I see a lot of articles talking about the white elephants that might be lost from public view, which is probably the biggest tragedy, using their KI-10 as an example.

The one I’m most worried about from that collection is that they have the last known operational CDC6000 series machine (Theirs is a slightly smaller CDC6500, the flagship CDC6600 is the machine that made Seymour Cray famous, it fucked so hard it was 3x as fast as the previous title holder when came out in 1964 and was still the fastest machine in the world until 1969… when it was replaced by the derived, upgraded CDC7600 from 1969-1975).

It’s a 12,000lb, 80" tall, 165" on a side monster that draws 30kW (at 208V/400Hz), I haven’t heard a plan for it, and there are very, very few possible long-term-secure homes for such a thing.

I guess it’s just not in the current auction so it isn’t drawing as much attention yet?

…I probably should have checked the byline before posting. It does still come from the same material, just a little more directly.

There’s some weirdness on that because she did some important but not-very-public work at IBM in the 60s with their ACS/“Project Y” effort that did what we later call superscalar/multi-issue processors like …20 years before those terms existed. As part of that she wrote a paper about “Dynamic Instruction Scheduling” in 1966 under her pre-transition identity that is a like retroactive first cause for a bunch of computer architecture ideas.

There was almost nothing about that work in public until Mark Smotherman was doing some history of computing work in the late 90s, put out a call for information about it, and she produced a huge trove of insider information after deciding it was worth exposing the provenance. There’s a neat long-form LATimes piece about the situation which is probably the primary source for the history in OP’s link.

That’s credible.

I find the hardware architecture and licensing situation with AMD much more appealing than Nivida and really want to like their cards for compute, but they sure make it challenging to recommend.

I had to do a little dead reckoning with the list of supported targets to find one that did the right thing with the 12CU RDNA2 680M.

I’ve been meaning to put my findings on the internet since it might be useful to someone else, this is a good a place as any.

On a fresh Xubuntu 22.04.4 LTS install doing the official ROCm 6.1 setup instructions, using a Minisforum UM690S Ryzen 9 6900HX/64GB/1TB box as the target, and after setting the GPU Memory to 8GB in the EFI before boot so it doesn’t OOM.

For OpenMP projects, you’ll probably need to install libstdc++-12-dev in addition to the documented stuff because HIP won’t see the cmath libs otherwise (bug), then the <CMakeConfig.txt> mods for adapting a project with accelerator directives to that target are

find_package(hip REQUIRED)
list(APPEND CMAKE_PREFIX_PATH /opt/rocm-6.1.0)
set(CMAKE_CXX_COMPILER ${HIP_HIPCC_EXECUTABLE})
set(CMAKE_CXX_LINKER   ${HIP_HIPCC_EXECUTABLE})
target_compile_options(yourtargetname PUBLIC "-lm;-fopenmp;-fopenmp-targets=amdgcn-amd-amdhsa;-Xopenmp-target=amdgcn-amd-amdhsa;-march=gfx1035"

And torch, because I was curious how that would go (after I watched the Docker based suggested method download 30GB of trash then fall over, and did the bare metal install instead) seems to work with PYTORCH_TEST_WITH_ROCM=1 HSA_OVERRIDE_GFX_VERSION=10.3.0 python3 testtorch.py which is the most confidence inspiring.

Also amdgpu_top is your friend for figuring out if you actually have something on the GPU compute pipes or if it’s just lying and running on the CPU.

Neat.

I set up some basic compute stuff with the ROCm stack on a 6900HX-based mini computer the other week (mostly to see if it was possible as there are some image processing workloads a colleague was hoping to accelerate on a similar host) and noticed that the docs occasionally pretend you could use GTT dynamicly allocated memory for compute tasks, but there was no evidence of it ever having worked for anyone.

That machine had flexible firmware and 64GB of RAM stuffed in it so I just shuffled the boot time allocation in the EFI to give 8GB to the GPU to make it work, but it’s not elegant.

It’s also pretty clumsy to actually make things run, lot of “set the magic environment variable because the tool chain will mis-detect the architecture of your unsupported card” and “Inject this wall of text into your CMake list to override libraries with our cooked versions” to make things work. Then it performs like an old GTX1060, which is on one hand impressive for an integrated part in a fairly low wattage machine, and on the other hand is competing with a low-mid range card from 2016.

Pretty on brand really, they’ve been fucking up their compute stack since before any other vendor was doing the GPGPU thing (abandoning CTM for Stream in like a year).

I think the OpenMP situation was the least jank of the ways I tried getting something to offload on an APU, but it was also one of the later attempts so maybe I was just getting used to it’s shit.

Don’t trust that they’re 100% compatible with mainline Linux, ChromeOS carries some weird patches and proprietary stuff up-stack.

I have a little Dell Chromebook 11 3189 that I did the Mr.Chromebox Coreboot + Linux thing on, a couple years ago I couldn’t get the (weird i2c) input devices to work right, that has since been fixed in upstream coreboot tables and/or Linux but (as of a couple months ago) still don’t play nice with smaller alternative OSes like NetBSD or a Haiku nightly.

The Audio situation is technically functional but still a little rough, the way the codec in bay/cherry trail devices is half chipset half external occasionally leads to the audio configuration crapping itself in ways that take some patience and/or expertise to deal with (Why do I suddenly have 20 inoperable sound cards in my pulse audio settings?).

This particular machine also does some goofy bullshit with 2 IMUs in the halves instead of a fold-back sensor, so the rotation/folding stuff via iio sensors is a little quirky.

But, they absolutely are fun, cheap hacker toys that are generally easy targets.

You can still readily get crank hand drills, I have a (vaguely) modern one that I use for situations where I want the control/tactile feedback and/or have restricted access or the like. It covers a different set of problems than the standard cordless.

Mine is Fiskars branded and a little plasticky (and not the version they sell currently). I like it enough that I’ll get a nicer one if I kill it.

Sure, drop me a note with the details and I’ll see if I can give you a hand. I’m not super expert in all the specifics of the Chromebook ecosystem, but I have good general computer/Unix skills and have hacked a couple so I know where to look for resources.

Relevant place to ask: I’ve been trying to find a reference for the earliest Emacs that could host a terminal emulator or subshell in a window.

Multics emacs appears to have had both split windows and a character-at-a-time input and output mode as far back as 1978 for use as a SUPDUP and/or TELNET client, which is currently the earliest I’m aware of. Ancient ITS TECO EMACS had splits pretty early on, and may have sprung the necessary character plumbing earlier - but I’ve never found any reference material to confirm/deny.

It’s a fringe to a larger interest, which is that I’ve been trying to document the history of terminal multiplexers, especially in the Window (1986)-Screen(1987)-Tmux(2007) tradition (as opposed to the historical meaning which we’d call terminal servers). I’m slowly becoming convinced they came about after the advent of floating window GUIs hosting multiple terminal emulators. If you were super connected and could get access to one, sometime fairly early in the window between the 1973 introduction of the Alto and the surviving 1979 manuals the Alto program “Chat” could run multiple telnet sessions in floating windows (I’m also looking for a more precise date for when Bob Sproull made Chat able to do that trick). Several other early graphical systems like Blit terminals (1982 inside Bell, commercial as the 5620 in 1984) and early Sun Windowing System of early SunOS (1983) could also do multiple floating terminal emulators, so they were common by the early 80s.

Because the 36-bit DEC lineage had pretty robust psuedoterminals all the way back into the mid 1960s ref, a lot of hackers did a lot of fun shit on PDP-10s with ITS and TENEX and WAITS, and Stanford and MIT had PDP-10s connected to fancy video terminals by the mid 70s, it’s IMO the most likely place for the first terminal multiplexers to emerge… if I could just find some documentation or dated code or accounts.

I have an ESP32 set up with Zimodem which just makes the ESP32 act like a Hayes modem that talks IP instead of phone numbers to the serial port. They’re a ton of fun.
FujiNet appears to offer a little bit more in the way of high level service translation.

IIRC, the Ultra 1 and 2 are strictly SBus machines, the all the later Ultra 5/10/30/60/80 are PCI machines, plus most but not all members of the family have UPA slots with that freaky two rows of card edge connector for fancy video boards?

For readers not exposed to lots of Sun lore, Ultras were distinguished from SparcStations because they host 64 bit SPARCv9 parts branded “UltraSPARC,” as opposed to the 4m SparcStations which were based on 32-bit SPARCv8 processors.

I’ll also add that, if you don’t want to fuck around with large pieces of aging hardware and just want to marinate yourself in a retro Solaris environment, the qemu sparc support is really good. Folks restoring Sun stuff with disc issues often do their installs via netboot from an emulated server. Adafruit even has a beginner click-by-click tutorial for spinning your own emulated Sun4m system.

Selecting Suns is easy because there aren’t many bad choices in the era you’re talking about, but a little weird because the internal names and the package label names don’t always match in obvious ways. Most of the “classic era” Sparc boxes are Sun-4 variants, with SparcStatons mostly being Sun-4c or Sun-4m and Ultras mostly being Sun-4u machines. The Sun-4* name is more important to knowing what you are looking at than the case badge. For example, I have a “SparcServer 20” that some previous owner installed a TurboGX (cgsix) video board in, so it’s almost exactly a similarly-spec’d SparcStation20 with different badges.

Pre-SparcStation Sun-3 and Sun-4 VME based machines are quite a bit more exotic to source parts for in a modern context, and newer stuff are PCs (remember they did go and re-use the Ultra name for a family of x86 boxes a couple years later, so watch model numbers if you’re trying to buy a SPARC Ultra).
SparcStations are a little more bespoke and workstation-y (SBus cards, SCSI discs) and Ultras are generally a little more PC-like (mostly PCI cards, ATA discs), but neither are particularly hard to work on these days since the common SBus peripherals aren’t terribly expensive and SCSI disc emulators like BlueSCSIs have come down in price and up in performance. IIRC, in all cases you have to be kind of specific with RAM, some older machines use memory modules unique to the family and Ultras mostly take 168pin PC style DIMMs but are picky about the exact details.

IMO the SS10/SS20/SS5 Sun-4m machines are pretty nice to work with because they are still “workstation grade” high reliability parts but were made in HUGE quantities and are extremely modular within the family so it’s easy to work on them and get parts/upgrades/documentation/etc. They also have 10baseT Ethernet onboard (careful about degrading your whole switch), while the older SS1/SS2 need an AUI transceiver.

Peripherals:

Remember that older Suns use their own protocol over MiniDIN-8 for keyboard and mouse and 13-W3 video cables. You’ll need a suitable Sun keyboard (probably a Type 5 or Type 6) and mouse, and those can be expensive on their own if not bundled because keyboard people. They’re not as bad as some of the more exotic and/or desirable to keyboard enthusiast bespoke keyboards, but still pay attention when considering a machine to buy. Video is a little easier because 13W3-to-VGA cables are a thing, (I have one of these with switches so you can configure for Sun or SGI or Next or IBM’s particular signaling). You still need a monitor or scan converter that works with Sync-On-Green to accept the signal… most modern LCDs with VGA ports actually can, but the labeling is typically not very clear about that. Sun video adapters are generally a little more willing to negotiate video modes than some of the other workstations (eg. My SS20 has talked to almost everything I’ve plugged it into, my HP Apollo 9000/735 and its absurd CRX-24z video board will talk to the Dell P2314H on my real work desk and has spurned every other monitor I’ve tried it with).

NVRAM:

Most older Suns have a chip on the motherboard - typically with a yellow barcode sticker if it’s original - which contains a small battery-backed NVRAM storing the serial number, the Ethernet MAC, and various configuration parameters, and a RTC (Real Time Clock). At this point the internal batteries on all of them should be presumed dead. The M48Txx line of chips Suns use were originally made by Mostek, who was absorbed by SGS-Thompson, who became STMicro. Ref for NVRAM chips. Once it dies the machine loses its machine ID and MAC address and such. Fortunately, they can be reprogrammed from OpenFirmware, either with original values read from stickers and the like, or suitable made-up replacements. There are a lot of surviving Suns hand-assigned MAC addresses containing amusing strings like DEAD, BEEF, CAFE, C0FFEE etc. as people have made up suitable numbers. Sun’s factory MAC addresses have a 08:00:20 prefix if you want networking tools that notice that sort of thing to assume it’s a Sun.

Generally there are 3(and a half) options for dealing with them:

1. Modern production compatibles are still available though you have to be a bit careful about model compatibility, and they’re rather expensive these days, something like $25 a piece (eg. Mauser has a small stock of MT48T08s for $26.50+S&H ).

2. You can also grind an end and attach a 3.3v coincell battery holder yourself - some folks say you should always cut the old battery all the way out because there may be unwanted effects to having the dead battery in parallel with the good one.

3. You can crack the whole top of the module with the battery and crystal off and solder on a module with a replacement crystal and user-serviceable battery holder in place.

4. For rarely-used machines, you can just do the reprogramming procedure (in the first ref) at the OpenFirmware OK prompt by hand each time you start the machine, it will hold while the computer is powered.

It’s not a huge deal, but it is a thing to expect to have to deal with.

Software:

Remember that the OS nomenclature is a little weird because Solaris started out being versioned on top of SunOS (eg. SunOS 5.1 hosts Solaris 2.1), and at they dropped the SunOS name then leading “2” from Solaris versions so you have Solaris 2.5->2.6->7->8. The Wikipedia version history table is straightforward enough to work through, and has decent notes on supported systems. You’ll generally be between 2.1 and 9 on the era of systems you’re talking about, and those are the ones that “feel” like old commercial workstation Unix with OpenWindows and CDE and whatnot - I’m partial to 7 as “peak Solaris” but I’m sure that’s because I helped maintain a bunch of 7 boxes at one point, it’s a fully mature SVR4 with all the commercial Unix-isms before it started to converge with the modern Free Unix-likes. Many of the usual suspects like Tenox and WinWorldPC have install media and/or software.

Edited to add from downthread:

Emulation:

If you don’t want to fuck around with large pieces of aging hardware and just want to marinate yourself in a retro Solaris environment, the qemu sparc support is really good. Folks restoring Sun stuff with disc issues often do their installs via netboot from an emulated server. Adafruit even has a beginner click-by-click tutorial for spinning your own emulated Sun4m system.

Most Chromebook’s firmware is Coreboot, but it’s running a Depthcharge payload instead of UEFI (or BIOS or whatever). Mr. Chromebox maintains UEFI Coreboot payloads and install tools for a wide variety of (x86) Chromebooks, which can be used to flash a normal UEFI payload and boot normal OSes. It’s strictly possible to boot normal Linux systems on a the Depthcharge payload modern Chromebooks use, but uh… here’s the gentoo wiki on it, it’s a substantial pain in the ass.

S9

I’m still using my S9. Size is about as big as I want to deal with. Indicator LED is great. 3.5mm jack is great. SD socket for local storage. Camera is still fine. Qi charging is one of the few gimmicks that hasn’t turned out to be useless. Screen is drastic overkill. Design is a stupid friction-less glass egg, but that’s easily fixable with a minimalist case. Performance is still perfectly adequate.

It’s long out of support, but I’m finding the market wildly un-compelling, and will probably just roll with it until something renders it unusable.

1. Neat, I never knew what that system was called. I have fond memories of my local libraries DEC terminals.

2. Ooh, it looks like that ran on Pick, which is a neat early database/operating system/programming environment …thing named for one of its primary authors (I shit you not) Dick Pick. Later typically the UniVerse hosted variant, which is proprietary up to it’s eyeballs, and still sold by Rocket Software as U2. I don’t think I’ve ever come across a copy of native Pick or UniVerse for a 90s Unix or NT in a vintage software archive, but it has been widely used essentially forever, and it is virtually search proof, so it might be out there.

3. Dynix itself is pretty search proof since it was also the name of an influential multiprocessor Unix from Sequent, which, like Pick, was at some point bought by IBM.

4. Holy shit they’re still a thing https://www.sirsidynix.com/

Have you seen that Jason Scott recently received and started digitizing 200 issues of Computer Shopper? The February 1986 issue is already uploaded as a pilot test, and it’s pure nostalgia fuel to flip through. You lose a little bit of the experience not having to wrangle a ridiculously large piece of low grade paper, but it’s still delightful.

He doesn’t currently have a complete set, if anyone has a cache they might be willing to contribute to the effort, check the bottom of the post for missing issues.

In the same kind of vein as imagemagick, Dave Coffin’s dcraw tool at least partly underlies almost every non-proprietary RAW image decoder, and some of the commercial ones (if they don’t use code, they use constant matrices and such).

He’s not a sole maintainer to any of his major projects anymore, but honorable mention to Fabrice Bellard who initiated both ffmpeg and qemu among other notable activities.

IIRC the Expat XML parser that’s embedded everywhere was basically on spare-time maintenance by Clark Cooper and Fred Drake for a couple decades, but I think they have a little more resources now.

SQLite is a BDFL situation more than single-maintainer, but D. Richard Hipp still has his hands on everything, and there are only a relatively small number of folks with commit access.

I’m somewhat surprised to learn that each Snowmobile unit is only about 100PB in a 45x8x9.6ft highwall shipping container

Gonna ballpark in stupid units to see how wrong my intuition that that’s not very dense is:

Assuming dense but not hot new thing spinning rust, 16TB per 34.5 cubic inch standard 3.5" disc.

(100PB/16TB)*2 (assume at least two spindle redundant) is about 12,000 discs, so about 414,000 cubic inches of just discs without any of the supporting equipment.

A highwall shipping container is like 5,900,000 cubic inches, so only like 7% of that thing would be discs.

Or, accommodating a little bit of the support, let’s say it’s just full of those commodity 90-bay 4U storage servers. Those are 19" x 7" x 26.4" (3511.2 cubic inches) for 1440 raw TB each, again 2 spindle redundancy so you’d need about 140 of the things for 100PB, round up to 500,000 cubic inches of those… still less than 9%.

Yeah, unless I did my math radically wrong, that’s surprisingly not very dense.

Oh, it is, I just meant my instance has had its keys rebound so I haven’t run into that problem.

When last I looked miniflux wasn’t quite as featured a solution as I’m looking for (I keep a lot of notes and such embedded in my DB), but it is pleasingly simple.