Appendix A. Performance Data
GFK-2222AD
April 2018
297
A-3.8
I/O Interrupt Performance and Sweep Impact
There are several important performance numbers for I/O interrupt blocks. The sweep impact of an I/O
interrupt invoking an empty block measures the overall time of fielding the interrupt, starting up the
block, exiting the block, and restarting the interrupted task. The time to execute the logic contained in
the interrupt block affects the limit by causing the CPU to spend more time servicing I/O interrupts and
thus reduce the maximum I/O interrupt rate.
The minimum, typical, and maximum interrupt response times reflect the time from when a single I/O
module sees the input pulse until the first line of ladder logic is executed in the I/O interrupt block.
Minimum response time reflects a 300 µs input card filter time + time from interrupt occurrence to first
line of ladder logic in I/O interrupt block. The minimum response time can only be achieved when no
intelligent option modules are present in the system and the programmer is not attached. Typical
response time is the minimum response time plus a maximum interrupt latency of 2.0 ms. This interrupt
latency time is valid, except when one of the following operations occurs:
▪
The programmer is attached.
▪
A store of logic,
RUN Mode Store
, or word-for-word change occurs.
▪
A fault condition (logging of a fault) occurs.
▪
Another I/O interrupt occurs.
▪
The CPU is transferring a large amount of input (or output) data from an I/O controller (such as a
GBC). Heavily loaded I/O controllers should be placed in the main rack whenever possible.
▪
An event that has higher priority and requires a response occurs. An example of this type of event is
clearing the I/O fault table.
Any one of these events extends the interrupt latency (the time from when the interrupt card signals the
interrupt to the CPU to when the CPU services the interrupt) beyond the typical value. However, the
latency of an interrupt occurring during the processing of a preceding I/O interrupt is unbounded. I/O
interrupts are processed sequentially so that the interrupt latency of a single I/O interrupt is affected by
the duration of the execution time of all preceding interrupt blocks. (The worst case is that every I/O
interrupt in the system occurs at the same time so that one of them has to wait for all others to complete
before it starts.)
The maximum response times shown below do not include the two unbounded events.
