fucking low-level computers

Discussion in 'OT Technology' started by piratepenguin, Sep 6, 2007.

  1. piratepenguin

    piratepenguin New Member

    Joined:
    Jun 18, 2006
    Messages:
    1,067
    Likes Received:
    0
    Location:
    Ireland
    I've had this idea for a long time and I'm telling you all now because I'm wondering about it again and I'm expecting one of these years to hear about it being executed, and to probably be very successful. Probably by Sun because they have Java and their CPUs. IIRC I haven't told of this before on this board.

    Why, why are computers so low-level in this day and age? Why aren't we running Python programs on Python machines instead of a Python interpretor on an Intel machine, for example?

    If I was to take this idea into my own hands, I'd actually be more interested in creating a Mozilla machine, or a web machine. It's the same concept though. (Anyways there are hardware XML parsers (mostly for embedded machines, I'm sure), and all the other hardware work would likely be useful for releasing more modular chips)

    I realize that it would add probably another dimension of complexity to designing the machine. But think about all the work that would be so much simpler - no need to write a billion line operating system - in a Mozilla machine, no need to write a web-browser (Mozilla's codebase is one of the biggest of any software project ever, but that includes Thunderbird etc), email client, etc, in a Python machine, these things will be that much simpler to do anyways. How many tens or hundred of thousands of web programmers do we have?

    I'm not so sure about speed, should be at least as fast as web/Python apps on a not-so-old Intel machine, though I don't think any of us can be sure. If you can give an educated opinion, please do and explain.

    Security would be far more critical, needing to be far better tested before shipping as hardware can't be updated as simply. Penetrating the market would obviously be difficult, and it would cost SO MUCH at least at the beginning, but I'd say it'd be a much better of a way of working. Probably worth research for electronics students (which I'm thinking of going into).
     
  2. skinjob

    skinjob Active Member

    Joined:
    Jan 6, 2001
    Messages:
    2,337
    Likes Received:
    0
    Location:
    Aztlán
    So, instead of billions of lines of code written for a general purpose machine, there will be billions of lines of hardware description langauge to create a specialized piece of hardware? I don't think there is any major difference in complexity in implementing either case.

    Or, are you saying we should stop solving the same problem over and over again, and use one solution set in stone (silicon)?
     
  3. piratepenguin

    piratepenguin New Member

    Joined:
    Jun 18, 2006
    Messages:
    1,067
    Likes Received:
    0
    Location:
    Ireland
    I think it's certainly the right idea to be looking at it and heading that direction.

    Like if I was to build a Mozilla machine, I'd be trying to get out a hardware XML parser, XSLT processor, Javascript interpreter, etc.... And these would have uses today in themselves.

    I actually came across a piece of hardware for optimizing Java not that long ago, done as a research project. Could've been an almost complete Java VM, can't find it right now.
     
  4. EvilSS

    EvilSS New Member

    Joined:
    Jun 11, 2003
    Messages:
    5,104
    Likes Received:
    0
    Location:
    STL
    You do realize that you would exponentially increase the cost of a computer while at the same time reducing it's versatility and creating massive lock-in.
     
  5. piratepenguin

    piratepenguin New Member

    Joined:
    Jun 18, 2006
    Messages:
    1,067
    Likes Received:
    0
    Location:
    Ireland
    Massive lock-in, nope. Python and web apps would still run as they do in their environments ontop of intel computers. That's the beauty of it.
     
    Last edited: Sep 6, 2007
  6. ady

    ady New Member

    Joined:
    Mar 3, 2005
    Messages:
    10,959
    Likes Received:
    0
    Location:
    UK
    Because it is highly unlikely it would be financially viable for any company to do so.
     
  7. piratepenguin

    piratepenguin New Member

    Joined:
    Jun 18, 2006
    Messages:
    1,067
    Likes Received:
    0
    Location:
    Ireland
    You're right ady, but what about academics researching it? Might be something I'll look into a lot more.
     
  8. deusexaethera

    deusexaethera OT Supporter

    Joined:
    Jan 27, 2005
    Messages:
    19,712
    Likes Received:
    0
    90% of the computers in the world are embedded machines with specialized hardware, software, and command sets. And the ones that are general purpose, or "low-level" as you call it, are that way because it makes them infinitely more flexible. Not that it isn't fun trying to get my microwave's countdown timer to run Halo, of course, but it's not for most people.

    No matter how you work it, you'll have to translate the language into binary code so it can be executed mathematically. It doesn't matter if there's a hardware interpreter; the hardware interpreter will run much faster than a software interpreter, to be sure, but it will all get downconverted into assembly code before it will get executed, anyway. It's just not practical to make a microchip that can speak English as its native language.

    There are many ways to respond to your opinion, but I'm going to go with "tens of thousands of people who get paid a lot of money for decades at a time have arrived at the solution of general-purpose machines running special-purpose emulators, because it's the best practical solution to the problem".

    - - -

    EDIT: Now, if you wanted to make a PCI card with a hardware Python/XML/Java/VBScript/etc interpreter, that would be a brilliant idea, so long as the command set is flash-upgradeable.
     
  9. EvilSS

    EvilSS New Member

    Joined:
    Jun 11, 2003
    Messages:
    5,104
    Likes Received:
    0
    Location:
    STL
    Yes, but you would be locked in on that machine to THAT vendor's version of PYTHON. You would be locked in to a vendor and a language. That's one of the reason the java accelerators are not all that popular. You also severely limit upgrade-ability since you can't flash this type of processor with new firmware (or microcode in this case), so as the language evolved the hardware becomes more and more out of date.

    What you are talking about are not computers so much as appliance devices. Which are great, if you have a specific need for them. Hardware SSL is popular with large websites for example. But trying to take that same concept and make it global across computing just won't work. The reason a PC is so useful is because it's generalized. You can run any code compiled in any language that is compatible with the processor. And processors really don't change all that often. Look how long the x86 instruction sets have been around.

    It's just not a model that expands to a "global" (as in wide reaching, not globe) scale well.
     
  10. copiertalk

    copiertalk Secure Our Borders! OT Supporter

    Joined:
    Sep 21, 2003
    Messages:
    5,695
    Likes Received:
    0
    Location:
    west texas
    What about the cross platform. You the code interpreted so it can tun in different environments.

    on a windows machine, the compiled code would run fine but with the added cost of changing the code to then run on macos, unix, linux, centos.. it is much more cost effective to have an interpreter relay to the end user than to build a compiler and change the code for each operating system.
     
  11. EvilSS

    EvilSS New Member

    Joined:
    Jun 11, 2003
    Messages:
    5,104
    Likes Received:
    0
    Location:
    STL
    Not really. With that type of solution you just shift the compiler functions from software to hardware and put the burden of dealing with the OS on the drivers for the device. Either way, you are putting the same thing in and getting the same thing out. Only with the hardware solution you have to tack on fab and component costs.
     
  12. piratepenguin

    piratepenguin New Member

    Joined:
    Jun 18, 2006
    Messages:
    1,067
    Likes Received:
    0
    Location:
    Ireland
    Thanks for your points guys!
    Yeah. Would need to start with a whole load of research. Modern CPUs can't be easy to beat with too simple of specialized circuitry.

    Well, if I do go the electronics route, I'll be checking this out. I still think it'd be worth that.
     
  13. EvilSS

    EvilSS New Member

    Joined:
    Jun 11, 2003
    Messages:
    5,104
    Likes Received:
    0
    Location:
    STL
    It is worth looking into, just for the right application. These types of devices are very useful in appliances or dedicated devices where you don't want to spend the money or use a lot or energy on a full blown processor. Hardware DivX, for example, never took off on the PC (Plextor tried it, but CPU's outpaced it) but it is used in a number of dedicated devices such as stand alone video recorders and home theater components. J2ME accelerators are marketed to cell phone manufactures. and on and on.
     
  14. copiertalk

    copiertalk Secure Our Borders! OT Supporter

    Joined:
    Sep 21, 2003
    Messages:
    5,695
    Likes Received:
    0
    Location:
    west texas

    Alot of copiers use vmx? vmworks? I will have to look over my notes from my last class to find the operating system. There are several appliances that use it. Refridgerators, microwaves, copiers, fax machines. I will have to look at my notes.
     
  15. CodeX

    CodeX Guest

    Anything implemented in software can be implemented in hardware with better performance. There is one reason is doesn't happen:

    COST
     
  16. Penguin Man

    Penguin Man Protect Your Digital Liberties

    Joined:
    Apr 27, 2002
    Messages:
    21,696
    Likes Received:
    0
    Location:
    Edmonton, AB
    But Python could never change. You can't reasonably expect the programming language of choice to stay the same for any length of time. These days we see a new language fad almost every year, whereas instruction sets gain and lose popularity much more slowly (x86 has dominated desktops for, what, 15 years?). Also, different developers use different languages, and for good reason: different languages are good for different things.

    In any case, this type of things has been done before. LISP machines, anyone?
     

Share This Page