life2.i is a revised version of my Life program, differing in two
ways from the original.  First, throughout the program I have
streamlined the loops by using COME FROM statements and combining
initialization and re-initialization phases where practical.  Second,
instead of just printing the coordinates of live cells, I use the new
binary output feature to print out the board as an array of .'s and
O's.  This new output section is made into a subroutine---entry point
(300)---which I call from within the main loop as well as at the end of
the program.  If you don't want to see the position at every time step,
just remove the call to (300) from within the main loop, and only the
final position will be printed.

Notes for programmers:  The output routine prints the state matrix a
full line at a time, using a 1-D array one cell longer than a row of
the matrix.  This final cell is used to print a line feed (ASCII 012)
at the end of each line.  Live cells are represented as O's (0117),
while dead cells are shown as .'s (056).  If your machine does not
use ASCII or does not use LF to terminate lines this may not work.
The INTERCAL output routine reverses the order of bits in each of
these characters, so the constants hard-coded into the program are
#242 for 'O', #116 for '.', and #80 for '\n'.  The value of the last
character output, which is needed by the Turing Text output model,
is stored in .2.  This variable is initialized to #0 and then STASHed
at the beginning of the program, then RETRIEVEd by the output routine
each time it is called.  This is opposite from the usual way STASH is
used, but I was feeling a little twisted the day I wrote all this.

There is still considerable room for improvement in the (300) output
algorithm.  With printing at each time step the program spends
roughly two-thirds of its time doing output, and I suspect that
most of that is squandered in the (1010) subtraction routine.  It
should be possible to greatly speed up the process by precomputing
and storing all the possible differences.  Except for initialization
there are only six to consider: "LF-.", ".-LF", "O-.", ".-O", ".-.",
and "O-O".  Of these the first two occur predictably at the beginning
and end of each line, and the last two are trivially zero.  The
necessary program logic would be a little more complex; perhaps it
could be done with array indexing?  Consider it a challenge!

Despite all its inefficiencies, the INTERCAL version on a Sparc 1 is
still faster than the first Life program I ever wrote, in BASIC on
a TRS-80, by a factor of about 50.  Sobering thought.

                           Louis Howell
                           December 15, 1991