PS3 Open Platform - some big news

[popper]

But on the other hand, the access to the framebuffer might have a lot higher bandwidth than nearly any PC you can think of? From the vids I've seen so far, I don't get the impression it's anywhere near slow in that regard.

heres a divx video showing it in action, doesnt seem to shabby to me, and thats useing bog standard PPC port , no ALtivec and no spe AFAIK.
http://stage6.divx.com/members/246437/videos/1042281,
OC if someone here were to just take one of the OpenGL codebases and tweek it for the Altivec and one? of the Spe's, it would be so much nicer perhaps.


hell even one of the mobile GL ES might be fun for someone to try and port/patch it on PPC gentoo and Lu_Zero's Cell toolchain see: #gentoo-ppc64.
http://planet.gentoo.org/developers/lu_zero
http://planet.gentoo.org/developers/ranger/2007/01/03/

http://www.khronos.org/
http://www.khronos.org/podcasts/
"Dec 07, 2006
OpenMAX DL: Media Codec Portability
OpenMAX DL (Development Layer) APIs contains a comprehensive set of audio, video and imaging functions that can be optimized by platform developers for new CPUs , hardware engines, and DSPs. Developers can then use the components to create portable accelerated codec functionality such as MPEG-4, H.264, MP3, AAC and JPEG." they mean the old Mpeg4-ASP and the new and better Mpeg4-AVC.

seems perfect to potter about useing the ps3 as your DVB-H(2) server plaything dev system after you get it working there.

plus dont forget (i have to keep mentioning this) currently no ones bothered to patch up the PPC linux to take advantage of Altivec in any shape or form.

theres that libfreevec Markos could do with help filling out etc, and Lu_zero had done a great job for your PPC linux Video apps using Altivec but they are few and you are many, go help them and help the users in return, theres even a new MOL (JoseJX) website being worked on to bring that project back into the light of new PPC/CELL devs that you could help on if your willing and able http://mac-on-linux.sourceforge.net/
again see: #gentoo-ppc64 and the http://www.powerdeveloper.org/forums/index.php site /altivec section.

 

[SPM]

Two more PS3 Linux How Tos

PS3 Ubuntu full install howto
http://www.louiscandell.com/ps3/

PS3 Gentoo full install howto
http://ps3.qj.net/Gentoo-Linux-on-your-PS3-With-full-install-instructions-/pg/49/aid/78739
(Gentoo allows you to compile the entire OS for a different CPU optimisations if you are that way inclined).

 

[Titanio]

RapidMind offers evaluation dev tools for PS3 Linux

This could be interesting to play around with..not sure what the limitations are on the evaluation version:

http://home.businesswire.com/portal...d=news_view&newsId=20070111005836&newsLang=en

RapidMind and Terra Soft Partner to Unleash PlayStation®3 for Linux Developers

WATERLOO, Ontario--(BUSINESS WIRE)--RapidMind and Terra Soft announced today that they have teamed up to make application development for the PlayStation®3 easier than ever. As part of an agreement with Sony Computer Entertainment Incorporated, Terra Soft Solutions announced last month the release of Yellow Dog Linux for the PlayStation 3. With the RapidMind Development Platform, developers can easily create applications that will run on the PlayStation 3, as well as other hardware solutions based on the Cell Broadband Engine™ (Cell BE).

RapidMind will be providing a full-featured evaluation version of the RapidMind Development Platform v2.0 for distribution from Terra Soft's YDL.net Enhanced accounts. “It’s exciting to imagine what will happen when you put a supercomputer-class PlayStation 3 with a great OS into the hands of developers,” said Ray DePaul, RapidMind President and CEO. “Developers are eager to take advantage of the Cell BE and the PlayStation 3 without the need to understand the underlying architecture of this complex multi-core processor. The RapidMind platform makes that possible.”

PlayStation 3 is the latest interactive entertainment console from Sony, designed for a cutting-edge gaming experience. The PlayStation 3 incorporates the Cell BE processor which was created by IBM, Sony, and Toshiba for more than gaming. The Cell BE is already available from IBM in the form of a blade for the IBM® BladeCenter® Server as well as from Mercury Computer Systems in several form factors.

“The PlayStation 3 places a supercomputer in the home,” said Kai Staats, Terra Soft CEO. “Yellow Dog Linux provides a complete Linux OS for the PlayStation 3 resulting in a very powerful computing platform. We are thrilled to be working with RapidMind to make this platform more accessible for professional developers and hobbyists alike.” Developer’s now have an excellent development platform to develop applications for the Cell BE, whether for deployment on the PlayStation 3 or as a jumping off point for deployments to other Cell BE based products, including those from IBM, Sony, or Mercury.

RapidMind will also be participating in a Hack-a-thon (www.hpc-consortium.net) from January 20th to the 26th for the optimization of applications for the Cell BE, hosted by Terra Soft at their Loveland, Colorado headquarters. RapidMind, along with organizations such as Argonne, Oak Ridge, Colorado State, IBM, Mercury, and Tungsten Graphics have committed resources with dozens of other labs, universities, and individuals planning to participate. The intent of this event is to create a set of application examples to showcase the potential of the Cell BE.

You may recall that they did that 'thousands of sheep' Cell flocking demo some time ago, and IBM has been promoting them fairly heavily w.r.t. Cell development. Their site is here:

http://www.rapidmind.net/

Some whitepapers:

Writing Applications for the Cell BE Using RapidMind
This document provides a description of how the RapidMind Development Platform makes programming the Cell Broadband Engine accessible and effective.

http://media.rapidmind.net/wp/cell.pdf

Cell BE Porting and Tuning with RapidMind: A Case Study
This document shows how an application was ported to and tuned for the Cell Broadband Engine processor with RapidMind, achieving impressive speed results.

http://www.rapidmind.net/cellbe-porting-qjulia

Crowd Simulation: Cell BE Development Made Simple

Developing a simulation of many thousand characters, each with a mind of their own, is usually a complex task. It requires a large amount of computing power and normally involves a serious effort from the developer in order to achieve reasonable performance. Utilizing the RapidMind Development Platform and the Cell Broadband Engine (Cell BE), achieving excellent performance of this simulation is made simple.

http://media.rapidmind.net/samples/crowds.pdf

Real-time Ray Tracing

RTT AG provides a high-performance, real-time ray tracer, RealTrace, as part of DeltaGen, their software for complex 3D visualization. A leading global provider of visualization in the automotive, aerospace and consumer goods sector, RTT created the ray tracer using the RapidMind Development Platform running on graphics processing units (GPUs). This application was then demonstrated at SIGGRAPH 2006 on the Cell Broadband Engine, achieving excellent performance.

http://media.rapidmind.net/samples/RapidMind_Raytracer_Demo.pdf

 

[Panajev2001a]

I am getting a little bit less excited about homebrew gaming development compared to XNA on Xbox 360 because of a thought that has been creeping up in my mind... the option of having either MesaGL running unoptimized on the PPE or doing Software rendering (T&L, Shader programs, Textyring, Culling, Clipping, Scanline rasterization, etc... all by hand...) on one or more SPE's...

You are getting way better CPU access and way more available CPU power with PLAYSTATION 3 Linux though, but boy having a total of less than 200 MB of RAM available and Software rendering are starting to get on my nerves...

Yes, I am lazy, I know .

 

[Arwin]

I partly agree. Sony should really allow at the very least RSX OpenGL drivers to be released, and preferably just allow full RSX access under Linux.

I do find the Cell extremely interesting, but it would be much nicer if the RSX could be fed Cell manipulated vertices rather than that the Cell has to take care of those things itself.

 

[macabre]

IBM is asking for feedback from PS3 users for better support:

http://www-128.ibm.com/developerworks/forums/dw_thread.jsp?forum=739&thread=147708&cat=46

 

[Shifty Geezer]

What does RapidMind offer, and how is the free evaluation version they're offering limited? Can you write and publish with it for, say, freeware, but sellable products need a proper license for RapidMind?

@ Panajev : I've never thought of PS3 homebrew as suited for games. I still say by far the best homebrew game development platform is PC because of the wide range of different engines, tools, etc. that elliminate a lot of the hard grind of creating the frameworks. XB360 offers a nice advantage over PC with standardised hardware, but it's far more complicated for what's needed for most simple homebrew efforts. For me, homebrew on PS3 is (or rather, would be if I had one ) working on algorithms and processes, and media apps etc. But as I mentioned in some other thread I think, what PS3 needs is a basic rasterizer for people to work with. If someone could provide a...2 SPE renderer and graphics library, that'd be a lot folks wouldn't have to worry about. As for the RAM limit, is 200 MB really a limit for homebrew game development? You must have some pretty big aspirations if you're looking to fill up >200 MB!

 

[archangelmorph]

What does RapidMind offer, and how is the free evaluation version they're offering limited? Can you write and publish with it for, say, freeware, but sellable products need a proper license for RapidMind?

@ Panajev : I've never thought of PS3 homebrew as suited for games. I still say by far the best homebrew game development platform is PC because of the wide range of different engines, tools, etc. that elliminate a lot of the hard grind of creating the frameworks. XB360 offers a nice advantage over PC with standardised hardware, but it's far more complicated for what's needed for most simple homebrew efforts. For me, homebrew on PS3 is (or rather, would be if I had one ) working on algorithms and processes, and media apps etc. But as I mentioned in some other thread I think, what PS3 needs is a basic rasterizer for people to work with. If someone could provide a...2 SPE renderer and graphics library, that'd be a lot folks wouldn't have to worry about. As for the RAM limit, is 200 MB really a limit for homebrew game development? You must have some pretty big aspirations if you're looking to fill up >200 MB!

That's a good point!

Especially considering the RAM size/speed of the PS2 and you only have to look at God of War to see what's capable from the hands of very clever people on restricted hardware.. Plus the PS2 never had a dedicated GPU so no access to the RSX via linux shouldn't be too big of a problem provided Cell is still wide open..

 

[Panajev2001a]

That's a good point!

Especially considering the RAM size/speed of the PS2 and you only have to look at God of War to see what's capable from the hands of very clever people on restricted hardware..

Look, it is simple really: if I wanted to develop on PlayStation 2 Linux with all the limitations that such an environment brings forward... I can just keep doing that.

I know that is fun and a good testament of your skills if you can achieve AWESOME things doing TONS of clever work yourself.

I like programming, but I would like not to have to concentrate on EVERY aspect of the equation at once by myself: I do not work in a 5 people team not even in a two people team...

If asking to have even a simple rasterizer that took care of taking my triangles and drawing them on screen (without asking me to provide the code for the scan conversion of triangles, lines, etc... as well as manually texture mapping each surface) is too much... well we have to agree to disagree.

Plus the PS2 never had a dedicated GPU so no access to the RSX via linux shouldn't be too big of a problem provided Cell is still wide open..

On PlayStation 2 Linux you HAD to-the-metal access to a as dedicate as you can think of GPU which was actually BLAZINGLY fast at what it did. Sure mip-mapping is borked and it does not perform clipping for you...

People, really... besides the I/O CPU and the SPU2 (the system's sound processor) you had UBER-LOW-LEVEL access to EVERYTHING about the CPU and the GPU of PlayStation 2 and a detailed documentation of both chips.

I know, I am lazy, but one reason why I liked a LOT the idea of PLAYSTATION 3 Linux was because I thought (amongst the other things):

1.) Finally, Vector Units with a unified register file, proper (well, IMHO it can be considered at least better) integer/boolean math support and scalars support.

2.) Finally, the Vector Units can do DMA's on their own (DMAC programming on PlayStation 2 was kinda fun I have to say even if painful when you start... I felt like drowning in the beginning :lol).

3.) Finally, the DMA engines "understand" Virtual addresses and are not restricted to only Physical addresses (it avoids some of the pain involved with DMA transfers specifically on PlayStation 2 Linux... you would end up doing the same on commercial gains, but as a performance optimizations, but having the limitation of the libraries being able to allocate a chunk of max of 4 KB of physically contiguous and unswappable memory [allocated memory was Virtually Contiguous, but EE's DMAC does not "understand "Virtual addresses] was a pain when starting DMAC coding although solving it is not too hard, once you know how to do it and nice people give you good hints).

4.) Finally, a fully featured GPU that handles a lot of the boring work for you (culling, clipping, mip-mapping, etc...).

Granted that besides point 4.) much of the excitement remains, but hey... think about THIS:

All the cycles you spend doing software rendering on the SPE's are less cycles to be used for other things (whatever those other things might be for you).

 

[SPM]

Look, it is simple really: if I wanted to develop on PlayStation 2 Linux with all the limitations that such an environment brings forward... I can just keep doing that.

I know that is fun and a good testament of your skills if you can achieve AWESOME things doing TONS of clever work yourself.

I like programming, but I would like not to have to concentrate on EVERY aspect of the equation at once by myself: I do not work in a 5 people team not even in a two people team...

If asking to have even a simple rasterizer that took care of taking my triangles and drawing them on screen (without asking me to provide the code for the scan conversion of triangles, lines, etc... as well as manually texture mapping each surface) is too much... well we have to agree to disagree.



On PlayStation 2 Linux you HAD to-the-metal access to a as dedicate as you can think of GPU which was actually BLAZINGLY fast at what it did. Sure mip-mapping is borked and it does not perform clipping for you...

People, really... besides the I/O CPU and the SPU2 (the system's sound processor) you had UBER-LOW-LEVEL access to EVERYTHING about the CPU and the GPU of PlayStation 2 and a detailed documentation of both chips.

I know, I am lazy, but one reason why I liked a LOT the idea of PLAYSTATION 3 Linux was because I thought (amongst the other things):

1.) Finally, Vector Units with a unified register file, proper (well, IMHO it can be considered at least better) integer/boolean math support and scalars support.

2.) Finally, the Vector Units can do DMA's on their own (DMAC programming on PlayStation 2 was kinda fun I have to say even if painful when you start... I felt like drowning in the beginning :lol).

3.) Finally, the DMA engines "understand" Virtual addresses and are not restricted to only Physical addresses (it avoids some of the pain involved with DMA transfers specifically on PlayStation 2 Linux... you would end up doing the same on commercial gains, but as a performance optimizations, but having the limitation of the libraries being able to allocate a chunk of max of 4 KB of physically contiguous and unswappable memory [allocated memory was Virtually Contiguous, but EE's DMAC does not "understand "Virtual addresses] was a pain when starting DMAC coding although solving it is not too hard, once you know how to do it and nice people give you good hints).

4.) Finally, a fully featured GPU that handles a lot of the boring work for you (culling, clipping, mip-mapping, etc...).

Granted that besides point 4.) much of the excitement remains, but hey... think about THIS:

All the cycles you spend doing software rendering on the SPE's are less cycles to be used for other things (whatever those other things might be for you).

The question really is will Sony consider that allowing full RSX access to Linux (which it can easily do if it chooses to do so) will allow PS3 games to be sold bypassing Sony. Microsoft doesn't allow access to boot other OSes off the Xbox 360 either for the same reason, or any other console manufacturer. I think Sony can reasonably be asked to do two things - 1) allow Linux to access an accelerated OpenGL library which may itself be protected by the DRM. I don't care if Sony can slows down all OpenGL calls not used by Compiz/Beryl/and X to discourage game use, so long as all Linux desktop application graphics, sound and movies run at full speed. 2) match Microsoft's home brew dev kit offerings (without Linux) for full access to RSX and Cell. This homebrew dev kit could boot into a rather more closed environment than provided for Linux to keep Sony happy.

 

[Fafalada]

will allow PS3 games to be sold bypassing Sony

To sell games to 5% of userbase? And if we're to believe local XBox/Nintendo brigades that userbase will already be tiny to begin with.

Linux isn't even a viable game market on PC, I fail to see how people see that becoming on a console where chances are majority of users won't even know what "OS" means, let alone install it.

I think Sony can reasonably be asked to do two things

I still maintain that "the scene" on other products, has already made sure to dissuade Sony from wanting to open things on their platforms further. Unless you choose to believe conspiracy theorists and people like DA have been fed inside info by Sony on purpose (I have no doubt that inside info has been exchanging hands anyhow).

 

[Fox5]

To sell games to 5% of userbase? And if we're to believe local XBox/Nintendo brigades that userbase will already be tiny to begin with.

Linux isn't even a viable game market on PC, I fail to see how people see that becoming on a console where chances are majority of users won't even know what "OS" means, let alone install it.


I still maintain that "the scene" on other products, has already made sure to dissuade Sony from wanting to open things on their platforms further. Unless you choose to believe conspiracy theorists and people like DA have been fed inside info by Sony on purpose (I have no doubt that inside info has been exchanging hands anyhow).

Theoretically, could a game be made that includes its own linux os that would boot up, something like a live cd?

 

[Shifty Geezer]

If asking to have even a simple rasterizer that took care of taking my triangles and drawing them on screen (without asking me to provide the code for the scan conversion of triangles, lines, etc... as well as manually texture mapping each surface) is too much... well we have to agree to disagree.

archangelmorph was talking about the RAM limits, not the extra development efforts. "Why is 200MB a limit to creating a PS3 Linux game when games like GOW fit into 32 MB?" is the argument here.

Writing a whole rasterizer etc. is a point archangelmorph didn't talk about.

All the cycles you spend doing software rendering on the SPE's are less cycles to be used for other things
(whatever those other things might be for you).

For sure, but again I ask what are homebrew devs doing that's going to miss a couple of SPEs? If you think of the Cell BBE as a six-placement dinner, if you have a dev team of 12, they're going to run out of food and be wanting more. But a lone developer has 6 meals all to himself. Isn't that enough?! Myself, I can't say I look at Cell and instantly see limits and restrictions in what can be done. I see loads of potential that'd take a lot to fill up before I'd ever start lamenting the absence of more power and RAM. Comparing it to XNA, I can't obviously see any development the majority homebrew developer can create that'll be hampered by 200 MB RAM versus XB360's 500 MB. XB360 could certainly benefit from things like normal mapping that a Cell rasterizer won't be so good at. PS3 homebrew games will probably need to be more simple in graphics. That doesn't hamper what you can do gamewise though. I dunno how I'd go about filling 200 MB of RAM with game content, especially when creating all that content myself!

 

[Panajev2001a]

archangelmorph was talking about the RAM limits, not the extra development efforts. "Why is 200MB a limit to creating a PS3 Linux game when games like GOW fit into 32 MB?" is the argument here.

Writing a whole rasterizer etc. is a point archangelmorph didn't talk about.

The kind of optimization to keep all the data in RAM in very compact form and the technology to efficiently stream and de-compress the data from the DVD drive is NOT trivial.

Also, for development purposes you might want a lot of more RAM than the memory pool you are targeting: PlayStation 2's TOOL has like what... 4x the available main RAM (128 MB vs 32 MB of the retail version) to better debug your code?

For sure, but again I ask what are homebrew devs doing that's going to miss a couple of SPEs? If you think of the Cell BBE as a six-placement dinner, if you have a dev team of 12, they're going to run out of food and be wanting more. But a lone developer has 6 meals all to himself. Isn't that enough?! Myself, I can't say I look at Cell and instantly see limits and restrictions in what can be done. I see loads of potential that'd take a lot to fill up before I'd ever start lamenting the absence of more power and RAM. Comparing it to XNA, I can't obviously see any development the majority homebrew developer can create that'll be hampered by 200 MB RAM versus XB360's 500 MB. XB360 could certainly benefit from things like normal mapping that a Cell rasterizer won't be so good at. PS3 homebrew games will probably need to be more simple in graphics. That doesn't hamper what you can do gamewise though. I dunno how I'd go about filling 200 MB of RAM with game content, especially when creating all that content myself!

It is very easy to waste RAM, very easy if you are not careful :lol. You underestimate how hungry a single person can be .

GPU wise the XNA environment blows away what you can achieve currently with PLAYSTATION 3 Linux, there is no point of debating it IMHO.

The situation is not an obstacle impossible to overcome, but it all also depends on what kind of graphics you do want to push: if you want tri-linear filtering or at least mip-mapped bi-linear filtering, several texture layers, anti-aliasing, normal and/or parallax mapping and per-pixel lighting, shadow mapping, etc... it is easy to burn LOTS of SPE cycles running the graphics code in software.

I do not think that all homebrew game developers are like me, working alone or with maybe few colleagues, there are going to be many other developers that are not going to like having to waste many SPE's doing purely graphics related code.

 

[assen]

Linux on the PS3, with the fast SPUs and the disabled RSX, are a once-in-a-lifetime chance of raytracing proponents to put up or shut up

 

[Shifty Geezer]

And in addition, it's a great incentive to develop alternative renderers too. eg. A 2D renderer that renders AA'd vectors a la Flash for impeccable quality, perhaps from 3D source objects, with depth based effects like blur or fog (used in flOw). A pure ray-tracer would be ideal. A simple Snooker/Pool game would be a great image-quality centred target game, throwing everything at tracing photorealistic balls. And how's about a Voxel renderer using a lot of procedurally generated content to avoid the large memory requirements? I dare say it's such experimentation that Sony are really keen to take off, creating engines and techniques rather than homebrew games. This is where the demo scene ought to be picked up. Sony should throw some big prizes and publicity at it to attract more input, while stepping back enough to keep it independent and not tread on toes. Leave the homebrew games to other platforms and PSN, and promote PS3 Linux as the creative centre of coders who want to stretch their algorithms more than their game design.

 

[Arwin]

Sure, but at the same time, we can do that stuff anyway. No need to withhold all the extra GDDR3 RAM from us that we could use for double buffers, mixed render targets, texture storage and what not. I'm still hoping they will open it up. They didn't open up everything on PS2 Linux right away either. In this case, they may want to keep the others guessing for a littel while longer, or maybe they are still testing security, or maybe even they have an agreement with Nvidia that prevents them from opening up stuff outside of NDAs at this point in time.

 

[one]

If they release RSX and related user development it'll be on GameOS/CellOS. But apparently the security is their concern. I guess a software VM like that of the XNA development is one of the solutions, but to offer such a thing they have to ponder what they really should do because offering a half-baked middleware and its clumsy scripting language is pointless for them. They must be thinking what the best option is to maximize the opportunity within their limited development resources. For Microsoft, pushing .NET and managed languages across platforms is a well thought-out plan and the big established cause. SCE needs something like that IMO.

If SCE offers a new software development environment for an end-user I hope it's a new language or an abstraction layer that enables people to write a parallel/distributed code for Cell/RSX easily.

 

[popper]

And in addition, it's a great incentive to develop alternative renderers too. eg. A 2D renderer that renders AA'd vectors a la Flash for impeccable quality, perhaps from 3D source objects, with depth based effects like blur or fog (used in flOw). A pure ray-tracer would be ideal. A simple Snooker/Pool game would be a great image-quality centred target game, throwing everything at tracing photorealistic balls. And how's about a Voxel renderer using a lot of procedurally generated content to avoid the large memory requirements? I dare say it's such experimentation that Sony are really keen to take off, creating engines and techniques rather than homebrew games. This is where the demo scene ought to be picked up. Sony should throw some big prizes and publicity at it to attract more input, while stepping back enough to keep it independent and not tread on toes. Leave the homebrew games to other platforms and PSN, and promote PS3 Linux as the creative centre of coders who want to stretch their algorithms more than their game design.

if their going to go do all that though Shifty, it might also be nice to once again see some updated spe hardware assisted amiga style parallel/parallax scrolling and Blitter that can once again do amiga screens/dragging , you can explain that to them windows devs better than i shifty LOL.


were are all the old Amiga Demo coders these days?, perhaps they can put a few days in on this spe Gfx problem and show these new PC devs how it can be done with 256k of code....

on a side note:
i just noticed that the The Interchange File Format (IFF) is listed on the http://www-128.ibm.com/developerworks/power/library/pa-spec16/ page and its dated 13 Jun 2006.

it seems its coming back into fashion again and good for them.

 

[SPM]

To sell games to 5% of userbase? And if we're to believe local XBox/Nintendo brigades that userbase will already be tiny to begin with.

Linux isn't even a viable game market on PC, I fail to see how people see that becoming on a console where chances are majority of users won't even know what "OS" means, let alone install it.

The PC would be a difficult base to develop for in this way because the hardware varies, but since every PS3 would be able to run Linux, Linux based games would be a viable alternative on every PS3 if Sony permitted full access. The Linux kernel of course takes up some space, but is is very flexible and can be and is trimmed down to a minimum to use in mobile phones and other embedded applications for example. In addition the loader that boots Linux on the PS3, can be used to load any code, not just Linux, and that includes games - the PS3 can't tell whether it is Linux or another OS or a commercial game that you are loading up using the "other OS" loader.

I still maintain that "the scene" on other products, has already made sure to dissuade Sony from wanting to open things on their platforms further. Unless you choose to believe conspiracy theorists and people like DA have been fed inside info by Sony on purpose (I have no doubt that inside info has been exchanging hands anyhow).

I can't understand what you are saying here, but as for myself, I can understand why Sony would be reluctant to allow full RSX access for the other OS loader environment, and I am pragmatic enough to accept it. Personally I am quite happy to run PS3 Linux as a personal computer and use the PS3 to run official Sony games. What I am saying is that general OS and games use graphics differently, and please Sony can you make the OpenGL API bits that are used specifically by the X-server, Compiz/Beryl, and sound and video player acceleration available to Linux even if the other parts of the OpenGL API is slowed down to discourage cutting edge commercial games from using the other OS loader. I think Sony should also provide a separate homebrew dev kit for those who want to get into RSX+Cell game development. This doesn't need to have any restrictions if the games cannot run without the devkit and Sony charges say $100 for the devkit instead of allowing free distribution like Linux.

What it boils down to is that as a development platform, PS3 Linux is not a suitable platform for developing console games. It is a good platform for learning Cell programming, developing Linux applications for the PS3 for which there will definitely be a commercial market, and maybe PC desktop type games like the SIMs and various non graphics intensive online games (for which there is also a definite commercial market), but for the typical graphically intensive console game, it certainly isn't suitable. For that Sony needs to release a specific homebrew games devkit which allows full access to RSX and has a similar environment and some of the dev tools that come with the commercial devkit.