Achieve Pidsley Level (Rape my RAM)

[GekkoP]

Yep, just want to learn. Otherwise, my Elektra is better than good as it is and there's no point working on RAM usage.

Here's my pstree right after boot.

[pidsley]

OK then, just start looking at the processes in the tree and decide whether you need them running as daemons. Look them up if you don't know what they do. For example, avahi is used to automatically "discover" hardware like printers on a local network. Do you need that running all the time? cupsd? tclsh?

You can shave a little more by making sh your login shell and only running zsh in interactive terminals, and a little more by using a static IP address instead of dhcp. But we're starting to split hairs there, and you may need dhcp on your laptop.

That's actually a pretty clean tree compared to some I've seen.

[GekkoP]

Thank you. I already tried disabling cupsd: 48MB instead of 50. I think avahi can go, I got only one wireless network printer and I have already put that on cups, so it should be ok. tclsh goes with i8kmon, and on my Dell is helping me with temperature and fan usage.
Might try sh at login, then.

[pidsley]

Well there you go. You understand what everything is doing and why it's running, so you've optimized your system. That's great. SomeMost people have 64 ConsoleKit processes running doing absolutely nothing.

(edit) fixed for dkeg :)

[dkeg]

Thread got moved anyway. Probably best

some might be an understatement.

Edit edit, looks like pids and I both edited at same time. :D

[dkeg]

so to chime in with numbers. 23 no X, 43 in X. But more importantly, no unnecessary processes. I have d/l colin, so next free time can start to install and configurate that guy.

[rhowaldt]

hihihi, tried some pstree to see what's up, but got this:

Code: Select all

[22:08:03]$ pstree
systemdââ¬âFAHClientââ¬âFAHClientââ¬âFAHCoreWrapperââ¬âFahCore_a4âââ3*[{FahCore_a4}]
        â           â           â                ââ{FAHCoreWrapper}
        â           â           ââ6*[{FAHClient}]
        â           ââ{FAHClient}
        ââacpi_fakekeyd
        ââacpid
        ââapache2âââ5*[apache2]
        ââbitlbee
        ââconkyâââ5*[{conky}]
        ââcron
        ââcupsdâââlpd
        ââ2*[dbus-daemon]
        ââdbus-launch
        ââgconfd-2
        ââgpm
        ââirqbalance
        ââloginâââbashâââstartxâââxinitââ¬âXorg
        â                               ââshâââopenboxââ¬âfirefox-binââ¬âplugin-containeâââ{plugin-containe}
        â                                              â             ââ22*[{firefox-bin}]
        â                                              ââurxvtâââbashâââpstree
        âântpd
        âârsyslogdâââ3*[{rsyslogd}]
        ââsystemd-journal
        ââsystemd-logind
        ââtint2
        ââudevdâââ2*[udevd]
        ââvnstatd
        ââwpa_supplicant

okay okay, which LANG/LOCALE variable is set wrong according to you guys?

Code: Select all

[22:09:28]$ echo $LANG
en_US.UTF-8

[22:09:32]$ echo $LOCALE

hmmz. since the thread is done, i felt safe hijacking it. still don't see any Consolekit in there, though, so i suppose i'm doing great :D

[ivanovnegro]

Maybe change the locale/encoding (preferably UTF-8) in your terminal emulator?

[rhowaldt]

^ wow.. that did it. fuck me for never thinking of something so simple...

[machinebacon]

rhosey, vnstatd you can remove if you don't need a detailed listing of how much you have up/downloaded, the CONKY stats work without it.
gpm is only interesting if you need a mouse in tty, else it can even cause 'problems' (not really, but happened) in an X session
ntpd is a daemon to sync to network time. I'd run it manually once every now and then instead of giving it a service.

Just ideas :)

[DebianJoe]

To be sincere, this thread has left me thinking about what are the MOST extreme answers to the the common "I'm trying to achieve Pidsley-level RAM usage".

Given two very minimal WM's, which one will generally pull ahead? Is conky more/less bloated than a piped bash script? Are dynamically linked c libraries significantly heavier on resource usage than statically linked ones? (Okay, that last one is just something I'm wondering, but it's an important question.)

We have a good deal of personal experience here, but I'd like to take the time to experiment with a SINGLE option at a time in an attempt to at least put numbers beside some of our subjective opinions on the matter. If you guys have requests in regards to what all get's tested, then simply ask.

BRB: Setting up a "science" partition.

[rhowaldt]

@jules: thanks! i'll consider those :)
@DJ: interesting experiment, you make me curious :)

[DebianJoe]

Okay, experiment #1: Terminal Emulators.

I installed a base "LinuxBBQ Tiny", upg'd, and pulled in git/build-essential/xorg-xserver. From here, I hacked up a very fancy .xinitrc that only included "exec <terminal_name>" and did a fresh reboot before running "free" to see what the used RAM would be.

Since we're going to need a system to measure the results with, I have decided that by taking the MOST bloated example, and subtracting the usage by all of the others, we get a nice whole number in KB that should be representative of the difference. A higher score == less used RAM. Just for fun, we'll call this the "Pidsley Score."

So, on to the terminal emulators and their subsequent Pidsley-score.

Code: Select all

Xterm(base)=  0
urxvt      =  5996
evilvte    =  6724
st         =  7096
Aterm      =  7512

So, aterm pulls ahead of the rest by a narrow margin of 1/2 of a MB. (Note, take the Pidsley-score, divide by 1024, that's your RAM saved in MB). I tried installing Eterm, but couldn't remember the package name to get 75dpi adobe-helevetica fonts...and I don't have a webbrowser installed yet on tiny. Rather than skew the results by adding packages, I decided to let it be. I won't change anything else about the tiny install until the next experiment, so if you have a terminal emulator you want to score...let me know what it is.

[machinebacon]

Bravo. I guess eterm is slightly "heavier" than aterm, so somewhere between aterm and st. What about evilvte? Should be there in RAM usage, too.

[machinebacon]

BTW, tiny should have xlinks2, or elinks-lite, I don't remember. try with 'bbq' from command line, in the network section.

[DebianJoe]

BTW, tiny should have xlinks2, or elinks-lite, I don't remember. try with 'bbq' from command line, in the network section.

I simply assumed that you hadn't put one in there MB, sorry. I'll get evilvte up and add its ranking to the list.

[machinebacon]

If you assume
you make an ASS of U and ME

:D

Kidding. Thanks for the ranking!

[DebianJoe]

Okay, so evilvte is in there. Install, hack .xinitrc, fresh reboot...login + "free". Exact same procedure as the others.

Aterm is still the winner. (I don't have time for looking for browsers when there's SCIENCE to be done.) I will get it up and working if you think that it has any chance against aterm, or if there are other challengers.

What needs tested after terms? WMs?

[machinebacon]

I'd say esetroot/hsetroot/xsetroot/feh

[rhowaldt]

i just dropped by to say that i luvluvluv the Pidsley Score. this is going to get a Wiki Entry for sure, along with the results of DJ's grand sexperiments.