN Reevaluates proj or progr no as octal.
.p -8,0,2;SDALLO Allocate space in DC1 list.
.p -8,0,2;SDHID HIDDEN PROTECTED specification.
.p -8,0,2;SDPRO Protect attributes at classend.
.p -8,2,8;MODULE SH
.p -8,0,1;---------
.p -8,1,2;SH Transform identifier from SIXBIT to internal identifier no.
.p -8,0,2;SHNEW Append a new identifier to symbol table.
.p -8,2,8;MODULE SM
.p -8,0,1;---------
.p -8,1,2;SMERR Generate error message for syntax error.
.p -8,0,2;SMLIND Output line symbol preceding declaration.
.p -8,0,2;SMLINE Output line symbol preceding executable statement.
.p -8,0,2;SMUID Output referenced identifier to IC1 and XRF.
.p -8,2,8;MODULE SP
.p -8,0,1;--------
.p -8,1,2;SP Initialise low segment and start syntax parsing.
.p -8,0,2;SPDCIN Initialise DC1 area.
.p -8,2,4;MODULE SYS1
.p -8,0,1;-----------
.p -8,1,2;Contains no executable code.
.p -8,2,8;MODULE SR
.p -8,0,1;---------
.lm 9;.p 0,1,3;
Labels in this procedure are formed from the recognition routine
names prefixed by SR. The recognition routines are:
.lm 17
.p -8,1,2;APAR Actual parameter part.
.p -8,0,2;ARY Array declarations.
.p -8,0,2;ARYS Array subscripts.
.p -8,0,2;BLK Block.
.p -8,0,2;CLAS Class declaration.
.p -8,0,2;ECLA Check switches for external class.
.p -8,0,2;ELST IF-statement.
.p -8,0,2;EPRO Check switches for external procedure.
.p -8,0,2;EXPR Expression.
.p -8,0,2;EXT External declaration.
.p -8,0,2;FLST FOR list element.
.p -8,0,2;FPAR Formal parameter part.
.p -8,0,2;LAB Labels.
.p -8,0,2;LST Specification identifier list.
.p -8,0,2;MAIN Check switches for main program.
.p -8,0,2;PBLK Prefixed block.
.p -8,0,2;PRG Program starting with current symbol.
.p -8,0,2;PROC Procedure declaration.
.p -8,0,2;PROG SIMULA program.
.p -8,0,2;SEXP Simple expression.
.p -8,0,2;SIST FOR, INSPECT or WHILE statement.
.p -8,0,2;SPEC Procedure specification.
.p -8,0,2;STAT Statement.
.p -8,0,2;TYPE Type.
.p -8,0,2;UNST Unconditional statement.
.p -8,0,2;VIRT Virtual specification.
.p -8,2,8;MODULE T1
.p -8,0,1;---------
.p -8,1,2;T1 Terminate pass 1 and enter pass 2.
.p -8,0,2;T1AB Termination error encountered. Enter pass 3 directly.
.p -8,2,8;MODULE P1DUMP
.p -8,0,1;-------------
.p -8,1,2;_OUTOP Open file nnnDEB.LST.
.p -8,0,2;_.OUT Output a buffer to nnnDEB.LST.
.p -8,0,2;_OUTCL Close nnnDEB.LST.
.p -8,0,2;_.IN Read a buffer from temporary file.
.p -8,0,2;_.NEWPA Output new page and page header to nnnDEB.LST.
.p -8,0,2;_.NEWLI New line.
.p -8,0,2;_.PUTT Text output.
.p -8,0,2;_.PUTTAR Text array output.
.p -8,0,2;_.PUTID Identifier output.
.p -8,0,2;_.PUTSIX Sixbit text output.
.p -8,0,2;_.PUTOCT Octal number output.
.p -8,0,2;_.PUTDEC Decimal number output.
.p -8,0,2;_.PUTN Number output.
.p -8,0,2;_.ZH Output fields common to ZHE and ZHB.
.p -8,0,2;_.ZQU Output ZQU record.
.p -8,0,2;_.ZHE Output ZHE record.
.p -8,0,2;_.ZHB Output ZHB record.
.p -8,0,2;_DFDUMP Output symbolic dump of DF1 file.
.p -8,0,2;_ICDUMP Output symbolic dump of IC1 file.
.p -8,0,2;_LSTRAC Output trace of symbols from LS.
.p -8,0,2;_P1INIT Initialise debug output.
.p -8,0,2;_P1DUMP Dump of intermediate files. Symbol table output to nnnDEB.TMP.
.lm 9;.p -8,4,10;II.A.3##Low segment organisation
.br;------------------------
.s;The low segment of pass 1 consists of three parts:
.lm 13;.p -4,1,2;1)##A global part (see II.3).
.p -4,0,4;2)##A fixed pass 1 specific part. See LOWSEG.MAC and the
modules that refer to the data for layout descriptions.
.p -4,0,2;A dynamic part containing the DC1 structure (see II.6).
.p -4,1,6;The sizes of the parts are:
.s;Part 1: #4 K
.br;Part 2: 10 K
.br;Part 3: .5 K (initially)
.lm 9;.p -8,4,10;II.A.4##Pass 1 interface
.br;----------------
.p -8,2,8;II.A.4.1 Pass initialisation
.p 0,1,1;Pass 1 has five entry points:
.i 4;SIMULA Entered via .R SIMULA or .START
.i 4;SIMULA+1 Via COMPIL/SIMULA
.i 4;I1START-1 Entered from pass 3.
.i 4;I1START Entered when STARTed after first start.
.i 4;I1REN Entered via REENTER.
.br;The entry points are located in the I1 module.
.p -8,2,6;II.A.4.2 Pass 1 termination
.p 0,1,1;Pass#1 enters pass#3 if an error has occurred, otherwise
pass#2.
.p -8,2,10;II.A.4.3 Job data area
.p 0,1,1;The following locations are set specially in pass 1:
.lm 13;.s;.nf;_.JBSA(RH) SIMULA
_.JBREN I1REN
_.JBAPR I1OVFL
_.JB41 PUSHJ 17,DM
.f;.lm 9;.p -8,2,8;II.A.4.4 UUO's used
.p 0,1,1;The following user defined UUO's are used in pass 1:
.br;ERR, ERRLI, ERRT, ERRI1, ERRI2.
.br;They are defined in SIMMAC.MAC (universal file).
.lm 1;.p 0,4,30;.nf;II.A.4.5 Input/Output in pass 1
.s 3;
+----------+ -------------- ----------------------
!nnnSIM.TMP! ) Source code ) ) External attributes )
+-----+----+ -------------- ----------------------
Pass 1 - ! - - - - - - - -!- - - - - - - - ! - - - -
! V V V !
+----------+ +----------+ +-----------+
! ! Phase 1 !====>! Phase 2 !====>! Phase 3 !===!=> Next pass
+----------+ +----------+ +-----------+
! ! !!!! !!! !
- - - - -!- -- - - -!!!! - - - - - - - !!! - - - - -
! !!!! !!!
! !!!! ------------ !!! ------------
! !!!-> ) nnnIC1.TMP) !!+---> ) nnnDF1.TMP)
! !!! ------------ !! ------------
! !!! ------------ !! ------------
! !!+-> ) nnnLS1.TMP) !+----> ) nnnZSE.TMP)
! !! ------------ ! ------------
! !! ------------ !
! !+--> ) nnnXRF.TMP) !
! ! ------------ !
! ! +-----> ------------
! +-------------------------> ) nnnREL.TMP)
+------------------------------------> ------------
.f;.lm 9;.p -8,4,8;II.A.5##Common macro file
.br;-----------------
.s;Pass 1 code refers to the universal file SIMMC1.UNV which
defines macros, fields, switches and constants used in pass 1.
SIMMAC.UNV is also used. See listing of SIMMC1.MAC for details.
.p -8,4,8;II.A.6##Testing
.br;-------
.s;The debug version of pass 1 contains some built-in features
for testing. The special features are controlled by insertion
in the source code
of the special debug character (control-uparrow, octal 036)
followed by a code sequence.
The code sequences are:
.s;.nf;LS1 Start trace of lexical scanner output.
LS0 Stop trace of lexical scanner output.
DF1 Create symbolic dump of DF1.
IC1 Create symbolic dump of IC1.
.f;.p 0,1,5;Examples of test output are given in VIII.7. In the trace
of LS1 output and in the dump of IC1 all internal compiler symbols
start with percent (%) and any values are given in octal radix.
Identifiers are written as in the source (only 12 characters).
The optional second value for the %LINE symbol is the semicolon count.
.p 0,1,5;The DF1 dump contains one line per record. The line starts
with the offset of this record in the current declaration segment and
is followed by the record type. The rest of the line contains the field
values of the record.