Gentoo is less responsive than other distros.[SOLVED]

Mr_Shameless · Last edited by Mr_Shameless on Wed Nov 08, 2006 4:29 pm; edited 1 time in total

Hi,
I guess this is just my own problem. Everybody knows that gentoo is one of the most responsive distro, or at least better than bloated ones like SUSE or Ubuntu. But it's not true in my case.

The first time I installed gentoo, I set my CFLAGS like this:

CFLAGS="-mcpu=pentium2 -O3 -pipe"
CFLAGS="-march=pentium2 -O3 -pipe"

Well, the programs started a little faster than in other distros, but when they had started, their responsiveness was poor. For example in firefox, sometimes it took 2 seconds to switch from one tab to another, and it often took 1 second for a page to start scrolling when I scroll my mouse wheel. The same problem appeared in Gaim and the file managers. There's always some weird "latency".

I have non of these problems in windows or SUSE, slackware. In SUSE, though it took quite long to start a program, I never had any problem like this.

After reading some article in this forum about "How to fly with Gentoo" or something, I started out. I editted the scripts, used hdparm, but it didn't help.

I was a little frustrated so I installed FreeBSD, but it broke my windows box after it'd been installed. So I decided to make one more try and reinstalled Gentoo :lol:

Well, this time I change my make.conf file a bit (after reading articles on Gentoo wiki and this forum). This is how it is:

CHOST="i686-pc-linux-gnu"
#CFLAGS="-mcpu=pentium3 -O2 -pipe"
CFLAGS="-march=pentium3 -O2 -pipe -fomit-frame-pointer -mmmx -msse"
CXXFLAGS="${CFLAGS}"

PORTAGE_TMPDIR=/var/tmp
PORTDIR=/usr/portage
DISTDIR=${PORTDIR}/distfiles
PKGDIR=${PORTDIR}/packages
PORT_LOGDIR=/var/log/portage
#PORTDIR_OVERLAY=/usr/local/portage

FETCHCOMMAND="/usr/bin/wget -t 5 -T 60 --passive-ftp \${URI} -P \${DISTDIR}"
RESUMECOMMAND="/usr/bin/wget -c -t 5 -T 60 --passive-ftp \${URI} -P \${DISTDIR}"

SYNC="rsync://rsync.asia.gentoo.org/gentoo-portage"
PORTAGE_RSYNC_RETRIES="3"
PORTAGE_RSYNC_EXTRA_OPTS=""

EMERGE_DEFAULT_OPTS=""
MAKEOPTS="-j2"

#PORTAGE_NICENESS=3
AUTOCLEAN="yes"
PORTAGE_TMPFS="/dev/shm"

#FEATURES="sandbox buildpkg ccache distcc userpriv usersandbox notitles noclean noauto cvs keeptemp keepwork autoaddcvs"
#FEATURES="sandbox ccache distcc distlocks autoaddcvs"
CCACHE_SIZE="2G"

DISTCC_DIR="${PORTAGE_TMPDIR}/.distcc"

PORTAGE_ELOG_CLASSES="warn error log"
PORTAGE_ELOG_SYSTEM="save"

#These lines are added by ME:
VIDEO_CARDS="i810"
INPUT_DEVICES="mouse keyboard"
LDFLAGS="-Wl,-O1"
CONFIG_PROTECT="-*"

But it seems like there's nothing different. I know it's me who is wrong, but I can't figure out how to make it. Please help me.

O, btw, this is my computer specs:
Celeron (coppermine) 667 MHz,
256MB RAM
i810 integrated graphic card

Thank you. :roll:

Headrush

Insanity5902 · Posted: Sat Oct 28, 2006 5:20 pm Post subject:

Headrush · Posted: Sat Oct 28, 2006 6:03 pm Post subject:

wrc1944 · Posted: Sun Oct 29, 2006 2:06 pm Post subject:

A few thoughts: A Celeron 667Mhz with 256mb is really a bit underpowered these days, so it might be related to frequent swapping out, and a slower cpu with a small cache. Maybe PORTAGE_NICENESS=3 needs to be 10, if he has more problems while emerging stuff in the background.

But I agree with Insanity5902, and it sounds more like a kernel config option problem. Let's see the emerge --info, hdparm -i /dev//hdx, and dmesg.

EDIT:
Forgot to mention: If your color depth is 24bit, try changing it to 16bit. This will not be visually noticable to most users, but significantly improve X performance, especially on lower powered systems like yours.
_________________
Main box- AsRock x370 Gaming K4
Ryzen 7 3700x, 3.6GHz, 16GB GSkill Flare DDR4 3200mhz
Samsung SATA 1000GB, Radeon HD R7 350 2GB DDR5
OpenRC Gentoo ~amd64 plasma, glibc-2.36-r7, gcc-13.2.1_p20230304
kernel-6.8.4 USE=experimental python3_11

padoor · Advocate Joined: 30 Dec 2005 Posts: 4185 Location: india

the slow kde behaviour is noted many times in other distros also
i dont remember exactly what but it has something in kde itself.
probably see the kdmrc setup.
i will look for it too
(try google}
fedora forum had a discussion on this. pclinuxos forum also.
the same problem has comeup here also once. i cant get the thread.
_________________
reach out a little bit more to catch it (DON'T BELIEVE the advocate part under my user name)

Mr_Shameless · Posted: Mon Oct 30, 2006 4:53 am Post subject:

Hi,

Thank you all for your help.

Im quite busy lately. Im taking my TOEFL this Friday, so i dont have much time to do anything else.

Once i've finished it, i will check my linux box, promise :wink:

O, btw, i use blackbox. I never use KDE.

Headrush · Posted: Mon Oct 30, 2006 1:04 pm Post subject:

DeepBass909 · Posted: Mon Oct 30, 2006 3:22 pm Post subject:

lma1980 · Tux's lil' helper Joined: 25 Sep 2005 Posts: 78

_droop_ · l33t Joined: 30 May 2004 Posts: 957

Hi,

Have you tried this ? (the MOZ_DISABLE_PANGO should help)

What option do you use in your kernel (preemption ? timer rate...)

(You should try LDFLAGS/prelink but it will not change the issue that you describe.)

yngwin · Posted: Tue Oct 31, 2006 2:50 pm Post subject:

zxy · Posted: Wed Nov 01, 2006 4:18 pm Post subject:

Did you configure X properly. It slowed down a machine of mine until I had it properly configured.

Another thing, try setting 1000MHz latency and preemptible kernel in kernel options.
Or you can try some of the unsupported kernels, like no-sources. They have cpu-shedulers that work great and latest ones have preemptible-rcu option included, too. This lowers latencies.
Gentoo-sources seemed quite slow to me, when I was using them.

For Celerons I would try -Os option for CFLAGS, as celerons have very small processor (L2) cache. It should make a difference.

And last, but not the least, search forums for hashstyle LDFLAG. There is a new way of linking and it's very quick. I think it's supported upstream too, but I'm not sure. You can find how to install it in nxty's overlay or in conrad install guide. Search the unsuported software in the forums.

Good luck.
_________________
Nature does not hurry, yet everything is accomplished.
Lao Tzu

Mr_Shameless · Posted: Wed Nov 08, 2006 9:49 am Post subject:

Hi, im back.

First, i tried the MOZ_DISABLE_PANGO thing, and that was awesum. Thank you :lol:

Then I had a look at my kernel settings and change it the way you guys told me to, i.e. enabling preemption using the model for low-latency desktops and changing time frequency to 1000Hz. Also, I noticed that my specific CPU type was not set correctly. It was set to Pentium Pro and I changed that into Pentium3/Celeron/Xeon PIII. Its amazing. Amazing!!! My problem has been solved now.

However, I have a few questions. Im just curious. What is 'preemption' and why is the setting for workstations/servers (i dont remember the word) yields more latency than the one for desktops? I thought a server would require more 'power' to do its jobs.
The second question is, why do I have to change the frequency to 1000Hz myself? Why isnt that default? Is there any risk of doing this?

Thank you. :wink:

Mr_Shameless · Posted: Wed Nov 08, 2006 10:14 am Post subject:

O, btw, my computer settings are quite a mess. So, now i should set my CFLAGS like this right?

CFLAGS="-march=pentium3 -O2 -pipe -fomit-frame-pointer" (should it be O2 or Os, as someone suggested?

And then to rebuild the whole system, I do this right?

emerge --newuse world

Thank you all

Mr_Shameless · Posted: Wed Nov 08, 2006 10:21 am Post subject:

Well, from what i read on gentoo wiki, Os is faster but less safe than O2 right?

Ok, but to me its worth trying.

Mr_Shameless · Posted: Wed Nov 08, 2006 11:43 am Post subject:

Oops, the --newuse things are for USE flags right? So, I just need to do this right?
emerge --update world

Sorry, i've made 4 successing posts :lol:

wrc1944 · Posted: Wed Nov 08, 2006 12:43 pm Post subject:

emerge --update world will only update those packages that are updatable.

To rebuild the entire system with your new settings, you need to use:

emerge --ep world (to check and see all the packages to be re-merged and/or updated).

Then just drop the p to actually do it. Be aware, this will take a LONG time, maybe 2-3 days with a Celeron 667mhz.

If a package fails (a few probably will), the emerge will stop, and you'll need to use emerge --resume --skipfirst each time, and then worry about fixing the failed packages later. It's no big deal, and normal procedure, and won't "uninstall" the failed packages.

There's a helpful script at https://forums.gentoo.org/viewtopic.php?p=3669312#3669312 to avoid having to do the --resume --skipfirst, manually, which avoids having the emege -e world quitting on you. Also avoids having to constantly monitor the huge and long emerge. Scroll down towards the end for the emerge -e world version.

Preemptible means that if you have a process running, and you wish to call another process, the kernel can do so without having to wait for the other one to finish. Lately, with more recent 2.6 kernels, this has become less of a problem in Linux, but most desktop users seem to find that setting the preemptible option is still best..

I wouldn't use -Os and select -O2 instead, as -Os is known to break some packages. However, with a low-cache cpu like yours, maybe zxy is correct and it's worth a shot. The only way to know is to try it. Bottom line, I'd use -O2, and not worry about it. Maybe some Celeron users can offer better advice on this one.

I'd also put mmx and sse in your USE flags, so any apps that can use them will compile with support for them.
_________________
Main box- AsRock x370 Gaming K4
Ryzen 7 3700x, 3.6GHz, 16GB GSkill Flare DDR4 3200mhz
Samsung SATA 1000GB, Radeon HD R7 350 2GB DDR5
OpenRC Gentoo ~amd64 plasma, glibc-2.36-r7, gcc-13.2.1_p20230304
kernel-6.8.4 USE=experimental python3_11

batistuta · Veteran Joined: 29 Jul 2005 Posts: 1384 Location: Aachen

Insanity5902 · Posted: Wed Nov 08, 2006 1:42 pm Post subject:

I am glad everything worked out for you. If you could put [SOLVED] in the subject line of the original post that will show others that find search for similar problems and fine this that we came to a solution that worked for you.

A very fast overview of way servers and desktops have different settings is b/c of how they are interacting. A server is a lot of internal interactions of scripts be executing and network connections being served. A desktop is all about user interaction. Those changes improve I/O with the user via keyboard, mouse and monitor output.

Again glad everything worked out , and throw that [SOLVED] in the title.
_________________
Join the adopt an unanswered post initiative today

Mr_Shameless · Posted: Mon Nov 13, 2006 1:27 pm Post subject:

Oh, well, just one more dum question

After I rebuilt my whole system (with the Os option, it didnt break anything), everything went well except my clock. It went 7 hours sooner than my local time. The clock in my windows box is still normal, so I guess my clock has been reset to display UTC time. How do I change it back to local time? I dont know where to look at, because I've always done this during installing linux.

Thank you so much, nice people :lol:

batistuta · Veteran Joined: 29 Jul 2005 Posts: 1384 Location: Aachen

go here and look at the section Hardware clock

Mr_Shameless · Posted: Mon Nov 13, 2006 4:02 pm Post subject:

Thank you.

mikegpitt · Advocate Joined: 22 May 2004 Posts: 3224

You also may want to take a look at your IO scheduler. I personally use CFQ which seems nice for a desktop. Deadline is nice for a server.

BTW - I realsed looking at my config that I'm running at 250Hz instead of 1000. I will have to make the switch and see how things feel.