cvs commit: src/sys/i386/i386 pmap.c

[Doug Rabson]

On Sun, 2003-10-26 at 06:41, Peter Wemm wrote:
> Jeff Roberson wrote:
> > On Sat, 25 Oct 2003, Peter Wemm wrote:
> 
> > Wow, pentium4 sucks.  Yes, I agree then, we should revert the change.
> I'll do it.
> > 
> > Intel looks more disappointing every day.
> 
> Well, think of their optimization goals... The pentium4 was designed for
> two things..  1) to increase MHz, since thats all dumbass customers and
> sales droids understand, and 2) to increase game framerate benchmarks.
> Anything that didn't contribute to that goal and consumed transistors
> started losing.

The trick is to find some way to make intel interested in your problems
(e.g. change a large site from using intel processors to amd). When the
marketing people start to care about an application, the technical
people start to collect instruction traces to use for optimising the
next generation.

In the mid 90s, 486 processors had terrible floating point performance
and intel didn't much care. Their instruction traces showed that the
applications they cared about (mainly word and excel) didn't use
floating point much. As soon as people started trying to use floating
point more intensively for games software, intel started profiling and
optimising for it and these days, their floating point performance is
reasonable for all applications (not just games).

The trick, I guess, is to make the right kind of case. If, for instance,
the engineers developing Longhorn started telling Intel that AMD
processors could e.g. context switch ten times faster than P4s and this
would affect the performance of some bogus Longhorn feature, then I
imagine things might change.