Slashdot Mirror


PS3 Linux Now Installable

Quinton writes "Around midnight Pacific time on the 17th, Sony updated their Open Platform website needed to install PPC Linux on the PS3. The FTP Site contains the CELL Linux ADDON CD image, which has the bootloader (kboot/otheros.bld) and instructions needed to install Fedora Core 5, PPC. A full install from DVD takes about two hours. Most all hardware is supported except for graphics accelerator support (framebuffer only, up to 1920x1200)."

14 of 109 comments (clear)

  1. PowerPC? by Constantine+XVI · · Score: 3, Interesting

    Did they say they were running standard PowerPC Linux on the PS3? In theory, what would stop us (besides Apple's legal dept.) putting the PowerPC Mac OSX on it?

    --
    "I think an etch-a-sketch with an ethernet port would beat IE7 in web standards compliance."
  2. Almost by androvsky · · Score: 5, Interesting

    There's a hypervisor running between the kernel and the hardware, so I don't think it's going to be an easy task to hack the nvidia ppc macintosh drivers to run on this thing. I got the impression from the documentation that the accelerator was pretty much locked off, but even if it wasn't, we're pretty much stuck waiting for nvidia to cough up a binary driver blob. Unless someone wants to port opengl to the cell spus. It couldn't be nearly as fast as the nvidia chip, geforce3 territory at best, but it could support any kind of shaders you throw at it.

    1. Re:Almost by Anonymous Coward · · Score: 3, Interesting

      Sony won't boot a non-approved OS so don't hold out hope for a Linux that uses the RSX. Besides, you're not missing much. The RSX is a severely cut down nv40-based chip. As people are just now finding out, it has no scaler and no video acceleration (as in "PureVideo"). It is just a vertex shading pipeline and a very simplified pixel shading pipeline. Sony expected people to do all their pixel effects on the Cell, but that isn't working out too well which is why people are noticing fullscreen effects looking worse on PS3 versions of games like NFS:Carbon. Sigh. It's the PS2 all over again. Lots of raw processing power that is wasted because it is crippled by an underpowered, feature-poor rasterizer. It's like Sony couldn't create a balanced hardware design to save its life.

  3. This is encouraging news. by LogicHoleFlaw · · Score: 2, Interesting

    Honestly, this is pretty cool. I'd love to get a PS3 to play around with the Cell processor. This is really a new thing - far from discouraging homebrew hacking on their system, Sony is making it possible to install a full-blown Linux system on *launch day*. The hardware in the system is all standard stuff so everything should Just Work. I imagine that the community will have the custom graphics processor up and running in no time. If I understand correctly, the seven Cell cores are already supported by Linux due to IBM's desire to have the Cell architecture be used in more than just PS3s. It's just the custom graphics chipset which needs to have drivers written.

    I can understand why Sony hasn't written such drivers - it looks like they've architected the system to be easy to install alternative OSes, but they haven't gone out of their way to write code for a particular 3rd party OS vendor. I look forward to seeing where this all goes.

    --
    -- Flaw
    1. Re:This is encouraging news. by Morphine007 · · Score: 2, Interesting

      funny... I write my ai code in C... game developers usually use whatever they're coding their games in... I still don't see how poor handling of branch prediction* (or no handling of it at all) is going to have any effect on whether or not a coder can code in an ai using a language that completely and utterly abstracts that low level crap away? There's a difference between not being able to easily code something, and not having your code run in quite the optimized way that it used to. The former simply isn't true from anything I've seen (hell... even ASSEMBLER abstracts that extremely low level shit away...) and the latter is a matter that's up for debate; having your code run a bit slower than normal is somewhat bad, but having it be able to run on a dedicated processor and hence much faster than normal more than makes up for it.

      So I'm still calling bullshit on this one so far, but only because no one has given any details... if you know them, please show 'em...

      * - it should be noted that branch prediction is just an OPTIMIZATION technique at the hardware level... code in whatever language you want, it'll still run fine, the processor just won't have already prefetched the instruction and associated memory blocks... if this is all the FUD boils down to, then it's massively busted since my (and anyone elses) ai code will still run perfectly fine on it....

  4. Re:Graphics Chip will never work by LWATCDR · · Score: 3, Interesting

    It's simple, without a graphic chips it impossible to use Linux to play pirated games.

    With out the graphics chip it is impossible to run any good games in Linux.

    If Sony opened up the graphics chip then people could create games without Sony's okay.

    On a bright note it opens up the critter to emulators :)
    MAME PS/3 anyone?
    When it calms down and the price drops I might get one.

    --
    See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
  5. How long for MythTV, MAME etc.? by DrXym · · Score: 3, Interesting

    Install Linux, install Myth, plugin a Haupage WinTV USB device - the PS3 becomes a PVR, plays DiVX etc. Or install MAME, UAE, Virtual Boy, SheepShaver, QEMU, Bochs and you have a pretty decent console / arcade gaming rig. Might even be able to play those SNES / N64 titles before they turn up on the Wii...

    1. Re:How long for MythTV, MAME etc.? by ClamIAm · · Score: 2, Interesting

      Unfortunately, there's no driver for the graphics card. If one never surfaces, just hope and pray that you can somehow wrangle those SPEs into doing all that video decoding.

    2. Re:How long for MythTV, MAME etc.? by DrXym · · Score: 2, Interesting
      I have no idea what the refresh rate & characteristics are the display buffer, but it should possible to offload codecs, zip/bzip2 compression and even parts of Mesa (shaders, transformations, etc.) onto the SPEs. The net result would be a system which while not stellar should be able to put out reasonable 3D and video performance.

      I expect a whole cottage industry will spring up to add functionality to the various libs that does just that.

  6. But OpenGL ES and Collada will (hopefully) by Anonymous Coward · · Score: 1, Interesting

    Graphics Chip will never work

    You're forgetting one rather important little point: Kutaragi (and also one other top Sony guy whose name I forget) both said that they want to encourage amateur games development on the console, and to foster a culture of community development for it.

    That doesn't mean that the graphics chips will be programmable directly of course, but it does mean that the graphics subsystem will be made available at least through an API. Since the PS3 uses OpenGL ES and Collada, it's easy to see how this could be achieved without too much effort.

    Unless of course Kutaragi et al were simply lying in the name of generating buzz.

  7. Gentoo by gentoofu · · Score: 2, Interesting

    I wonder how long it would take to install Gentoo and the stuff on it...

  8. Re:chirp chirp by Anonymous Coward · · Score: 1, Interesting

    Oh Dr Marks, you know as well as I do that Linux on PS3 is just a gimmick. With no GPU access, you're doing CPU rendering of the entire GUI ... using a PPC that doesn't have instruction reordering. Your customers can't do any interesting visualization stuff; people will have to write PS3-specific codec ports (I mean, port the things to the SPU architecture) to decode video on it; and isn't the audio on the GPU too, so no sound? A games/media machine with a version of Linux that can't do games or media. What a coup. Microsoft must be quaking in their boots.

    Of course people will still figure out how to hack it without Sony's support, but it'll probably require a mod-chip. Can you sue people for making mod-chips that merely unlock the GPU under Linux? I suspect not.

  9. Compute Cluster? by bill_mcgonigle · · Score: 2, Interesting

    It still sounds good for a target for a massive compute cluster, ala BigMac. If the algorithm is sufficiently simple and easily parallelized the lack of branch prediction and out of order execution might not be too bad, especially at a FLOP-per-$ or FLOP-per-square-foot or FLOP-per-Watt*$ basis. My uneducated guess is that this would work out OK using gcc 4 (with SSA trees) on things like BlueGene does or even just computing Rainbow tables. Without myrinet or inifiniband you'd need a good ram cache and more CPU than I/O.

    Anybody doing this for a day job care to comment/shred my guesses?

    --
    My God, it's Full of Source!
    OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
  10. Re:Most except... by hr.wien · · Score: 2, Interesting

    The Cell PPE is a dual issue, dual threaded, in-order core, which means it ain't all that. A normal G4 would probably kill it. Yellow Dog Linux on the Cell only runs on the PPE (leveraging the other cores would take some heavy re-engineering of just about every piece of software in it), so I think it's safe to say the PS3 Linux experience won't be anything special either. Fun to play with, sure, but certainly no speed demon.