I had to make performance forecasts for computer systems that did not yet exist. Moments such as the cycle time ratio can only give an approximate guideline. As a rule, I had to resort to simulation, the closest I could measure the actual equipment.
However, you can find numbers for tests similar to your code, which will at least give you a starting point.
I wouldn’t do this while working on one chip at a time - your code may have a problem that makes it too slow for any possible chip. I would try to find a chip that is fast enough and works if it is much faster than necessary.
Patricia Shanahan
source share