Re: Cycles per ASM instruction

On Feb 21, 12:32 am, Arthur T. <art...@[EMAIL PROTECTED]
> wrote:
>      I remember seeing them for the 370/115.  It's hard to forget
> that moving from a 360/22 to a 370/115 gave us a CVD timing worse
> by one to two orders of magnitude.  IIRC, the number of cycles for
> that instruction was approx 250.


Are you saying the 370-115 was slower than the 360-22?  (I didn't even
know there was a model 22).

When I had exclusive use of a machine (actually the similar Univac
90/30), I experimented with using binary vs. packed-decimal.  Binary
was faster internally, but the CVD and CVB instructions were killers.
You had to be careful they were outside any loops.  (This involved
COBOL subscripts).


To this day proper definition of subscripts and SEARCH arguments in a
COBOL program will save processing time if the files have any length
to them.  We learned the im****tance of which field to use, but I'm not
sure all programmers were taught it or remembered it from class.
(Many were taught but probably forgot.)