On Thu, Apr 22, 1999, Karsten Johansson <[EMAIL PROTECTED]> wrote:
>
> "Yates, Larry" wrote:
> >
> > Maybe bogomips are BOGUS!
>
> Heheh, you know when Linux boots, and it primes the random number generator?
> Maybe *that's* a bogomip, too. :)
>
> I'm gonna do a bit of research to see what is actually counted in a bogomip.
> There seem to be extreme variations from one machine to another.
No need. It's how fast the machine can do an idle loop. The reason for
the variations are because of the architecture of newer cpu's and how
they execute multiple instuctions in one clock cycle.
Older Pentiums took 1 clock cycle per instruction (2 instructions per
loop) so the bogomips would be half the clock speed.
Pentium Pro's (IIRC) could do 2 instructions per cycle, so they could do
the whole loop in one clock cycle. Bogomips were the same as the clock
speed.
Newer CPU's like the K6 and Pentium 2/3 can do 4 instructions per clock
cycle in some circustances. So it can do the whole loop twice in one
clock cycle. Bogomips were twice the clock speed.
There's FAQ out there explaining it in much greater detail, but suffice
to say, Bogomips are NOT a good measure of processor speed. They are
only used to time idle loops to wait for I/O, etc.
JE
--
To get out of this list, please send email to [EMAIL PROTECTED] with
this text in its body: unsubscribe suse-linux-e
Check out the SuSE-FAQ at http://www.suse.com/Support/Doku/FAQ/ and the
archive at http://www.suse.com/Mailinglists/suse-linux-e/index.html
