return to main page

Sorting Demo, using serial media, such as magnetic tapes

Sorting in a business environment is a "Good Thing"
When dealing with serial media, such as magnetic tapes (and decks of cards), sorting is often "a good thing".

Various card oriented tasks, such as billing operations, payroll, inventory control, are made faster and more convenient if the records are sorted by customer number, or stock number.

(Random access based storage, (such as disk) is a completely different matter, and not discussed here.)

We expect tapes to move and spin
In the movies, and oddly in real life, groups of magnetic tape drives moving tape in a jerky fashion, then suddenly the tape reels spinning backward, then jerking tape forward again, - that is tape sorting - is what business data processing did for hours a day.

With great perseverance by the tape crew, we may be able to get three tape drives working reliably :-))

Three tape drives is the absolute practical minimum for practical tape sorting. Doing it with less is a silly exercise.

Everyone wants to see tapes move, and hear them chatter like a machine gun,

and the rewind of a 729 tape drive is a wonder to behold.

One can do some random things with tape drives,

but everyone gets a sense of serious purpose watching a group of tape drives doing a sort. -

So - how can we provide a 3 tape sort, to thrill our guests, and of course ourselves?

Note: SORT 7 requires a minimum of 4 working tapes - currently only a dream -

Proposed:
The Royal "We" write a three tape sort. Royal in the sense of a King saying

" 'We' will work harder".
No one expects the King to swing a hammer, pull a plow, ... harder ;-))

Constraints & Problems:

Ideas:

Tools:
Writing the above is an "exercise for the student".

Dealing with imperfect tape and the imperfect world is more difficult !!
If 1401 tape I/O subroutines are not available,
Van Snyder offers the following (Aug 5, 2009) 
Ed:

Tape I/O and error handling on 1401 aren't very difficult.  See
A24-3069. [2.4 megabytes]
The buffer areas are addressed at the low-core end and continue until a
group mark with a word mark.

In the case of writing, the buffer is written from the addressed
character up to the next one having a group mark with a word mark.

In the case of reading, the record is read until it is completely read
or a character in core has a group mark with a word mark, whichever
happens first.  A group mark is deposited in core after the last
character read.

In either case, after the transfer the B address register is one after
the last character transferred.

You can read and write with or without word marks.

The op codes are M without word marks and L with word marks.

The A address is the tape drive address, %U1 .. %U6 for even parity, %
B1 .. %B6 for odd parity.  Overlap adds some complications so let's not
think about that now.  The B address is the buffer start address.

The D modifier is R to read and W to write, or you can use the Autocoder
mnemonics RT, RTW, WT, WTW (with these the A address is just the unit
number 1..6, not %U1..%U6).

Tape control uses the U op code (Autocoder mnemonic is CU).  The A
address is the unit number.  The D modifier is R for rewind, U for
rewind and unload, B for backspace, E for skip and blank tape (this
actually sets an "erase" switch that is examined during the next write
so it doesn't matter which unit number you select), M to write an
end-of-file mark, and A for "diagnostic read" which skips a record and
checks for error and end-of-file without transferring any data.
Autocoder mnemonics that don't need D modifiers are RWD, RWU, BSP, SKP
and WTM.  Diagnostic read doesn't have a special Autocoder mnemonic.

A read routine should check for noise records (usually defined as 12
characters or less), and clear the GM the TAU deposits if the buffer has
word marks.  You can do length checking if you feel like it by looking
for the GM where you expect it before clearing it.  I used to set aside
a buffer with an extra two characters to detect long records, then look
for the GM in the right place

Here are some simple I/O routines (no length checking).

     rdtape    sbr  rdexit&3
               mz   *-6,azone     make character the tape can't transfer
               mcw  azone,test&7
               lca  gmwm,buf&999
     rdagin    za   err
     rdinst    mcw  azone,buf&11
               rt   1,buf
               mcw  err           clobber gm
               bef  done
     test      bce  rdinst,buf&11,0  check for noise
               ber  rderr
     rdexit    b    0-0
     rderr     a    *-6,err
               bsp  1
               bce  rdinst,err-1,0  fewer than ten errors
               nop  666
               h
               b    rdagin
     azone     dcw  #1

If you know the records don't have GM in them, you can simplify the
noise test (you don't need AZONE).  If you don't have word marks in the
read area you can leave out "mcw  err":

     rdinst    rt   1,buf
               bef  done
               bce  rdinst,buf&12,}
               chain12


     wrtape    sbr  wrexit&3
     wragin    za   err
     wrinst    wt   2,buf
               bef  full   Tape reel is full, mount another one
               ber  wrerr
     wrexit    b    0-0
     wrerr     a    *-6,err
               bsp  2
               bce  wrinst,err-1,0  fewer than ten errors
               skp  1
               nop  1666
               h
               b    wragin

     err       dcw  #2

If you're reading into an area without word marks, you don't need to
clobber the GM the TAU deposits, so you don't need to worry about
clobbering the one that defines the end of the buffer, so you can do
that with DCW instead of LCA.

A package that checks and writes labels, and does length checking is
attached.  If you want to use it without label checking, clear the WM in
the second character of the packet before branching to XXTAPE.

I/O is a lot faster with blocked records.  Let me know if you want to do
this and I'll send some tips.

An alternative is to use IOCS (C24-1462).  IOCS can also do blocking.
You'd have to use Real Autocoder since my cross-assembler doesn't do
IOCS.  Let me know if you want to do this and I'll tell you which
Autocoder tapes have IOCS macros on them, and how to use them in SimH.

Dick Weaver is an expert on tape I/O in general and IOCS in particular.

Best regards,
Van

return to main page