There is - again as I recall - 2 cases: executing in "main" where the max delay is the time from phase x of the current instruction to phase x of the next instruction. In ADDITION if you happen to hit at phase x of an IRET the next instruction is protected so the execution time of the IRET will be added. Thus since all instructions take the same time to reach phase x just calculate instruction time. The minimum will be that the interrupt arrives just before an instruction reaches phase x.

Erik

Erik