Irix was the operating system developed by Silicon Graphics from 1988 to 1998. The OS supported the company’s high-end workstations and served in many serious roles. The company cut off support for the UNIX-based OS in 2006, but now a diehard community is looking to bring the ancient codebase back to life.
While SGI’s workstations once sold for five or six figures, surviving examples can now often be had for just a few hundred dollars on eBay. The MIPS-based hardware was potent for its time, often used for 3D rendering work for video games, films, or for scientific purposes. IRIX was SGI’s own OS built specifically to support these use cases.
The IRIX Network is a hobbyist community that loves these old machines and their software. The group hopes to raise $6,500 through crowdfunding to reverse-engineer IRIX. The hope is to use those learnings to create an open-source derivative version named IRIX-32, based on IRIX 5.3, the last 32-bit version of the OS.
It’s a monumental task, but admirable nonetheless. Whether we one day see IRIX reborn, akin to what happened to AmigaOS, remains to be seen.
Cool, maybe I can run it on my Indy (64-bit computer from the mid 1990s!!!) :-)
Why do they need to raise money to do the reverse engineering? Just do it!
Or do they lack the talent and need to pay some hired guns?
Not that the reverse engineering project would be anything trivial, but that is what the Hackaday folks do.
Licensing, probably. There’s still a vestige of SGI around.
They could reimplement it from scratch and call it Lirix, or something.
Attorney fees to make sure it is reverse engineered properly to avoid a lawsuit.
HP Enterprise has been pretty hands off, although I wonder if this team couldn’t buy the assets outright?
> Why do they need to raise money to do the reverse engineering?
Because some of the team are not able to do it under hopes and dreams, that’s why. This is a paltry sum and I (yes it’s Kazuo Kuroi/Raion) am funding it myself as well.
> Not that the reverse engineering project would be anything trivial, but that is what the Hackaday folks do.
Working for free is not something that I can ask the assembled team to do.
I’m not sure they have any idea what they want to do or what’s legally possible.
IRIX is just a modified version of Unix, licenced from AT&T. Unix derivatives like NetBSD already run on this hardware. If there’s some deficiencies in hardware maybe work on that, maybe do some reskinning work.
It makes no sense to “reverse engineer” some old crusty OS when you know it’s just Unix, and even less sense when actively maintained versions of Unix already have support.
I have one of those Indigos. Converted to PC Keyboard, mouse and VGA monitor. 10BASE-T to AUI adapter. I bought a SCSI-I drive and used my Linux system as a TFTP server to install Irix 5.3 off some CD images I “found” [nudge, nudge, wink, wink] on the Web.
I have (well, had) everything running except for a compiler. I know you can get GCC to run, but the Irix OS differs from standard Unix in that there is no “cc”…you have to pay for it and license it. And I wanted the original compiler and libraries. So I don’t have them…kinda waiting for it to be declared abandonware, I guess. I’m cheap.
Also…did you know that there are very few websites that don’t require some version of Javascript? Of course, Netscape Navigator doesn’t know what to do with Javascript. :-) The graphics demos are cool though, even on a VGA.
SGI released the c compiler suite for IRIX 5.3 as freeware in the late 90s. It should still be available online somewhere yet. Beware! 5.3 didn’t have a native c++ compiler, only an awful c++ to c translator.
Found it: iris-development-option-5.3.tardist
Thanks for that tip. Googling for that file brought up all kinds of SGI archive websites, hardware and software.
> Beware! 5.3 didn’t have a native c++ compiler, only an awful c++ to c translator.
Ehehehehe, ‘cfront’ I guess. It’s what I used for programming C++ at AT&T, a long, long time ago. But we ran it on IBM RS/6000 systems with AIX (and now you can guess how long ago it was ;)).
But it was not so horrible, imo. cfront added the C++ code as comment to it’s C output. And you could see it in dbg.
Nowadays, I would *never* *ever* want to work with such a primitive toolchain anymore. I have been throughly spoiled by all those IDEs out there (but hate that when they fail, they tend to fail miserably with a vengeance).
But at the time it wasn’t bad at all. I could even run the toolchain and debug directly from EMACS. Come to think of it: with EMACS it was almost an IDE.
But at the time there was also Turbo Pascal 3, and I would consider that a real IDE.
Ok. I’m done with my nostalgia trip. :)
You could try to cross compile gcc to MIPS. Word of warning it would be an older version of the tool chain.
You want the sgi freeware cd for IRIX 5.3 to get gcc. I built a lot of software with it, because I did not have the development distribution for 5.3.
I did have it on 6.5!
Also, on the IRIX 6.5 branch, MipsPRO will compile with nag screens from 6.5.10 and older, 6.5.9, .8… you get the idea. I cannot remember whether 6.5.10 was the last working for free release, or the one that killed the free lunch.
While it ran for free, the highest optimization: -O3 made a big difference over gcc at the time.
On 6.3, I built Mame once using both. I remember Smash TV would almost run on an Indy using gcc and ran solid using MipsPro.
Finally, it was slow, but I kept one Indy back on the 6.5.10 (last MipsPro freebie) to build software with. Worked out well for a good long time. Build on the Indy, then push to the daily drivers. Also, somewhere around that same release the GREAT online books got changed for the worse.
The good set were built to run local and had a spiffy interface. They contained almost everything, complete with many copy and paste examples. I learned a ton. Fantastic resource. It is worth keeping a machine back for those too. That is what I did with the MipsPRO machine.
SGI computing in the late 90’s through early 00’s was amazing. Best overall computing experience I had. Expensive though. I was lucky to work in an industry where I had good access to all that stuff and a lot of high end software.
Its you! Based on this article I did a search on restoring Indigos and ran into a post you wrote with some similar info on gcc and Indy’s from a few years ago. Hah.. its a sall world sometimes. Anyway nice work!
“Although generally “dirty room” reverse engineering efforts that occurred with direct access to code, they emerged thanks to an NSA-developed tool, Ghidra, which makes software-based reverse engineering easier.”
Thank you, NSA.
Ghidra is pretty cool. They even made a function in it that gives you a flowchart of the disassembled code base, which is very handy.
I perpetually don’t understand that community’s obsession with weirdly sexualized asian cartoons. I imagine they’d get more engineering help if they dropped that nonsense.
1 image on the front page is an obsession? And it’s probably a reference to Nekochan.net… which also had an anime character on the front page.
Haiku os and be os
I am a retired (semi disabled) pensioner living in South Africa. One of the residents at the retirement village here brought a working INDY PC to me in fully working condition! I had never seen or worked on one of these before and am amazed that it still works and also at the capabilities of such old hardware and software. I managed after some struggles to gain Root access (I do not know a great deal about Linux and even less about IRIX 5.3) The owner says that a particular function which could be selected off a menu just ‘disappeared’ one day. I would appreciate it very much if anyone could provide a (set of IRIX) command\s to do a recursive search (from the root directory) to find a specific text string in a command script or menu. I am trying to determine whether the utility still exists on the hard drive, or whether it was on a network resource which was removed. At startup, the PC tries to configure itself according to a network configuration but this obviously fails. Any advice at all will be greatly appreciated. Please contact me at ken.everett.sa@outlook.com
>It failed to make the leap to a new era when x86 architecture began to dominate the wider computing industry.
This is total rubbish.
The thing that killed SGI was that it chose to follow intel when intel ABANDONED the x86 architecture. SGI and many others bet hard on the VLIW-style 64-bit intel itanium, abandoning their own CPU designs that were all far more powerful than x86. Alpha at that point could emulate x86 faster than x86 could actually run. When itanium flopped it left all of them high and dry, leaving no choice but to fall back to intel’s x86.
SGI’s high-performance MIPS CPUs were hardly the only casualty. HP’s PA-RISC and DEC Alpha were also abandoned. Suddenly instead of 5+ competitors with faster RISC architectures (and growing), intel’s tired 1970’s x86 only had to contend with Sun’s SPARC (and lesser-known IBM POWER). Intel gathered up the market share from their now CPU-less competitors and used the added cash flow to simply burn transistors and pull ahead. (they moved to a “RISC-like” core which was marketing-speak for cumbersome ISA translation and a RISC-y superscalar microarchitecture that persists to this day)
The rest, as the say, is history.
PS. that cumbersome ISA translation also happens to be why arm beats the tar out of x86 when it comes to perf-per-watt. There’s a whole pile of extra work an x86 CPU has to do on every batch of instructions before it can get to the juicy superscalar bit. All in the name of compatibility with a design from 1978.
The thing that killed SGI was a business model that was completely reliant on their customers desperately clutching to their entire ecosystem while other systems grew up that could do the same work for less money. At some point the technical niftiness that an Idie could manipulate huge textures if you $100,000 in hardware and an ongoing Irix contract and an ongoing hardware contract plus other ecosystem costs wore off. Cause you could by a $5k computer and do some piece of that, and for many that piece was sufficient. Spend $20k and you could do the same thing only slower and without the ongoing subscription costs.
Nonsense. I was @ SGI at that time and simply put; they lacked the cash to continue developing MIPS processors further. In fact, SGI had to buy MIPS in the 90s because they were about to go out of business and they needed to secure their processor supplier until they could transition away from MIPS. Other than IBM POWER, every other major high end RISC architecture was done by the late 90s and early 2000s. Alpha literally killed DEC and similarly, SGI basically died because of MIPS. HP saw the writing on the wall by the early 90s, and planned the transition away from PA-RISC right then. SUN was also killed by SPARC going out of order.
I don’t think people understand that CPU architecture is on an exponential design cost. So unless you can leverage it via enough economy of scale, you will eventually run out of funds and customers who are willing to pay the premium.
The idea appears, at least at this point, to be to reverse-engineer drivers to the point where emulation is possible. I don’t see why that’s required – NetBSD already did all of that work with the sgimips port ( http://wiki.netbsd.org/ports/sgimips/ ). If this SGI “community” can’t figure out how to use fully working drivers to improve emulation, I don’t have a lot of hope for anything on any sort of larger scale.
What are the quotes around community for?
From your link, NetBSD doesn’t support hardly any of the hardware on these machines… isn’t that what they’re attempting to work on? You can’t use any of the 3D hardware, any of the video hardware, very little of the audio hardware it seems, forget any interesting option cards. NetBSD also doesn’t support all the 32 bit machines, just the more common ones. Where’s the drivers for Elan or XS graphics in the Indigo? What about Galileo Video? or IndyVideo? Or the XZ graphics in the Indy, or Extreme in the Indigo2? It appears you can’t even use the ICE hardware in the O2. Where are these sold called drivers that are “fully working”? The quotes are there because they don’t exist.
What’s wrong with people trying to save these old computer systems before they’re gone? And why comment on things you don’t know anything about?
“today SGI workstations can now be had on ebay for just a few hundred dollars” – where?! I’ve been after one for years but they’re always thousands of dollars when I look.
Some times you can find as-is husks on auction for a couple hundred, but the skins will be wasted and it will be missing key components. Sold my last SGI (Octane 1 original pristine skins, upgraded to Octane2) when the xbow and mobo died and I used the last of my spare parts to fix it. Couldn’t find cheap spare parts and this was before nekochan died.
I still have an OG high-end SGI Octane in my closet I bought from a homie in 2000. It wouldn’t boot at the time as my homie got wild with it, but it was an insane piece of hardware to be had for $1500 bucks back then, apparently it “fell off a truck” from some major movie studio in silicon valley back then. For years, it was useless, but made for an interesting doorstop and conversation piece for geeks in my crib.
Fast forward some 7 years later, the same buddy was scabbing together another Octane for nostalgia, and had the original Plextor scsi cd drive to get it bootable, and I bought as new of an irix cd copy set off ebay I could find, and we got it booting and to a desktop again. In circa 2007, it was anti-climactic, but for the first time I tweaked with irix.
Love these behemoths of technology, though it collects dust in a closet as I am too old to lift and move it around anymore.
I like the IRIX OS. If the IRIX can be rebuild into a high performance new system and compitable to new hardware, I think it will be spread out wildly、quickly.
Irix source code is still there on internet , so nevermind about reverse engineering. The problem is license and copyright owners Hewlett-Packard. It’s a Enhanced version os Unix System V it would be nice to see OpenIrix competing with Linux or FreeBSD.
You failed to read the article.
I hope some day this happens, but what’d really would be fun if they’d open source the kernel (less foreign copyright issues) and firmwares and ‘we’ could start porting it to newer MIPS CPUs.
Imagine a ThunderX in an Octane. I know, not realistically possible at all, but the same could have been told to the people who wrote Irix in the first place.