Re: 663.55 BogoMips (Win95 crash)

Aaron Tiensivu (tiensivu@pilot.msu.edu)
Tue, 29 Dec 1998 15:32:25 -0500

> > The BogoMips for an AMK K6 are twice the speed in MHz.
> > My K6 200MHz reports a BogoMips of 400.59.
> .... which tells you that the AMD engineers did a good job optimizing their
> branch prediction code for (such tiny) loops. Reminds me of the old story of
> the loop machine insn being much slower on recent intel CPUs than decl ecx;
> jnz/jns; (This was a deliberate decision by intel not shared by AMD and
> Cyrix.)

This is also why the K6 got bit in the ass by a Windows 95 bug that didn't show
up under Pentium-2 boxes. M$ did a simple timing loop and would take the result
of that timing loop and divide a number by it. Well, the K6 (350 or above) ran
it so fast that the result ended up being 0..

Boom.. kernel crash.. unchecked divide by zero on boot-up. Sometimes it would
boot, sometimes it would not.

The really annoying thing about that whole situation is that M$ made people pay
money to get it fixed (~$50 depending on the type of license you had).. well,
of course that file got spread around everywhere on the net not soon after but
that still angers me.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/