Well, I don't know how did you come to the conclusion that Richard uses 4-clocker '51 derivatives; nevertheless, a 4-clocker does not run 3 times faster at the same oscillator frequency, as the individual instructions' cycle count is different than at the 12-clocker.
Once again I refer you to that table of mine. The 12-, 6- and 2-clockers have the same instruction cycle count, the 4- and 1- have different (and different from each other, although they generally follow a relatively common scheme, having the simpler instructions have as much instruction cycles as the number of bytes they occupy in code space).


I won't comment on this practice.

I just want to lay down the following: for code timing, the best is of course to do it on the actual hardware, but that might not be good enough - for example, it might be hard to do a worst case analysis simply running it. In a simulator, you can usually direct the execution the "worst case way", but it is often a paintaking exercise. In (dis)assembler, you have to calculate and think to figure out the exact timing.

Jan Waclawek