Homebrew

idsn6

Regular
After posting in the RSX access thread, I had some thoughts that I decided to collect.

I really should open this with the disclaimer that I haven't coded for the Cell BE or the PS3. Instead of apologizing for this personal deficiency on the Cell programming board, I'd like to explain why it's so.

Like everyone else here, I was excited about the potential for Linux on the PS3. There was quite a lot of sound and fury about an astoundingly powerful open platform-within-a-platform that would foster a homebrew revolution under Sony's latent laissez-faire stewardship.

There's laissez-faire and then there's simply NOT CARING. Sony actually managed to go beyond even abandonment and only seems to step in when it looks like Linux programmers are starting to make progress in getting around their limitations, and then only in order to topple them down again. I have never seen a company so outrightly hostile toward its own users, especially for a touted feature. Every rare Sony-branded action of note since launch on this front has been to treat its PS3 Linux coders as the enemy; with this kind of Sisyphean relationship to look forward to, is it really any wonder that the platform is stagnant?

Contrasting with this anti-support, there was the significant initial push for Cell development from IBM (whose contributions here were incredible) et al., but even that has dried up now. All that's left behind is a monotonous trickle of H.264 decoders and academic-interest-only physics simulations or number crunchers. Headline-worthy achievements have been undertaken with clusters of PS3s (pictures of which always churn my stomach at the sheer amount of superfluous, totally wasted hardware in each racked box), but these have absolutely zero practical impact on the home user with a scant single console. The unvaried, unsurprising, and frankly nigh-uninteresting nature of these reports has only cemented the perception of Cell as an accelerator of mundanities and a poor performer at everything else.

Consider this: PS2 programming seems to me, even all these years later, attractively arcane. Using the Linux kit, you could drill down to the metal and play with a unique architecture at a level few others ever see. I don't have any clearly explainable rational basis at this moment for the following statement, but it is my honest gut reaction: the idea of porting a piece of software to run well on the PS2 seems like a CHALLENGE. Doing the same for the PS3 seems like a CHORE. It's almost like a variation on the uncanny valley theory (which is personally ironic): the further a platform gets from a highly specialized widget and the closer it approaches a generalized PC in terms of capabilities, the more frustratingly blatant the arbitrary limitations become. For instance, I have my own windowshopped issues with the XNA platform, but you have to give them this: at least Microsoft granted you access to the DAMNED GRAPHICS.

It's like the Amiga never existed, and the lessons gleaned from it and its community never learned. VISUALS MATTER. Visuals matter, and for fuck's sake they matter a hell of a lot more on a device whose raison d'être IS visuals. I've expressed before how utterly unappealing it is to me to either write or use an SPU-based renderer when an infinitely more capable processor is actually present, but locked away. It's ridiculous. It's like watching a recent episode of the Simpsons: an activity that leaves you feeling robbed of a portion of your life with absolutely nothing gained in return. Without an immediately viscerally appealing aspect to programming, the platform is simply poisoned to a large number of talented hobbyist coders who have better and more interesting things to do with their time. Without this central kernel to grow around, the platform never gathers enough mass to draw in many of those who could contribute to other aspects of the experience. Sony has now left the PS3 in the homebrew doldrums for so long that I can't imagine it ever escaping.

This is probably a predictable point to bring up these days, but just look at the meteoric success of the iPhone SDK and App Store. Granted, the comparison isn't perfect in the details of each program; the iPhone SDK is probably more closely analogous to XNA than full-blown OS support, but the broad goal of all of these endeavors is the exact same: to draw upon the strength of the community to provide value to the package. Both the iPhone and Cell have free, high-quality SDKs available for OSX and Fedora respectively. The devices even launched at the same price point. So why is PS3 Linux languishing more than two years later while the App Store is celebrating 500 million downloaded applications six months after launch? Why, when I decided to code recreationally, did I buy an iPod touch with a developer membership and not even consider the PS3 until this very post? Maybe these are fallacious questions. I certainly haven't pondered the details much yet, and it can be said that the aims of the companies are different from each other and orthogonal to my answers. But I will say that the fact that the iPhone exposes its GPU, even one as underpowered as its own, factored into my interest. A Cell-based software renderer would run circles around this MBX Lite, and yet I feel infinitely more satisfied programming this weaker device to its full potential than I would using someone's SPU rasterizer workaround.

I suppose I should have a concluding paragraph, but I've run out of steam.
 

patsu

Legend
With their recent bad experiences in PSP and Home hacking, Sony would be wary of its "userbase".

Sony and partners are also selling Cell-based Linux workstations.

In the best case, I'd say we won't see any RSX driver until Sony makes money (or does not lose money) from each PS3 hardware sold. Higher level access like Lua for Home may stand more chance, IMHO -- especially given LBP's success in this area.

To be honest, I'd rather they focus on the game and media side of the business. Lua for Home would be a good start.

What they have now on the Linux side is already incredibly useful for serious work (See the cracking of the PKI root cert, and assorted scientific computing progress). I believe PS3 has achieved much more in real scientific progress than the PS2 Linux; although the latter might have created more satisfaction for the homebrew folks.
 

Carl B

Friends call me xbd
Legend
You have to look at Linux on PS3 in two ways. I understand that the focus of your OP was homebrew related isdn6, but consider also the pure scientific side. As far as the Cell architecture is concerned, Linux on PS3 has been a tremendous boost for Cell, as I think a lot of the interest and actual application/implementation work that there has been has had its origins on PS3 testbeds. I wouldn't be dismissive of the academic/scientific (and professional) work that has gotten done on the console.

For Sony's part I think there was a fundamental shift in strategy when Kaz replaced Kutaragi, as part of the entire reason for doing so was in order to "refocus" on gaming. I think a lot of the assorted PS3 benefits have thus gone unheralded by Sony's marketing and strategy teams, but that's neither here nor there and topic inappropriate to this sub-forum.

In terms of homebrew, I was originally hopeful for a community created/optimized distro packed with several Cell-accelerated apps out of the box, but that doesn't seem to be on the immediate horizon. Still, even YDL 6.1 vs 6.0 shows that notable progress/work is still being done. Hopefully the RSX situation will change in the future. If not, I still believe there could be a lot more being done in terms of homebrew than is presently the case. It'll be interesting to see what model/degree of openness Sony carries forth with in future consoles; I certainly hope they don't change direction and abandon the open hardware concept.
 

Sdw

Newcomer
I agree totally with you, isdn6!

Even after the fact that we would not be getting a RSX-driver was known, I atleast thought there would be a good all-in-one homebrew media-player (something like XBMC, but in HD) utilizing the SPUs, but nothing has materialized.
Considering that I don't think you can even switch to 24p mode from Linux, it would have sucked for playing real HD mkv's anyway I guess.
Sony not only dropped the ball, the ball went through the floor into the basement on this one.
 

warrenn

Newcomer
I completely share the sentiments of the original poster and have such a bad taste in my mouth about everything from Sony that I will probably never buy anything from them again. Their approach to the PS3 has been extraordinarily short-sighted, but no need to beat that into the ground.

The upside is that if one isn't interested in games (like me) and has firmware <2.1, the accelerated 2D xorg works very well, as I have observed in a different thread. Before I installed the accelerated xorg, using the machine was very painful - window moving was slow, often with ugly redraw artifacts and with occasional multi-second lags when switching windows. With the "new" xorg, the machine behaves like a reasonable computer and is actually quite pleasant to use. I bought the PS3 to program small physics demo apps using the SPUs and OpenGL graphics (software rendered, unfortunately). So far, it has worked fine for me (though I haven't rewritten the apps to exploit the SPUs yet).

Just a thought (and too bad one cannot downgrade the firmware).

Warren Nagourney
 

rpg.314

Veteran
Are you saying if Sony had released Linux access to RSX, you'd buy PS3 games and movies eventually ?

Me, nopes. ;)

I think that their current situation in console market is making them paranoid so they don't want to take chances with their no piracy situation, nor do they want to risk unofficial games appearing. Which makes me think that they are perhaps waiting for it to become profitable. But given the conditions, I feel that by the time it happens, Cell+RSX will be quite un-competitive with CPUs+GPUs, atleast for my purposes. :cry:
 

patsu

Legend
SCE owes it to their investors to advance the business. It would serve their paying customers and investors best if they focus on their game business first and forth most. Linux's RSX access may be a highly requested feature but they need to tie it to some sort of business model since Sony is losing money as a whole (even if SCE is making money).

The closing of the old RSX holes may anger or disappoint homebrew folks. But if it's not in their current plan or priority (yet), then closing it asap is the right thing to do. It would be worse if Sony drag their feet and then close it when there is already a wealth of IPs built on top of the hole.
 

Crazyace

Regular
PS3 linux is way more open than PS2 linux ( or the PS1 Yaroze ) as there is no 'cost of ownership' - you could download linux and the install completely for free from the net.
( PS2 linux was limited to a small number of 'key' discs ... and Yaroze was even more limited )
 

Shifty Geezer

uber-Troll!
Moderator
Legend
I think that their current situation in console market is making them paranoid so they don't want to take chances with their no piracy situation, nor do they want to risk unofficial games appearing.
Nor do they want people buying their loss-leader hardware and not make up for that initial loss by buying software! If you bought a PS3 to run Linux in the early days, Sony will have footed a few hundred bucks for that. If you then don't buy any games, that's a big loss to them. Unless you are hoping Sony are feeling very charitable and happy to sell hardware under cost so those with an interest can use it as a cheap and powerful Linux box, there's no reason to expect a full and open system for coders to do what they want with. If the hardware become profitable with suitable margins, perhaps they could be a motion in that direction, a $200 Linux PC with open development. But as you allude to, Sony will still want to keep a software revenue stream.
 

idsn6

Regular
Nor do they want people buying their loss-leader hardware and not make up for that initial loss by buying software! If you bought a PS3 to run Linux in the early days, Sony will have footed a few hundred bucks for that. If you then don't buy any games, that's a big loss to them. Unless you are hoping Sony are feeling very charitable and happy to sell hardware under cost so those with an interest can use it as a cheap and powerful Linux box, there's no reason to expect a full and open system for coders to do what they want with. If the hardware become profitable with suitable margins, perhaps they could be a motion in that direction, a $200 Linux PC with open development. But as you allude to, Sony will still want to keep a software revenue stream.

At this point, the main few parties still interested in writing for Linux on PS3 are researchers in academia who are specifically taking advantage of Sony's hardware subsidization to get some (relatively) cheap FLOPS. They are buying multiple consoles at Sony's loss to string together for better numbers, and are far, far less likely to use the machines for games and movies than someone attempting to actually use Linux on the PS3 as an OS instead of a largely ignored environment from which to run SPU kernels. If this is some sort of PR win, it's a pyrrhic one at best.

The fact that OtherOS is seen as a burden to be thrown scraps in charity only as an afterthought (if at all) is illustrative of a serious lack of vision and support that seems to permeate Sony.
 

rpg.314

Veteran
My bigger question is that why bother opening up your PS3 linux when it so useless. Sony wants to make money, yes, by all means, try to make some. But then why bother opening up linux if it doesn't bring in money. If it was to build software support for Cell, then it is not going anywhere as the masses (and those with a lot of capability to do great things) are not doing it because of lack of RSX.

They must have had a reason to do it. Only I can't see what it is. As this hack shows, opening up some APIs can mean much more opportunity to attack security. I wonder why they took this risk. So far, no damage done, but why go this far?
 

Shifty Geezer

uber-Troll!
Moderator
Legend
At this point, the main few parties still interested in writing for Linux on PS3 are researchers in academia who are specifically taking advantage of Sony's hardware subsidization to get some (relatively) cheap FLOPS...If this is some sort of PR win, it's a pyrrhic one at best.
These PS3 systems are just cheap introductions to mainstream Cell-blade system and general Cell development. Academia fuels an industry for Cell development software and tools and best practice which should lead to intellectual returns for STI. Those research graduates will, STI hopes, go on to work in big budget companies recommending heavy investment in Cell hardware. Furthermore, if PS3 Linux was being used for Cell develppment, it'd have the same effect, which is what a lot of us were expecting from Linux on PS3. It's inclusion was as a spawning ground for Cell expertise that could be used in the professional industries. eg. I could homebrew an audio engine on Cell which would lead onto a Cell based hardware synthesizer. This, rpg.314, is your reason ;)

RSX plays no part in that. It offers no benefit to STI at all. GPU experience can be gained on PC. The only reason to open up RSX on PS3 is to enable graphics stuff, which has no benefit to Sony on the whole.
 

patsu

Legend
They sell Cell-based Linux workstations. STI will need to enable open source Cell Linux assess due to the GNU licenses anyway.

The real question is: Why allow PS3 to dual boot (not just Linux, other OSes are supposedly possible too) ? I think it's one of Kutaragi's "Cell everywhere" vision. Not sure how much of that is still around (probably not). Presumably a cheap Cell + Linux OS can power a sophisticated media systems someday.
 

rpg.314

Veteran
These PS3 systems are just cheap introductions to mainstream Cell-blade system and general Cell development. Academia fuels an industry for Cell development software and tools and best practice which should lead to intellectual returns for STI. Those research graduates will, STI hopes, go on to work in big budget companies recommending heavy investment in Cell hardware. Furthermore, if PS3 Linux was being used for Cell develppment, it'd have the same effect, which is what a lot of us were expecting from Linux on PS3. It's inclusion was as a spawning ground for Cell expertise that could be used in the professional industries. eg. I could homebrew an audio engine on Cell which would lead onto a Cell based hardware synthesizer. This, rpg.314, is your reason ;)

RSX plays no part in that. It offers no benefit to STI at all. GPU experience can be gained on PC. The only reason to open up RSX on PS3 is to enable graphics stuff, which has no benefit to Sony on the whole.

As I said the following before, it's not working out. RSX doesn't help sti, agreed. But from what I see, all I can look at is some random video decode projects and a SPU based 3d pipeline. And frankly, at this point I don't see any point in working on CELL alone. GPUs' have far better programming model, regularly increasing capabilities, wide industry support and are available at a lot of price points. Academia interest in GPUs >> Cell atm.

If it was to build software support for Cell, then it is not going anywhere as the masses (and those with a lot of capability to do great things) are not doing it because of lack of RSX.
 

Shifty Geezer

uber-Troll!
Moderator
Legend
GPUs' have far better programming model, regularly increasing capabilities, wide industry support and are available at a lot of price points. Academia interest in GPUs >> Cell atm.
Well, I have no idea how one measures this. Of course GPU investigations are going to be more mainstream because they are more established. However, a lot of high-spec companies are using Cell, even with the option for GPGPU, because GPGPU can't do everything Cell can. We've seen issues with GPGPU video decodes for example, which Cell can cruise through.

But as I say in the other thread, which is far too parallel to this one, what is your interest in developing on PS3? Why not just focus on GPGPU in the PC space?
 

patsu

Legend
As I said the following before, it's not working out. RSX doesn't help sti, agreed. But from what I see, all I can look at is some random video decode projects and a SPU based 3d pipeline. And frankly, at this point I don't see any point in working on CELL alone. GPUs' have far better programming model, regularly increasing capabilities, wide industry support and are available at a lot of price points. Academia interest in GPUs >> Cell atm.

The RSX architecture is outdated. If you want to program the GPU, there are easier and newer ways to do it. What's missing in PS3 Linux is a fast OpenGL implementation for (great) convenience. It would be very nice, yes, but not for the argument of better programming models. :D

For programming model, you can do OpenGL, OpenMPI, OpenCL, and any other new ones on Cell anyway. It's more flexible than the GPU in general. And yes, you can gain better performance for some problems, and less performance for others relative to a GPU. The thing is you don't really need world leading performance for homebrew. Afterall, PSP is not exactly a powerhouse but PSP homebrew is through the roof.

It may be more about self expression and coolness. Like I said in the other thread, a million consistently-spec'ed Cell nodes connected to HDTVs all around the world has its allure.

Could you clarify what you mean here?

I meant once STI implemented Linux on their Cell workstations/boards. The know-how and source will propagate into the Linux community anyway.

Since PS3 supports booting from another OS, it's only a matter of time for people to port the Cell workstation Linux to PS3 (even if Sony doesn't do it).
 

rpg.314

Veteran
But as I say in the other thread, which is far too parallel to this one, what is your interest in developing on PS3? Why not just focus on GPGPU in the PC space?

I am. It's just that ps3 isn't suitable enuf because of it's handicaps.
 
Top