Re: [IAEP] DC Photo Jam 2 - New batch of photos of SoaS running onnetbooks
On Tue, Mar 17, 2009 at 05:42:36AM -0400, John Watlington wrote: On Mar 17, 2009, at 5:34 AM, Tomeu Vizoso wrote: On Tue, Mar 17, 2009 at 04:48, John Watlington w...@laptop.org wrote: I would expect it to be faster on other processors. I don't think the processor is the biggest factor. In my profiling, I have seen bigger impact from jffs2/nand, the graphics subsystem and the lack of swap. If you run Soas-1 on a sd card with some swap space, you will see how the XO feels much faster. The difference here is that read access to data/programs on an SD card is about twice as fast as JFFS2 (and incurs little processor overhead, whereas JFFS2 tends to occupy the processor 100%), and write access is about 15 times faster than JFFS2. I doubt the speedup is due to the presence of swap. I seem to recall this is one of those conversations that comes up periodically, and that once someone / people created a wiki page with thoughts on a partitioning layout design that could be of broad use. Does anyone know where that wiki page is? I couldn't find it with some naive google searches... Martin pgp0b4Xbomf0N.pgp Description: PGP signature ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: [IAEP] DC Photo Jam 2 - New batch of photos of SoaS running onnetbooks
On Tue, Mar 17, 2009 at 8:58 AM, Martin Dengler mar...@martindengler.comwrote: I seem to recall this is one of those conversations that comes up periodically, and that once someone / people created a wiki page with thoughts on a partitioning layout design that could be of broad use. Does anyone know where that wiki page is? I couldn't find it with some naive google searches... http://wiki.laptop.org/go/Feature_roadmap/Replace_JFFS_file_system_with_better_oneand http://wiki.laptop.org/go/UBIFS -- Luke Faraone http://luke.faraone.cc ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Linux performance without swap (was: Re: [Sugar-devel] [IAEP] DC Photo Jam 2 - New batch of photos of SoaS running onnetbooks)
On Tue, Mar 17, 2009 at 12:58:29PM +, Martin Dengler wrote: On Tue, Mar 17, 2009 at 05:42:36AM -0400, John Watlington wrote: I doubt the speedup is due to the presence of swap. Didn't get the original mail (yet), so replying to this one instead. I don't know if the issue is already fixed in the kernel used on the XO-1, but Linux has a very old bug that causes the system to slow down to a crawl if no swap is enabled and memory is getting tight. I'm afraid I don't have any references, but still encountered it some weeks ago when installing Linux on my phone. Once I enabled swap everything was fine. PS: FUP2 set to sugar-devel CU Sascha -- http://sascha.silbe.org/ http://www.infra-silbe.de/ signature.asc Description: Digital signature ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: [IAEP] DC Photo Jam 2 - New batch of photos of SoaS running onnetbooks
2009/3/17 Martin Dengler mar...@martindengler.com On Tue, Mar 17, 2009 at 05:42:36AM -0400, John Watlington wrote: On Mar 17, 2009, at 5:34 AM, Tomeu Vizoso wrote: On Tue, Mar 17, 2009 at 04:48, John Watlington w...@laptop.org wrote: I would expect it to be faster on other processors. I don't think the processor is the biggest factor. In my profiling, I have seen bigger impact from jffs2/nand, the graphics subsystem and the lack of swap. If you run Soas-1 on a sd card with some swap space, you will see how the XO feels much faster. The difference here is that read access to data/programs on an SD card is about twice as fast as JFFS2 (and incurs little processor overhead, whereas JFFS2 tends to occupy the processor 100%), and write access is about 15 times faster than JFFS2. I doubt the speedup is due to the presence of swap. Adding SWAP on the standard XO distro running from NAND does speed things up a lot. While it may still be faster to use SoaS on an SD card, the difference of just adding a few MB of SWAP memory makes the system able to run more than one/two apps at a time. Last weekend, I managed to crash the XO by having browse open and adding up the Talk activity - this without SWAP - it just crashed the whole system, as the lack of SWAP made it constantly read from disk while removing some pages from memory. I waited 10min and couldn't even login from console, as the time it took to read the login details made it timeout before it could be completed. With SWAP everything was slow but nothing crashed and Talk actually talked to me. Best regards, Tiago Marques I seem to recall this is one of those conversations that comes up periodically, and that once someone / people created a wiki page with thoughts on a partitioning layout design that could be of broad use. Does anyone know where that wiki page is? I couldn't find it with some naive google searches... Martin ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
