Had some software conversion experience with the 360.  When LSU went from
the 7000 series machine to a  360, many programs had to be revised.  Decided
it was cheaper to buy time on a 7000 series machine than rewrite larger number
crunching programs for the 360.  Later, in the Oil Patch had a few encounters
with the big 370.  Even got permission to write and use an assembler routine.
The 360 operating system made constructing overlays and JCL for big multi-
step jobs a colossal pain.

Loved getting my very own VAX with VMS for the lab.  It was amusing to watch
someone at terminal when the big instrument burped out a scan of data --
everything else stopped until the data was decoded and stored.   Line printer
would stop dead, and the terminal would stop, too. Gave the data acquisition
task high hard and software priority, since a real time data transfer must
complete before the little controller computer on the instrument runs out of
memory, or the data is gone forever.  NASA had some concerns with satellite
data acquisition which led to DEC fixing VMS to eliminate some cases where
task switching wasn't fast enough.

FORTRAN in the 7000 series scientific machines was a natural.  With the 360,
it was necessary to go back and make all the real variables double precision
and adjust calling sequences accordingly.  Also had some problems with length
of integers.  Seems to me the major failing in the standard for FORTRAN was
not defining a minimum number of significant figures and magnitudes for single
precision reals and requiring that single precision integers occupy the same
size storage.  When FORTRAN originated, the word oriented machines had a
fixed length data word so the question probably never arose.

In the old days we were operating with very little fast memory and slow disk
reads and writes, so we squeezed as much as we could in memory and spent
much time organizing the computation to minimize input and output.  We also
used the EQUIVALENCE statement to efficiently use and reuse the same
memory at different steps.  Recall exceeding memory by one cell in a tricky
multiple overlay program.

Used a machine between the 650 and the 7044, can't remember the number,
that had variable word length.  Was very slow at number crunching, but did
enable me to balance rounding and truncation error for a subroutine that was
widely distributed.  It would examine ratios of input variables and determine
how which numerical integration formula would be adequate for a two
dimensional numerical integration.

Then there was the little IBM machine we called the CADET -- can't add,
doesn't even try.  Used table lookup for arithmetic.  Did pioneer the concept
of a subroutine loading on call, as far as I know

Choppy



At 08:49 AM 9/24/03 -0400, "Buck"  wrote:

>Hmmmm,
>
>I wonder if Choppy has ever worked on an IBM 360?  We had one at the 
>Shreveport Trade School on Hope Street back in the early 60's, LOL!

Reply via email to