Pandora 3d Chip Questions When Working?


kevcal said:
So 2d acceleration isn't accessible yet either ???

2D acceleration is flat 3D polygons, so no. It's not that essential to be honest though, as the ARM is powerful enough to do such things in software.

As for drivers, some developers have several versions along with demos and example code, but none of which can be made public or work on the current Pandora kernel.
 
Last edited by a moderator:
Laurent said:
mali said:
In my opinion, that's the reason why maciek_urbansky should be supported best possible. His work could be "Plan B". I really hope, he has gotten a dev-board.

Just remember Maciek has clearly stated he would not work on OpenGL ES drivers. He "only" wants to describe hardware registers and instruction set. Getting a driver out of his work will be a major task.


I know that, but I'm sure there are devs out there, who would be able to do it, if there is enough info reverse engineered by Maciek. I just think of it as a possible solution in a worst-case scenario. I followed the discussion about the linux driver for OMAP2 and there has never been any prove, that Imgtec is able to provide a good linux driver. There was a discussion about reversing the Symbian driver, but because of hope to get an official driver, the project was dropped.
 
Last edited by a moderator:
mali said:
I know that, but I'm sure there are devs out there, who would be able to do it, if there is enough info reverse engineered by Maciek. I just think of it as a possible solution in a worst-case scenario. I followed the discussion about the linux driver for OMAP2 and there has never been any prove, that Imgtec is able to provide a good linux driver.

All I can say is that providing the same level of what IMGtec driver is giving would require a huge amount of work.
For instance, OpenGL (ES) drivers require a compiler as they generate code on the fly. Do you really think you could assemble a team that could achieve that in a reasonable amount of time without access to real specs (Macek will surely provide good documentation, but will it be both error free and complete?)?
 
Last edited by a moderator:
Squidge said:
kevcal said:
So 2d acceleration isn't accessible yet either ???
2D acceleration is flat 3D polygons, so no. It's not that essential to be honest though, as the ARM is powerful enough to do such things in software. Yeah, I realise that; but obviously we'd like to see the hardware used to it's full potential. :)
I guess in the early stages it doesn't matter and (hopefully) acceleration will be available within a few months of pandora release.
 
Last edited by a moderator:
craigix said:
There is a driver, that is why there are 3D demos on youtube using the 3D hardware on OMAP3 devices. We just don't have one for the current kernel, it will rise to the top of the list once we are ready for it software wise.

ImgTech have asked for a Pandora board in the past so they are aware of the project.
So that means the SGX driver will only be usable on an outdated kernel version like it happened with Radeon binary drivers?

Or those drivers will be updated and maintained in a fast way so being available for the latest stable kernel release out there?
 
Last edited by a moderator:
timofonic said:
So that means the SGX driver will only be usable on an outdated kernel version like it happened with Radeon binary drivers?

Or those drivers will be updated and maintained in a fast way so being available for the latest stable kernel release out there?
No, IMGtec will provide an open source module for the kernel (ala nVidia). At least that's the plan, it's not been seen yet :)
 
Last edited by a moderator:
Laurent said:
timofonic said:
So that means the SGX driver will only be usable on an outdated kernel version like it happened with Radeon binary drivers?

Or those drivers will be updated and maintained in a fast way so being available for the latest stable kernel release out there?
No, IMGtec will provide an open source module for the kernel (ala nVidia). At least that's the plan, it's not been seen yet :)


The free part needs to be maintained too and sure it works under that damn protection against non-GPL modules that the Linux kernel has. It was a pain in the ass with ATI Radeon drivers that made me to use Open Source ones even if lacking proper 3D support...
 
Last edited by a moderator:
Laurent said:
All I can say is that providing the same level of what IMGtec driver is giving would require a huge amount of work.
I've looked up the quote, that you mentioned. I feel reassured, now :)
May I assume, that you've received a dev-board, after all?
 
Last edited by a moderator:
timofonic said:
The free part needs to be maintained too and sure it works under that damn protection against non-GPL modules that the Linux kernel has.

You seem to be implying that Linux has some sort of built in DRM. Do you have a cite for this rather astonishing fact?

I ask, because as far as I know it's completely false. The linux kernel is not even aware of what license its modules are distributed under.

Even Torvalds himself says that closed source drivers are (sometimes) fine legally and morally. (link) It would be bizarre if, after saying that, he went on to make it impossible for closed source drivers to function.
 
Last edited by a moderator:
craigix said:
ImgTech have asked for a Pandora board in the past so they are aware of the project.
Do they have one yet? ;)

Squidge said:
It's not that essential to be honest though, as the ARM is powerful enough to do such things in software.
True, but I think the point is: Hardware Acceleration->less MHz needed->less power consumption->longer Battery life->longer fun. ;)
So why using 500MHz in Software Mode, if we can have the same speed with 200MHz (or so) with Hardware Acceleration? ^_^
 
Last edited by a moderator:
fusion_power said:
True, but I think the point is: Hardware Acceleration->less MHz needed->less power consumption->longer Battery life->longer fun. ;)
So why using 500MHz in Software Mode, if we can have the same speed with 200MHz (or so) with Hardware Acceleration? ^_^
I hope the overkill of specs doesn't make coders lazy and leave low-key stuff not fully optimized.
 
Last edited by a moderator:
timofonic said:
The free part needs to be maintained too and sure it works under that damn protection against non-GPL modules that the Linux kernel has. It was a pain in the ass with ATI Radeon drivers that made me to use Open Source ones even if lacking proper 3D support...
You know...you should be cautious about remarks of that nature.

If it weren't for that GPL license on things, you wouldn't HAVE an OS to put on this. Serious.
 
Last edited by a moderator:
I do wish people would stop this "we have plenty of power to do xyz, so it doesn`t matter if the code sucks" mentality, I keep hearing people say things like that all over the place and it does my head in, just the other day I was talking to an ex IBM mainframe coder and member of mensa and he said exactly the same thing, where is this idiotic "it`s ok to write crap code cos I have a fast computer" meme coming from???

I was surprised to hear a code "oldie" say something like that, considering the machines he used to code for had less memory and cpu speed than my calculator you would have thought he would be a fervent disciple of "trim/optimise/polish"

this creeping mindset worries me, less performance==less features, and features (shine) matter in games, we always used to wring every last bit of polish out of code, not let things slide "cos we have plenty of power to run a basic invaders game", it`s bad practice to let tics get away from you.
 
True that. Someone on these forums has an epic signature comparing what we did with C64's back in the day to what we do with Core2 Duos now. Sad state of affairs.

Come out, dude with an epic signature! You are needed!
 
timofonic said:
So that means the SGX driver will only be usable on an outdated kernel version like it happened with Radeon binary drivers?
No, as that driver will never be public.
timofonic said:
Or those drivers will be updated and maintained in a fast way so being available for the latest stable kernel release out there?
Well the community will most likely be to update and maintain the the driver far faster than Imgtec can. After all, the full source of the kernel module will be released, so you can compile/port the driver for any kernel version (which means it will easily pass the kernels "Are you licensed under the GPL?" question). You will not care about the binary blob, as for all intents and purposes, it's just a data file.

hobbyman II said:
I do wish people would stop this "we have plenty of power to do xyz, so it doesn`t matter if the code sucks"
Its more along the lines of "We have no choice but to use processor power, as there is currently no alternative". Also, until there is a library for it, I don't see many people doing full OpenGL initialisation just to have multiple copies of an object on the screen.
 
Last edited by a moderator:
...also don't forget quake 2 runs just fine in SW.. so it's not like you are gonna be totally without even if drivers are a bit delayed
 
hobbyman II said:
I do wish people would stop this "we have plenty of power to do xyz, so it doesn`t matter if the code sucks" mentality, I keep hearing people say things like that all over the place and it does my head in, just the other day I was talking to an ex IBM mainframe coder and member of mensa and he said exactly the same thing, where is this idiotic "it`s ok to write crap code cos I have a fast computer" meme coming from???
I agree with the point you build up to in the second half of your post, but this first half is a bit shaky.

There has always been a trade-off between engineer-time and computer-time. It does not make sense to always favor one or the other. You have to make the decision on a case-by-case basis depending on how much you value those two resources.

There may be some situations where computer-time (aka performance) is so much more valuable than engineer-time, that you'll want the programmers to sit down and count assembly instructions, add up all the CPU cycles so they precisely line up with I/O-accesses, make sure every last memory access makes exactly optimal use of the cache, etc.

And there may be other situations, where programmer time is so much more important than performance that you can do whatever crazy thing you like so long as it compiles and runs, even if it takes ten seconds to respond to a mouse click. Whatever it takes to get you onto your next project.

Most real life situations will obviously fall between these two extremes.

The 'meme' that it's never OK to take shortcuts, and that it's somehow a grand programmer virtue to waste your time optimizing something that will already run in the blink of an eye, is rather irritating. (And common among beginners and non-professionals!)

I'll bet your old timer from IBM was shaking his head at how naive you were.


Squidge said:
(which means it will easily pass the kernels "Are you licensed under the GPL?" question.)
Wow, you mean the kernel does challenge its modules about their licensing?
I wasn't aware of that at all. Perhaps I'm being naive as well.
 
Last edited by a moderator:
@VRandy from what I understand it is not about the kernel rejecting it but that it wouldn't be legal to have a non free kernel to blob interface the blob itself can be licensed whatever but the interface must be open

just case you lack the will and skill to write fast code doesn't mean you should down the people that do yes there is a time trade off but you seem to fail to under stand the scope of the mentality .. for instance most modern programs in windows and even linux (open office... X11 etc...) had previous version that ram with MUCH similar functionality on FAR weaker hardware

if you write slow code and write more slow code on top of it... you get exponentially slower code which explains where we are now... GHz PCs proforming meanial tasks less effectively than 10 year old PCs did

besides... if you spent the money to get the fastest computer... don't you think it is pretty stupid to not speed a reasonable amount of time making sure you are getting your moneys worth out of it?
 
@javaJake ... so you quote someone else and respond to me :-|| LOL

also ... I don't think there is anything stoping you from adding such a module... except that it would be illegal to set said bit (is there really such a thing i wasn't aware) which probably isn't all that hard


a tainted kernel is a kernel that has non open source code running in kernel space... (or apparently even code for which the exact source code is not provided)
 
Back
Top