Default Powervr Sgx530 Technical Ref Manual


Before anyone gets too carried away, there's not really anything in that document other than an overview of the engine's capabilities. There's no really fun stuff in there:

QUOTE

1.4.2 SGX Register Description
For more information about register descriptions, contact your TI representative.



:(
 
Firefox said:
Nice little overview, though.
Indeed. And there's a bit more info than one would think there. Just nothing immediately useful past
the specs listings- something very useful in and of themselves.
 
Last edited by a moderator:
*sigh* There was me being all excited about this thread... Does anyone know whether it would be realistic to expect to get such documentation if we ask nicely for it? (I.e. does anyone know whether some small companies have been successful in that at all?)
 
Vladimir said:
*sigh* There was me being all excited about this thread... Does anyone know whether it would be realistic to expect to get such documentation if we ask nicely for it? (I.e. does anyone know whether some small companies have been successful in that at all?)
You have no chance unless you design a SoC using their chip :(
 
Last edited by a moderator:
Laurent said:
Vladimir said:
*sigh* There was me being all excited about this thread... Does anyone know whether it would be realistic to expect to get such documentation if we ask nicely for it? (I.e. does anyone know whether some small companies have been successful in that at all?)
You have no chance unless you design a SoC using their chip :(

That's worser than I thought, I mean, what is so secret about those little registers?
 
Last edited by a moderator:
Vladimir said:
Does anyone know whether it would be realistic to expect to get such documentation if we ask nicely for it? (I.e. does anyone know whether some small companies have been successful in that at all?)
Not very realistic I'd say. It would fall under the realm of PowerVR's IP. The most I'd expect at the moment would be something similar to nvidia's Linux driver, which is an open-source interface that loads the real driver blob. I believe this is in the works, which means the current situation is even worse.
 
Last edited by a moderator:
Vladimir said:
That's worser than I thought, I mean, what is so secret about those little registers?

Maybe nVidia can answer your question :D
 
Last edited by a moderator:
Laurent said:
Vladimir said:
That's worser than I thought, I mean, what is so secret about those little registers?

Maybe nVidia can answer your question :D

So when does the Pandora with the ATI chips come out? :p

Seriously though, no way? :blink:
 
Last edited by a moderator:
Vladimir said:
That's worser than I thought, I mean, what is so secret about those little registers?

absolutely nothing. in fact, they're so not secret that tomasz is proceeding very well with reverse-engineering of powervr. http://code.google.com/p/gles6410. the only bit he's having difficulty with is the USSE (User-programmable Shader Engine). you would do well to help him out, create a kernel driver based on the TI OMAP equivalent of what he's done (based on information that's a) already available in the form of the TI-published omap powervr driver and based on what's in spruff6a) and then test out the entirely Free-Software-Compliant OpenGL ES 1.1 library he's working on, and then help him to improve it.

it would be nice to be able to tell ImgTech to shove their Q3 2011 promise where the sun doth not shine:
http://ddevnet.net/entry/there-might-be-an-open-powervr-driver-in-q311-6-3910.html

l.
 
Last edited by a moderator:
Seeing as someone has already dug up this thread I'd point out that the 6410 doesn't seem to use an SGX, but rather an in house GPU

http://forum.beyond3d.com/archive/index.php/t-47776.html
 
Back
Top