7 ms·
The Architecture of the Burroughs B5000 (1982)
- dang 6y agoLink was originally http://www.ajwm.net/amayer/papers/B5000.html http://www.ajwm.net/amayer/papers/B5000.html but that one's dead so we changed it.
- hisham_hm 13y agoIncidentally, I just finished reading "The Mythical Man-Month", which also discusses another old large system, IBM System/360 and its operating system OS/360. It's really interesting to see how current the discussions on these old machines can be. From the OP: "For additional security, code and data were distinguished in memory by the use of a "flag bit", and the hardware would not execute data or alter code without first having explicitly changed the flag" We've been dealing with buffer-overruns and code injection in the stack until very very recently, and there are surely machines out there that still don't have support for the NX bit. We surely still have a lot to learn from the past.
- mikestew 13y agoWhat was it, _South Park_ that did the "Simpsons did it" episode? Thinking about some of the things that came along 20 or 30 years later remind of that. Virtual machines? OS/360 did it. Virtual memory? OS/360 did it. Non-executable data? OS/360 did it. I was a security lead for a product while at Microsoft when the NX was being implemented, and wondered at the time why it took 35 years to get around to that. (I've never worked on a 360, but did some work its successor the 370 and OS/370 some many years ago.) Even the Burroughs had multitasking before then, in the early 60s. And how can you not like a system with a "HEYU" instruction to interrupt other processors?
- da02 13y agoEver get an answer, or come close to one, as to why it took 35 years?
- cliffbean 13y agoBefore the Internet arrived, very few people had a big need for it. It makes some things marginally harder, and it's more work to implement, so it makes sense that it got left off. By the time the Internet generally arrived, and people generally realized what the Internet meant, we had accumulated enough backwards compatibility burden that sweeping architectural changes were difficult.
- mikestew 13y agoNever had access to those that could give an informed answer. As cliffban says, probably not much of a need. But if I had to pick a reason, and it would be a reason based on no inside knowledge, it's that the 8086 was a stop-gap design until the Intel iAPX 432 came out. The 432 had a lot of capabilities that mainframes took for granted, including segregation between data and executable code. Problem was that the 432 was so over-designed (which made it slow, for one problem) that it never saw production. So everyone just kind of settled on x86 and close to forty years later we're stuck with the 8086 legacy.
- derefr 13y agoMicroprocessors were one of those "disruptive" innovations, in that their architectures mostly threw out the previous 30 years of good Engineering design ideas in order to give customers something cheap and immediately useful. Oddly, it's taken another 30 years for microprocessors to reabsorb those ideas.
- Taniwha 13y agoI disagree - I was a B6700 hack, loved the machine - but code for it was just so non-portable because the machine model was just so non-standard - remember this is a machine (B5000 et al) that depended on the compiler creating correct and sane code not just for system security but system integrity - only 'root' could write compilers and one bad instruction would take the entire system down - it was fragile. Microprocessors when they first came out were also feeling their way into architecture, Intel sort of headed down the Burroughs tagged capability architecture route - 286 segments, the 960, the 432 et al But in the long term the PDP11/Vax were more the machines that people loved and their simple memory model (a simple array of bytes) and generic identical registers allowed portable languages and operating systems to flourish - prior to this operating systems were designed for particular hardware platforms - Burroughs MCP was never going to run on any other machine. People forget that the idea of a portable operating system in a time where operating systems were written and were the closed property of hardware manufacturers was a pretty radical idea at the time. RISC, where simple was beautiful, loved the simple memory/execution models, Unix having grown up on PDP11s and Vaxen was the same. Intel's through this was bit of a weird duck - they tried everything and failed at many of them - the CISCiest thing imaginable (the 432 a true capability machine) and the 960/860 (both kind of RISCy) all appeared at about the same time - but the 'x86 architecture has walked a pretty magically fine line - in a time of VAX derived CISCy architectures (68k 32k etc) they had a relatively simple memory addressing model (no indirect stuff) and a pretty simply instruction set - in retrospect much more RISCY than it's contemporaries, though missing the generic register file, the 386 eventually adopted the simple linear memory model and tried to forget about all that silly segment stuff and the 64-bit instruction set has moved to a RISCy world - I used to really hate the x86, especially when I wrote compilers, but in retrospect it's mellowed The thing that microprocessors and the architectural trends they adopted have really done is they've taken us from a proprietary world where nothing was portable, where once you'd bought a system you were a captive customer - to a marketplace with competitive manufacturing and happier consumers
- protomyth 13y agoLooks like there is at least one emulation project http://retro-b5500.blogspot.com http://retro-b5500.blogspot.com
- ams6110 13y agoThere are still systems descended from this architecture being produced, the Unisys Libra servers, that can still run the MCP OS and programs.
- nitrogen 13y agoInteresting. What industries benefit more from preserving this old architecture than they would from reimplementing for a more mainstream system?
- ams6110 13y agoThe usual suspects: government/military, banking, insurance, transportation. Large legacy systems, source code possibly lost, that need to keep running.
- protomyth 13y agoThere is a lot of software that they don't want to rewrite because it is stable, fast enough, and makes money transactions. That is the general reason a lot of mainframe software is still with us.
- derefr 13y agoA more interesting question to me is, what is the benefit of maintaining the architecture in silicon, rather than as a software emulation/compatibility layer on top of modern commodity hardware (e.g. x86)? Keeps all the lock-in and services contracts, loses the supply-chain costs.
- Someone 13y agoApparently, performance: http://www.theregister.co.uk/2012/06/19/unisys_libra_460_mainframe/ http://www.theregister.co.uk/2012/06/19/unisys_libra_460_mai...: "Unisys has two different families of mainframes in the ClearPath line. From its Burroughs heritage comes the Libra family with their Master Control Program (MCP) operating system, and from the Sperry-Univac side comes the Dorado family with their OS 2200 operating system. At the high-end and upper midrange of the Unisys product line are custom CMOS processors designed by Unisys and manufactured by IBM Microelectronics that have the high single-threaded performance on batch jobs that mainframe customers need. In the midrange, there are smaller CMOS-based mainframes that bear the Dorado or Libra labels as well as a mix of machines that have ports of the MCP and OS 2200 operating systems and special virtualization and emulation sauce that lets CMOS binaries run atop x86 processors. The entry Libra and Dorado products are all based on Xeon machinery."
- girvo 13y agoSomewhat related, but how feasible is it to experiment with building interesting architectures like this using an FPGA? I've been trying to convince myself to get one to fiddle with, and making experimental CPU architectures would be something that would do so :)
- klodolph 13y agoDefinitely feasible. The original was done in discrete logic, so the FPGA makes the process like 100x cheaper and 100x simpler. There's actually a Wikipedia article about doing this, http://en.wikipedia.org/wiki/Soft_microprocessor http://en.wikipedia.org/wiki/Soft_microprocessor Although you might want to start with something simpler, like an 8-bit design so you can get the hang of sequencing before you start adding things like a MMU. It's also easy to simulate FPGAs on a computer, so you could get a simulation running before you go out and start spending your beer money buying Xilinx boards. The simulation will also help you choose a reasonable board, since the big ones are expensive and the small ones might not run your CPU. Or they might.
- girvo 13y agoAn 8-bit design is what I actually want to fiddle with to begin with. I've been looking at OpenCores lately to get a feel for what it's like, and it's quite exciting. I'm thinking of getting a Spartan-3 or -6 board to play with, but yeah running some simulations is a great idea! Thanks for the awesome answer :)
- strlen 13y agoReminds me somewhat of the Minsk- series of indigeneously developed Soviet mainframes -- stack architecture, RISC-like ISA, far ahead of its time. Due to pressure from the higher ups -- or as my parents bluntly put it, "комуняки" (the commies) -- it was abandoned in favour of reverse engineering and cloning the IBM 360 (an enormous feat of engineering) an inferior architecture (my dad called this a setback for Computer Science[1] -- not just in USSR, but in general). Read more here: http://www.computer-museum.ru/english/minsk0.htm http://www.computer-museum.ru/english/minsk0.htm I guess "no one got fired for choosing IBM" held true not only under capitalism, but under communism as well. (In a "small world coincidence", I happened to work with the son of the designer of the Minsk-32 -- the most advanced machine in the series -- at Yahoo. Belarus was one Soviet Union's Silicon Valleys, and incidentally the Belarusian State University Institute's Radio Technical Institute still has an excellent EE/CS program (hint: I think you'll have a lot less competition recruiting there than at Stanford!) [1] Personally I'm a bit skeptical of stack architectures despite their huge popularity (JVM, CLR, Forth, etc...):I understand the cachet, but (to also quote my dad...) "programs can play chess very well", i.e., register windowing is one of the most well understood and mature algorithms, let the compiler handle that job. LLVM's use of a register-based IR is a good step in that direction.
- agumonkey 13y agoFor people interested in Stack computers, Eric LaForest did a thesis on them : http://is.uwaterloo.ca/Eric_LaForest_Thesis.pdf http://is.uwaterloo.ca/Eric_LaForest_Thesis.pdf (from Hamblin RPN to Forth processors)
- strlen 13y agoThanks for sharing this! Very interesting.
- ajuc 13y agoSoviets did it quite often. Production of one of the first minicomputers http://en.wikipedia.org/wiki/K-202 http://en.wikipedia.org/wiki/K-202 designed by Poles in cooperation with British was also canceled despite being very advanced for the time.