Messages in this thread |  | | Date | Sat, 11 Jan 1997 00:34:59 -0500 | From | "David S. Miller" <> | Subject | Re: Since no one else has stepped forward: 'ZeroD' patch |
| |
From: Jes Degn Soerensen <jds@kom.auc.dk> Date: 09 Jan 1997 13:30:28 +0100
Zerod may also be a bad idea for other small-cache architectures such as the sun4c - but I'm not sure about this, guess one of the Sparc hackers will be better to answer this one.
Sun4c has 64k pure virtual cache, it gets flushed so frequently just do to the fact that it is pure virtual that zerod cannot make things any worse. Most of the sun4m caches are either 8k to 16k a piece of split I/D pure physical caches, most have some huge 1MB of so L2 cache sitting underneath.
Therefore I think it might be a good idea to make zerod a config-option or at least an option to disable it?
If this is truly idle time in which zerod is running, and as you mention you have the 256byte caches, anyone doing quick sleeps and the processing for the device driver performing the "event" it is waiting for is going to blow the tasks resident set anyways (I see this on MicroSparc's with 4k data caches as well, any sleep loses your resident set in the cache completely no matter how small the data being touched by the processing is for the sleep nor how short the time you go to sleep is)
---------------------------------------------//// Yow! 11.26 MB/s remote host TCP bandwidth & //// 199 usec remote TCP latency over 100Mb/s //// ethernet. Beat that! //// -----------------------------------------////__________ o David S. Miller, davem@caip.rutgers.edu /_____________/ / // /_/ ><
|  |