[kaigan]

Hey, everyone,

I had a very long post that I accidentally deleted by pressing the wrong button before submitting it. Oops.

So, short version: inspired by the recent Hackaday article "Could Seven Gamers Play Quake on Just One 1996 SGI Machine?", I decided that I want to try it with my Onyx2 deskside. The IO6G and CADduos in the PCI card cage all show up as input devices, currently giving me five sets. So far, so good.

However, I can't figure out how to configure Xsgi to divide a single graphics board's channels up into displays instead of screens (in the X sense of the word). All of the documentation I can find is either for the Onyx4's XFree86 implementation or how to do it for the CADduo on an Octane with two discrete graphics boards.

I'd appreciate any advice on if this is possible on a single graphics board (the Onyx2's IR3 DG5-8 in this case) and how to do it. Given that the IO6G has two keyboard and mouse inputs itself and the desksides only support one set of IR graphics, I'd think it's possible. I just haven't figured out how.

Thanks!

[Shiunbird]

Hi Kaigan,

I'm very interested in such thing.
I will try in my HP-UX box. The setup shouldn't be very different.

I have two DVI outputs and can hook up multiple USB input devices.

This guide is for Linux, but it should help:
https://linuxgazette.net/124/smith.html

[kaigan]

Thanks for the info! I'll have to see what I can adapt over to Xsgi, since there isn't a traditional Xorg.conf file or the like. I imagine it will just take some more reading, along with a little bit of trial and error, to figure this out.

[Shiunbird]

In this article, they are using one GPU per user, so it's back to zero.
But, at least on HP-UX, it sees my two outputs as different devices, so I think it should work.

I will try it out and let you know if it works.

[kaigan]

All of the documentation I've been able to find on the Onyx2's "multi-seat mode" as SGI called it, refers to the rack systems which can have multiple Infinite Reality pipelines. Since that's the case, it seems like running seven/eight players off of a single SGI machine would require at least a three-rack system, with four Onyx2 graphics modules and a compute module. You'd also likely be able to run Quake at far higher resolutions than pretty much any PC user would have been able to do in 1996.

I'll keep poking around here and there, but I've lost confidence that a single graphics pipe can be divided this way.

[vishnu]

I would say just telnet to your Onyx2 and set the DISPLAY variable to open its windows on each player's workstation, but I've only ever been able to get that to work when all the involved workstations were running IRIX. I just tried it again from my Linux box and received the ubiquitous "Error: Can't open display" reply. Thank you xorg, that's a very helpful error message, NOT!

EDIT: And yes, I did do `xhost +` on my Linux box first...

[kaigan]

I would say just telnet to your Onyx2 and set the DISPLAY variable to open its windows on each player's workstation, but I've only ever been able to get that to work when all the involved workstations were running IRIX. I just tried it again from my Linux box and received the ubiquitous "Error: Can't open display" reply. Thank you xorg, that's a very helpful error message, NOT!

EDIT: And yes, I did do `xhost  +` on my Linux box first...

In this case, the goal is to run each display/keyboard/mouse solely from the Onyx2, with no other hardware or workstations involved. And, unfortunately, it looks like it would take a graphics pipe per player to pull that off.

[jan-jaap]

EDIT: And yes, I did do `xhost  +` on my Linux box first...

Ah, but did you add '-listen tcp' to the X server command line? Otherwise your network-transparent window system only works for 127.0.0.1 ...

[Shiunbird]

Hello everyone,

I took my IBM T221 out of storage (the bloody thing synchronises with anything, any resolution, with proper scaling, and never fails). Here is my progress:

These are my guides:
http://h10032.www1.hp.com/ctg/Manual/lpv38356.pdf
https://linuxgazette.net/124/smith.html

To operate, I use my laptop via serial console.

I modified the default XF86Config as follows.

1. I created a full duplicate of my existing configuration within the same file.
2. I modified the second section to have name MainLayout2 and I renamed the original session MainLayout.
3. Now by issuing "X -layout MainLayout" or "X -layout MainLayout2", I could run sessions.
4. The table under "Step 3" of https://linuxgazette.net/124/smith.html was helpful. I ran a ioscan and managed to identify my 2nd keyboard and 2nd mouse.
5. I pointed the second keyboard and second mouse to MainLayout2. Now, when I do "X -layout MainLayout2", my 2nd keyboard and mouse are active.

Here is where I am stuck.
My original X configuration pointed my graphics path to be:

Section "Device"
Identifier "Console"
DeviceFile "/dev/gvid"
EndSection

A quick ls /dev pointed out that I have /dev/gvid, /dev/gvid0 and /dev/gvid1.
I tried /dev/gvid0 and it works. However, /dev/gvid1 complains that there are no displays connected.

Maybe I need to change both the device path and the screen information at the same time, or maybe change only the screen information and keep the device path.

If I try changing via SAM (HP-UX configuration tool), it says that having two independent screens (canvases) is not supported, but it allows me to create a single logical canvas with two screens, and it works ok. I'm not sure if this is an artificial limitation due to CDE or if it's really X.

I'm hopeful this could work, because for some reason the OS sees the second port as being in a completely different bus!

bash-2.04# ioscan | grep 0/4
0/4 ba Local AGP Bus Adapter (784)
0/4/0/0 graphics AGP Display (10024a4d)
0/4/0/1 graphics PCI Display (10024a6d)

The funny thing is... whatever I do, my T221 is the console screen and when I start X, it goes blank and the other screen takes over.
(I always start and stop X from the serial console).