Quote:
Originally Posted by georgewr3
Our system often has problems that occur only in real time.
|
One of the problems, apart from the detail of running the processor with a faster clock (which is probably easy if you want 5%, somewhat difficult if you want 10 - 15%, and extrapolate yourself from there) is that if you don't want to introduce new bugs, you have to speed everything up.
That probably includes odd things that you haven't thought about like interrupts and timer/counters and the hard disk. And, of course, the external machinery.
Now if you do find a way of speeding up the hard disk, do please let me know, but I think you'll run into real difficulty when it comes to speeding up the rest of the universe (simplification: you may be able to get away with only speeding up the bit of the universe to which the system directly interfaces).
In general, this is one of the problems with real time. It is real and it is time and that makes it harder to debug stuff.
There are things that make the system faster (i.e., have greater throughput) but it is unclear whether these are likely to make the existing errors happen after shorter time intervals or introduce completely new errors or make the old errors go away. If you are really lucky, you make the old errors go away and introduce new errors, so that you now spend all of your time debugging stuff that has absolutely nothing to do with the original problem.