Getting X on Linux/SGI

Getting X on Linux/SGI

[Ariel Faigon]

[note the change in subject]

First a note to all SGI linux subscribers:
There are about 10 interested parties external to SGI
on this list.  I just say this so you're all aware of it.


:
:Given that it's unlikely we'd release the source code to our gfx drivers,
:
Well, after getting some hardware, that's our next hurdle :-)

I believe it would be a very good idea to release the Indy low-level
graphics source (even under NDA, although personally, I wouldn't
use NDAs) to XFree developers.  I'm not even talking amazing 3D or
Octane stuff here.  Let's first get basic X11 running on Indys
then worry about OpenGL / O2s etc.

Note that this means getting the basic 2D stuff running.  Heck, I can't
understand the logic of anyone being so protective about 5 year
technology, it is available on every PC with mid-range level graphics
by now.


:
:I realize this comes off as fairly negative, but I'm just trying to explain the
:issues involved once gfx gets added to the mix.  There would need to be a
:buy-in at a very high level of SGI mgmt. before we could start making the
:hardware details of our graphics subsystems available (at least the more modern
:ones, such as O2, Impact, etc.).
:
John, lest I sound negative, I don't mean to.  I hope all the
people on this list can agree on such obvious things.  I hope
that that fuzzy cloud called "upper level management" will
somehow transform into a person I can talk to.  My experience
is that once you get to the right upper level person, and
you state your case sensibly, you get what you want.

If anyone on the list knows the people to talk to to get this
happen please share.


P.S:
It is interesting to note how the SPARC port happened despite Sun
never releasing low-level stuff (as if they had anything to lose
by that) and David's reverse engineering all their stuff.  SGI
has nothing to lose and everything to gain from cooperation with
the hacker's community.  Linux is running on Alpha and SPARC
by now.  We don't want to see Linux running on HP and IBM
before it does on our iron, do we ?

-- 
Peace, Ariel

Getting X on Linux/SGI

[Ariel Faigon]

[note the change in subject]

First a note to all SGI linux subscribers:
There are about 10 interested parties external to SGI
on this list.  I just say this so you're all aware of it.


:
:Given that it's unlikely we'd release the source code to our gfx drivers,
:
Well, after getting some hardware, that's our next hurdle :-)

I believe it would be a very good idea to release the Indy low-level
graphics source (even under NDA, although personally, I wouldn't
use NDAs) to XFree developers.  I'm not even talking amazing 3D or
Octane stuff here.  Let's first get basic X11 running on Indys
then worry about OpenGL / O2s etc.

Note that this means getting the basic 2D stuff running.  Heck, I can't
understand the logic of anyone being so protective about 5 year
technology, it is available on every PC with mid-range level graphics
by now.


:
:I realize this comes off as fairly negative, but I'm just trying to explain the
:issues involved once gfx gets added to the mix.  There would need to be a
:buy-in at a very high level of SGI mgmt. before we could start making the
:hardware details of our graphics subsystems available (at least the more modern
:ones, such as O2, Impact, etc.).
:
John, lest I sound negative, I don't mean to.  I hope all the
people on this list can agree on such obvious things.  I hope
that that fuzzy cloud called "upper level management" will
somehow transform into a person I can talk to.  My experience
is that once you get to the right upper level person, and
you state your case sensibly, you get what you want.

If anyone on the list knows the people to talk to to get this
happen please share.


P.S:
It is interesting to note how the SPARC port happened despite Sun
never releasing low-level stuff (as if they had anything to lose
by that) and David's reverse engineering all their stuff.  SGI
has nothing to lose and everything to gain from cooperation with
the hacker's community.  Linux is running on Alpha and SPARC
by now.  We don't want to see Linux running on HP and IBM
before it does on our iron, do we ?

-- 
Peace, Ariel

Re: Getting X on Linux/SGI

[Christopher W. Carlson]

On Jun 17,  5:34pm, Ariel Faigon wrote:
>
> P.S:
> It is interesting to note how the SPARC port happened despite Sun
> never releasing low-level stuff (as if they had anything to lose
> by that) and David's reverse engineering all their stuff.  SGI
> has nothing to lose and everything to gain from cooperation with
> the hacker's community.  Linux is running on Alpha and SPARC
> by now.  We don't want to see Linux running on HP and IBM
> before it does on our iron, do we ?
>
> --
> Peace, Ariel
>-- End of excerpt from Ariel Faigon


I agree.  I'm not sure, but how much are we really giving away if we
provide some of our software to the linux community?  Imagine the
proliferation of OpenGL software that would be generated and the
number of computers we would sell if Joe Shmoe could buy an O2 with
Linux and OpenGL support?  SGI is a hardware company!  It surprises me
that much of our company has forgotten that.

-- 

		Chris Carlson

	+------------------------------------------------------+
	| Also, carlson@sgi.com                                |
	|   Work:   (714) 224-4530                             |
	|   Vnet:       6-678-4530     FAX:    (714) 833-9503  |
	|                                                      |
	| Trivia fact: an electroencephalogram shows that a    |
	| human brain and a bowl of quivering lime Jell-O have |
	| the same waves.  [Time Magazine, Mar 17, 1997]       |
	+------------------------------------------------------+

Re: Getting X on Linux/SGI

[Nigel Gamble]

On Wed, 18 Jun 1997, Christopher W. Carlson wrote:
>                            SGI is a hardware company!  It surprises me
> that much of our company has forgotten that.

If it was ever true, it's certainly not true any longer.  SGI is
a systems company.  By which I mean that our "edge" over the
competition comes from a combination of hardware and software
working very closely together.


Nigel Gamble       "Are we going to push the edge of the envelope, Brain?"
Silicon Graphics   "No, Pinky, but we may get to the sticky part."
nigel@sgi.com
(415) 933-3109

Re: Getting X on Linux/SGI

[John Chen]

On Jun 18,  9:41am, Ken Klingman wrote:
> Subject: Getting X on Linux/SGI
> John,
>
> There's an effort going to port Linux to SGI systems.  They've
> got the basic kernel up on an Indy, but are now trying to figure
> out what to do with X.  Could you reply to the mail alias (linux@engr)
> with a few words on the effort involved and what kind of specs they
> need?  I know that our X server architecture is complicated by the
> support of OpenGL, but wouldn't it be a lot simpler to just do X
> without any OpenGL support?
>
> ...
>
> Ken
>
> ...
>
> Forwarded message:
> > Larry McVoy wrote:
> > >
> > > : There are some very serious issues which come up even getting
> > > : Xfree to a moderate level of acceleration.
> > >
> > > How about to a simple level of working?  Without any acceleration?
> > > For most people, just having xterms and netscape working is enough.
> > >
> > > I'm not a graphics or X person.  Could someone who knows SGI's gfx
> > > devices tell us how hard it would be to make the basics work?

The fastest path to make basic X works is probably by implementing
a device dependent layer (DDX) that controls all pixels going in/out
framebuffer, opens and closes devices like keyboard and mouse,
interpret input events and control graphic display backend.
Since frame buffer on Indy system is not directly accessible.
all pixels in/out frame buffer has to be via this DDX layer.
A lot of sample code under cfb directory should be
reuseable with minor changes. To begin with non-accelerated
rendering, only setpixel and getpixel are required.

A generic DDX layer which described in "The X Window System Server" book
(authors: Elias Israel and Erik Fortune) can be FTP from export.lcs.mit.edu

I think you also need following stuff for Indy system with Newport graphic:

1) Newport graphics spec.
2) A basic graphic driver that map RE chip to Xsgi's address
   space, so X can program RE registers
3) This graphic driver also needs to set up graphic backend display
   id table and display mode registers (if X support only one visual,
   this step is simple), to program Cmap for cursor color
   and to program cursor registers for glyph and location.
   All of these work can also be done in X if driver map backend to X.

> > >
> >
> > ...
> >
> >  Questions:
> >
> > - How much HW dependent stuff is in Xsgi itself?

Xsgi contains something specific to IRIX, e.g. share memory input
queue stuff which may not be wanted by Linux.

> > - Which of the DSOs in /usr/lib/X11/dyDDX are minimally
> >   needed to bring up an non-GLX Xserver?

All DSOs in dyDDX support GLX. If the system is Indy,
rex3.so is used for Newport graphic and exp.so is used for
EXPRESS graphic. By trimming down Xsgi and DSO's, you should be able
to run Xsgi on Linux without GLX.

-John

> > - How much efforts would it cost to compile the dyDDX
> >   stuff for Linux and distribute the binaries only
> >   (assuming that there is not to much HW stuff in Xsgi itself)?
> >
> >
> >  And of course, we probably have to provide the microcode and
> > loader for the different GFX cards.
> >
> >  I definitely agree with Ariel, that this is the most important
> > topic once we have Linux running stable.
> >
> > Martin
> >
>-- End of excerpt from Ken Klingman

Re: Getting X on Linux/SGI

[John Chen]

On Jun 18,  9:41am, Ken Klingman wrote:
> Subject: Getting X on Linux/SGI
> John,
>
> There's an effort going to port Linux to SGI systems.  They've
> got the basic kernel up on an Indy, but are now trying to figure
> out what to do with X.  Could you reply to the mail alias (linux@engr)
> with a few words on the effort involved and what kind of specs they
> need?  I know that our X server architecture is complicated by the
> support of OpenGL, but wouldn't it be a lot simpler to just do X
> without any OpenGL support?
>
> ...
>
> Ken
>
> ...
>
> Forwarded message:
> > Larry McVoy wrote:
> > >
> > > : There are some very serious issues which come up even getting
> > > : Xfree to a moderate level of acceleration.
> > >
> > > How about to a simple level of working?  Without any acceleration?
> > > For most people, just having xterms and netscape working is enough.
> > >
> > > I'm not a graphics or X person.  Could someone who knows SGI's gfx
> > > devices tell us how hard it would be to make the basics work?

The fastest path to make basic X works is probably by implementing
a device dependent layer (DDX) that controls all pixels going in/out
framebuffer, opens and closes devices like keyboard and mouse,
interpret input events and control graphic display backend.
Since frame buffer on Indy system is not directly accessible.
all pixels in/out frame buffer has to be via this DDX layer.
A lot of sample code under cfb directory should be
reuseable with minor changes. To begin with non-accelerated
rendering, only setpixel and getpixel are required.

A generic DDX layer which described in "The X Window System Server" book
(authors: Elias Israel and Erik Fortune) can be FTP from export.lcs.mit.edu

I think you also need following stuff for Indy system with Newport graphic:

1) Newport graphics spec.
2) A basic graphic driver that map RE chip to Xsgi's address
   space, so X can program RE registers
3) This graphic driver also needs to set up graphic backend display
   id table and display mode registers (if X support only one visual,
   this step is simple), to program Cmap for cursor color
   and to program cursor registers for glyph and location.
   All of these work can also be done in X if driver map backend to X.

> > >
> >
> > ...
> >
> >  Questions:
> >
> > - How much HW dependent stuff is in Xsgi itself?

Xsgi contains something specific to IRIX, e.g. share memory input
queue stuff which may not be wanted by Linux.

> > - Which of the DSOs in /usr/lib/X11/dyDDX are minimally
> >   needed to bring up an non-GLX Xserver?

All DSOs in dyDDX support GLX. If the system is Indy,
rex3.so is used for Newport graphic and exp.so is used for
EXPRESS graphic. By trimming down Xsgi and DSO's, you should be able
to run Xsgi on Linux without GLX.

-John

> > - How much efforts would it cost to compile the dyDDX
> >   stuff for Linux and distribute the binaries only
> >   (assuming that there is not to much HW stuff in Xsgi itself)?
> >
> >
> >  And of course, we probably have to provide the microcode and
> > loader for the different GFX cards.
> >
> >  I definitely agree with Ariel, that this is the most important
> > topic once we have Linux running stable.
> >
> > Martin
> >
>-- End of excerpt from Ken Klingman

Re: Getting X on Linux/SGI

[David S. Miller]

   From: "John Chen" <chen@betty.esd.sgi.com>
   Date: Wed, 18 Jun 1997 12:41:23 -0700

   I think you also need following stuff for Indy system with Newport graphic:

   1) Newport graphics spec.

I have this.

   2) A basic graphic driver that map RE chip to Xsgi's address
      space, so X can program RE registers

15 minutes of coding...

   3) This graphic driver also needs to set up graphic backend display
      id table and display mode registers (if X support only one visual,
      this step is simple), to program Cmap for cursor color
      and to program cursor registers for glyph and location.
      All of these work can also be done in X if driver map backend to X.

I can do this since I have #1, in fact I might be setting up the
backend display for one visual already in the text console driver I
wrote.  The way to manipulate the cursor and cmap is pretty much
documented in my text console driver as well, but I think some of the
actual cursor code is just pound define'd out but it is/was there.
(worse case you have to sift through the CVS history for the driver
and check out a version right before I snipped the code out if I in
fact did remove it at some point)

As for the input queue stuff, this has already been implemented on the
Sparc port because we were using SunOS Xsun binaries long ago, the
stock X11R6 sources use this mechanism anyways in the Sun frame buffer
support code, and their interface is very similar to IRIX's I think
(they call it VUID events).

Re: Getting X on Linux/SGI

[Alex deVries]

On Wed, 18 Jun 1997, David S. Miller wrote:

>    From: "John Chen" <chen@betty.esd.sgi.com>
>    Date: Wed, 18 Jun 1997 12:41:23 -0700
>    I think you also need following stuff for Indy system with Newport graphic:
>    1) Newport graphics spec.
> I have this.

Q: is the newport graphics spec available outside of SGI?  If so, is there
any way for me to get my hands on a copy of it, either electronically or
physically?

- Alex

Re: Getting X on Linux/SGI

[Larry McVoy]

: On Wed, 18 Jun 1997, Christopher W. Carlson wrote:
: >                            SGI is a hardware company!  It surprises me
: > that much of our company has forgotten that.
: 
: If it was ever true, it's certainly not true any longer.  SGI is
: a systems company.  By which I mean that our "edge" over the
: competition comes from a combination of hardware and software
: working very closely together.

We can argue this all day.  It's pointless.

Something that is not pointless, something that the nay-sayers should
take to heart, is that more than 10% of Alpha sales last year were Linux.
I confirmed this with Digital this morning - the number is somewhere
between 25 and 30 thousand Alphas running Linux.

Add to that the 4-6 million PCs running Linux and you start to have
a story.

Add to that the 19 million embedded MIPS chips and you start to ask
yourself why we aren't more serious about this market.

So, while we could sit around and argue with those who disagree, I say
that there is more than enough evidence that there is a market there.
If there's a market, we do it.  That's the SGI way.

Re: Getting X on Linux/SGI

[Nigel Gamble]

On Wed, 18 Jun 1997, Larry McVoy wrote:
> : On Wed, 18 Jun 1997, Christopher W. Carlson wrote:
> : >                            SGI is a hardware company!  It surprises me
> : > that much of our company has forgotten that.
> : 
> : If it was ever true, it's certainly not true any longer.  SGI is
> : a systems company.  By which I mean that our "edge" over the
> : competition comes from a combination of hardware and software
> : working very closely together.
> 
> We can argue this all day.  It's pointless.

Who is arguing?  I'm just trying to point out that belittling
the efforts of all the software engineers at SGI is hardly likely
to endear your cause to them.  There are those of us who think
there is a place for both Linux and IRIX (although I know that
you, Larry, are not one of them); so please try to understand
where we are coming from.


Nigel Gamble       "Are we going to push the edge of the envelope, Brain?"
Silicon Graphics   "No, Pinky, but we may get to the sticky part."
nigel@sgi.com
(415) 933-3109

Re: Getting X on Linux/SGI

[Ralf Baechle]

Larry McVoy wrote:

> We can argue this all day.  It's pointless.
> 
> Something that is not pointless, something that the nay-sayers should
> take to heart, is that more than 10% of Alpha sales last year were Linux.
> I confirmed this with Digital this morning - the number is somewhere
> between 25 and 30 thousand Alphas running Linux.

Jon "Maddog" Hall says that the existence of Linux for Alphas helped
Digital to sell more Digital UNIX systems than they would have sold
without Linux.  His explanation is simple.  First Linux pulls their
customers into the UNIX board.  Somewhen the customers have a problem
that cannot be solved by Linux but a Digital UNIX system.  So it's a
logical next step for them to use such a systems.

SGI is mostly in the systems bussines and the market of PC class MIPS
machines is carefully expressed not doing very well, so this does not
100% apply to their situation.  The point is that this not a Linux vs.
IRIX situation.

> Add to that the 4-6 million PCs running Linux and you start to have
> a story.
> 
> Add to that the 19 million embedded MIPS chips and you start to ask
> yourself why we aren't more serious about this market.

A large fraction of these embedded MIPS chips is able to run Linux from
it's hardware specs.  And I stopped counting the number of companies
that have contaced me about embedded Linux.  I admit that my and probably
most other people's core interest is in "real computers" ...

> So, while we could sit around and argue with those who disagree, I say
> that there is more than enough evidence that there is a market there.
> If there's a market, we do it.  That's the SGI way.

  Ralf

Re: Getting X on Linux/SGI

[Ralf Baechle]

Larry McVoy wrote:

> We can argue this all day.  It's pointless.
> 
> Something that is not pointless, something that the nay-sayers should
> take to heart, is that more than 10% of Alpha sales last year were Linux.
> I confirmed this with Digital this morning - the number is somewhere
> between 25 and 30 thousand Alphas running Linux.

Jon "Maddog" Hall says that the existence of Linux for Alphas helped
Digital to sell more Digital UNIX systems than they would have sold
without Linux.  His explanation is simple.  First Linux pulls their
customers into the UNIX board.  Somewhen the customers have a problem
that cannot be solved by Linux but a Digital UNIX system.  So it's a
logical next step for them to use such a systems.

SGI is mostly in the systems bussines and the market of PC class MIPS
machines is carefully expressed not doing very well, so this does not
100% apply to their situation.  The point is that this not a Linux vs.
IRIX situation.

> Add to that the 4-6 million PCs running Linux and you start to have
> a story.
> 
> Add to that the 19 million embedded MIPS chips and you start to ask
> yourself why we aren't more serious about this market.

A large fraction of these embedded MIPS chips is able to run Linux from
it's hardware specs.  And I stopped counting the number of companies
that have contaced me about embedded Linux.  I admit that my and probably
most other people's core interest is in "real computers" ...

> So, while we could sit around and argue with those who disagree, I say
> that there is more than enough evidence that there is a market there.
> If there's a market, we do it.  That's the SGI way.

  Ralf

Re: Getting X on Linux/SGI

[Christopher W. Carlson]

On Jun 18,  4:42pm, Nigel Gamble wrote:
> Subject: Re: Getting X on Linux/SGI
> On Wed, 18 Jun 1997, Larry McVoy wrote:
> > : On Wed, 18 Jun 1997, Christopher W. Carlson wrote:
> > : >                            SGI is a hardware company!  It surprises me
> > : > that much of our company has forgotten that.
> > : 
> > : If it was ever true, it's certainly not true any longer.  SGI is
> > : a systems company.  By which I mean that our "edge" over the
> > : competition comes from a combination of hardware and software
> > : working very closely together.
> > 
> > We can argue this all day.  It's pointless.
> 
> Who is arguing?  I'm just trying to point out that belittling
> the efforts of all the software engineers at SGI is hardly likely
> to endear your cause to them.  There are those of us who think
> there is a place for both Linux and IRIX (although I know that
> you, Larry, are not one of them); so please try to understand
> where we are coming from.
> 
> 
>-- End of excerpt from Nigel Gamble


Please!  Don't think that I am belittling the software engineers!  I
had no intent on doing that.  I *am* one of them.

My point was that SGI is into making hardware.  That's how it started
and that is still its main focus.  Yes, it is true that our software
is very important because it takes advantage of the proprietary stuff
in the hardware, but the main income for SGI is hardware.  Note in any
press releases and stock holder's documents how much is said about the
wonderful software SGI produces (Alias/Wavefront is still a separate
entity).  You won't find much.

-- 

		Chris Carlson

	+------------------------------------------------------+
	| Also, carlson@sgi.com                                |
	|   Work:   (714) 224-4530                             |
	|   Vnet:       6-678-4530     FAX:    (714) 833-9503  |
	|                                                      |
	| Trivia fact: an electroencephalogram shows that a    |
	| human brain and a bowl of quivering lime Jell-O have |
	| the same waves.  [Time Magazine, Mar 17, 1997]       |
	+------------------------------------------------------+

Re: Getting X on Linux/SGI

[William J. Earl]

Alex deVries writes:
 > 
 > On Wed, 18 Jun 1997, David S. Miller wrote:
 > 
 > >    From: "John Chen" <chen@betty.esd.sgi.com>
 > >    Date: Wed, 18 Jun 1997 12:41:23 -0700
 > >    I think you also need following stuff for Indy system with Newport graphic:
 > >    1) Newport graphics spec.
 > > I have this.
 > 
 > Q: is the newport graphics spec available outside of SGI?  If so, is there
 > any way for me to get my hands on a copy of it, either electronically or
 > physically?

     With the approval of Tom Furlong (then the SGI VP/GM in charge of Indy),
I provided the specification to the Linux-on-SGI project (in the person
of David Miller :-) ).  It is only available in paper, and David may be
the quickest source for it.  (I am now in a different division, far from
the home of the paper documents.)

Re: Getting X on Linux/SGI

[Steve Alexander]

"Christopher W. Carlson" <carlson@heaven.newport.sgi.com> writes:
>My point was that SGI is into making hardware.  That's how it started
>and that is still its main focus.  Yes, it is true that our software
>is very important because it takes advantage of the proprietary stuff
>in the hardware, but the main income for SGI is hardware.

How many people buy the hardware and program the bare metal?

>Note in any
>press releases and stock holder's documents how much is said about the
>wonderful software SGI produces (Alias/Wavefront is still a separate
>entity).  You won't find much.

That's a separate issue, which is that very few people here value software.
That will probably change with time if the company is going to continue to
be successful as the industry shifts.

-- Steve

Re: Getting X on Linux/SGI

[Larry McVoy]

: "Christopher W. Carlson" <carlson@heaven.newport.sgi.com> writes:
: >My point was that SGI is into making hardware.  That's how it started
: >and that is still its main focus.  Yes, it is true that our software
: >is very important because it takes advantage of the proprietary stuff
: >in the hardware, but the main income for SGI is hardware.
: 
: How many people buy the hardware and program the bare metal?

We own MIPS.  MIPS sold or licensed 19.2 million chips last year.  About
19 million of those were of the type "program the bare metal".

Much of the Linux/MIPS interest is for embedded systems.  We should have
done this years ago, there is no reason why not to do it other than NIH.

Re: Getting X on Linux/SGI

[Steve Alexander]

lm@neteng (Larry McVoy) writes:
>We own MIPS.  MIPS sold or licensed 19.2 million chips last year.  About
>19 million of those were of the type "program the bare metal".

MIPS Chips != SGI systems.  My question still stands.

-- Steve

Re: Getting X on Linux/SGI

[Steve Alexander]

lm@neteng (Larry McVoy) writes:
>We own MIPS.  MIPS sold or licensed 19.2 million chips last year.  About
>19 million of those were of the type "program the bare metal".

MIPS Chips != SGI systems.  My question still stands.

-- Steve

Re: Getting X on Linux/SGI

[David S. Miller]

   Date: Fri, 20 Jun 1997 00:59:06 -0700
   From: Steve Alexander <sca@refugee.engr.sgi.com>

   lm@neteng (Larry McVoy) writes:
   >We own MIPS.  MIPS sold or licensed 19.2 million chips last year.  About
   >19 million of those were of the type "program the bare metal".

   MIPS Chips != SGI systems.  My question still stands.

All your assertion proves is that today IRIX is the "enabling"
software technology which puts "SGI systems" out the door.

Don't get me wrong, I was constantly reminded what keeps the lights on
in bldg. 9 when I was there last summer, but when and if Linux begins
to do some (not all) of that "enabling", people might begin to
perceive the situation a bit differently.

Re: Getting X on Linux/SGI

[Steve Alexander]

"David S. Miller" <davem@jenolan.rutgers.edu> writes:
>Don't get me wrong, I was constantly reminded what keeps the lights on
>in bldg. 9 when I was there last summer, but when and if Linux begins
>to do some (not all) of that "enabling", people might begin to
>perceive the situation a bit differently.

Don't get me wrong.  I'm not opposed to Linux on SGI.  In fact, I'm in favor
of it, at least on some set of systems.  I don't believe that Linux on 64P
O2000s makes a lot of sense yet, but I think it would be nice to have Linux
running on O200s and Indys.

Personally, I wouldn't buy an O2 to run Linux.

-- Steve

Re: Getting X on Linux/SGI

[David S. Miller]

   From: fisher@hollywood.engr.sgi.com (William Fisher)
   Date: Fri, 20 Jun 1997 01:55:31 -0800 (PDT)

I'm glad someone stepped in this.

	   Nearly all of those systems were nearly bare metal software
	   environments since we had numerous talks with Chet at MIPS
	   about software features that would propell the business.
	   NONE of those customers wanted anything to do with any
	   flavor of Unix, Linux, BSD, System V, etc.

	   In fact Hunter-Ready and the other real-time kernels
	   weren't that popular with the hard core real-time/embedded
	   folks.

	   Hence I would like to here more about larry's definition of
	   "embedded systems" because they sure don't sound like any
	   that our semiconductor partners sell to.

Allow me...

Let me relay a story a co-Sparc developer of mine told (and showed!)
me when I saw him in person in Germany the other week.

His name is Eddie C. Dost and he works for a software/hardware house
there where they do nothing but embedded work.  About two weeks ago he
was asked to tackle two projects.  One involved an i386SX based board
which controlled lighting systems for skyscrapers, the other was a
m68k based board which was meant to be deployed by the hundreds to
control campus wide key card entry systems for a bunch of corporations
and universities in the German speaking countries.

He was given QNX to write his drivers and get the box going.  Since it
is a micro-kernel, you have to perform a task switch to handle even an
interrupt, and you have to compile your interrupt handlers with a
special compiler and compiler options using QNX compilers just to cope
with this bullshit.  This was on both systems.

Interrupt response was so slow, that even when he coded the drivers in
raw optimized assembly he could not keep up and would drop characters
easily on his serial ports, the ISDN performance sucked balls as well.

So eddie got so frustrated one night that he took both the m68k and
Intel ports of Linux, in about an hour added kernel build time
configuration options such as "CONFIG_NO_MEMORY_MANAGEMENT",
"CONFIG_NO_FANCY_SYSCALLS", "CONFIG_NO_USELESS_FEATURES" and the like
to the point where he was able to get a 120k sized Linux kernel with
his drivers and the specialized code to run the control systems he
needed to deploy, and he got full over the serial line KGDB source
level debugging of his kernel as well.

The next night he got it completely working and debugged, needless to
say this thing didn't have the interrupt performance problems QNX
did.  The next evening he blew the first revisions of the PROM's the
boxes would eventually use in production when these things got sent
to the customers.

Now here comes the interesting part...

He had to then approach his boss, he explained the issues and how he
had solved the problem.  The boss went "what, use Linux, why the fuck
would we even want to do such a thing?"

His response was: "Firstly, QNX doesn't even fucking work, at best the
product would be delayed if I could get it to work with QNX at all,
and thus you might even lose these sales.  Secondly, I just saved you
a shitload of money and future engineering costs.  Not only did I just
save you the QNX license on all the hundreds of these fuckers you are
going to sell to people (you can sell it at the same prive, the
customer won't know the difference), I have a usable mini-Linux kernel
that we can use (and thus save the licensing fee's + the engineering
all over again) as a base for a lot of our future embedded products."

His boss was blown away, the rest is history...

So you see, the real issue is that people don't know they want it...

Oh btw, Eddie will in fact be working on some MIPS based embedded
boards in the near future, and I can tell you he sure as hell isn't
gonna be putting QNX or any other "embedded OS" in those things.

Later,
David "Sparc" Miller
davem@caip.rutgers.edu

Re: Getting X on Linux/SGI

[William Fisher]

> 
> : "Christopher W. Carlson" <carlson@heaven.newport.sgi.com> writes:
> : >My point was that SGI is into making hardware.  That's how it started
> : >and that is still its main focus.  Yes, it is true that our software
> : >is very important because it takes advantage of the proprietary stuff
> : >in the hardware, but the main income for SGI is hardware.
> : 
> : How many people buy the hardware and program the bare metal?
> 
> We own MIPS.  MIPS sold or licensed 19.2 million chips last year.  About
> 19 million of those were of the type "program the bare metal".
> 
> Much of the Linux/MIPS interest is for embedded systems.  We should have
> done this years ago, there is no reason why not to do it other than NIH.
> 
	If the demand for Linux on MIPS chips is comming from the "embedded"
	systems then things sure have changed since the early days of
	MIPS Computer Systems. The embedded market in 1992-1996 was
	NOT asking us for Linux, since it was very sensitive to running
	small real-time kernels. In fact nearly all of the original
	chips vendors who licensed the chips used the PROM code, remote
	dbx, compilers, etc. which were sold/licensed to the chip vendors
	for resell to the embedded customers.

	If fact the original VP of embedded system marketing at MIPS was
	Chet Silvestri, now VP of SPARC embedded systems at SUN. Chet
	was very familar with embedded systems having worked at Intel
	on numerous automobile systems, real-time process control, etc.
	
	Nearly all of those systems were nearly bare metal software
	environments since we had numerous talks with Chet at MIPS
	about software features that would propell the business.
	NONE of those customers wanted anything to do with any flavor
	of Unix, Linux, BSD, System V, etc.

	In fact Hunter-Ready and the other real-time kernels weren't that
	popular with the hard core real-time/embedded folks.

	Hence I would like to here more about larry's definition of
	"embedded systems" because they sure don't sound like any
	that our semiconductor partners sell to.

	I would like to see Linux on the low end of the SGI workstations
	for two reasons; One as an existance proof that a lightweigth
	Unix can be shown to run very fast on our hardware and secondly
	for sale to either universities or other research labs that want
	source code and want to hack. I don't think it will sell that many
	machines but it can't hurt.

	I personally believe that the popularity of Linux on ALPHA is
	pushed by the Digital Semiconductor Division and they will do
	anything to sell chips and Digital systems. I think it is
	similar behavior that you see in both IBM's Semiconductor Division
	and at the Motorola's Semiconductor Division with the PowerPC.
	That being they have no particular religion on any software or
	OS, anything that moves chips is alright with them.

-- Bill

Re: Getting X on Linux/SGI

[William Fisher]

> 
> : "Christopher W. Carlson" <carlson@heaven.newport.sgi.com> writes:
> : >My point was that SGI is into making hardware.  That's how it started
> : >and that is still its main focus.  Yes, it is true that our software
> : >is very important because it takes advantage of the proprietary stuff
> : >in the hardware, but the main income for SGI is hardware.
> : 
> : How many people buy the hardware and program the bare metal?
> 
> We own MIPS.  MIPS sold or licensed 19.2 million chips last year.  About
> 19 million of those were of the type "program the bare metal".
> 
> Much of the Linux/MIPS interest is for embedded systems.  We should have
> done this years ago, there is no reason why not to do it other than NIH.
> 
	If the demand for Linux on MIPS chips is comming from the "embedded"
	systems then things sure have changed since the early days of
	MIPS Computer Systems. The embedded market in 1992-1996 was
	NOT asking us for Linux, since it was very sensitive to running
	small real-time kernels. In fact nearly all of the original
	chips vendors who licensed the chips used the PROM code, remote
	dbx, compilers, etc. which were sold/licensed to the chip vendors
	for resell to the embedded customers.

	If fact the original VP of embedded system marketing at MIPS was
	Chet Silvestri, now VP of SPARC embedded systems at SUN. Chet
	was very familar with embedded systems having worked at Intel
	on numerous automobile systems, real-time process control, etc.
	
	Nearly all of those systems were nearly bare metal software
	environments since we had numerous talks with Chet at MIPS
	about software features that would propell the business.
	NONE of those customers wanted anything to do with any flavor
	of Unix, Linux, BSD, System V, etc.

	In fact Hunter-Ready and the other real-time kernels weren't that
	popular with the hard core real-time/embedded folks.

	Hence I would like to here more about larry's definition of
	"embedded systems" because they sure don't sound like any
	that our semiconductor partners sell to.

	I would like to see Linux on the low end of the SGI workstations
	for two reasons; One as an existance proof that a lightweigth
	Unix can be shown to run very fast on our hardware and secondly
	for sale to either universities or other research labs that want
	source code and want to hack. I don't think it will sell that many
	machines but it can't hurt.

	I personally believe that the popularity of Linux on ALPHA is
	pushed by the Digital Semiconductor Division and they will do
	anything to sell chips and Digital systems. I think it is
	similar behavior that you see in both IBM's Semiconductor Division
	and at the Motorola's Semiconductor Division with the PowerPC.
	That being they have no particular religion on any software or
	OS, anything that moves chips is alright with them.

-- Bill

Re: Getting X on Linux/SGI

[Ralf Baechle]

> So you see, the real issue is that people don't know they want it...
> 
> Oh btw, Eddie will in fact be working on some MIPS based embedded
> boards in the near future, and I can tell you he sure as hell isn't
> gonna be putting QNX or any other "embedded OS" in those things.

Three people/companies have already started basing embedded systems
based on the MIPS/Linux [1] at times when I personally wouldn't have
trusted the entire MIPS port five minutes without peeking at it.

Toshiba is currently porting Linux to a new R-family type processor
for verification of the processor.  Hey, it does run Linux even before
there is Silicon at all!

Dave is right:

> So you see, the real issue is that people don't know they want it...

You see, we have not yet really taken off with the MIPS port and we
already have some impact on the market.

Another interesting embedded project will be done by Alan Cox.  He
intends to port Linux to the R4650, a CPU without a TLB but just a
pair of base/bounds registers.  Not an ELKS port but an almost full
port of Linux just without certain features like mmap-ing.

  Ralf

[1] Not going to be M/Linux ;-)

Re: Getting X on Linux/SGI

[Ralf Baechle]

> So you see, the real issue is that people don't know they want it...
> 
> Oh btw, Eddie will in fact be working on some MIPS based embedded
> boards in the near future, and I can tell you he sure as hell isn't
> gonna be putting QNX or any other "embedded OS" in those things.

Three people/companies have already started basing embedded systems
based on the MIPS/Linux [1] at times when I personally wouldn't have
trusted the entire MIPS port five minutes without peeking at it.

Toshiba is currently porting Linux to a new R-family type processor
for verification of the processor.  Hey, it does run Linux even before
there is Silicon at all!

Dave is right:

> So you see, the real issue is that people don't know they want it...

You see, we have not yet really taken off with the MIPS port and we
already have some impact on the market.

Another interesting embedded project will be done by Alan Cox.  He
intends to port Linux to the R4650, a CPU without a TLB but just a
pair of base/bounds registers.  Not an ELKS port but an almost full
port of Linux just without certain features like mmap-ing.

  Ralf

[1] Not going to be M/Linux ;-)

Re: Getting X on Linux/SGI

[Todd Shrider]

My take on things:

On Fri, 20 Jun 1997, William Fisher wrote:

> 	If the demand for Linux on MIPS chips is comming from the "embedded"
> 	systems then things sure have changed since the early days of
> 	MIPS Computer Systems. The embedded market in 1992-1996 was
> 	NOT asking us for Linux, since it was very sensitive to running
> 	small real-time kernels. In fact nearly all of the original
> 	chips vendors who licensed the chips used the PROM code, remote
> 	dbx, compilers, etc. which were sold/licensed to the chip vendors
> 	for resell to the embedded customers.

OK, point taken, but I think that the definition of an embeded system HAS
changed draticly from five years ago. Then you were thinking chips in
cars, elevators, etc, etc, etc. Probably, at the time, the most kerenl
heavy processes were running on chips in smart houses and there weren't a
whole hell of a lot of those.

Today, at least when I think of embeded systems, I think of small game
machines, set-top-boxes, real smart-house systems, etc. Systems that
aren't performing one or two tasks, but that are doing multiples of things
that can be very different and somewhat unreleated (except maybe in an
abstract sense). If you take an "ideal" set-top-box you think of something
that can hook up to the network, via cable or phone or whatever, browse
the web, ineract with the tv video feeds, receive and send mail (even if
only operating as a pop-client), browse news feeds, maybe even run
someting like Point-Cast as your "TV Screen Saver". 

When I think of systems, runnig chips like the 4300i, that can do things
like this, I think a pared down Linux is a very justified choice to run
them with. I'd rather it than Windows CE! If your still thinking of Linux
as a traditional Unix, you may still ask why, but if you think of it a
simplistic behind the scenes consumer app, than I think that also resolves
alot of problems.

> 	I would like to see Linux on the low end of the SGI workstations
> 	for two reasons; One as an existance proof that a lightweigth
> 	Unix can be shown to run very fast on our hardware and secondly
> 	for sale to either universities or other research labs that want
> 	source code and want to hack. I don't think it will sell that many
> 	machines but it can't hurt.

I'd disagree and point you to Digital. At ALE Maddog was talking about a
trip through Southern America where half-way digital actually advised him
to stop pitching Digital Unix and start talking about Linux. I admit that
right away it probably won't be precieved as the OS of choice right away,
even Alpha Linux is still not quite the pretty picture the Intel Linux is,
but with time it will gain support from the die hard fans (those people
who run Linux at home and SGI/IRIX at work). Then the OS get stable and
you find a way to run all those wonderful Irix apps on it (see EM86). Then
major companies start looking at it, and you get calls (like digital has
reported) where people make orders ($$$) and ask you to have Linux over
Irix. 

Look at it this way. Linux has a lot of support in the college community.
In five years a majority of these kids will be in industry, do you think
there gonna want Irix on there SGIs? 

> 	I personally believe that the popularity of Linux on ALPHA is
> 	pushed by the Digital Semiconductor Division and they will do
> 	anything to sell chips and Digital systems. I think it is
> 	similar behavior that you see in both IBM's Semiconductor Division
> 	and at the Motorola's Semiconductor Division with the PowerPC.
> 	That being they have no particular religion on any software or
> 	OS, anything that moves chips is alright with them.

I bought a system to runn it on, having never talked to anyone at digital.
You gotta love the 500mHz!!! And isn't that the way it should be? If you
let the users define what runs on the chips your going to have much
happier users.

Oh well, just my thoughts... Cheers!

---
Todd M. Shrider         Oops, My brain just hit a bad sector!
todds@ontko.com         http://www.ontko.com/~todds

Re: Getting X on Linux/SGI

[Ralf Baechle]

Hi,

> that can hook up to the network, via cable or phone or whatever, browse
> the web, ineract with the tv video feeds, receive and send mail (even if
> only operating as a pop-client), browse news feeds, maybe even run
> someting like Point-Cast as your "TV Screen Saver". 
> 
> When I think of systems, runnig chips like the 4300i, that can do things
> like this, I think a pared down Linux is a very justified choice to run
> them with. I'd rather it than Windows CE! If your still thinking of Linux

The Algorithmics P4032 which I recently mentioned I'm also running Linux
on has a R4300i.  The patches are too dirty and still based on 2.1.14,
so they're not in the CVS tree ...

I'd love to see one of the MIPS based palmtop class machines to run Linux ...

  Ralf

Re: Getting X on Linux/SGI

[Ralf Baechle]

Hi,

> that can hook up to the network, via cable or phone or whatever, browse
> the web, ineract with the tv video feeds, receive and send mail (even if
> only operating as a pop-client), browse news feeds, maybe even run
> someting like Point-Cast as your "TV Screen Saver". 
> 
> When I think of systems, runnig chips like the 4300i, that can do things
> like this, I think a pared down Linux is a very justified choice to run
> them with. I'd rather it than Windows CE! If your still thinking of Linux

The Algorithmics P4032 which I recently mentioned I'm also running Linux
on has a R4300i.  The patches are too dirty and still based on 2.1.14,
so they're not in the CVS tree ...

I'd love to see one of the MIPS based palmtop class machines to run Linux ...

  Ralf

Re: Getting X on Linux/SGI

[Martin Knoblauch]

Steve Alexander wrote:
> 
> 
> Personally, I wouldn't buy an O2 to run Linux.
> 

  I personally would not buy an O2 to run IRIX either. If I
buy one, then for running some "cool" apps that let me do
my job (better), or have let me have some real fun. The OS
finally involved is pretty much irrelevant (OK, this statement
is not true for NT, or is it).

 Which just means that whatever OS gives me more "cool" apps
will be more useful for me.

 OK. Enough marketing speak.

Martin
-- 
+---------------------------------+-----------------------------------+
|Martin Knoblauch                 | Silicon Graphics GmbH             |
|Manager Technical Marketing      | Am Hochacker 3 - Technopark       |
|Silicon Graphics Computer Systems| D-85630 Grasbrunn-Neukeferloh, FRG|
|---------------------------------| Phone: (+int) 89 46108-179 or -0  |
|http://reality.sgi.com/knobi     | Fax:   (+int) 89 46107-179        |
+---------------------------------+-----------------------------------+
|e-mail: <knobi@munich.sgi.com>   | VM: 6-333-8197 | M/S: IDE-3150    |
+---------------------------------------------------------------------+

Re: Getting X on Linux/SGI

[Larry McVoy]

: Personally, I wouldn't buy an O2 to run Linux.

OK, we won't try and sell you one.  What about the people that would buy
an O2 to run Linux?  

You != <the entire customer base>.

Re: Getting X on Linux/SGI

[Larry McVoy]

: lm@neteng (Larry McVoy) writes:
: >We own MIPS.  MIPS sold or licensed 19.2 million chips last year.  About
: >19 million of those were of the type "program the bare metal".
: 
: MIPS Chips != SGI systems.  My question still stands.

Have you forgotten Nintendo 64 sales?

MIPS Chips == $$$ for SGI.  $$$ for SGI is what it is all about, that's what
feeds the light bills.   If making a bunch of hackers happy turns into $$$
for SGI in the long term, then we do it.  At this point, I believe it is 
obvious that that is the case.

Re: Getting X on Linux/SGI

[Mike Shaver]

Thus spake Larry McVoy:
> : Personally, I wouldn't buy an O2 to run Linux.
> 
> OK, we won't try and sell you one.  What about the people that would buy
> an O2 to run Linux?  
>
> You != <the entire customer base>.

FWIW, there's a _lot_ of interest at my company in having one OS that
runs on all our boxes (Intel, SP*RC, Alpha, now MIPS), and Linux is
the preferred choice.  The reasons don't necessarily have anything
to do with disliking the native OS (except for Solaris, perhaps), but
we _would_ like:
- common file structure and apps
- source compatibility
- access to source for the firewall machines

If the Solaris binary compatibility was up to hosting Oracle and the
Netscape WWW servers[*], we'd probably have already gone that route on
all but the 4d monster.

I know that having Linux on SGI hardware would make it much easier to
`sell' it into our environment, which includes solutions we deliver to
clients.  (And I don't get to make all that many purchasing
decisions. =) )

[*] At the DevCon last week, the question of support for Linux was
raised at the executive Q&A panel.  There was a show of hands, and a
_lot_ of people in the audience were interested in seeing SuiteSpot on
Linux; Netscape's asked me to write them a little paper explaining why
`people who aren't willing to pay for Linux would be willing to pay
for SuiteSpot', and I'll send a copy around here when it's done.

Mike

-- 
#> Mike Shaver (shaver@ingenia.com) Ingenia Communications Corporation 
#>              Commando Developer - Whatever It Takes
#>                                                                     
#> "See, you not only have to be a good coder to create a system like
#>    Linux, you have to be a sneaky bastard too." - Linus Torvalds

Re: Getting X on Linux/SGI

[Mike Shaver]

Thus spake Larry McVoy:
> : Personally, I wouldn't buy an O2 to run Linux.
> 
> OK, we won't try and sell you one.  What about the people that would buy
> an O2 to run Linux?  
>
> You != <the entire customer base>.

FWIW, there's a _lot_ of interest at my company in having one OS that
runs on all our boxes (Intel, SP*RC, Alpha, now MIPS), and Linux is
the preferred choice.  The reasons don't necessarily have anything
to do with disliking the native OS (except for Solaris, perhaps), but
we _would_ like:
- common file structure and apps
- source compatibility
- access to source for the firewall machines

If the Solaris binary compatibility was up to hosting Oracle and the
Netscape WWW servers[*], we'd probably have already gone that route on
all but the 4d monster.

I know that having Linux on SGI hardware would make it much easier to
`sell' it into our environment, which includes solutions we deliver to
clients.  (And I don't get to make all that many purchasing
decisions. =) )

[*] At the DevCon last week, the question of support for Linux was
raised at the executive Q&A panel.  There was a show of hands, and a
_lot_ of people in the audience were interested in seeing SuiteSpot on
Linux; Netscape's asked me to write them a little paper explaining why
`people who aren't willing to pay for Linux would be willing to pay
for SuiteSpot', and I'll send a copy around here when it's done.

Mike

-- 
#> Mike Shaver (shaver@ingenia.com) Ingenia Communications Corporation 
#>              Commando Developer - Whatever It Takes
#>                                                                     
#> "See, you not only have to be a good coder to create a system like
#>    Linux, you have to be a sneaky bastard too." - Linus Torvalds

Re: Getting X on Linux/SGI

[richard offer]

* $ from lm@neteng at "20-Jun: 9:40am" | sed "1,$s/^/* /"
*
*
* : Personally, I wouldn't buy an O2 to run Linux.
*
* OK, we won't try and sell you one.  What about the people that would buy
* an O2 to run Linux?
*
* You != <the entire customer base>.
*

This is once again turning into an advocacy war.

We all at some point want to see the same thing, Linux on SGI. Some people are
raging Linux advocates that think its the best thing since sliced bread and
that all other OSs are inherently a Bad Thing[TM], others think it would just
be neat/interesting thing to do. Others may even think that it could possible
damage our competative advantage if we make too much of our hardware details
public.

Unfortunetly, the people that maybe able to help us often fall into the last
category.

If you want a favour from someone you don't start my telling them that what
they've worked on for the last x years is dead and your way is the future.



richard---pragmatist.




_________________________________________________________________________

A Guest Signature by Michael Berube <dxc4 at po.cwru.edu>

    "Yes, we plot no less than the destruction of the West. Just the
    other day a friend and I came up with the most pernicious academic
    scheme to date for toppling the West: He will kneel behind the West
    on all fours. I will push it backwards over him."

Re: Getting X on Linux/SGI

[Todd Shrider]

On Fri, 20 Jun 1997, Larry McVoy wrote:
> Have you forgotten Nintendo 64 sales?

I'm not sure as of what date, probably 3/31/97:

   N64 2.6 Million units (US)
   Super NES 40 million
   Gameboy 45 million
   Original NES 62 million

Projected sales for the 12 months starting 4/1/97: 12 million N64
system units.

I'd say that if the 64 was extended (with linux) to a set-top-box type
application that this could be more like 18-20 million ($$$). If it's true
that there are about 3 million linux users (estimates taken from the 97
ALE conference) how many of these people do you think would spend $200 for
another xterm? I bet 1 million at least. So that's 19-21 million estimated
sales for the next 12 months alone! With the expandibility of the N64 and
the Linux OS I could see the 64 sky-rocketing past the original NES. I can
just hear it know "But mom, I need a nintendo 64 with linux so I can get
on the schools web page to get homework and stuff finished." Now the
parents don't have anymore excuses not to buy a system for the kid. (Why
do I feel like to tabaco industry now?) 


> MIPS Chips == $$$ for SGI.  $$$ for SGI is what it is all about, that's what
> feeds the light bills.   If making a bunch of hackers happy turns into $$$
> for SGI in the long term, then we do it.  At this point, I believe it is 
> obvious that that is the case.
> 

---
Todd M. Shrider         Oops, My brain just hit a bad sector!
todds@ontko.com         http://www.ontko.com/~todds

Re: Getting X on Linux/SGI

[Christopher W. Carlson]

On Jun 20, 12:21am, Steve Alexander wrote:
> Subject: Re: Getting X on Linux/SGI
> "Christopher W. Carlson" <carlson@heaven.newport.sgi.com> writes:
> >My point was that SGI is into making hardware.  That's how it started
> >and that is still its main focus.  Yes, it is true that our software
> >is very important because it takes advantage of the proprietary stuff
> >in the hardware, but the main income for SGI is hardware.
>
> How many people buy the hardware and program the bare metal?
>
> >Note in any
> >press releases and stock holder's documents how much is said about the
> >wonderful software SGI produces (Alias/Wavefront is still a separate
> >entity).  You won't find much.
>
> That's a separate issue, which is that very few people here value software.
> That will probably change with time if the company is going to continue to
> be successful as the industry shifts.
>
> -- Steve
>-- End of excerpt from Steve Alexander


Ah!  That's my point.  As a company, the hardware is valued but the
software isn't.  Just because we *give* software with our hardware
doesn't mean that we are a software oriented company.  We are
definitely hardware oriented.

I've been arguing the point that software is as much of value as
hardware for years.  Look at Bill Gates!  He didn't make it by selling
hardware.  Unfortunately, I've finally succumbed to the realization
that SGI is a hardware company first.

So, if one wants to make any points to upper management, it has to be
in terms of "how much hardware are we going to sell?"  If developing a
really powerful *free* operating system, such as Linux, will sell lots
of boxes, you can bet upper management will be interested.

-- 

		Chris Carlson

	+------------------------------------------------------+
	| Also, carlson@sgi.com                                |
	|   Work:   (714) 224-4530                             |
	|   Vnet:       6-678-4530     FAX:    (714) 833-9503  |
	|                                                      |
	| Trivia fact: an electroencephalogram shows that a    |
	| human brain and a bowl of quivering lime Jell-O have |
	| the same waves.  [Time Magazine, Mar 17, 1997]       |
	+------------------------------------------------------+

Re: Getting X on Linux/SGI

[Nigel Gamble]

On Fri, 20 Jun 1997, David S. Miller wrote:
> He was given QNX to write his drivers and get the box going.  Since it
> is a micro-kernel, you have to perform a task switch to handle even an
> interrupt, and you have to compile your interrupt handlers with a
> special compiler and compiler options using QNX compilers just to cope
> with this bullshit.  This was on both systems.
> 
> Interrupt response was so slow, that even when he coded the drivers in
> raw optimized assembly he could not keep up and would drop characters
> easily on his serial ports, the ISDN performance sucked balls as well.
> 
> So eddie got so frustrated one night that he took both the m68k and
> Intel ports of Linux, in about an hour added kernel build time
> configuration options such as "CONFIG_NO_MEMORY_MANAGEMENT",
> "CONFIG_NO_FANCY_SYSCALLS", "CONFIG_NO_USELESS_FEATURES" and the like
> to the point where he was able to get a 120k sized Linux kernel with
> his drivers and the specialized code to run the control systems he
> needed to deploy, and he got full over the serial line KGDB source
> level debugging of his kernel as well.
> 
> The next night he got it completely working and debugged, needless to
> say this thing didn't have the interrupt performance problems QNX
> did.  The next evening he blew the first revisions of the PROM's the
> boxes would eventually use in production when these things got sent
> to the customers.

This tells me more about this guy's determination to use Linux, come
what may, than it does about QNX.  If he'd put half the effort into
learning the QNX device driver model that he did into hacking Linux,
I bet he could have solved his problem with QNX.  I certainly could.

When I first implemented a version of the Linux parallel port
printer driver that used interrupts (because the polling driver was
only printing one line every 30 seconds on my old dot-matrix printer),
I discovered that my driver couldn't send characters fast enough
to keep up with a laser printer.  Does this imply that Linux's
(non-threaded) interrupt performance sucked?  No, it just meant that
my naive first attempt was taking an interrupt for every character.


Nigel Gamble       "Are we going to push the edge of the envelope, Brain?"
Silicon Graphics   "No, Pinky, but we may get to the sticky part."
nigel@sgi.com
(415) 933-3109

Re: Getting X on Linux/SGI

[Steve Alexander]

lm@neteng (Larry McVoy) writes:
>You != <the entire customer base>.

Really?  Thanks for the tip.

-- Steve

Re: Getting X on Linux/SGI

[Steve Alexander]

lm@neteng (Larry McVoy) writes:
>You != <the entire customer base>.

Really?  Thanks for the tip.

-- Steve

Re: Getting X on Linux/SGI

[Larry McVoy]

: This tells me more about this guy's determination to use Linux, come
: what may, than it does about QNX.  If he'd put half the effort into
: learning the QNX device driver model that he did into hacking Linux,
: I bet he could have solved his problem with QNX.  I certainly could.

Why are you so sure that you could have made it work?  I know Eddie,
he's pretty bright and he isn't the sort of guy to use a hammer where
you need a screwdriver, as you suggest.  Even disregarding his approach,
why are you so sure you could have made it work with QNX?  Have you solved
that problem before in QNX?  Have you developed similar systems with
QNX?  If so, was it back before QNX had VM so it was quite lightweight?
Have you revisted the issues recently?  It sounds cocky of you to suggest
that you could trivially solved the problem but I'm sure you'll be able
to back it up with all the tecnhical details.

As to your comments about the line printer driver, yes, we've heard that
story before.  It's great that you figured out how to do it right.

However, what bearing does any of this have on our topic?  The original
topic, unless I'm mistaken, was questioning whether there was a reasonable
market for Linux on embedded MIPS.  While QNX is a great system, it
doesn't run on MIPS and is therefor pretty uninteresting to us.  While 
your experiences with the line printer are cool, does the fact that you
had a learning experience with Linux driver writing have any bearing on
Linux/MIPS sales?  I'm a little lost as to where you were going with all
this.

If you want to spend time and energy here, it would be nice if you could
help by figuring out the ways that Linux could be useful, and there are
many to choose from, on MIPS.

Re: Getting X on Linux/SGI

[Larry McVoy]

: 	If the demand for Linux on MIPS chips is comming from the "embedded"
: 	systems then things sure have changed since the early days of
: 	MIPS Computer Systems. 

I think it is safe to confirm that yes, indeeedy, the market has changed.

: 	If fact the original VP of embedded system marketing at MIPS was
: 	Chet Silvestri, now VP of SPARC embedded systems at SUN. 

I know Chet, and Chet was real interested in getting some free OS on 
embedded SPARC when I was at Sun.

: 	Hence I would like to here more about larry's definition of
: 	"embedded systems" because they sure don't sound like any
: 	that our semiconductor partners sell to.

I think almost 100% of what I would call embedded Linux on MIPS is for
game platforms.  What they really want is a full on Unix environment
that they can use for development/debugging.  When they get done,
they want to rip out Linux/Unix and run an app on the bare hardware.
They may eventually want to leave a stripped down Linux in there but I
don't see why today.

What they really want is a reasonable development environment and they are
smart enough to realize they get there fastest with Linux.

Re: Getting X on Linux/SGI

[Larry McVoy]

: 	If the demand for Linux on MIPS chips is comming from the "embedded"
: 	systems then things sure have changed since the early days of
: 	MIPS Computer Systems. 

I think it is safe to confirm that yes, indeeedy, the market has changed.

: 	If fact the original VP of embedded system marketing at MIPS was
: 	Chet Silvestri, now VP of SPARC embedded systems at SUN. 

I know Chet, and Chet was real interested in getting some free OS on 
embedded SPARC when I was at Sun.

: 	Hence I would like to here more about larry's definition of
: 	"embedded systems" because they sure don't sound like any
: 	that our semiconductor partners sell to.

I think almost 100% of what I would call embedded Linux on MIPS is for
game platforms.  What they really want is a full on Unix environment
that they can use for development/debugging.  When they get done,
they want to rip out Linux/Unix and run an app on the bare hardware.
They may eventually want to leave a stripped down Linux in there but I
don't see why today.

What they really want is a reasonable development environment and they are
smart enough to realize they get there fastest with Linux.

Re: Getting X on Linux/SGI

[Martin Knoblauch]

Christopher W. Carlson wrote:
> 
> 
> Ah!  That's my point.  As a company, the hardware is valued but the
> software isn't.  Just because we *give* software with our hardware
> doesn't mean that we are a software oriented company.  We are
> definitely hardware oriented.
>

 I have to agree with Chris (a rarely do :-). Past behaviour
of SGI has been that of a "HW oriented Systems Vendor". Which
does not mean that the SW we ship is not important (to the
contrary), but revenue (and profit) come mostly from the sales of
HW.

 Only recently we seem to have discovered that some of our SW
products make us real profit. Now, should we transform ourselves
into a software company? I don't think so. I would like to
see us as a solutions provider (HW+base-SW+Added-Value+Support).
Long way to go for us.

> I've been arguing the point that software is as much of value as
> hardware for years.  Look at Bill Gates!  He didn't make it by selling
> hardware.  Unfortunately, I've finally succumbed to the realization
> that SGI is a hardware company first.
>

 Of course, he did not make his wealth by giving away the
software :-)
 
> So, if one wants to make any points to upper management, it has
> to be in terms of "how much hardware are we going to sell?"  If
> developing a really powerful *free* operating system, such as
> Linux, will sell lots of boxes, you can bet upper management
> will be interested.
> 

  Yes, the almighty god of the Businesscase :-)

Martin
-- 
+---------------------------------+-----------------------------------+
|Martin Knoblauch                 | Silicon Graphics GmbH             |
|Manager Technical Marketing      | Am Hochacker 3 - Technopark       |
|Silicon Graphics Computer Systems| D-85630 Grasbrunn-Neukeferloh, FRG|
|---------------------------------| Phone: (+int) 89 46108-179 or -0  |
|http://reality.sgi.com/knobi     | Fax:   (+int) 89 46107-179        |
+---------------------------------+-----------------------------------+
|e-mail: <knobi@munich.sgi.com>   | VM: 6-333-8197 | M/S: IDE-3150    |
+---------------------------------------------------------------------+

Re: Getting X on Linux/SGI

[Christopher W. Carlson]

On Jun 20,  1:20pm, Steve Alexander wrote:
> Subject: Re: Getting X on Linux/SGI
> "Christopher W. Carlson" <carlson@heaven.newport.sgi.com> writes:
> >Initially, I wouldn't buy an O2 to run Linux, either.  But, after a
> >while, I might consider puting Linux on it.  If Linux was pretty well
> >tuned (maybe not as well as IRIX but good enough) and I wanted to do
> >some kernel tinkering, you can bet Linux would at least be on a
> >partition.
>
> Why not just a PC?  Cheaper, more expansion, probably better performance for
> just plain kernel hacking.
>
> Unless you're doing graphics things, there's just no point in having an O2,
> IMO.
>
> -- Steve
>-- End of excerpt from Steve Alexander


Yes, that's why I might want to run Linux on an O2.  Since it isn't
even an option yet, I'd run IRIX (if I had an O2).  Also, if there was
some bug in IRIX that kept me from doing something but the bug didn't
exist in Linux, I might want to switch.  It is far easier (and
cheaper) to get a bug fixed in Linux than in IRIX.  Especially if you
can't afford the maintenance fees.

-- 

		Chris Carlson

	+------------------------------------------------------+
	| Also, carlson@sgi.com                                |
	|   Work:   (714) 224-4530                             |
	|   Vnet:       6-678-4530     FAX:    (714) 833-9503  |
	|                                                      |
	| Trivia fact: an electroencephalogram shows that a    |
	| human brain and a bowl of quivering lime Jell-O have |
	| the same waves.  [Time Magazine, Mar 17, 1997]       |
	+------------------------------------------------------+

Re: Getting X on Linux/SGI

[Christopher W. Carlson]

On Jun 22, 10:50am, Larry McVoy wrote:
>
> I think almost 100% of what I would call embedded Linux on MIPS is for
> game platforms.  What they really want is a full on Unix environment
> that they can use for development/debugging.  When they get done,
> they want to rip out Linux/Unix and run an app on the bare hardware.
> They may eventually want to leave a stripped down Linux in there but I
> don't see why today.
>
> What they really want is a reasonable development environment and they are
> smart enough to realize they get there fastest with Linux.
>-- End of excerpt from Larry McVoy


I would think that the reason they might want to have Linux running on
some embedded system is the same as the reason Microsoft wants to run
Windows on your microwave oven/television/VCR/etc...

I'm sure it isn't "Windows", it's just the core drivers and stuff.  I
think the ultimate goal of Microsoft is to allow people to have
everything electrical in their house networked together.  Then they
could turn on the microwave by logging into their home computer and
telling it to turn on the microwave (of course that's just a
pedagogical example).

FYI, according to an article I read in New Yorker, Microsoft's goal is
to get money every time one of their products is used.  Thus, with the
above scenerio, they could get a penny because you logged into your
home computer and another penny because you enabled your microwave.

I'd prefer to have Linux as the embedded system.  Then my microwave
can run some program using X windows and display it on my home
computer's X server and Microsoft doesn't get anything!  :-)


-- 

		Chris Carlson

	+------------------------------------------------------+
	| Also, carlson@sgi.com                                |
	|   Work:   (714) 224-4530                             |
	|   Vnet:       6-678-4530     FAX:    (714) 833-9503  |
	|                                                      |
	| Trivia fact: an electroencephalogram shows that a    |
	| human brain and a bowl of quivering lime Jell-O have |
	| the same waves.  [Time Magazine, Mar 17, 1997]       |
	+------------------------------------------------------+

Re: Getting X on Linux/SGI

[Steve Alexander]

"Christopher W. Carlson" <carlson@heaven.newport.sgi.com> writes:
>It is far easier (and cheaper) to get a bug fixed in Linux than in IRIX.

No doubt.

>Especially if you can't afford the maintenance fees.

I feel compelled to point out that you no longer need to be on support to get
the recommended patches for IRIX.  That's only a partial solution, but it's a
good first step.

-- Steve