[rescue] E3000's

[vance at neurotica.com]

On Tue, 5 Aug 2003, Jonathan C. Patschke wrote:

> > How about the Patschke?
>
> That'd be silly (and hard to spell).
>
> Also, now that I think about it, it's pretty hard to clearly define a
> VUP on a non-DEC system.  I still like my idea of clockrate x avg IPC.

But 1MIP ~ equals 1VUP anyway.  But even then, we have to take into
account how complex an operation that the machine can execute in one
clock.  Doing matrix transforms in a single clock should have more weight
that doing an integer add in a single clock.

> But neither of these take into account memory bandwidth or other
> niftiness, and we can't have PCs winning a rescue-defined benchmark, now
> can we?

Indeed.

Peace...  Sridhar