Will The Gpu/cpu Of The Pandora Have Open Source Drivers?

[castlefox]

Will the gpu/cpu of the pandora have open source drivers?

I will be even more excited in getting a Pandora if they have open source drivers. I've wanting to support some hardware that does this for a long time now.

I thought I was reading that Pandora would be supporting this BUT I could not find an official link to support that.

 

[Exophase]

The 3D accelerator won't have open source drivers. Sorry. I don't know what you mean by "CPU drivers" in this context (CPU is the one thing that doesn't need a driver), but most of the other drivers on Pandora will be open.

 

[castlefox]

The 3D accelerator won't have open source drivers. Sorry. I don't know what you mean by "CPU drivers" in this context (CPU is the one thing that doesn't need a driver), but most of the other drivers on Pandora will be open.

Thanks for the quick response.

I dont think I know what I meant by cpu drivers. (whoops)

 

[WizardStan]

Someone was working on reverse engineering the GPU drivers, but I think the project is on extended hiatus.
I think the WiFi drivers were also almost closed source, but if I recall correctly, someone found an open source alternative that seems to be mostly working.

 

[Mr B]

The TMS320C64x+ DSP has open drivers, and the audio/video chip is supposed to be "based" on this.
The 3D chip is talked about here.
B!

 

[Exophase]

The TMS320C64x+ DSP has open drivers, and the audio/video chip is supposed to be "based" on this.
The 3D chip is talked about here.
B!

Four things: the GPU is the closed source SGX, not the DSP. The audio is going out over a serial connection on the SoC to a DAC, the DSP is not interfacing audio. And while the DSP can be used to assist in audio/video decoding/encoding, there's also a video acceleration block next to it whose functionality is completely unknown. Finally, while drivers for accessing the DSP are open, actual audio/video codec code to be ran on it doesn't exist as far as I know, much less is it open source.

 

[ldesnogu]

Are you sure TI doesn't provide any codec using the DSP?
I thought this was applicable: http://focus.ti.com/docs/toolsw/folders/print/linuxdvsdk-omap.html

 

[Squidge]

The DVSDK contains demonstration and evaluation software only. The evaluation version codecs included are closed source.

Which means that someone can write there own open-source codec, but TI are not going to give you the source code to theres.

It also means that you can't use the provided codecs for anything other than evaluation purposes.

 

[ldesnogu]

That's very disappointing...

 

[Squidge]

At least they are giving you source code and documentation for how to write your own codec, which is better than a lot of companies do. "Here's a closed source codec, figure out how to write your own by reversing it."

 

[skeezix]

You have to love the state of engineering.. all these hw companies producing hw they don't want people to actually use... *sigh*

jeff

 

[Exophase]

Is hardware vendors releasing peripheral devices that can only be interfaced with proprietary closed source drivers supposed to be a new thing? Hasn't this always been happening?

What's a little worse is how CPUs are being made today where the instruction sets aren't fully immediately available for some reason, much less timing information (although I guess ARM does provide that). Now that's just strange.

 

[Kramy]

What I find strange is a software company (like Microsoft) focusing on the UI rather than creating innovative new frameworks for GPU acceleration and stuff. Sure can tell where their priorities are.

Optimization is not one of them.

 

[ldesnogu]

What's a little worse is how CPUs are being made today where the instruction sets aren't fully immediately available for some reason, much less timing information (although I guess ARM does provide that). Now that's just strange.

That's not strange as far as CPUs go: it's become almost impossible to exactly describe what's happening in an OoO CPU. One can give high-level rules (such as AMD and Intel providing SW optimization guides), but giving exact rules is simply not possible except by providing RTL code

 

[fusion_power]

You have to love the state of engineering.. all these hw companies producing hw they don't want people to actually use... *sigh*

jeff

Sounds really like the 3D Hardware in the Pandora is not more easy to use than the 3D Hardware in any other (closed) mobile device like iPhone, Palm Pre and so on...
What can a coder do with the OMAP if all the nice 3D Features are so hard to reach? I am not a coder but I think even with a closed source driver you could do alot IF there is a good Driver for the Hardware. I mean a Hardware company should make not only good hardware but also good drivers, otherwise the hardware is quite useless and nobody buys the Hardware.
When I look at my (wonderful ) PC, I use Nvidia Graphics cards since Years and they always use Closed Source Drivers and look, what a Coder can do with this (in combination with OpenGL / DirectX etc...), I never had problems with this Drivers or this Graphic Hardware even if everything is closed. There is still a good bunch of Homebrew using all of the modern PC Hardware features and nobody seems to care about all the closed Stuff onto PCs.
I hope it will not be a real problem if the Pandora only has closed source (3D) Drivers. ^^ It would be a huge waste of Hardware power if everything has to be done in Software only and all the great SGX and DSP and idontknow - Units onto the Chipset are bored to death.

 

[dflemstr]

You have to love the state of engineering.. all these hw companies producing hw they don't want people to actually use... *sigh*

jeff

Sounds really like the 3D Hardware in the Pandora is not more easy to use than the 3D Hardware in any other (closed) mobile device like iPhone, Palm Pre and so on...
What can a coder do with the OMAP if all the nice 3D Features are so hard to reach? I am not a coder but I think even with a closed source driver you could do alot IF there is a good Driver for the Hardware. I mean a Hardware company should make not only good hardware but also good drivers, otherwise the hardware is quite useless and nobody buys the Hardware.
When I look at my (wonderful ) PC, I use Nvidia Graphics cards since Years and they always use Closed Source Drivers and look, what a Coder can do with this (in combination with OpenGL / DirectX etc...), I never had problems with this Drivers or this Graphic Hardware even if everything is closed. There is still a good bunch of Homebrew using all of the modern PC Hardware features and nobody seems to care about all the closed Stuff onto PCs.
I hope it will not be a real problem if the Pandora only has closed source (3D) Drivers. ^^ It would be a huge waste of Hardware power if everything has to be done in Software only and all the great SGX and DSP and idontknow - Units onto the Chipset are bored to death.

With open source drivers, we could have had full OpenGL 2.0 (And probably 3.0 because we have ES 2.0 already) support on the Pandora.

Also, if you've been following various discussions in various forums recently, you will know that because of licensing issues, only open-source graphics drivers will have access to the kernel KMS infrastructure, which really sucks for us with binary blob drivers.

Basically, you can do a lot of stuff with closed-source drivers, but open-source drivers give you the ability to expand that usability spectrum.

 

[Exophase]

IMO it's good that people are forced to use OpenGL ES 2 instead of OpenGL 2, because the shaders in OGL2 do not support precision modifiers except as decorations. SGX is much more suited for processing colors with lowp precision than full 32bit float precision.

 

[ldesnogu]

On top of that my understanding is that OpenGL 3.n is getting closer to OpenGL ES 2.0 rather than the reverse.

 

[Enverex]

How did this all pan out? - http://www.gp32x.de/board/index.php?showtopic=45664
I see it's a mention that the drivers are open source, but the part that is open isn't all that useful and it uses binary sections in userspace that you obviously can't do much with.

Did maciek_urbanski have any luck reverse engineering them at all? Can't say I've seen his name around here lately.

 

[darkblu]

On top of that my understanding is that OpenGL 3.n is getting closer to OpenGL ES 2.0 rather than the reverse.

well, duh. gl es2 is the more sane API. when you don't have to support generations upon generations of abandonware (i.e. anything closed-source and without sound maintenance) in your API, things all of a sudden start looking, well, brighter.

ps:

Did maciek_urbanski have any luck reverse engineering them at all? Can't say I've seen his name around here lately.

he did reverse-engineer the driver to considerable depths, but his interests lay mainly in the ISA (USSE), and that's where he focused. i believe that if we asked him politely, he'd either (a) continue with the work (once he has his pandora), or (b ) pass on the project to others.