github.com/PDP-10.its
1
0
Fork 0
mirror of https://github.com/PDP-10/its.git synced 2026-01-25 19:56:53 +00:00
Code Issues Releases Wiki Activity

PALX - PDP-11 cross assembler.

Browse Source
This commit is contained in:
Lars Brinkhoff
2016-12-09 20:35:34 +01:00
parent fbcecf6092
commit 02f60d544c
8 changed files with 8231 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

170
doc/_info_/palx.format Executable file
View File

@@ -0,0 +1,170 @@
Format of PDP11 binary files as stored on ITS
Output produced by PALX and :ABSDMP in 11simulator starts with is
PDP11 paper tape absolute loader format, stored 1 byte per pdp10 word.
The symbol table starts immediately after the end of pdp11 core.
The entire file is divided into blocks, usually the PDP11 core image
is a series of blocks, increasing in core address. Blocks are generally
less than 120 bytes in length, although the format theoretically can
allow nearly an entire address space (actually 65530 bytes) to
be stored as one block.
A typical block: (Note each PDP11 byte is stored right justified,
that is in the low-order end, of a PDP10 word)
------------------------
| 000000000001 | always there, means absloader format block
| 000000000000 | always there
| low order byte | of count of number of bytes in block
| high order byte | includes the initial 1 and 0 but not checksum
| low order byte | of address of this data
| high order byte | of address
| _ _ _ _ _ _ _ |
|/ \/ \/ \/ \/ \/ \/ \/
.
. Byte count minus 6 bytes of data,
. low order byte first then high order
.
_/\_/\_/\_/\_/\_/\_/\_/|
| |
| |
| checksum | of all bytes of block including the 1,0,count
------------------------ and the address. Sum of all bytes + checksum
should equal 0 modulo 256.
The last block of data to be loaded into the 11 is has a byte count
of 6. The load address of that block is the starting address for the
program, unless it is odd in which case the processor is supposed to halt.
The symbol table block always seems to be after the PDP11 core image
data and looks like:
------------------------
| 000000000002 |
| sixbit symbol |
| flags,,value |
| sixbit symbol |
| flags,,value |
| _ _ _ _ _ _ _ |
|/ \/ \/ \/ \/ \/ \/ \/
.
. any number of symbols followed by value cells...
.
_/\_/\_/\_/\_/\_/\_/\_/|
| |
| 000000000000 | apparently ended just by a word of zero
------------------------ in the word the symbol is usually in.
The flags appear in the left half, I've only found the first 4 useful
but just in case PALX's internal definitions for the rest of the bits is
given.
HKLSYM==020000 ;HALF KILLED SYMBOL
UNDSYM==010000 ;UNDEFINED SYMBOL FLAG
REGSYM==004000 ;REGISTER
LBLSYM==001000 ;LABEL
INDSYM==000040 ; VALUE OF SYMBOL DEPENDS ON ANOTHER SYMBOL
ENTSYM==000400 ; SYMBOL IS AN ENTRY POINT
EXTSYM==000200 ; SYMBOL IS EXTERNAL
RELSYM==000100 ; SYMBOL HAS RELOCATABLE VALUE
MDLSYM==002000 ;MULTIPLY DEFINED LABEL FLAG
SUPSYM==040000 ;DON'T OUTPUT THIS SYMBOL.
NCRSYM==100000 ;DON'T CREF THIS SYM.
INISYM==200000 ;PDP-11 INSTRUCTION, ERROR IF REDEFINED (BUT OK TO EXPUNGE).
PALX "Newbin" format
The disk format used by the ITS and Sail version of Palx are rather
inefficient of disk space and take quite a bit of interpetation (e.g.
converting symbols from sixbit to rad50) in order to load into a pdp11
debugger such as Rug. The "newbin" format is produced by a palx which has
been generated with the newbin flag on. This is default for the Twenex
version. There is an ITS version under the name NPALX that does this
also. The newbin format is exactly the pdp11 loading format described
below stored with four 8 bit bytes high order justified in a pdp10 word
exactly as you would expect for use with pdp10 byte pointer instructions.
PDP11 Loading Protocol
The format of the data used when loading a PDP11, is essentially
the absloader format, except that the data path is only 7 bits wide.
The PDP11 must send a positive or negative acknowledge after each block.
If positive the next block is sent, if not the last block is repeated
until a postive acknowledge is recieved.
The stream of characters as it comes to a PDP11 looks like:
------------------------
| 001 | always there, the 1st 2 bytes are meant to cause the
| 002 | PDP11 Consol program to go into "load" mode.
------------------------
------------------------
| 001 | \
| 000 | ]
| _ _ _ _ _ _ _ | ]
|/ \/ \/ \/ \/ \/ \/ \/ \
> absloader format blocks
_/\_/\_/\_/\_/\_/\_/\_/| /
| | ]
| checksum | /
------------------------
If PDP11 sends back a 4, the next block is sent,
if anything else (usually a 6) is recieved, the last
block is re-sent.
After the last abloader format block (indicated by
a block with a byte count of 6) the PDP11 sends back
a 5 if it wants symbols sent, or a 47 if it doesn't.
If symbols are to be sent they are sent as data to be loaded into
Rug's symbol table. Symbol table entries consist of 2 rad50 words
for the symbol name followed by 1 word of symbol value. After
16 symbols, two words containing bit flags specify which of the 16
symbols are half-killed and which are register symbols. (Low-order
bit corresponds to the symbol entry stored at the highest end of
the 16 symbols.) Loder sends symbols in groups of 16 followed by
these flag words, each group of 16 loading below the previous group.
Thus, Rug's symbol table grows downward in chunks of 100 decimal bytes.
The load address used for each group of 16 symbols is a negative offset
from the top of the symbol table (i.e. the 1st group of symbols will
load from -100 to 0). Rug will add the symbol-table top to this
offset while loading symbols.
------------------------
| 001 | \
| 000 | ]
| _ _ _ _ _ _ _ | ]
|/ \/ \/ \/ \/ \/ \/ \/ \
> absloader format block, 106. byte count.
_/\_/\_/\_/\_/\_/\_/\_/| / Data is 100. bytes, 16 entries of 3 words
| | ] each, plus two flag words. Load address
| checksum | / is -100., -200., etc.
------------------------
Symbol table:
(low address)
_/\_/\_/\_/\_/\_/\_/\_/|
| | (register flag word)
|----------------------|
| rad50 | 1st 3 letters of 1st symbol name
|----------------------|
| rad50 | last 3 letters of 1st symbol name
|----------------------|
| value | 1st symbol value
|----------------------|
|/\_/\_/\_/\_/\_/\_/\_/|
(15 more symbol entries)
_/\_/\_/\_/\_/\_/\_/\_/|
| | (16th symbol value)
|----------------------|
| hk flag | half-kill flag (bit 0 => 16th symbol)
|----------------------|
| reg flag | register flag
|----------------------|
| | (1st 3 letters of 1st symbol name for
|/\_/\_/\_/\_/\_/\_/\_/| next 16 symbols)
This file last updated 3:30pm Sunday, 8 July 1984 by CBF.

468
doc/_info_/palx.order Executable file
View File

@@ -0,0 +1,468 @@
PALX command string format:
BIN, LST, ERR _ SRC1, SRC2,...,SRCN
The SRC files are effectively concatenated and assembled.
If the BIN spec is empty, no bin file is generated, unless the BIN spec is
followed by the "_".
If the LST spec is empty, a LST file is generated with default names iff there
is no ERR spec.
If no LST is wanted, the comma should be omitted as well.
If the ERR spec is present (that is, the LST spec is followed by a comma) an
ERR file is produced with the default names. (If you don't want one, omit the
comma between the LST spec and the "_")
The device defaults to DSK for the BIN, LST and 1st SRC files.
The 1st name for BIN, LST and ERR files defaults to that of 1st SRC file.
The 2nd name for the BIN file defaults to BIN;
That of the lst file, LIST if /c isn't given, CREF if it is
That of the ERR file, to ERRORS
That of the SRC files, to > .
:_SRC1,SRC2,...,SRCN can be abbreviated
SRC1,SRC2,...,SRCN
Switches do not constitute a non-null spec. (":" or ";" will do that)
The switches are:
/B - Suppress binary. (useful with short format)
/C - Output cref listing. forces /L.
/D - Give back tty and run disowned. Logout when finished. Forces /E.
/E - Force error file to be written.
/H - Double size of symbol table for each /H seen.
/L - Output listing even if no LST spec. (useful with abbreviated format)
If /L appears twice in the command string, listing will be done on both
passes.
/M - Suppress listing of lines generated by macros.
/N - Suppress error messages on tty. Forces /E
/P - Output bin on Ptp:. Forces /S.
/R - Reproduce source in listing. Forces /L.
/S - No symbols in BIN. Forces BIN to be generated unless /B.
/T - Cause normal listing if LST is on tty, cause tty-type listing if LST isn't
on tty.
/V - Sets the number of lines per page to be used in the listing. It should be
followed by an L for LPT, an X forf XGP, or a decimal number terminated by
a comma, which should be the desired number of lines per page including
space for margins. (That is, 60. for the LPT, etc.) This is the only
switch which affects more than one command string - the value set by the
switch will be used for each assembly until another occurrence of the V
switch alters the value.
Command files:
If "@FILSPEC" is placed in a command string, the contents of the file FILSPEC
are effectively inserted in the command string, surrounded by spaces. The
character following "@FILSPEC" is lost, and should be a comma unless "@FILSPEC"
is at the end of the command string, when the same ^M that ends the command
string will end "@FILSPEC".
^M and ^J in the file are treated as spaces; ^L and ^C terminate the file.
Command files may refer to other files to a depth limited by the iopdl.
Command file names default independently to DSK: >
They have no effect on the defaulting of other file names.
The character ^U cancels an entire command string, just as a sufficient number
of rubouts would.
in DDT,
:PALX command
is equivalent to
PALX^K
command^C
On finishing a command, PALX will normally ask for another. But, if the
command was terminated by ^C instead of ^M, and no fatal errors occurred during
execution, PALX will do a .BREAK 16,140000 to delete itself without flushing
the typein buffer. Fatal errors do flush it. Thus, one can type DDT commands
after a command terminated by a ^C. (And also commands to other programs that
might be run by the DDT commands).
Macro-argument scan:
If the first character read when a macro-arg is expected is ^, the next
character is used as the delimiter and all successive characters up to the next
appearance of the delimiter go in the macro arg. After that, spaces are
skipped, and a comma, CR or ; should follow.
If the first character seen is <, all characters between it and the matching >
(not including the < and > themselves) go in the macro arg. After the >,
spaces are skipped and a comma, CR or ; should follow.
If the first character is \, an expression is read in and its value is
converted to a string in base 8, which becomes the macro argument. After the
expression, spaces are skipped, etc.
Otherwise, all characters up to but not including the first comma, CR or ;
go in the macro arg, except that trailing spaces (and tabs, of course)
before a ; will be ignored.
If the scan of the argument stopped on a CR or ;, there are no more arguments.
Any more arguments wanted will be made null. The CR or ; will remain to be
re-read after the expansion of the macro, etc. If the scan stopped on a comma,
the next argument's scan will start with the character after the comma.
Pseudo-ops:
Ordinary pseudo-ops are legal only as the first symbol in a line. Each use of
a pseudo-op uses a whole line.
Value-returning pseudo-ops may be used anywhere an ordinary symbol may be used.
Such a pseudo will usually skip over any comma following its last arg. If
there is to be anything on the line after the psuedo's last arg the comma
should be used; eg:
foo==.length ^/ gg,,hh /,+1 ;sets foo to 9.
.ABS Sets %ABSADR to nonzero. Exactly the same as %ABSADR=-1
.ADRMD ADDR is a value-returning pseudo whose value is the addressing mode
of ADDR (the 6 bit quantity that would go in the SRC or DEST
field of an ordinary instruction given ADDR as an arg). If you
want to follow this construction with an arithmetic operator
put a comma in between; otherwise the operator may be included
in ADDR.
.ALSO Assembles the following code up to the matching .ENDC iff the
last conditional succeeded; i.e. if %SUCCESS is non-zero
.ASCII /TEXT/ Generates several bytes from the chars of TEXT. "/" may be
replaced by any char not in TEXT.
.ASCIZ /TEXT/ Similar but puts a zero byte after the TEXT.
.AUXIL Does nothing. This operator gives information to the @ lister.
.BLKB N is like .=.+N
.BLKW N Moves to even addr, then leaves N words of space. Like
.=.+<.&1>+<2*n>
.BYTE A,B,C Generates 3 bytes containing A,B,C respectively. Any number of
args may be given. Only commas separate args. Null args are
treated as 0.
.EJECT Starts a new listing page. Useful in macros, etc. where ^L's
can't be used.
.ELSE Assembles the following lines up the the matching .ENDC iff the
last conditional failed; i.e. iff %SUCCESS is zero. %SUCCESS
is complemented.
.END ADDR Ends a pass. Remaining input is ignored. Sets the starting
address to ADDR. ADDR may be absent, in which case 1 is used.
.ENDC Terminates a conditional. Error if unmatched.
.ENDM Terminates a .MACR or .TTYMA .
.ENDR Terminates a .REPT .
.EOT Causes whatever remains in current SRC file to be ignored.
.ERROR MSG Causes an error whose message is just the source line ".ERROR
MSG". As with other errors, the PC, page and line are printed.
.EVEN Like ".=<.&1>+." -- moves up to even addr.
.EXPUNGE Makes the symbols in the list following it undefined, like
.KILL in MIDAS: .EXPUNGE A,B,C
.IALSO STATEMENT
Assemble STATEMENT iff %SUCCESS is non-zero. See .ALSO .
.IELSE STATEMENT
Assemble STATEMENT iff %SUCCESS is zero. See .ELSE .
.IF CONDITION
Assemble lines up to following .ENDC only if CONDITION is true.
The conditions are:
1) Arithmetic conditionals:
NAME EXP
where NAME can be E,EQ,L,LE,LT,G,GE,GT,NE,NG,NL,NZ,Z
and any expression can be used. E.g., .IF EQ X-Y
2) String comparison:
DIF ARG1,ARG2
where ARG1 and ARG2 are macro-type args. E.g.,
IDN ARG1,ARG2
3) String comparison with null string.
B ARG true if ARG (a macro-type arg) is null
NB ARG true otherwise.
4) Definition testing
DF EXP true if all syms in EXP are defined.
NDF EXP the opposite.
5) Pass number testing
P1 true on pass 1
P2 true on pass 2
.IFF A subconditional. May appear only inside a conditional.
Switches the sense of the conditional for the code that
follows, so that it will assembled iff the conditional failed.
.IFT A subconditional that restores the original sense of the
conditional for the code that follows.
.IFTF A subconditional that causes the following code in the
conditional to be assembled unconditionally. This is different
from ending the conditional because a .IFT may appear later and
make assembly conditional again, for instance.
.IF1 Assemble lines up to following .ENDC only on pass 1.
.IF2 ...only on pass 2.
.IFB LINE ...only if 1st nonblank in LINE is CR, ; or comma.
.IFDF EXP ...only if all symbols in EXP are defined.
.IFG EXP ...only if EXP is positive.
.IFGE EXP ...only if EXP is nonneg.
.IFL EXP ...only if EXP is negative.
.IFLE EXP ...only if EXP is nonpositive.
.IFNB LINE ...only if 1st nonblank in LINE isn't comma, CR or ; .
.IFNDF EXP ...only if some symbol in EXP is undefined.
.IFNZ EXP ...only if EXP isn't 0.
.IFZ EXP ...only if EXP is 0.
.IIF CONDITION STATEMENT
Assemble STATEMENT only if CONDITION true. See .IF for
documentation of conditions.
.INSRT FILSPC Pushes the current source file and/or repeat, macro, etc. and
starts reading from the specified file. After the end of that
file, the file or macro containing the .INSRT will be resumed
starting with the next line. The characters "/@,_" are not
special in .INSRT FILSPC's - they go right in the filenames.
If the spec'd file isn't found, a non-fatal error occurs and
assembly of the .INSRT-ing file resumes. %FNAM2 is set to the
2nd name of the .INSRTED file and remains set after the file is
finished (unless that file has done a .INSRT, of course).
.IRP DUMMY,<A1,A2,A3>
Repeats all the following text up to the matching .ENDM
substituting A1 for DUMMY the first time, A2 for DUMMY the
second time, etc. The stuff following the comma is a macro-arg
and might also be delimited by ^/ ... /, for example. If the
arg is delimited with < and > then <'s and >'s within the arg
are significant; i.e. ".IRP X,<FOO,<BAR,FAH>>" repeats over two
args, "FOO" and "<BAR,FAH>", whereas ".IRP X,^|FOO,<BAR,FAH>|
repeats over three args, "FOO", "<BAR", and "FAH>".
.IRPC DUMMY,STRING
Repeats the text up to a matching .ENDM once for each character
in STRING, each time substituting that character for DUMMY.
DUMMY must be a symbol; STRING is read in as a macro-argument.
.LENGTH STRING Where STRING is read in as a macro arg, returns the number of
characters in STRING. This is a value-returning pseudo and
arithmetic can be done with it (but remember to put a comma
between the arg and any following arithmetic operators).
.LALSO
ANY-STATEMENT-ON-NEXT-LINE
Assembles the next line iff %SUCCESS is non-zero. See .ALSO .
.LELSE
ANY-STATEMENT-ON-NEXT-LINE
Assembles the next line iff %SUCCESS is zero. See .ELSE .
.LIF CONDITION
ANY-STATEMENT-ON-NEXT-LINE
Assembles the next line only if CONDITION is true. If
CONDITION fails, lists the next line but does nothing with it.
.LIST Decrement XLIST count if it isn't 0. Listing takes place only
if XLIST count is 0. The .LIST itself lists only if XLIST
count was 0.
.MACR M A,B,C Define M as a macro with args named A,B,C . The macro
definition is on the following lines, up to a .ENDM
(intervening matched .MACR'S and .ENDM's go into the macro
being defined).
.MACRO M A,B,C Same as .MACR M A,B,C .
.MEXIT Pops out of the innermost .REPT, macro-call, .IRP or .IRPC .
.MSG LINE Like .ERROR LINE, but doesn't increment the error count.
.NARG SYM Sets SYM to the number of args in the innermost macro
invocation (that is, does SYM==%NARG). The symbol will be
halfkilled.
.NCHR SYM,ARG Sets the symbol SYM to the number of characters in ARG. The
syntax for ARG is the same as for macro args. The symbol will
be halfkilled.
.NTYPE SYM,ARG Sets SYM to the addressing mode of ARG. For example, if ARG is
200(4), SYM will be set to 64. The symbol will be halfkilled.
.NLIST Increments the XLIST count. same as .XLIST .
.ODD Moves to next odd address -- same as .=.+1-<.&1>
.OFFSET EXP Sets the offset to EXP. Whenever the symbol "." is used, in
relative addressing, and in defining labels, the offset is
added to the real location counter. The offset is 0 at the
start of a pass.
.PAGE Starts new page in listing. same as .EJECT .
.PRINT /TEXT/ Prints TEXT on the tty.
.RAD50 /TEXT/ Generates several words of radix 50 code, one for every 3 chars
in TEXT. Zeroes fill up last word, if necessary.
.REM MSG Same as .MSG MSG . Don't ask me why RCC added this.
- Look, if we're going to put comments in here: don't ask me
why you wrote this documentation all in upper case. It's so
hard on the eyes that I spent 2 hours fixing it. --PGS
.REPT EXP Starts a repeat. The body of the repeat consists of all
following lines up to a matching .ENDR . The body of the
repeat is assembled EXP times.
.SBTTL LINE Same as .STITL LINE .
.SEE SYM May go anywhere a symbol would be evaluated. Has no effect
except to CREF the symbol SYM. The @ lister also will see that
SYM is used.
.STITL LINE Sets the subtitle to LINE. The subtitle lists on the 2nd line
of each header. If the .STITL is the first line of a page, it
takes effect on that page; else, on the next page.
.TITLE LINE Sets the title to LINE, also types it on tty. The title
appears on the 1st line of each header. Takes effect as .STITL
does.
.TTYMA A,B,C Reads a line from the tty, then defines A,B,C from that line
using the macro-arg scanning rules. Within the scope of the
.TTYMAC (up to a matching .ENDM) A,B,C will be replaced by the
strings obtained by scanning what was read from the tty.
.WORD A,B,C Like .BYTE A,B,C but makes a word from each arg, instead of a
byte.
.XCREF A,B,C Prevents CREF output from being generated for symbols A, B, C.
.XLIST Increments the XLIST count.
Special symbols:
These may be used as normal symbols, or set with "=".
. The location counter (including offset). . always equals
%.+%OFFSET . Setting .=A is equivalent to %.=A-%OFFSET.
%. The unoffset location counter (where code will actually be
loaded).
%ABSAD If nonzero, PALX will use absolute addressing as the default,
instead of relative addressing. That is, "MOV FOO,%0" will
assemble in this mode the way "MOV @#FOO,%0" always does. Note
that "@FOO" will still use indexing by the PC. Explicit
indexing by the PC will not be overridden.
%COMPAT If nonzero (as is the default), PALX gives an error message
when an instruction that executes differently on different
model PDP-11's is assembled.
%DAY The number in the month of the day on which the assembly is
done. (see %MONTH, %YEAR) [not Tenex]
%FNAM2 The result of taking the current source file's second name, and
turning it into a decimal number, ignoring non-digits.
%MONTH The number in the year of the month on which the assembly is
done. (see %DAY, %YEAR) [not Tenex].
%NARG The number of args given to the innermost macro invocation.
%OFFSE The value of the offset.
%SUCCESS Non-zero if the last conditional succeeded, i.e. the the
condition was true; zero otherwise. This is set both when the
conditional is evaluated and when its scope is terminated (e.g.
by a .ENDC).
%TTYFL Initially 0. when nonzero, tty typeout is inhibited. Cleared
by any attempt to do tty input.
%XCREF Stops CREFfing if not 0.
%XLIST The value of the XLIST count.
%YEAR The number of the year in which the assembly is done. This is
only that last two digits, so in 1980 %YEAR is 80., and in 2001
%YEAR is 1. (see %DAY, %MONTH) [not Tenex]
.RPCNT Normally 0 except in repeats. In repeats, is 0 the 1st time
through, 1 the next, etc.
.IRPCN Like .RPCNT but for IRPs instead of repeats.
Symbols of the form 69$ are "local tags." The scope of these symbols is
limited to the area between two ordinary labels. These are typically used for
tags in branches that skip over a few lines within a single routine. These
symbols behave the same as in MACRO-11.
The following symbols are predefined in PALX:
For teletype: %TKS, %TKB, %TPS, %TPB and trap vectors %TKV, %TPV
For paper tape: %PRS, %PRB, %PPS, %PPB, and vectors %PRV, %PPV
For 60-cycle clock: %LKS and vector %LKV
For programmable clock: %PKC, %PKCSB, %PKCSR and %PKV
For line printer: %LPS, %LPB, and %LPV
For RF-11 disk: %DCS, %DWC (also %WC), %DCA (also %CMA), %DAR, %DAE, %DBR, %DSA
(also %ADS), and %DV
For RK-11: %RKDS, %RKER, %RKCS, %RKWC, %RKBA, %RKDA and %RKV
For Knight display: %NGCSR, %NGREL
For DR-11: %DRS, %DRIB, %DROB, and %DRV1, %DRV2
For DC-11: %RCSR, %RBUF, %TSCR, %TBUF
For DM-11: %CSR, %BAR, %BCR, %TBR
For EAE: %DIV, %AC, %MQ, %MUL, %SC, %SR, %NOR, %LGS, %ARS
For processor traps: %ERRV(=4), %BPTV(=14), %IOTV, %PWRV, %EMTV, %TRPV
For processor regs: %PS, %SWR (switches)
For use with PDP11^K (the ITS PDP-11 simulator): %CSX
I will define others on request.
Error message format:
An error message has the following components:
REL-LOCTR ABS.-LOCTR. page line DESCRIPTIVE-MESSAGE.
The REL-LOCTR has the form LABEL+DISP. It gives the value of the unoffset
location counter, relative to the most recently defined label. It does not
appear if no labels have yet been defined (tabs are printed instead).
The ABS.-LOCTR is the unoffset location counter in octal.
The page and line numbers are in the source file.
In addition, on the first error in a source file other than the first file, the
message
FILE such-and-such
will be printed.