| |||||||||||||
|
| |||||||||||||
Memory management for graphics processorsMemory management for graphics processorsPosted Nov 9, 2007 1:11 UTC (Fri) by zlynx (subscriber, #2285)In reply to: Memory management for graphics processors by jwb Parent article: Memory management for graphics processors
I believe that the Linux kernel DRM (direct rendering) modules are supposed to check the command stream to ensure that cannot happen. I suppose that with a very optimized DMA design, this might not be possible. In that case you'd have to trust the X server and OpenGL driver to take care of it.
(Log in to post comments)
Memory management for graphics processors Posted Nov 9, 2007 19:05 UTC (Fri) by nix (subscriber, #2304) [Link] Unless the commands are so simple they aren't Turing-complete, or their form is extremely restricted and stereotyped, I suspect this reduces to solving Rice's theorem, which is of course impossible in the general case for just the same reason solving the halting problem is, and is ridiculously difficult even in most useful special cases.
|
|
||||||||||||