Linked by Howard Fosdick on Sat 24th Nov 2012 04:12 UTC
Thread beginning with comment 543087
To view parent comment, click here.
To read all comments associated with this story, please click here.
To view parent comment, click here.
To read all comments associated with this story, please click here.
RE[2]: Just in time for a successor?
by ssokolow on Sat 24th Nov 2012 13:06
in reply to "RE: Just in time for a successor?"
Not sure if I understand this about the drivers... I had gotten the impression that they've already made the drivers open-source. The part that's still proprietary is the chip's internal firmware, which is something you don't need to access its functions.
I suppose it means you can't fix Broadcom's bugs for them, and you can't repurpose the GPU for other computing tasks (OpenCL?), but those seem more like quibbles than real obstacles, to me.
I suppose it means you can't fix Broadcom's bugs for them, and you can't repurpose the GPU for other computing tasks (OpenCL?), but those seem more like quibbles than real obstacles, to me.
The argument is that, if a firmware is so complex that there's a GLSL compiler in it, then it's a co-processor with a closed-source OS of its own, not mere firmware on a subordinate processor.
Heck, on the Pi, the GPU is responsible for loading the image off the SD card and setting up initial state before handing off to the CPU so, if you're of the tinfoil hat persuasion, you could plausibly rant about the possibility of backdoors and the risk of techniques similar to Ken Thompson's hypothetical compromised C compiler.
http://cm.bell-labs.com/who/ken/trust.html
Edited 2012-11-24 13:11 UTC
RE[3]: Just in time for a successor?
by _txf_ on Sat 24th Nov 2012 13:22
in reply to "RE[2]: Just in time for a successor?"
The argument is that, if a firmware is so complex that there's a GLSL compiler in it, then it's a co-processor with a closed-source OS of its own, not mere firmware on a subordinate processor.
As further clarification. The GPU runs what you'd consider a graphics driver on an RTOS (threadX). The application processor running linux then sends commands to the GPU much like an application would send commands to the OGL driver.
Things that require direct access to the hardware like OpenCL, or even X acceleration (Render) cannot be done (without broadcom offering updated firmware).
Objectively, it is quite an interesting encapsulation, but ultimately a useless one if you want access to the hardware. Ultimately, the Rpi foundation was mischaracterising the open nature of their driver.
Edited 2012-11-24 13:25 UTC





Member since:
2012-04-28
Not sure if I understand this about the drivers... I had gotten the impression that they've already made the drivers open-source. The part that's still proprietary is the chip's internal firmware, which is something you don't need to access its functions.
I suppose it means you can't fix Broadcom's bugs for them, and you can't repurpose the GPU for other computing tasks (OpenCL?), but those seem more like quibbles than real obstacles, to me.