Erik,

can you please do me a favour: rip off a piece of a *real life* code - it can be part of a bigger binary - and you or I will run it through Jeff's disassembler with the cycle counting on. That would reveal the "truth".

That most of the 255 '51 instructions in the cygnal core take 1-2 cycles by far does not mean that most of the instructions in real programs run in 1-2 cycles. Most programs jump a lot, and jumps take 3-5 cycles, for example. External and code memory access takes 3 cycles, and so does mov dptr,#const16 - both of which are very freqent in bigger programs where external RAM is unavoidable. From this, often, the "straightforward" counting of cycles of instructions present in a piece of code is unrepresentative for what a routine takes when it runs, as the percentage of executed jumps might be higher.

And, in the 50 and 100MHz derivatives things got a bit worse because of the caching.

JW