Subject: Re: HPC devices drop sh and mips for arm?
To: Brett Lymn <>
From: John Utz <>
List: port-hpcarm
Date: 09/10/2001 00:48:48
comments inline

On Mon, 10 Sep 2001, Brett Lymn wrote:

> According to Warner Losh:
> >
> >My guess would be over the next 4 to 9 months.  Basically, no one is
> >going to do a new design based on MIPS.
> >
> Hmmmm does anyone else find it odd that M$ always seem to start off
> supporting multiple architectures and then back down to one? (c.f. NT
> which is down to i386)

here is my quasi-informed opinion ( i worked at BSQUARE for 3+ years, with
the most notable projects i contributed to being the super-secret
cleanroom port of HP's chaivm to wince that never saw the light of day and
the v3.0 kernel debugger )

note that arm is an *intel* product ( i know it wasnt *always* but it is
now! )

bsquare did all the compilers except x86 and was supposed to work on all
of the cpu prjects *except* arm.

why? 'cause the intel relationship is the alpha and the omega of M$FT's
existence - sort of like a very, very jealous boyfriend.

if anybody else where to work with intel then they might screw up M$FT's
Good Thing(tm)

so, wince has serioulsy failed to meet expectations and msft's thrown in
the towel on the multiplatform support, probably because hitachi, toshiba,
etc where unwilling to fork out the big bux to continue funding

for v1, there was *insane* amounts of expensive reference hardware, with
the culmination being the 'odo' which was this amazing contraption that
had a humoungous fpga that you could flash with a new 'cpu personality'.
yup. same hardware, completely different cpu. what an engineering marvel
at us$8K per item!

so, any company that wanted wince ported to it's cpu had to fork out for a
few hundred of these.

after the disappointing sales of v1, this didnt sit real well. :-)

anyway, M$FT had to walk a tightrope between:

 alienating intel, who helps them make fat margins on desktop boxen,

 cutting into M$FT's desktop hegemony ( notice a distinct lack of
 wince harddrives even tho wince supported ataflash? ),

 pissing off hitachi, casio, et all, since they where funding wince...

in the end, i am sure that the semi's where beating M$FT over the head
with linux as a way to get the license cost down ( recall that the vtech
helio was going to be a wince device until they concluded that the license
costs made it impractical ).

 This has come back to bite the semi's in the ass because they neglected
to actually *ask* any open source devs if they would be willing to write a
kernel from scratch for them out of the goodness of their hearts!

so, let's be clear here, the strongarm is an astoundingly good design. i
think that it's probably the first thing since the x86 that is likely to
become ubiquitous in personal computing the way that the x86 did - good
code size, flexible execution modes ( makes for a real challege to write a
debugger fwiw), *competent* manufacturing and marketing....

DO NOT FLAME ME :-)  mips's has no champion because nobody really owns it,
hitachi doesnt like to make big noises and motorola should have gotten out
of the cpu business after the 68hc11 because those people just can not
execute. they can design, they can write great documentation, but they
just cant sell anything at a cost that makes sense.

so, once again, M$FT philandered, maybe even slept around a bit. but when
times got tuff and the other flashy romances didnt pan out, it returned to
the dysfucntional but cozy embrace of it's true lover and enabler, intel.

John L. Utz III

Idiocy is the Impulse Function in the Convolution of Life