Hi Peter! I hope you don't mind if I cc:ed the reply to kernelnewbies as well.... On Sun, Jul 27, 2008 at 4:34 PM, Peter Teoh <htmldeveloper@xxxxxxxxx> wrote: > BTW....there is a talk on PAT today...(page attribute table)...really puzzle me: > > > > the key api is set_memory_wb.....needed by X window for > performance....which he mentioned that writing to devices should not > cache in memory.......know anything about memory aliasing....??? OK, I have trouble understanding it too, so first thing to do seems like reading http://en.wikipedia.org/wiki/Aliasing_(computing) for both of us :D :D :D But logically, I guess by making I/O mmapped memory region (am I using right term guys?) uncached, you simply make any modification goes directly to the device... and that certainly beneficial in some cases (video card's memory is modified in order to manipulate certain bits...it happens instantly rather than delayed). It's unlike block access where you might get more benefit when doing delayed access (think about possibility of merging adjacent block writes). regards, Mulyadi. -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx Please read the FAQ at http://kernelnewbies.org/FAQ
