SIMMON

In this mode, each instruction in the guest program was simulated without ever passing control directly to the guest. As an Instruction Set Simulator, SIMMON was unusual in that it simulated the same architecture as that on which it was running, i.e. that of the IBM System/360/370. While an order of magnitude slower than Interrupt mode (below), it allowed close attention to the operation of the guest. This would be the mode used by various instruction trace test tools.

Interrupt mode (a/k/a Bump mode) constrained the guest program to run in user program state, with the SIMMON kernel handling all hardware interrupts and simulating all privileged instructions the guest attempted to execute. This mode could be used, for example, by a test tool to simulate a hardware device.

(ERror GENeration and Test): This test tool was developed to test the device support error recovery in IBM's PCP (Primary Control Program) operating system, then being developed. It used a novel and very efficient table-driven finite state machine (FSM) to inject simulated errors and verify that the operating system followed the detailed specifications of actions to be taken to attempt recovery.

The table driven FSM aspect was granted U.S. Patent in October, 1972.

MAPPER (not to be confused with the Unisys product of the same name) was a statistical performance analysis tool. It operated by allowing the program under test to run in Interrupt mode, but also used the system timer to periodically interrupt it. The addresses where the tested program was interrupted were recorded and later summarized and tabulated in the form of a map, showing the density of interrupts over the memory addresses. The result resembled nuclear scintigraphy images, showing the parts of the program most frequently used under the test conditions.

HOTSPOTS was an instruction trace tool written to help identify performance problem areas in IBM's MFT operating system. Branch trace data was written to tape, then summarized. The report took the form of a listing similar to a storage dump, with program entry points and exit points identified, including frequency of use for each instruction sequence.

These data identified the Memory Management component as consuming about 20% of CPU resources, and was used to justify a task force to try to improve the performance.

While not a specific test tool, the distorted timing relationships while running under SIMMON found a number of problems, particularly in the input/output sections. Unless a SIMMON tool was put in place to normalize and delay I/O events, these would appear to the guest program as happening unnaturally quickly.