Good news: no more begging for HW

Good news: no more begging for HW

[Ariel Faigon]

[Some good news I wanted to share with all]

Bill Fisher and I are just returning from a meeting w/
Todd Johnson.  Bottom line is that he agreed to commit
for 5 machines/year (or "even 10 if you need it") for
free-software initiatives including Linux.  It is up to
us (mainly this mailing list subscribers) to find the right
people who are willing to work on SGI ports and projects and
Todd will fund the hardware we need.  Of course every system
will need to be justified based on the developer's reputation
and ability and the suggested project at hand.

So, if you know of good people who can sign-up for
real useful projects in return for hardware donations
please share it with me.  The three machines we have out
are just a start.

Looks like I won't have to spend so much time begging
for donations in the coming months.

-- 
Peace, Ariel

Re: Good news: no more begging for HW

[Christopher W. Carlson]

On Jun 16,  4:25pm, Ariel Faigon wrote:
> Subject: Good news: no more begging for HW
> [Some good news I wanted to share with all]
> 
> Bill Fisher and I are just returning from a meeting w/
> Todd Johnson.  Bottom line is that he agreed to commit
> for 5 machines/year (or "even 10 if you need it") for
> free-software initiatives including Linux.  It is up to
> us (mainly this mailing list subscribers) to find the right
> people who are willing to work on SGI ports and projects and
> Todd will fund the hardware we need.  Of course every system
> will need to be justified based on the developer's reputation
> and ability and the suggested project at hand.
> 
> So, if you know of good people who can sign-up for
> real useful projects in return for hardware donations
> please share it with me.  The three machines we have out
> are just a start.
> 
> Looks like I won't have to spend so much time begging
> for donations in the coming months.
> 
> -- 
> Peace, Ariel
>-- End of excerpt from Ariel Faigon


One group that I always think could use an SGI machine is the Free
Software Foundation.  They would make sure things worked on SGI before
releasing it.  That way, all of GNU would be tested on SGI by them.

-- 

		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: Good news: no more begging for HW

[Martin Knoblauch]

Ok,

 so let me chime in. There are the people with the KDE desktop
(http://www.kde.org/) environment. They gave a presentation at
the Linux Konference in Wuerzburg. If we want to do some non-kernel
stuff, they might be a good choice. Maybe even for IRIX :-)
I will have to contact them to find out whether they are interested.

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: Good news: no more begging for HW

[Miguel de Icaza]

Talking about userland nice applications, two extra ideas:

>  so let me chime in. There are the people with the KDE desktop
> (http://www.kde.org/) environment. They gave a presentation at
> the Linux Konference in Wuerzburg. If we want to do some non-kernel
> stuff, they might be a good choice. Maybe even for IRIX :-)

GNUstep:

There are a couple of extra free software projects that may benefit.
One of them is the GNUstep project (do not pay attention to the web
pages for the project the maintainer updates them once every six
months).  

They have similar goals to the KDE project, but they are going for an
OpenStep compliant API.  

scottc@net-community.com is the person who is doing the GUI code for
Unix. 

The GIMP:

There are a couple of students at UC Berkeley that wrote the GIMP (the
GNU Image Manipulation Program), which is a PhotoShop on steroids.

They cloned most of the functionality of Photoshop: plugins, ability
to handle large images, layers, channels.  And on top of that they
embedded a scheme interpreter, so you can create complex graphic art
by just fillling a form.  The Scheme scripts take care of the rest.

You can check their web page at: http://scam.xcf.berkeley.edu/~gimp

Miguel.

Re: Good news: no more begging for HW

[Ariel Faigon]

:
:Ariel,
:
:that's great news.  Just send all the hardware to me ;-)
:
:Jokes aside, I was thinking since some time if the FSF wouldn't be a
:great place to install some machines.  Another suggestion would be
:Thomas Bogendoerfer who already wrote a network driver for the Mips
:Magnum 4000 and others for Intel/Alpha.  I guess he might also be
:interested.
:
If you know interested people in the FSF (real names)
please have them email me with a short justification
why they want the machines and what will they use them for.
Same for Thomas.

I cannot promise anything since there may be some
oversubscription to the service :-)

I think the fairest way would be to publish all these
requests on this forum and have the people who care (us)
vote on who should get them.  I certainly don't want to
be the fascist person who decides who gets what.


-- 
Peace, Ariel

Re: Good news: no more begging for HW

[Ariel Faigon]

:
:Ariel,
:
:that's great news.  Just send all the hardware to me ;-)
:
:Jokes aside, I was thinking since some time if the FSF wouldn't be a
:great place to install some machines.  Another suggestion would be
:Thomas Bogendoerfer who already wrote a network driver for the Mips
:Magnum 4000 and others for Intel/Alpha.  I guess he might also be
:interested.
:
If you know interested people in the FSF (real names)
please have them email me with a short justification
why they want the machines and what will they use them for.
Same for Thomas.

I cannot promise anything since there may be some
oversubscription to the service :-)

I think the fairest way would be to publish all these
requests on this forum and have the people who care (us)
vote on who should get them.  I certainly don't want to
be the fascist person who decides who gets what.


-- 
Peace, Ariel

Re: Good news: no more begging for HW

[Alex deVries]

Ariel wrote:
> I think the fairest way would be to publish all these
> requests on this forum and have the people who care (us)
> vote on who should get them.  I certainly don't want to
> be the fascist person who decides who gets what.

Well... I hate to be the first to jump in pleading for hardware, but I
could really use it. To date, the only SGI I have access to is
bogomips.ingenia.ca, which is Mike Shaver's machine in Ottawa. This is
difficult, as I can't do anything hardware related, or anything involving
rebooting, etc. Plus, bandwidth is a frequent problem.

I've always been interested in working with the application side of things
(particularly, porting all of RedHat 4.2), but I'd like to get involved in
other things, such as an X server and filesystem support.

Despite my email address, I actually reside in Boston.  I just administer
engsoc.carleton.ca in Ottawa remotely in my spare time (and quite
creatively, too, might I add). 

Oh, and my vote is on getting Ralf an SGI, if he doesn't have one already.
I'm not quite sure what the future is like for bogomips.ingenia.com.

- Alex

      Alex deVries           "Alex can cut a mean rug."
  System Administrator       - M. Dittberner <shabby@engsoc.carleton.ca>
   The EngSoc Project     

Re: Good news: no more begging for HW

[David S. Miller]

   From: ariel@yon.engr.sgi.com (Ariel Faigon)
   Date: Mon, 16 Jun 1997 18:06:41 -0700 (PDT)

   I cannot promise anything since there may be some oversubscription
   to the service :-)

   I think the fairest way would be to publish all these requests on
   this forum and have the people who care (us) vote on who should get
   them.  I certainly don't want to be the fascist person who decides
   who gets what.

Before this gets out of control, I just want to express one sentiment
of high caution.

Although it may seem desirable to contribute most of the donation
hardware to kernel level hackers, this can be a mistake in the making.
At this stage in the game it is just as important to get userland/libc
developers machines.

Therefore I suggest that at least one person who knows GNU libc,
binutils, _and_ gcc internals backwards and forwards be on the top of
the donation list.  If I were asked for such a candidate, I would
recommend Richard Henderson (rth@stommel.tamu.edu) He has done the
Alpha/Linux port, he designed an ELF standard for 64-bit Alpha from
scratch with no existing standard available, he is doing the same
exact thing for 64-bit SparcLinux at the moment as well.

Not having a good libc/userland person in this port is why I lost half
my summer last year and was not able to hack the kernel as much as I
really would have liked to at all...

---------------------------------------------////
Yow! 11.26 MB/s remote host TCP bandwidth & ////
199 usec remote TCP latency over 100Mb/s   ////
ethernet.  Beat that!                     ////
-----------------------------------------////__________  o
David S. Miller, davem@caip.rutgers.edu /_____________/ / // /_/ ><

Re: Good news: no more begging for HW

[richard offer]

* $ from davem@jenolan.rutgers.edu at "16-Jun:10:01pm" | sed "1,$s/^/* /"
*
*
*    From: ariel@yon.engr.sgi.com (Ariel Faigon)
*    Date: Mon, 16 Jun 1997 18:06:41 -0700 (PDT)
*
*    I cannot promise anything since there may be some oversubscription
*    to the service :-)
*
*    I think the fairest way would be to publish all these requests on
*    this forum and have the people who care (us) vote on who should get
*    them.  I certainly don't want to be the fascist person who decides
*    who gets what.
*
* Before this gets out of control, I just want to express one sentiment
* of high caution.
*
* Although it may seem desirable to contribute most of the donation
* hardware to kernel level hackers, this can be a mistake in the making.
* At this stage in the game it is just as important to get userland/libc
* developers machines.
*
* Therefore I suggest that at least one person who knows GNU libc,
* binutils, _and_ gcc internals backwards and forwards be on the top of
* the donation list.

We need an X server if we are ever going to get it usable by real users---or is
everyone assuming its for headles machines only. I would put an X server above
most applications in terms of priority (just below native gcc/libc).

Perhaps one ought to go to Xfree86/someone who knows our hardware (not to start
this thread all over again).

This should be very tighly focused, since an X server is a lot of work and we
need a lot of them, one per board (okay in version one we could only support
1280x1024x8, but that seems a waste of all our spiffy hardware).

I used to have a contact in Xfree86, but I haven't heard from him for a while.

If someone takes the server, I'll try and get the clients libraries done
(assuming that I can get remote access to a box).

richard.

____________________________________________________________________

A Guest Signature from Laurent Duperval <Laurent at Grafnetix.com>

  I don't understand why people break up and then get back together.
  It's like going to the fridge, taking a carton of milk that has
  gone bad, then saying: "I'll put it back and see if it's better
  tomorrow."

Re: Good news: no more begging for HW

[Miguel de Icaza]

> If someone takes the server, I'll try and get the clients libraries done
> (assuming that I can get remote access to a box).

I was interested in working on the X server for the SGI.  I already
did that for the Linux/SPARC, and I had a couple of questions to make,
so this seems like a good time to ask them (please note that I haven't
actually traced my SGI X server to see what it does).

1. On the SPARC, the X server mmap()s the frame buffer into its
   address space and uses a couple of ioctls to talk with the kernel
   (to ask the kernel to change the palette and the hardware cursor,
   on later versions, with got rid of that, and we just poked at the
   frame buffer control registers from the X server).

   How does this work on the SGI?  Is the video card just a thing that
   can be mapped into the X server address space?  

   If this is the case, getting the X11R6 server to work will just take
   a couple of days of coding.
   

2. What kind of acceleration features are available on the SGI
   machines?  The X11R6 server has hooks for different set of
   features, so for example, bitblit can be easily hacked into the X
   server. 

   But I imagine the SGI has more acceleration features that I can
   dream of.


3. How does OpenGL work on the SGIs?  Is the OpenGL engine embedded in
   the X server, or it is something that is present on the video card?
   
   I looked yesterday at a program called glxinfo, which led me to
   believe that applications may have some of the GL code linked in
   trough the libraries and the other part resides on the X server.

   So, in this case, what are the specs for what needs to be on the X
   server to be able to run OpenGL applications.


4. Would it be possible for a free software company to redistribute
   the SGI's X server?  In that case, we could concentrate on getting
   the IRIX emulation as good as possible and just use the SGI X
   server and let Red Hat/Debian/GNU ship the cd with that binary.


Cheers,
Miguel.

Re: Good news: no more begging for HW

[Mike Shaver]

Thus spake Miguel de Icaza:
> 4. Would it be possible for a free software company to redistribute
>    the SGI's X server?  In that case, we could concentrate on getting
>    the IRIX emulation as good as possible and just use the SGI X
>    server and let Red Hat/Debian/GNU ship the cd with that binary.

FWIW, I'm quite interested in working on the IRIX emulation.
I picked up a bunch of stuff about the MIPS ABI at the SGI DevCon
(after arriving late for Ariel's talk =(), which also looks
interesting.

Mike

-- 
#> Mike Shaver (shaver@ingenia.com) Ingenia Communications Corporation 
#>                 UNIX medicine man -- dark magick, cheap!            
#>                                                                     
#>  When the going gets tough, the tough give cryptic error messages.  
#>          "We believe in rough consensus and running code."          

Re: Good news: no more begging for HW

[Mike Shaver]

Thus spake Miguel de Icaza:
> 4. Would it be possible for a free software company to redistribute
>    the SGI's X server?  In that case, we could concentrate on getting
>    the IRIX emulation as good as possible and just use the SGI X
>    server and let Red Hat/Debian/GNU ship the cd with that binary.

FWIW, I'm quite interested in working on the IRIX emulation.
I picked up a bunch of stuff about the MIPS ABI at the SGI DevCon
(after arriving late for Ariel's talk =(), which also looks
interesting.

Mike

-- 
#> Mike Shaver (shaver@ingenia.com) Ingenia Communications Corporation 
#>                 UNIX medicine man -- dark magick, cheap!            
#>                                                                     
#>  When the going gets tough, the tough give cryptic error messages.  
#>          "We believe in rough consensus and running code."          

Re: Good news: no more begging for HW

[richard offer]

* $ from miguel@nuclecu.unam.mx at "17-Jun:12:10pm" | sed "1,$s/^/* /"
*
*
*
* > If someone takes the server, I'll try and get the clients libraries done
* > (assuming that I can get remote access to a box).
*
* I was interested in working on the X server for the SGI.  I already
* did that for the Linux/SPARC, and I had a couple of questions to make,
* so this seems like a good time to ask them (please note that I haven't
* actually traced my SGI X server to see what it does).

The questions are good, but I'm not the person who can answer them, me I'm a
Motif hacker, anything at a level lower than Xt is akin to assembler for me.

So with that in mind take the rest with a pinch of salt.

*
* 1. On the SPARC, the X server mmap()s the frame buffer into its
*    address space and uses a couple of ioctls to talk with the kernel
*    (to ask the kernel to change the palette and the hardware cursor,
*    on later versions, with got rid of that, and we just poked at the
*    frame buffer control registers from the X server).
*
*    How does this work on the SGI?  Is the video card just a thing that
*    can be mapped into the X server address space?
*
*    If this is the case, getting the X11R6 server to work will just take
*    a couple of days of coding.
*
*
* 2. What kind of acceleration features are available on the SGI
*    machines?  The X11R6 server has hooks for different set of
*    features, so for example, bitblit can be easily hacked into the X
*    server.
*
*    But I imagine the SGI has more acceleration features that I can
*    dream of.

The problem is that (I think) we have so many graphics cards that its done
differently in every one (some cards are simple frame buffers (8/24bit), then
there are some with multiple GE, oh and we also have extra visuals for overlay
and pop-ups.

The O2 has no graphics memory, everything is done in main memory.

*
*
* 3. How does OpenGL work on the SGIs?  Is the OpenGL engine embedded in
*    the X server, or it is something that is present on the video card?
*
*    I looked yesterday at a program called glxinfo, which led me to
*    believe that applications may have some of the GL code linked in
*    trough the libraries and the other part resides on the X server.

Both :-) We don't do things the way easy here.

The is a GLX extension in the X server which allows GL to run in an X window.

There is also I think a DSO that holds the hardware specific GL calls.

*
*    So, in this case, what are the specs for what needs to be on the X
*    server to be able to run OpenGL applications.
*
*
* 4. Would it be possible for a free software company to redistribute
*    the SGI's X server?  In that case, we could concentrate on getting
*    the IRIX emulation as good as possible and just use the SGI X
*    server and let Red Hat/Debian/GNU ship the cd with that binary.

This would be my preferred solution, but I've had many an argument on this
subject that I felt very dubious about bringing it up again.

To me the quickest (and the best) way of getting an X server would be if we
could simply port the existing Irix X server to Linux/SGI. My suggestion would
be, now that we have backing for hardware to get official backing for software.
I don't think we should neccesarrily release the source code for the ddx part
of the X server to the public, but we should at least be able to get backing to
release .o files so the user could re-link the X server if they needed to (Sun
have done this before).

*
*
* Cheers,
* Miguel.
*
*


richard


-----------------------------------------------------------------------
Richard M. Offer        Widget FAQ --> http://reality.sgi.com/widgetFAQ
MTS-Core Design (Motif)
___________________________________________http://reality.sgi.com/offer

Re: Good news: no more begging for HW

[Miguel de Icaza]

> * 2. What kind of acceleration features are available on the SGI
> *    machines?  The X11R6 server has hooks for different set of
> *    features, so for example, bitblit can be easily hacked into the X
> *    server.
> *
> *    But I imagine the SGI has more acceleration features that I can
> *    dream of.
> 
> The problem is that (I think) we have so many graphics cards that its done
> differently in every one (some cards are simple frame buffers (8/24bit), then
> there are some with multiple GE, oh and we also have extra visuals for overlay
> and pop-ups.

What does "multiple GE" stand for?  

Supporting a wide variety of devices in X11R6 should be quite easy.
This X server also can support multiple visuals on a display, so this
should be easy to hack on as well.

> *    I looked yesterday at a program called glxinfo, which led me to
> *    believe that applications may have some of the GL code linked in
> *    trough the libraries and the other part resides on the X server.
> 
> Both :-) We don't do things the way easy here.

So OpenGL applications can run without an X server, or they have code
to bypass the X server if they need to?

> The is a GLX extension in the X server which allows GL to run in an X window.
> 
> There is also I think a DSO that holds the hardware specific GL calls.

Sorry, but what does DSO stand for?

> To me the quickest (and the best) way of getting an X server would be if we
> could simply port the existing Irix X server to Linux/SGI. My suggestion would
> be, now that we have backing for hardware to get official backing for software.
> I don't think we should neccesarrily release the source code for the ddx part
> of the X server to the public, but we should at least be able to get backing to
> release .o files so the user could re-link the X server if they needed to (Sun
> have done this before).

Ok.  This sound good enough.  

Miguel.

Re: Good news: no more begging for HW

[Martin Knoblauch]

Miguel de Icaza wrote:
> 
> 
> What does "multiple GE" stand for?
> 

 GE == Geometry engine. The part of the OpenGL pipeline that
does the 3D transformations, lighting and other FPU intensive
stuff.

 Some of our GFX "card" do this stuff on the CPU (your Indy,
O2, some Indigo2), but most adaptors have separate GEs. In
all cases (except O2), you have no direct mapping from virtual
memory into the frame buffer.

> 
> So OpenGL applications can run without an X server, or they have code
> to bypass the X server if they need to?
>

 The apps use the Xserver to create the windows and do the
window and event managment. On fast, HW accellerated adapters,
they bypass the server when drawing. But they also can draw through
the server (as in the remote display case).

> 
> Sorry, but what does DSO stand for?
> 

 Dynamic Shared Object. Basically our term for Dynamic Shared
Library.

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: Good news: no more begging for HW

[David S. Miller]

   Date: Tue, 17 Jun 1997 20:30:36 +0200
   From: Martin Knoblauch <knobi@munich.sgi.com>

    GE == Geometry engine. The part of the OpenGL pipeline that
   does the 3D transformations, lighting and other FPU intensive
   stuff.

How fast can these suckers assemble components along an interpolated
line?  Is it something like 1 component per clock at 300Mhz?

(some of you might know where that rate comes from, if you do,
 two points for you ;-)

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

Re: Good news: no more begging for HW

[John Wiederhirn]

On Jun 17, 10:53am, richard offer wrote:
> * $ from miguel@nuclecu.unam.mx at "17-Jun:12:10pm" | sed "1,$s/^/* /"
> *
> * 4. Would it be possible for a free software company to redistribute
> *    the SGI's X server?  In that case, we could concentrate on getting
> *    the IRIX emulation as good as possible and just use the SGI X
> *    server and let Red Hat/Debian/GNU ship the cd with that binary.
>
> This would be my preferred solution, but I've had many an argument on this
> subject that I felt very dubious about bringing it up again.
>
> To me the quickest (and the best) way of getting an X server would be if we
> could simply port the existing Irix X server to Linux/SGI. My suggestion
would
> be, now that we have backing for hardware to get official backing for
software.
> I don't think we should neccesarrily release the source code for the ddx part
> of the X server to the public, but we should at least be able to get backing
to
> release .o files so the user could re-link the X server if they needed to
(Sun
> have done this before).

While this appears to be an ideal solution on the surface, it has some obvious
and immediate problems as well.  Namely, the complete lack of the driver
infrastructure to support the device-dependant layer of the Xsgi server.

Given that it's unlikely we'd release the source code to our gfx drivers, there
is no easily viable way to produce the drivers in the linux kernel image.
 Lacking the drivers, Xsgi would need some _major_ redesigns, which brings us
back to Xfree-porting-level efforts needing to be expended.

There are some very serious issues which come up even getting Xfree to a
moderate level of acceleration.  By the time you've got all the pieces in
place, you are probably looking at the same device-independant ->
device-dependant -> gfx-device layering that Xsgi has in place.  OpenGL adds an
order of magnitude of complexity to the issue.

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 Wiederhirn (DSD, Graphics Kernel MTS)        jwiede@engr.sgi.com
       "Smithers, unleash the human insight and creativity."

Re: Good news: no more begging for HW

[richard offer]

* $ from jwiede@blammo at "17-Jun: 5:00pm" | sed "1,$s/^/* /"
*
*
* 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.).

I don't think this is being negitive, it might not be what we want to hear, but
at least someone who knows about our underlying hardware is making comments.

Probably the first thing to do would be to get the go ahead for a proof of
concept only (under NDA) probably using an old indy. If this proves impossible
from either the technical or political pov's then we need to re-think.

I personaly don't think we have much chance unless we go to NDA and even then
its going to be really hard to get buy-in.

*
*
* --
* John Wiederhirn (DSD, Graphics Kernel MTS)        jwiede@engr.sgi.com
*        "Smithers, unleash the human insight and creativity."
*

richard.

________________________________________________________________________

  Dogbert:    "From now on I will not try to reason with the idiots I 
              encounter. I will dismiss them by waving my paw and saying 
              'BAH'".

  Dilbert:    "Just because someone thinks differently from you doesn't 
              mean he's an idiot, Dogbert".

  Dogbert:    [waves paw], "BAH".

                                                   Dilbert: 27-July-96.

Re: Good news: no more begging for HW

[Ralf Baechle]

> Probably the first thing to do would be to get the go ahead for a proof of
> concept only (under NDA) probably using an old indy. If this proves impossible
> from either the technical or political pov's then we need to re-think.
> 
> I personaly don't think we have much chance unless we go to NDA and even then
> its going to be really hard to get buy-in.

I guess a workable model for the current technology like O2 etc. would be
that SGI gives documentation out under NDA to the developer(s).  After
a certain time (I think about 18 month or two years) after which a product
generation is obsoleted by successors the source and hopefully even the
documentation may be published freely.

  Ralf

Re: Good news: no more begging for HW

[Ralf Baechle]

> Probably the first thing to do would be to get the go ahead for a proof of
> concept only (under NDA) probably using an old indy. If this proves impossible
> from either the technical or political pov's then we need to re-think.
> 
> I personaly don't think we have much chance unless we go to NDA and even then
> its going to be really hard to get buy-in.

I guess a workable model for the current technology like O2 etc. would be
that SGI gives documentation out under NDA to the developer(s).  After
a certain time (I think about 18 month or two years) after which a product
generation is obsoleted by successors the source and hopefully even the
documentation may be published freely.

  Ralf

Re: Good news: no more begging for HW

[William J. Earl]

Ralf Baechle writes:
 > > Probably the first thing to do would be to get the go ahead for a proof of
 > > concept only (under NDA) probably using an old indy. If this proves impossible
 > > from either the technical or political pov's then we need to re-think.
 > > 
 > > I personaly don't think we have much chance unless we go to NDA and even then
 > > its going to be really hard to get buy-in.
 > 
 > I guess a workable model for the current technology like O2 etc. would be
 > that SGI gives documentation out under NDA to the developer(s).  After
 > a certain time (I think about 18 month or two years) after which a product
 > generation is obsoleted by successors the source and hopefully even the
 > documentation may be published freely.
...

    O2 is actually simpler.  We would have to document how to set up
the display engine (which copies the frame buffer to the screen), but,
since O2 is UMA, a purely software X rendering engine will work just
fine.  I doubt there will be much concern about X, and the software
OpenGL reference implementation is widely available.  I believe that
most internal concerns are about the optimized hardware-dependent GL
implementations, where software advances are a competitive issue.
(For example, we just released an updated set of GL libraries for O2
which increased performance by 30% on the same hardware.)  In that
case, since the graphics pipeline is conceptually the same on all our
platforms, the optimized algorithms do not become obsolete, just
because a particular product is obsolete.

    I expect that there is room for a reasonable compromise here, without
having to get into the complications of NDAs and the like, given that
the initial goal, at least, is to get a simple X port done.

Re: Good news: no more begging for HW

[Miguel de Icaza]

> While this appears to be an ideal solution on the surface, it has some obvious
> and immediate problems as well.  Namely, the complete lack of the driver
> infrastructure to support the device-dependant layer of the Xsgi server.

I personally would like to provide the same interface to the userland
independently of what X server we end up using (the IRIX X server or
the ported X11R6 server).  

And from the rest of your mail, it seems like the easier approach will
be to run the stock IRIX Xsgi server on Linux. 

I have been doing my homework, and have a list of device drivers that
are used by the X server.  Most of the devices are trivial to code
(keyboard, mouse, input, semaphore drver), some are more interesting (the shmiq will be
an interesting driver, since it seems the only user of this driver is
the X server, and it is nowhere documented in the man pages) and
finally the hard device driver to write is the /dev/opengl driver.

Some of the ioctls that are performed on the opengl should be trivial
to implemnt. 

There is particularly one interesting ioctl: the GFX_ATTACH_BOARD
which appears to take a (struct gfx_attach_board_args *).  This
structure is:

struct gfx_attach_board_args {
        unsigned int board;
        void        *vaddr;     /* this is a user space address */
};

On my machine, the ioctl on /dev/opengl is being called with vaddr set
to 0x02000000.  I wonder what exactly is being done at this address
space?    I know it does not do any mmap on this address (my test
program showed me this).  Probably I need to have allocated this
memory before hand?

Anyways, just after the X server calls this ioctl, it start calling a
bunch of ioctl, for which I could not figure out much:

0x530c, 0x520f, 0x520e, 0x5401, 0x5302, 0x5303,
0x5208, 0x5308, 0x5208, 0x5203, 0x5401, 0x5203

There are no ioctls in /usr/include that would make any sense for
these.  No IO*_ 'S' nor 'T' are documented there.

Put personally, I do not believe codign the /dev/opengl device will be
very hard either.  

Now, back to disassembling /unix opengl_ioctl and try to figure out
those ioctls :-)

Cheers,
Miguel.

Re: Good news: no more begging for HW

[William J. Earl]

John Wiederhirn writes:
 > On Jun 17, 10:53am, richard offer wrote:
 > > * $ from miguel@nuclecu.unam.mx at "17-Jun:12:10pm" | sed "1,$s/^/* /"
 > > *
 > > * 4. Would it be possible for a free software company to redistribute
 > > *    the SGI's X server?  In that case, we could concentrate on getting
 > > *    the IRIX emulation as good as possible and just use the SGI X
 > > *    server and let Red Hat/Debian/GNU ship the cd with that binary.
 > >
 > > This would be my preferred solution, but I've had many an argument on this
 > > subject that I felt very dubious about bringing it up again.
...
     I talked with David Miller about this area last year.  I suspect that
one can run any of the cards with the firmware loaded by the PROM, in "dumb"
mode.  One would basically render by hand, or with very limited acceleration
and just DMA the data into the frame buffer (since the frame buffer is generally
not memory-mapped).  The result would not be really fast, but it would be functional.
I am travelling today, but I will check tomorrow.

Re: Good news: no more begging for HW

[William J. Earl]

John Wiederhirn writes:
 > On Jun 17, 10:53am, richard offer wrote:
 > > * $ from miguel@nuclecu.unam.mx at "17-Jun:12:10pm" | sed "1,$s/^/* /"
 > > *
 > > * 4. Would it be possible for a free software company to redistribute
 > > *    the SGI's X server?  In that case, we could concentrate on getting
 > > *    the IRIX emulation as good as possible and just use the SGI X
 > > *    server and let Red Hat/Debian/GNU ship the cd with that binary.
 > >
 > > This would be my preferred solution, but I've had many an argument on this
 > > subject that I felt very dubious about bringing it up again.
...
     I talked with David Miller about this area last year.  I suspect that
one can run any of the cards with the firmware loaded by the PROM, in "dumb"
mode.  One would basically render by hand, or with very limited acceleration
and just DMA the data into the frame buffer (since the frame buffer is generally
not memory-mapped).  The result would not be really fast, but it would be functional.
I am travelling today, but I will check tomorrow.

Re: Good news: no more begging for HW

[Larry McVoy]

: I think the fairest way would be to publish all these
: requests on this forum and have the people who care (us)
: vote on who should get them.  I certainly don't want to
: be the fascist person who decides who gets what.

A nice facist is a good thing.  Voting assumes that everyone has enough info
which may or may not be the case.  I'd suggest that each person that wants
a machine fills out a little form that says

	who am i?

	what will i do w/ the machine?

	when will it be done?

--lm

Re: Good news: no more begging for HW

[Larry McVoy]

: I think the fairest way would be to publish all these
: requests on this forum and have the people who care (us)
: vote on who should get them.  I certainly don't want to
: be the fascist person who decides who gets what.

A nice facist is a good thing.  Voting assumes that everyone has enough info
which may or may not be the case.  I'd suggest that each person that wants
a machine fills out a little form that says

	who am i?

	what will i do w/ the machine?

	when will it be done?

--lm

Re: Good news: no more begging for HW

[Alex deVries]

Although I've posted before, I'll fill out the 'form' for reference.
Excuse the verbosity.

On Mon, 16 Jun 1997, Larry McVoy wrote:
> 	who am i?

Professionally, I'm a software engineer for IAC in Boston, MA, where I do
work with large databases.  I've also worked with TimeStep in Ottawa doing
VPN's and IPSEC, and NorTel/BNR doing very quirkly large scale revision
control tools. Lots of Unix and IP experience in this.

But, Linux experience comes from a project called EngSoc, which I started
and maintain almost solely at Carleton University in Ottawa two years ago. 
It provides full Linux access for 1,500 users on 20 machines in under
$8,000US annually. We have done a lot of hardware recovery, and, er,
creative implementations.

I've donated 30 hours a week maintaining and expanding the system for the
last two years, but it's time to move on to something newer, like
SGI-Linux. 

I have a lot of varied experience with Unices (Solaris, SunOS, Linux,
Net/FreeBSD, HPUX, OSx (oooh... SVR3 and BSD, in the same OS), Irix, etc)
and VMS, and the porting that goes along with it. 

So, clearly, I don't have nearly the experience that some people on the
list do with low level kernel authority (eg. Ralf, who must dream in MIPS 
assembler), but I am committed to this project and have been since I first
heard of it.

> 	what will i do w/ the machine?

Mostly, concentrate on userspace applications, although certainly I will
help test and debug kernels, gcc and libc. I'd like to concentrate on
porting all RPM's I can get my hands on, with a concentration on RedHat
4.2.

I'm also interested in getting an X server running smoothly on it, as well
as getting native file system support running well.

So far, my work has been on Mike's bogomips.ingenia.ca, but I can't do
much more without having physical access.

> 	when will it be done?

It depends. I suspect most of the common userland ports can be done over
three months, but there'll be subsequent releases, updates, etc.  I will
be happy when RedHat releases an official Linux/SGI CD.

- Alex

Re: Good news: no more begging for HW

[Larry McVoy]

: One group that I always think could use an SGI machine is the Free
: Software Foundation.  They would make sure things worked on SGI before
: releasing it.  That way, all of GNU would be tested on SGI by them.

This is a good idea.  Do we know if they don't have one?

Re: Good news: no more begging for HW

[Ralf Baechle]

> : One group that I always think could use an SGI machine is the Free
> : Software Foundation.  They would make sure things worked on SGI before
> : releasing it.  That way, all of GNU would be tested on SGI by them.
> 
> This is a good idea.  Do we know if they don't have one?

Somewhen RMS mentioned that they don't have MIPS machines.  Aside of PCs
their equipment seems to consinst of ancient 68k Sony NeWS machines,
an Amiga, old HP machines and other exotic equipment.  I think they also
have an Alpha.

But all in all they are for shure among the people that will make best
possible use of the hardware.

  Ralf

Re: Good news: no more begging for HW

[Ralf Baechle]

> : One group that I always think could use an SGI machine is the Free
> : Software Foundation.  They would make sure things worked on SGI before
> : releasing it.  That way, all of GNU would be tested on SGI by them.
> 
> This is a good idea.  Do we know if they don't have one?

Somewhen RMS mentioned that they don't have MIPS machines.  Aside of PCs
their equipment seems to consinst of ancient 68k Sony NeWS machines,
an Amiga, old HP machines and other exotic equipment.  I think they also
have an Alpha.

But all in all they are for shure among the people that will make best
possible use of the hardware.

  Ralf

Re: Good news: no more begging for HW

[Miguel de Icaza]

> : One group that I always think could use an SGI machine is the Free
> : Software Foundation.  They would make sure things worked on SGI before
> : releasing it.  That way, all of GNU would be tested on SGI by them.
> 
> This is a good idea.  Do we know if they don't have one?

An Indy machine may become the most powerfull machine at the FSF
nowadays.  I think their high end computers nowadays are the 486 PCs
running Linux.

I am not sure if the people at the FSF in Boston are actually doing
any developement besides Hurd and Emacs.

Miguel.

Re: Good news: no more begging for HW

[Larry McVoy]

: 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?

Thanks.

Re: Good news: no more begging for HW

[Larry McVoy]

: 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?

Thanks.