🚀 go-pugleaf

In article <4422@cruzio.santa-cruz.ca.us> aki@cruzio.santa-cruz.ca.us writes:
>
>I beg to differ from you on this point.  I have designed a MHz
>counting function that (so far) has returned correct or close-to
>correct results on all PC's it has been run on.  The largest error
>margin was on a 4.77 MHz XT that clocked as 3.5 MHz.  Usually the
>function is right on the money.
>
>What _IS_ impossible is to take into account and adjust for all the
>billions of possible combinations of cache/chipset/CPU to arrive
>at the exact correct number all the time.

This is just my point.  If your code has such a variance on a 4.77 MHz
PC for example, how would you know if a value returned should indicate
a 25 MHz CPU verses a 33 MHz CPU.  With the appropriate cache, the 25
MHz machine may perform closely to an uncached 33 MHz machine.  If the
33 MHz machine has very slow memory it may only perform a little better
than most 25 MHz machines.

There are just too many variables to claim that a piece of code can
figure out the clock frequency of a CPU chip by a timming loop.  If you
add the fact that some machines use the wrong frequency of crystal on
the timer chip, then you don't even have a standard to compare
against.

-- 

rap.

-----------------------------------------------------------------------
Robert A. Pease, Novell Inc.
Internet: rap@novell.COM
-- 

rap.

-----------------------------------------------------------------------
Robert A. Pease, Novell Inc.
Internet: rap@novell.COM