On Tue, 3 Jun 2014 23:51:45 +1000, Robin Vowels wrote: >From: "Tom Marchant" <[email protected]> >Sent: Tuesday, June 03, 2014 10:17 PM > > >> On Tue, 3 Jun 2014 20:01:24 +1000, Robin Vowels wrote: >> >>>And while I have the manual out, BCTR is from 40% to 70% faster than BCT. >> >> What manual are you looking at that has any relevance at all to >> the instruction timing for modern processors? > >Do you have something better?
Better than what? You didn't answer the question. As far as I know there is no manual that gives instruction timings for any IBM processor that was available for at least the last three or four decades. That's why I asked what manual you are using. For you to make claims like "BCTR is from 40% to 70% faster than BCT" and reference "the manual" is misleading at best. To answer your question, I don't have any useful reference. Timing information was published for most of the System/360 models, but not for the higher end models. The model 195 Functional Characteristics manual says, "For a parallel system like the Model 195, however, no average times are meaningful" It is not valid to extrapolate the timing information for such ancient processors to modern processors. Pipelining has already been mentioned in this thread. Other factors include the fact that those older processors were microprogrammed. On modern processors, the majority of instructions are implemented entirely in hardware. Many instructions have special circuitry to optimize their performance. -- Tom Marchant
