1401 Autocoder

I have keyed in the code from that listing, assembled it with an assembler I wrote, and compared the output with a deck from a tape that Dick Weaver lent me. There are links below for my assembler, the source code for Autocoder, the output "decks" and "listings," and the Autocoder distribution deck that was on the tape Dick Weaver lent me.

I have transcribed the modification letters, too. There's a link below.

All of the files except the distribution deck for Autocoder that I got from Dick Weaver are available as gnuzip-compressed "tar" files or "zip" archives. The distribution deck for Autocoder is gnuzip compressed and zip archived. It's not a tarball because it's only one file. The .zip files were all prepared with the MSDOS end-of-line sequences.

• My Autocoder assembler (.tgz, .zip) is written in Fortran 90. It comes with a Makefile that uses the Lahey/Fujitsu Linux compiler. It works with GNU Fortran 95, and should work with any compiler that supports the full Fortran 90 standard (I don't think I used anything that's in Fortran 95 but not in Fortran 90). It doesn't do macros, IOCS, or Disk I/O instructions (yet).
• The source code for Autocoder, in Autocoder (.tgz, .zip), is equivalent to modification zero of level 3 of 1401-AU-037, plus some modifications that removed two CC instructions (thereby causing Autocoder to use less paper). There are a few differences:
  • Macros were expanded manually (because my assembler doesn't handle them -- yet)
  • There is a seven-character BCE at sequence number 1866 (line 1669) in phase 3. My assembler doesn't allow this, so I used "machine code" form -- there is a B op code in column 19 instead of a BCE mnemonic in column 16.
  • My assembler puts the suffix in the first blank character of a label; 1401-AU-037 puts it in the sixth character if it's blank. There are references to "TDF6 X", at pg/lin 13/25 and 13/29 in pass 2; notice the blank in the fifth character and the suffix in the sixth. My assembler doesn't allow blanks in the labels, so I had to put SFX X before and SFX B after. This causes sequence numbers to be too large by four starting at sequence number 1350 (pg/lin 13/30). I can't just put TDF6X, because this is in a SFX B region, so the result would be TDF6XB.
  • There are some questionable instructions
    • MA &13,LIMAD appears at sequence number 340 (line 219) of autocoder5.s. This should be MA &013,LIMAD. It's repaired in modification letter #4.
    • The NUREC routine at sequence number 1211 (line 1027) of autocoder3.s branches to PUT, which is a subroutine. When PUT returns, the first instruction of GET stores the B-address register at GETXX. At this point, it will be the address of the first instruction after PUTXT. Branches to NUREC end up at CKCOM, not by returning to the next instruction. This may be what was intended, but it's bizarre.
    • Sequence numbers 596 and 597 (lines 466 and 467) of autocoder5.s are missing from the listing in the CE manual. They were apparently comment lines (because the object decks match).
    Questionable lines have ??? on them.
  • There may be typos in the comments -- I proofread the code by comparing my assembler's output to the "deck" on the tape that Dick sent me.
• "Deck" output from my assembler (.tgz, .zip). There is a separate deck for each of phases 1, 2, 7 and 8. Phases 3 and 4 are in a single deck, and phases 5 and 6 are in a single deck. There's a C-shell script "makedeck" that combines them, together with control cards in the file "control.cd" to produce a distribution deck, which is also here. If you run this deck, it makes an operational tape. This deck has no macros in it.
• "Listing" output from my assembler (.tgz, .zip). This uses column 1 for vertical format control: blank = single space, 0 = double space, 1 = new page. This is the convention that was used by 1401 tape-to-print utilities, and Fortran. Here's my "ansi" filter that turns these into C "new line" and "new page" character sequences:
  

  #include <stdio.h>
  
  main ( int argc, char* argv[] )
  { char c, d;    /* characters being copied */
    int first;    /* 1 = first character of a line, zero = not */
  
    setvbuf ( stdout, (char*)NULL, _IONBF, 1 );
    first = 1;
    while ( 1 )
    { c = getchar();
      while ( ( c == '\n' ) | ( c == '\r' ) )
      { d = getchar();
        if ( (c == '\r' & d == '\n') |
             (c == '\n' & d == '\r') ) d = getchar();
        c = d;
        first = 1;
      }
      if ( c == EOF ) break;
      if ( first )
      { if ( c == '0' ) { putchar ( '\n' ); putchar ( '\n' ); }
        else if ( c == '1' ) { putchar ( '\n' ); putchar ( '\f' ); }
        else if ( c == '+' ) putchar ( '\r' );
        else putchar ( '\n' );
        first = 0;
      }
      else putchar(c);
    }
    putchar ( '\n' );
  }
  

• Postscript rendering of the listings (.tgz, .zip). These were made from the listings using "ansi", then "enscript" with "line printer," "landscape" and "Courier8 font" options, and then piped through "psnup -2" to get two per page. Here's my script for it:
  

  ansi < autocoder$1.lst | enscript -l -r --font Courier8 -o - | psnup -2 > autocoder$1.2.ps
  

  Hmmm, Courier9 works for all but the first page. Weird.
• Distribution "deck" for Autocoder version 3, probably modification 11, from the tape from Dick Weaver (.gz .zip). There is also a "custom" modification by A. B. Platt here, to print A- and B-addresses in decimal as well as machine form. This is a gnuzip-compressed plain text file. Some of the transliterations from BCD to ASCII are: group mark = ", record mark = ', lozenge = ). I can send you the original EBCDIC if you don't like my transliterations.
• Modification letters from IBM (.tgz, .zip), transcribed by hand. Also available in PostScript (.tgz, .zip), landscape and one-up in enscript's default line printer font.
• Paul Pierce has an operational tape for Autocoder, which does not appear to include the update letters included in the one from Dick Weaver, but it does have the IOCS macros. It also has a modification that sends it back to the 1410 operating system when it finishes.