IBM 1130 User Manual page 536

OVERLAPPED INPUT/OUTPUT
Introduction
As a machine, the IBM 1130 Computing System is
capable of performing many tasks simultaneously.
For example, it can print, type, read a card, and
compute, all at the same time. This can be done
through its "cycle-stealing" I/O channels and the
priority interrupt system. Each I/O device may,
through an interrupt, signal the CPU that it requires
service, and steal a cycle (3.6 or 2.2 microseconds)
from some other process to do what it needs done.
This process is commonly referred to as "over~
lapping',' .
For example, in the case of the disk, one data
word travels past the read/write heads every 27. 8
microseconds. However, it only takes one cycle
(3.2 or 2.2 microseconds) to transfer that word
from core storage to the disk (if it is being written)
or from the disk to core storage
(if
it is being read).
Section Subsections
Page
70
20
I
01
01
This means that only a little more than 10% of the
CPU time is required to read and write on the disk;
the remainder is available for other use.
Although most of the 1130's I/o devices can be
overlapped, standard 1130 FORTRAN permits only
two of them to operate in this fashion: the disk and
the 1403 Printer. There are several good reasons
for this limitation. For example, suppose you
wrote a program to read two numbers from a card,
add them together, and print the result. With full
overlap, the addition could conceivably be under
way before the two numbers had even been placed in
core. Obviously, this would not be satisfactory.
To take full advantage of the potential of the
machine, in FORTRAN, it would be necessary to
develop a special FORTRAN, which would then
violate the USASI standards set up for that pro-
gramming language. Avoiding this, IBM has
developed the Commercial Subroutine Package
(CSP) -- a set of subroutines operating within the
FORTRAN system, rather than as part of the
FORTRAN language itself.