December 2005 - WLUG - wlug.mailman3.com

Re: [Wlug] getting that cpu to sleep ...

by Bradley Noyes

On Mon, Dec 12, 2005 at 06:17:35PM -0500, Jeff Moyer wrote: > ==> Regarding [Wlug] getting that cpu to sleep ...; brad <maitre(a)ccs.neu.edu> adds: > > maitre> greetings, > > maitre> I have a question for those linux programmers out there. i'm > maitre> writing a program on linux (kubuntu to be specific) that requires > maitre> high-precision timing, somewhere on the order of 100 to 1000 micro > maitre> seconds. I get very good timing out of a loop written such that a > maitre> while() loop checks the system clock after every iteration and > maitre> comparing that to the desired wait time, like so (pseudo_code): > maitre> while(time_passed < wait_time) { time_passed = get_time() } ... > > maitre> The problem with this approach is that the process uses 99% of the > maitre> CPU since it just spins the CPU until enough time has expired. My > > The other problem with this is that you can be scheduled if your timeslice > runs out. > umm, i didn't think about that one. > maitre> question: is there a way on linux (in C) to have the CPU sleep or > maitre> send it a no-op such that it's not just eating up cycles until time > maitre> expires? I have looked at nanosleep and usleep, however they appear > maitre> to only have precision to 10ms which too latent for my needs. > > As Mike pointed out, in more recent kernels you can tune the value of HZ to > get better granularity from select/poll. > Actually Mike's response reminded me of an article i bumped into about nanosleep on dragonflyBSD http://www.dragonflybsd.org/docs/nanosleep/ > What, precisely, are you trying to do and on what kernel? Perhaps with > more information we could provide better guidance. > The immediate task at hand is that i want to get readings from an instrument (a light sensor) at a uniform interval, say every 1000 micro seconds. I have been able to achieve this task by setting up the loop i described above. However the loop above takes 99% of the CPU. I'd like to be able to achieve this without pegging the CPU at 99%. Does this make sense? if not i can try explaining it again, perhaps with some code. The kernel i'm running is 2.6.12. (well uname -r reports '2.6.12-9-686-smp.' its ubuntu's stock smp kernel.) -brad

Dec. 13, 2005

Re: [Wlug] getting that cpu to sleep ...

by Jeff Moyer

==> Regarding [Wlug] getting that cpu to sleep ...; brad <maitre(a)ccs.neu.edu> adds: maitre> greetings, maitre> I have a question for those linux programmers out there. i'm maitre> writing a program on linux (kubuntu to be specific) that requires maitre> high-precision timing, somewhere on the order of 100 to 1000 micro maitre> seconds. I get very good timing out of a loop written such that a maitre> while() loop checks the system clock after every iteration and maitre> comparing that to the desired wait time, like so (pseudo_code): maitre> while(time_passed < wait_time) { time_passed = get_time() } ... maitre> The problem with this approach is that the process uses 99% of the maitre> CPU since it just spins the CPU until enough time has expired. My The other problem with this is that you can be scheduled if your timeslice runs out. maitre> question: is there a way on linux (in C) to have the CPU sleep or maitre> send it a no-op such that it's not just eating up cycles until time maitre> expires? I have looked at nanosleep and usleep, however they appear maitre> to only have precision to 10ms which too latent for my needs. As Mike pointed out, in more recent kernels you can tune the value of HZ to get better granularity from select/poll. What, precisely, are you trying to do and on what kernel? Perhaps with more information we could provide better guidance. -Jeff

Dec. 12, 2005

Re: [Wlug] getting that cpu to sleep ...

by Mike Frysinger

On Mon, Dec 12, 2005 at 05:48:24PM -0500, brad wrote: > My question: is here a way on linux (in C) to have the CPU sleep or > send it a no-op such that it's not just eating up cycles until time > expires? I have looked at nanosleep and usleep, however they appear > to only have precision to 10ms which too latent for my needs. you cant unless you tweak the kernel HZ settings read this page: http://kerneltrap.org/node/464 -mike

Dec. 12, 2005

getting that cpu to sleep ...

by brad

greetings, I have a question for those linux programmers out there. i'm writing a program on linux (kubuntu to be specific) that requires high-precision timing, somewhere on the order of 100 to 1000 micro seconds. I get very good timing out of a loop written such that a while() loop checks the system clock after every iteration and comparing that to the desired wait time, like so (pseudo_code): while(time_passed < wait_time) { time_passed = get_time() } ... The problem with this approach is that the process uses 99% of the CPU since it just spins the CPU until enough time has expired. My question: is there a way on linux (in C) to have the CPU sleep or send it a no-op such that it's not just eating up cycles until time expires? I have looked at nanosleep and usleep, however they appear to only have precision to 10ms which too latent for my needs. Thanks, --Brad

Dec. 12, 2005

Success, almost, with GNOME

by Ken Jones

Per the suggestion of Chuck Anderson I logged in under a new account. GNOME desktop worked just as it should. So I removed all of the /home/kjones/.gconf directory and logged in again as myself. Still it did not work correctly. In particular the right click on empty space did not work. I did a "diff -r" on the two (now default versions I suppose) .gconf directories. They seemed the same save for unique identifiers. Where do I look to find the differences? One other question. The Applications launcher on GNOME does not contain the LOG OUT feature. I know the command "shutdown", but I can not find a command to Log Out. Is there one? What is it? Ken Jones

Dec. 12, 2005

Re: [Wlug] Problem with gnome, Fedora core 3

by Chuck Anderson

On Sat, Dec 10, 2005 at 03:58:05PM -0500, Ken Jones wrote: > Any ideas what controls this behavior and ow I can fix it? On a > different machine I have a Core4 system running GNOME perfectly. Try creating a new account and logging in to that one to see if the problem exists there. If not, then it is definately something in your user account, such as your gconf settings.

Dec. 10, 2005

Problem with gnome, Fedora core 3

by Ken Jones

Gentlemen and ladies, My Fedora Core 3 - which has been yum'd up to date - behaves strangely when using GNOME as the windows manager. The right click on blank desktop space does not work. The right click on a panel to make a new panel does not work. The result is that I can not set up any launching icons on my desktop. There is no unused space on my one (bottom) panel. The panel-disappear icon is missing. Any ideas what controls this behavior and ow I can fix it? On a different machine I have a Core4 system running GNOME perfectly. Ken Jones

Dec. 10, 2005

Wanted: used laptop power supply

by Andy Stewart

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 HI gang, I'm looking for a used (functioning) laptop power supply. I won't be using it for a laptop. Input: 120VAC 60Hz (standard wall socket) Output: anywhere from 12 VDC - 48 VDC at about 2 amps I need about 24 watts, so the current goes up as V goes down 48 VDC @ 0.5 A, 24 VDC @ 1.0 A, 12 VDC @ 2.0 A, etc the barrel connector will look like this neg ---C. ---pos that pic won't look right, but it is neg on outside pos on the inside Please send email if you have one, and we can negotiate a price, trade, or whatever you'd like. Ideally, you could bring it to the upcoming WLUG meeting on Wednesday, Dec 14th. Thanks! Andy - -- Andy Stewart, Founder Worcester Linux Users' Group Worcester, MA, USA http://www.wlug.org -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDmm/vHl0iXDssISsRAkEPAJ9FapCV+aOZmtS1wUT2VJ4/C14fCQCeI//x 9H6CbNJklxMzQbepGvSBpa0= =1Gi+ -----END PGP SIGNATURE-----

Dec. 10, 2005

Re: [Wlug] Grub problems after imaging drive

by Brett Russ

On 12/7/05, Dick Goodman <wlug(a)goodman1.net> wrote: > root (hd0,0) > setup (hd0) > quit > > The messages after setup did not look like error messages. The behavior is I need to see those messages. > How do I resolve this? What additional information do you need? The messages above and the output of the grub "geometry (hd0)" command, assuming everything is on the first HDD in the system. Also let us know the intended mount point for each of those partitions. Andy is correct in that the best way to figure this out is to work with the grub command line and walk through the steps it would take trying to boot your system. And Eric is correct in that the grub config file would also be handy. Since the disks are partitioned the same my guess is that the problem lies in the grub stage image installation. I believe some filesystems can get away without a stage1_5 (ext2 being one example, likely used on your RH box) and since you didn't supply an explicit image device to the setup command it's possible that the install wasn't able to find them on the root (hd0,0) that you specified. > Is there a better way to clone Linux drive -> Linux drive? Yes, boot from a Linux CD (like the knoppix that you mentioned) and `dd if=/dev/<good> of=/dev/<blank>` <good> might be hda for example. Obviously blank needs to be as big or bigger than good or you risk an incomplete copy and therefore chopped filesystems. BR

Dec. 9, 2005

Re: [Wlug] Grub problems after imaging drive

by Eric Martin

dick, can you post you /boot/grub/grub.conf (or /boot/grub/menu.lst). menu.lst is the old way, it's probably a symlink now... On 12/7/05, Andy Stewart <andystewart(a)comcast.net> wrote: > > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Dick Goodman wrote: > > I recently configured a RedHat 9 system with OpenVPN . The original > > configuration was done on an old junk box. This is a single boot w/grub > > system > > > > As of last week I knew nothing about grub; I now know marginally more. > > > > As it was working successfully, it was decided to build a new box for > it. > > The hard drive was imaged with Ghost Corporate 8.x and restored onto the > > new drive. There are some known issues with 8.x and grub > > > > The grub screen comes up on booting the new box, but when selecting the > > (one) entry for RH9, after a short pause it comes back to the grub > > screen. I read the grub manual, and from the grub screen did c-command, > > then > > > > root (hd0,0) > > setup (hd0) > > quit > > > > The messages after setup did not look like error messages. The behavior > > is the same, although I see a message about stage 1.5 flicker by. > > I figured out how to look at the hard drive with Knoppix 3.8.2 Live CD > > and the partitions and contents look "ok", so it appears to just be a > > boot issue. > > > > How do I resolve this? What additional information do you need? Is there > > a better way to clone Linux drive -> Linux drive? > > > > Dick > > _______________________________________________ > > Wlug mailing list > > Wlug(a)mail.wlug.org > > http://mail.wlug.org/mailman/listinfo/wlug > > > > Here's what I use for boot commands in GRUB on my system. Its SuSE, but > I don't think it should make a difference. > > root (hd0,0)/vmlinuz <boot parameters> > initrd (hd0,0)/initrd > > - From a GRUB command line, make sure (by using the TAB key) that (hd0,0) > is a legitimate partition, and that it can find /vmlinuz (or whatever > kernel image you are using). If you need initrd, you can use the same > trick with the TAB key to get a poor man's directory listing. > > After you enter those lines, or edit the existing lines to make them > look as above, boot it and you should at least get the kernel to start. > You may need to specify "root=/dev/xxx" as one of the boot parameters > on the "root" command. If you get a panic when it tries to mount root, > it is either because you didn't specify "root=/dev/xxx", or because the > code for the root device isn't present in the kernel (for example, a > SCSI driver or a reiserfs module, as appropriate). I hope root isn't > part of LVM...that's a can of worms you should avoid, but I digress. > > I hope this helps, > > Andy > > - -- > Andy Stewart, Founder > Worcester Linux Users' Group > Worcester, MA, USA > http://www.wlug.org > > -----BEGIN PGP SIGNATURE----- > Version: GnuPG v1.2.5 (GNU/Linux) > Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org > > iD8DBQFDlnkCHl0iXDssISsRAuiaAJ0ZhX3Il4sFM2Yt2cCG1IQBZMy11wCfZIcn > bPoThROtkjrTtXGgaHq2Vms= > =J3lU > -----END PGP SIGNATURE----- > _______________________________________________ > Wlug mailing list > Wlug(a)mail.wlug.org > http://mail.wlug.org/mailman/listinfo/wlug >

Dec. 8, 2005