SGI: hinv

In spite of what gfxinfo seems to imply, one 3200x2400 desktop (if you can zoom in enough to read it, gfxinfo is running in the open window): I elected to use OpenGL Multipipe with SGI Xinerama as the X Proxy Layer . OpenGL MP run by itself uses its own xserver (Xdmx). Xdmx works to transparently allow OpenGL visuals on both pipes/all four monitors, while Xinerama (without OGL MP) only allows OpenGL to run on the primary pipe (the primary pipe being two-DCD-connected displays in my case).

Running OpenGL MP with Xinerama as the X proxy layer is a decent compromise - and I can still take advantage of OGL MP by calling OGL apps from the command line with the "omprun" command. That will let me take my time to import the parts of my current desktop environment into Xdmx, which in as-installed configuration provides only a single winterm and a Toolchest. Here's the SGI Cube run on the Xinerama-managed desktop, called from the command line with 'omprun': Without OpenGL MultiPipe only the half of the SGI Cube displayed on the primary pipe/pair of monitors would be visible - the half of the OGL window on the secondary pipe/pair of monitors would be black. Just went and took a pretty close look and wasn't able to find any evidence of tearing. I wonder if you might be seeing it because the 2@1920x2400 config we set up for your T221 is probably a couple of notches north of the intended capabilities of a single V12/DCD?

Didn't mean to upset east-west detente..... I swear this is solely for peaceful domestic purposes and not as an escalation of the sino-american graphics race.

Thanks! Not currently, but I might try it if I ever pickup another 1920x1200 IPS panel - the DVI port on the one I have is connected elsewhere.

Cool cool cool. Mikey wants !


I don't think it is, really ... I mean, the intended capabilities are limited to the 165mhz single-link dvi bandwidth (an oversimplification but convenient) and we aren't exceeding that ...

However, I still bet you are getting tearing. You just can't see it because of the bars in the way. The only time I see it is if I have something that crosses over the tangency border. Then if you move that window you can see that the dvd refreshes one side at a time, right-left right-left right-left.

It's not obnoxious but it is there. I wonder if some sort of sync device would overcome that, or if it is just unavoidable due to the way the dcd works ?

The image appears on the adjacent monitor as soon as it touches the edge of the LCD panel, so none of the visible image is actually blocked by the monitor frame. I haven't tried it yet but you can set up the displays to have a user defined amount of overlap, so that x number of pixels are displayed on both panels at the same time.
The OGL MP Release Notes obliquely touch on the intra-pipe synchronization:
So it looks like I'm probably getting some software sync from the default software buffering. I still have the 75-ohm BNC cabling that linked the 'swap ready' and 'genlock' ports when the Compositor was connected. Next time I shut down I'll reconnect/re-enable and try omprun with the -nosync switch to see what if connecting the hardware swap-ready/genlock/sync makes any noticeable difference.

Compared to how gfxinfo shows the display configuration with xinerama enabled, Xdmx is a little more savvy about the size of vitrual display. I was poking around in the log files created while I was testing Xdmx and noticed this:
Tried the Performer Town Demo as an easily accessible test of how the Xinerama/omprun combo handle an OGL app run full-screen (both V12s/all four panels):

I also shot a brief video of the demo auto-running in perfly mode - if there's anyone who hasn't already seen the Performer Town Demo in motion I can post a copy on youtube.

I don't think you would see this tho, unless the two sides were actually touching ?

Just wondering, do the two portions of each dcd synch to a single screen automatically for you every time on startup or do you ever have to diddle with them to get them to play nice ?

On mine, about half the time I have to do xsetmon from side-by-side flashing doublemint screens - two mice, two toolchests, two sets of everything -- choose the res it's already running at, click "okay" then they will settle down. Once in a while have to go through the sequence a couple times.

Curiosity killed the cat What do you get when you do a screenshit ? My guess is a single 3200 x 2400 rgb but who knows what evil lurks in the heart of man ?


Edit : Ah. Read more carefully, grashoppa. If you only have one pipe, you can turn on xinerama till you're blue in the face but it won't really do anything. So the software sync which you mentioned earlier is not functioning for me cuz it ain't there until the setup has more than a single pipe. Channel != pipe.

Would be interesting to see if the software sync does actually work when the various portions of the screen are touching, considering how they say the DCD works. Want to send over the test equipment ? i will gladly return it as soon as all the certification sequences are finished

*Never* experienced anything even remotely similar on any of the DCD equipped systems here - the Tezro and Octane2 running 2@1600x1200 and now the O350 running double 2@1600x1200.

Sounds like either the T221 is having difficulty syncing to the signal, or the V12/DCD is having trouble consistently generating it. The quick-n-dirty fix might be to treat the O350 like an O2 and *never* turn it off/restart the graphics sync process from scratch. A more involved solution might be to try some incremental changes in your 2@ refresh rate.... or better yet, put graphics in another O350 and spread that 3840x2400 graphics feed across four stripes instead of two.

Somebody buy the man a cigar.

I mentioned that as one possible reason why I'm not seeing the tearing when I cross the DCD or pipe boundary, but if you want to try, you could skip xinerama and try pure OGL MP/Xdmx set up, or better-yet-put-graphics-in-another-O350......

I think lend-lease to china ended when you were just a wee lad, but there's an unloved O350 sitting in the rack with hamei written on the toe tag (and you won't have to worry about past-due fees with that one).

It looks a little similar to what you get when you connect a monitor to the middle socket while running a 2@ resolution ...


With a T221 you never see the part that syncs ... no matter what the incoming signal is, the internals of the T221 supply a 48hz refresh to the panel. So the backy-forthy is strange. It's not a sync problem, the monitor is actually showing what it is receiving, alternating refreshes.

Also, during boot or if you go to the prom, everything is fine except doubled. It's only when you pick up the graphics of the desktop that the graphics do their dance. XSgi problem ?

Strange ...


Too noisy ... it's not a big deal - the first time scared the crap out of me but it's only a few clicks to get xsweatmon up, choose the refresh that it's already running at, then close xsetmon. Once in a great while takes two tries but that's better than the howl of an angry fan all night


I am wondering if I will have to do double sync-ups with two graphics cards ! Perhaps Xinerama will solve that problem. Some day hope to find out ...


Yeah, but now you have to resize all your screenshots, heh heh


Is it as big as an Octane V12 ?

If it walks and talks like a duck.... Sure sounds and looks like something in the vfo-V12-DCD-monitor conglomerate is half a notch off plumb.
The boot-time-prom behavior is normal - neither has any knowledge of the 2@ format magic IRIX performs on the DCD. Maybe, but there was the issue of VFC going into danger-will-robinson-mode when the format has a screen height greater than 2048 pixels . Which is why I was thinkin' four heads might be better than two - won't eliminate the-greater-than-2048 issue, but cutting the horizontal resolution in half might ease the the strain of your single V12/DCD livin' on the bleeding edge.
Angry howling all night? Sounds like your fan club is made up of CEOs and MBAs. ...it's as big as the promise, the promise of a coming day. If he's gonna bear the weight of the Southern Cross you better tell sherwood to bring the *big* suitcase.

Well, something's happenin' here. But what is it ain't exactly clear ...


You get an image of the booting process and Onyx Infinite Performance splash screens in quadruplicate on my desk first thing in the morning when you start up, I assume ?



Okay, I give in. Send over that top O350 with the V12 and DCD and I'll give 4@ 1920x1200 a try

Oh wait ! I have a better idea ! Keep the DCD's, just send over both O350's with the V12's. Ill add them to mine then we can skip the backy-forthy flickering problem entirely !


Sherwood has left the forest He's now in your neck of the woods wishing he were back here in the land of the free, the solvent*, the not-few.

* Except for me. Anyone depending on the honesty of the American businessman is an imbecile. Sigh.

It's threads like this that make me realize how little I've accomplished with my collection.

hey, fella ... you ever hear of ten-four fox ? I'd say the contest to see who can stick the most monitor cables onto the back of their computer pales in comparison to what you've done ...

Sherwood or not, you aren't completely Helpless. There are plenty of Big birds flying across the sky.

You'd just have to tote your own jumbo-sized suitcase.

Yep

Looks like it's digressed into an opportunity for hamei and I to play In-Line Jeopardy with Famous Songwriters of the 60's..... but you're more than welcome to join either endeavor.

BTW, how's the Firewire stuff going with your Fuel? I've been looking around for up an iSight to try with the pseudo dm10.

So, in the end Recondas, is it really usable, or just for fun?

Is there a difference?

My personal perspective aside, OpenGL MultiPipe (and the Onyx350 IP) work as advertised. The four-segment desktop is fully (and transparently) functional in two dimensions, and is equally capable with (the limited number) of 3D/OGL apps I have available to try on it.

If I had four projectors and an equal number of adjacent/empty walls I'm pretty sure it wouldn't be too hard to set up a cave.

For those that have four walls/projectors but no Cave suitable app, the OpenGL MultiPipe Software Dev Kit comes with the IRIX source code for Cave Quake Arena III and a molecular visualization app (if you have a smaller adrenaline addiction to feed).

Even if you don't have four walls, CQA3 would probably still look decent on a semi-circular array of four LCD panels. It'd also have the advantage of requiring very little preliminary configuration, left-to-right horizontal is the as-installed/out-of-the-box OGLMP set up (if you want a horizontal monitor configuration you'll have to tweak xservers).

So mia, where do you draw the line between usability and fun?

Right, allow me to elaborate.

For 2D applications (coding, vnc/rdesktop, administration, photo retouching with barco creator, etc.) and perhaps the occasional movie watching; is it perfectly usable? I'm interested building a similar setup.

I'm not a vnc/rdesktop/barco creator user (the much less capable PhotoShop 3.05 works just fine), but I don't see why they or any of the traditional IRIX 2D desktop apps wouldn't work in the same fashion they always have. The advantage there would be the extra desktop real estate - but for non-OGL apps the same multi-V12-DCD desktop expansion could be achieved without the need for OGLMP.

Not to belabor what I'm sure you're more than aware of as the obvious, but the underlying advantage to OpenGL MultiPipe is the ability to run apps with OpenGL visuals across multiple graphics pipes (something that isn't possible if the pipes are linked using Xinerama w/o OGLMP).

I found the installation instructions in the OpenGL MultiPipe User Guide more than adequate. If you're adding a (new) graphics pipe, I'd suggest configuring the new pipe (to match your existing pipes) before the installation of OGLMP.

For anyone who might not have had the opportunity to add a graphics pipe to an existing system, the first time you boot that system the new pipe will display a separate Desktop and Toolchest; any res/refresh changes needed in the new pipe can easily be made by opening xsetmon with the Toolchest attached to the new pipe (or use the new-pipe Toolchest to open winterm and call xsetmon as root if logged in as a non-privileged user).

Actually, I didn't know that. I've never really used any OGLMP application, and the ones I might have, I wasn't aware of this differentiation; my area of concern was storage, which is something so distant from visualization.

Now, I would love to know how this renders on a T221 or similar high-rez display. Do you think the frequency limitation of those (~48Khz) might be the clincher?

Most OpenGL apps should be transparently compatible with OpenGL MultiPipe. Several (of what I consider to be) essential IRIX tools have graphical interfaces that use OGL visuals. Out of habit I keep a copy of gr_osview re-sized to the same width and open immediately below the Desktop Toolchest (a habit acquired when I all-too-frequently wanted to know what had yet again slowed an O2 to a crawl ).

Unless I've (yet again) misunderstood, adding a graphics pipe/extra display won't have any effect on any storage (pre-existing or to-be-added).

It seems to work well enough for hamei. I suspect most of the tearing-while-crossing-the-DCD-boundry issues he's seeing with his set up are caused because he's feeding a *single* DCD-equipped V12 about as much as (or maybe slightly more than) it'll handle.

If you're considering using your four-pipe Onyx4 with a T221, SGI was kind enough to have published a fair amount of the necessary configuration data - something they didn't do for Tezro/T221 combination from the same era, which SGI didn't provide set up/configuration info, VFO files, etc.

I got a 4300B card with a TI controller and managed to drag the Fuel up to 6.5.30 and most of the Nekoware essentials. It's recognized as a DM10, so when I get a chance I'll unplug the iSight from the G5 and see how it works. What's your recommended app for testing capture?

I do not have any DCD on any of my Onyx350 (all single DVI out), so here are my options, please let me know if I missed something, should I want to go with a T221:

- Get DCD on each Onyx350 (how much do they go for? - the DCD, not the Onyx350 of course- )
- Use a G2n-brick (I have a few unused), hooked up to Onyx/Origin 350s; those have the unfamous (and unstable) ATI FireGL cards.

I do not have any Router, only directly numa connected servers.

Which route is best? Am I missing an obvious alternative?

Do I need a composer (I do not own any)?

You need at least a V12, the V10 doesn't have enough memory even at the lowest bit depth.


This is what I would try first. Free ! and should work decently for what you want to do.


With dcd's you'd only need two O350's. You could also do this with an O300 setup, an X/V-brick and two Octane V-12's plus DCD's. Or some O300's and two x/v-bricks plus four Octane V6/8/10/12's and four analog-to-digital converters. This would be another way to get graphics into an O300

Or a single Octane2 with the dualler V12-DCD combo setup. The Octane is still the nicest, most capable desktop computer ever built. Stupid-ass SGI took a great computer and shit on it.

To be honest ? If I were starting from scratch ? this might be the smartest and cleanest way to get massive graphics capabilities on a single desktop. One Octane is so much more convenient on a desk than a stack of O3xx's. And you can't put two graphics heads in a Tezro. Smooth move, ex-lax. (Unless maybe you could hack the Tez to bypass the check for that.)

Infinite Reality would also be interesting ... from the specs, it could be possible but you'd need the very hugest amount of memory you can stuff into one. And four analog-to-digital converters.

Or you could consider a Reality Monster

No matter what, it's not a cheap setup.


You have a G2n ? Try that first. For what you describe (mostly terminal and text use ?) the Infinite Disaster graphics should work fine. It was with OpenGL that they fell down (reportedly.) And (again reportedly) towards the end, they got a lot better. And, since it doesn't use Xsgi, the setup should be easier since all the Linux stuff should be applicable.

Plus it would be free for you to play with, to see if you liked it.

And the Recondas big-screen version would always be a backup option ...


The Composer goes in the other direction. It accepts several inputs to output an equal-or-lower resolution, higher-memory display. It won't go to 1920 x 1200, either.


You have to write your own

Search around for the thread where nekonoko got his Fuel to display a window upside-down of a camera input. That's the farthest I've heard anyone got with the supported firewire input.


Yeah, but it's a lot more than 500 miles over to your place and by the time I get to Phoenix, highway 61 will probably be closed


I'm on a jet plane



My PhotoSnot crapped out again, was going to try Eclipse one more time. Have you ever managed to figure out the weirdities in that program ? And maybe now that we have a reason to put up with gtk2, the gimp is worth another look ?