Raspberry Pi GPU unlocking.

[firewalker]

Have anyone received a Raspberry Pi yet?

I was wondering if it is possible to unlock all the extra features of the GPU like mpeg2 decoding. Is it "hackable"?

People over the Raspberry Pi believe such a thing is almost impossible.

Alexander.

[free_electron]

Hah ! if you'd find someone in the open-sauce community that is willing to shell out for the licence and you can convince broadcomm to release the core ... maybe ...
my guess is chances are slim.

[mikeselectricstuff]

Considering the numbers that will be in the field i think it is inevitable that a lot of currently-hidden info will lbe reverse-enineered out eventually - the only unknown is when.

[bullet308]

One of the main reasons I have no interest in the Raspberry Pi is the fact that Broadcomm is the chipset maker. I have already fought enough battles over Broadcomm chips and Linux, thanks. No more for me.

[T4P]

Yes, broadcomm won't so easily let info out, they take the money and keep quiet about the info
That's why i fought a battle with a broadcomm wifi chip from a laptop and gave up 
It's still in my box

[Gall]

If it is like other similar chips (Samsung etc.) then its main GPU has nothing to do with mpeg. There are independent accelerators for 2d, 3d, video decoding, windowing, rotation and jpeg, and they technically may be present in any combination. They are controlled with completely different register sets mapped to different i/o addresses. If one of them is not present, it is just not present.

The PowerVR 3d core is theoretically capable of doing 2d on its shaders in OpenGL too, like any GPGPU used for computations, but this is not the usual way. Probably nobody ever tried it.

[free_electron]

the problem sits in accessing the registers for the mpeg functionality. despite popular belief MPEG needs payment of royalties...
And you can bet the house that broadcomm will NOT release info on how to use these advanced features ( especially since they have HDMI which has crypto ... ) unless NDA's are signed and money is being forked over ... As the open-sauce community is penny-less and 'NDA' doesn't work very well with 'Open' .... fat chance of this ever happening.

[mikeselectricstuff]

I can see one approach - This chip must be used in other products that use all the functionality.
Reverse-engineering the code that runs on those products would be a possible way in.

[Leo Bodnar]

I have to ask this!

What are you trying to achieve?

Leo

P.S. I already massively don't like Raspberry because everybody else falls head over heels to have one.  Just another wannabeduino.
If someone wants to leave a dent in the Universe give African kids a 4-bit microcontroller demoboard with an LED and a pushbutton, not Wi-Fing and HDMIng *nix based MP3 player.

[tom66]

P.S. I already massively don't like Raspberry because everybody else falls head over heels to have one.  Just another wannabeduino.
If someone wants to leave a dent in the Universe give African kids a 4-bit microcontroller demoboard with an LED and a pushbutton, not Wi-Fing and HDMIng *nix based MP3 player.

I don't think it's intended for African schoolchildren - its original intention was to help UK Computing students. As one of them, let me tell you, the computer science lessons we have are dire, boring, uninspiring, and do not teach you how to actually program properly. I would like to see how they could be integrated into a coursework module; perhaps the low processing power would encourage optimised applications and limited resources would lead to unique solutions - and maybe if younger programmers learn this then the next Windows OS won't need 8 GB of RAM?

[Leo Bodnar]

perhaps the low processing power would encourage optimised applications and limited resources would lead to unique solutions - and maybe if younger programmers learn this then the next Windows OS won't need 8 GB of RAM?

32-bit RISC core running at 700MHz with hardware FP and 256MB RAM is a limited resource?! 
It's exactly the same spec as eMac http://en.wikipedia.org/wiki/EMac that was purpose-built for education market.  All the schools had them (i.e. cost to students was zero.)  How many students went beyond using it as a fancy jukebox?

Leo

[AntiProtonBoy]

The PowerVR 3d core is theoretically capable of doing 2d on its shaders in OpenGL too, like any GPGPU used for computations, but this is not the usual way. Probably nobody ever tried it.

PowerVR chips are really not that brilliant at GPGPU. In shaders, 128-bit vector operations are executed as four independent 32-bit scalar ops, for instance. And memory bandwidth is really limited.

[Bored@Work]

its original intention was to help UK Computing students. As one of them, let me tell you, the computer science lessons we have are dire, boring, uninspiring, and do not teach you how to actually program properly.

And how should a board like the RPi fix this? It is a question of proper selection of tools, like programming languages, proper teaching and proper supervising. Nothing of this requires an RPi. The RPi won't make your teachers better.

[free_electron]

Not only that. the RPi will also promot bloatware and mindless library reuse.
Teach programming fits on a simple 8 bit machine with limited resources like 8 k rom en 128 bytes ram and give students something to chew on...
We already have truckload of 'library using' codemonkeys. what we don't have enough off are people that can actually code something without an operating system and massive prechewed libraries , or without wasting 5000 calls to a library to configure some i/o pins ...

programming is not about a language.. it is about mechanism and learning to use the cpu and its hardware effectively to perform work.

People wrote whole bookkeeping and payroll programs in a PET with 20k ram and a casette ...  They landed on the moon with 14K of code ... and that thing was multitasking , flying the lander ,doing radar ranging and tons of other things. all at the same time... with a clock in the kilohertz range ... Now we have cellphones with quadcore 2 GHz and six core graphics accelerators .. to play 'angry birds'.... i foresee a VERY grim future ... Forget global warming ...the global dumbing down is coming... go wathc the movie 'idiocracy'.. we're not far off ...

Now even the bootloader won't fit in that anymore .. and turning on a LCD or plasma tv and staring at the manufacturer splashscreen for 10 seconds is perfectly acceptable ... because it has to 'boot'. And don't get me started on blu-ray and dvd players. The people who write that code these days should be publicly forced to eat that tv....

I have an old daewoo dvd player. Power-on to first image is under a second. pop in a dvd disc , it spins up and 2 seconds later i have the dvd menu. it has a braindead mips processor running at 27MHz for the video processing and decoding and a 8051 to do remote control ,display and operate the drive mechanism. Works beautifully.

Compare this to my latest blu ray player. Power on to first splash screen is also 2 seconds and then there is the scroll bar with 'initializing... that takes 30 seconds because it wants to go all wonky and see if it can get an update from the internet ( i have no cable plugged in... why do you keep waiting you piece of shit. can't you do a simple 'cable detect' that would take 1 microsecond. ) and it needs to boot some open-sores / broken-source operating system.

Then you put in a disc and you stare at a black screen for another 20 seconds while the drive is grinding away trying to load the java code to run on the virtual machine. What is wrong with the processor that is in the box ? why do you need to run a virtual machine on top of it ? you blithering codemonkeys! Just write code for the real processor you flaming-turd-code-shitters. Sony should rename themselves to ShitCo.

And then it goes through some handshaking on the crypto and if you are really lucky you may get the menu ( if it hasn't crashed by now ) , if you are a bit less lucky it comes with a message that it needs internet connection to check for an update. Here you have a carrier medium that can hold 50 gigabytes of data , and you tell me there is no room for a bit of code ?

and then you finally have gone through all the hoops , select 'main feature' and press play and it first goes off for another 10 second 'grinding session' ... aaaaargh !

If i want to enjoy a movie that i bought i have to sit there jumping through updates , endless waiting , and all sorts of misery before i can see something ( and then i haven't even started ranting an raving about all the 'coming soon drab' they inject on the disc... don't you understand that , if i watch this disc 1 year later that stuff is old and i don't want to see the 'coming ssoon crap anymore ? put it on the disc if you must but do not start playing that shit by default ! Friggin shit-for-brains movie studio's )

And this is acceptable practice these days ? Anyone that writes and release such code should be banished for eternity and locked in a room full of eproms, each loaded with a single byte of their code... Eproms that are in ceramic dil packages so they always land on the floor with their pins turned upward ... and they go in barefoot ..  if you know what i mean ... ( if you have ever stepped barefoot on an upturned dil package , you do know what i mean... multiply that one package by a few million , one chip per byte of crap code... my guess is about 16 megabytes of code in a blueray ... eternity isn't long enough for those nitwits.

[tom66]

perhaps the low processing power would encourage optimised applications and limited resources would lead to unique solutions - and maybe if younger programmers learn this then the next Windows OS won't need 8 GB of RAM?

32-bit RISC core running at 700MHz with hardware FP and 256MB RAM is a limited resource?! 
It's exactly the same spec as eMac http://en.wikipedia.org/wiki/EMac that was purpose-built for education market.  All the schools had them (i.e. cost to students was zero.)  How many students went beyond using it as a fancy jukebox?

Leo

The processor is a RISC architecture, so the performance is closer to a 300 MHz PIII/PII than anything else. 256 MB of RAM is shared between the CPU and GPU and try getting an OS written in the last 10 years to run within 128 MB.

[Bored@Work]

try getting an OS written in the last 10 years to run within 128 MB.

For what reason? If non fits, use one that was written in the last 20 years.

Uhmm, what is that sudden smell? Ah, kids all around shitting their pants on the suggestion of using something "old". What a horrible though for the generation that throws away their electronics every half year, because it got "old".

[firewalker]

We shouldn't say the same things again and again for "What Raspberry P" is or is not". It wasn't my intention.

Lets keep only the engineering aspect of it.

Alexander.

[codeboy2k]

People wrote whole bookkeeping and payroll programs in a PET with 20k ram and a casette ...

That was me. I did that when I was 17. I worked at BPI Microsystems and we wrote G/L, A/R, A/P, payroll, inventory and cost accounting for CBM/PET and Apple ][, later the VIC20 and C64, C128, Apple IIc, IIgs, Apple III ( or Apple ///) and the Olivetti M20 (they paid us to port our code).  That was all done in BASIC.  ugh!  I did some 6502 coding on the 9090 hard drive firmware back then, too.  I think I still have the hard drive schematics and firmware source code; it was a lot of the same code as was in the PET ROMs  -- bit banging the IEEE-488 bus, similar internal layout as the PETs. That made it pretty easy to wrap your head around the drive firmware, it just looked like another PET inside.

...Eproms that are in ceramic dil packages so they always land on the floor with their pins turned upward ... and they go in barefoot ..  if you know what i mean ... ( if you have ever stepped barefoot on an upturned dil package , you do know what i mean..

you made me cringe, for a pain I once knew... now I can't even get a tickle from a SO-8 pins up on the floor.

[T4P]

 ( if you have ever stepped barefoot on an upturned dil package , you do know what i mean...

It hurt and bled a lot

[Leo Bodnar]

Uhmm, what is that sudden smell? Ah, kids all around shitting their pants on the suggestion of using something "old". What a horrible though for the generation that throws away their electronics every half year, because it got "old".

If you look at any of my development Windows PC screens most of the time you will see Norton Commander v4.0 (released January 22 1993) in its full blue glory.  By the time kids click through animated Windows start menu crap to locate File manager (or is it Program Manager (or is it Explorer?)) I can navigate down ten subfolders, open a file in hex editor, change the contents, save it and go back to where I was.

Leo

[amyk]

I can see one approach - This chip must be used in other products that use all the functionality.
Reverse-engineering the code that runs on those products would be a possible way in.

Or decap and get die photos, then reverse-engineer the layout. SoCs are usually laid out in well-defined blocks with connections to a common bus, so there is only a need to focus on the desired sections. There are various companies in China that can do this for (relatively) cheaply if the process size isn't too small.

I've read that Broadcom uses hidden tricks like different register addresses and datasheets for each customer, so that leaked info could easily be traced back to its source. This is a strong deterrent against anyone doing so.

[Mechatrommer]

We already have truckload of 'library using' codemonkeys. what we don't have enough off are people that can actually code something without an operating system and massive prechewed libraries...
Then you put in a disc and you stare at a black screen for another 20 seconds while the drive is grinding away trying to load the java code to run on the virtual machine...
And this is acceptable practice these days ? Anyone that writes and release such code should be banished for eternity and locked in a room full of eproms

i think this is my thought as well these days. and there's another buzz... an application built on top of another application or "platform". i dont know why my recently installed app that can only do a simple thing load much slower than my older professional grade app. i think ramping up cpu clock is just another waste of effort. people should build "AI" applications by now, not making it dumber.

Now we have cellphones with quadcore 2 GHz and six core graphics accelerators .. to play 'angry birds'.... i foresee a VERY grim...

the swmbo just unplugged the atom cpu upstairs, the kiddos always sit infront and forgetting whats inside their schoolbag. sigh! i should build a simple arithmetic program with angry bird as background.

[T4P]

You thought you've seen the worst.
how about the people who buy iPhones or Galaxy S III's just 'cause it's the most expensive
and those idiots who claim their iPhone 4S's has a much better then my Sensation
HOW about those 'fanboys' we know as abominations

All of them are blind to what their phone is actually made to do
but in my opinion you know what's the worst abomination of a "invention" ?
Smart TV's. Yeah. Pay a big premium and still you don't even get to be able to play flash apps and videos
Those media boxes have android and costs like 200$ onwards and you only get a meagre 800MHz Dual Core ARM chip!
This would be okay if this was 2005!
Heck, for the price premiums one pays for smart TVs i can easily build a entire PC with a full featured HD5450 and an Intel chip!

My parents were on the hunt for a new LED-backlit TV, (I'm not gonna talk about the cheap ones, POS)
saw a few but only one 46" caught my eye but again, the 'abomination' is in it
and phillips ain't making a non-smart tv version.
Why i say it caught my eye? A standard 46" cheapie (the good ones? None. no one has a 46" non smart TV)
would already cost 999$
The phillips has a 400Hz panel and it has those "Smart/Stupid TV things" and a 14bit video DAC
for only 1399$ but i must say ... who the heck needs a 14bit DAC on a TV?
A samsung ( no one else sold a 400Hz that day ... ) 400Hz goes for 1799$ without ANYTHING

[IanB]

I don't think it's intended for African schoolchildren - its original intention was to help UK Computing students. As one of them, let me tell you, the computer science lessons we have are dire, boring, uninspiring, and do not teach you how to actually program properly. I would like to see how they could be integrated into a coursework module; perhaps the low processing power would encourage optimised applications and limited resources would lead to unique solutions - and maybe if younger programmers learn this then the next Windows OS won't need 8 GB of RAM?

I'm not quite sure how the hardware can influence this? Surely the problem with the lesson content is the lesson content? As I understand it, almost every schoolchild/student needs to have access to a PC today to write their essays and assignments? I have the impression that hand written essays are no more. If that is the case, then there are many inspiring programming environments available on PCs, not least of which is Visual Studio Express.

Just for fun, here are the machines I first learned programming on when I was at school:

http://en.wikipedia.org/wiki/Research_Machines_380Z (Ooh! Shiny!)
http://en.wikipedia.org/wiki/PDP-11
http://en.wikipedia.org/wiki/PDP-10

And note that the PDP-11 and PDP-10 ran proper multi-user, multitasking, time sharing operating systems in a memory space generally measured in kilobytes. (Megabytes were for millionaires.)

On the downside, software engineering concepts like structured programming and modularity were not part of the lesson plans. Would they even be today, I wonder?

[free_electron]

The phillips has a 400Hz panel and it has those "Smart/Stupid TV things" and a 14bit video DAC

can anyone explain me what the purpose of a DAC is in a flatpanel TV ? especially since the pixels are digitally controlled... why do you need a dac ? marketing bullshittery ?