I am afraid, that the nature of problem/program is what matters here, rather than the expertise of the programmer - as far as he/she is able to make the judgement. The occurence of problems where usage of this technique is appropriate is so rare...

Incidentally, one of such has been discussed recently on a related czech mailing list (hw-list@list.hw.cz), where a simple DSO was made from a mcu (a PIC, but that's really not the point). To achieve the highest possible sampling rate, the author made an unrolled, loooong sampling sequence with no testing for trigger or sample counter/pointer testing etc, just acquiring a sample and storing it using an autoincremented pointer. The trick was, a trigger event started a timer, expiring which caused an interrupt, and return from the isr was to outside the sampling sequence. Here, this technique is - I think - fully justified: the "main" here makes no calls nor pushes, no other interrupts are enabled, and the trick is employed to achieve functionality which is impossible by other means.

Jan Waclawek

Jan Waclawek