Messages in this thread |  | | Date | Mon, 11 Sep 2000 13:54:59 -0600 | From | "Jeff V. Merkey" <> | Subject | Re: Availability of kdb |
| |
I'll give it some thought. Most of Linux has better paging than NetWare already -- NetWare is pretty simple in this respect. THe process mapping stuff in Netware (PCB's are mapped to recursively point to themselves with a global brach table) is unique, but not as good as what's in Linux.
:-)
Jeff
Jamie Lokier wrote: > > Jeff V. Merkey wrote: > > One of the best references that describes the bus transaction model for > > Intel in "plain english" is the Pentium Pro Processor System > > Architecture Manual by Tom Shanley of Mindshare, Inc., Addison Wesley, > > ISBN: 0-201-47953-2. It explains a very good detail how the cache > > controllers and MESI works on Intel. > > Agreed, it is a very informative book. More detail than you need unless > you're cloning the chips ;-) Full of good ideas, especially for anyone > interested in memory hierarchies. > > > In NetWare, we didn't care if the page was touched or not since we > > used our own bits in field bits 11-9 to store page specific stuff, > > like whether the page was dirty or not. > > Linux does actually look at both bits, but the optimisation still > applies. Accessed in particular -- ptes are mapped on page faults so > you know the page is about to be accessed. > > > We saw a 4% performance improvement for Network I/O by avoiding the > > "hidden" locking in Intel's architecture. You should check if you need > > this bit, and if not, always create the PTE with it set. I don't know > > how much it would help Linux. NetWare is well optimized for Intel > > processors and any little thing that increased bus memory utilization > > was very noticeable on NetWare, since it supports such huge user loads > > already. > > > > :-) > > Use Linux. Grok Linux. Read the source and weep :-) > But thanks for the tip :-) > > Btw, any good tips you have on effective paging heuristics -- now they > _would_ be useful I'm sure. Linux paging is still rather experimental > after all these years. As in, it works but could do better. > > -- Jamie - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org Please read the FAQ at http://www.tux.org/lkml/
|  |