Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

It works, but I have the feeling pulseaudio process uses slightly more CPU than stock. I'll play a bit with the compiler flags to see what can be done.

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

Well, I hope you are also intelligent enough to post here "uname -a" output when asked three times.

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

Linux Nokia-N900 2.6.28.10-power52 #1 PREEMPT Wed Sep 12 14:17:10 CEST 2012 armv7l GNU/Linux

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

You are right, when playing a .mp3 file (192 kBit/s) pulseaudio uses about 7 - 12% cpu load. Average is 9 - 10%.

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

I would start pointing at NEON as the culprit - IIRC some parts of Pulse were NEON-optimized, as mentioned in one of the config files.

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

new pulseaudio and berkeley (aka contacts) DB is in the repo, I would like to ask those who can risk a bit to test them and report.

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

Updated - one oddity after this one - pulseaudio started to use 15%-17% of CPU time (@500 mhz) just after updating, without any sound file being played. It continue to do so, for last 15 minutes.

Obviously, nothing suspicious in dmesg and syslog - I'll search a little for what may be causing this, and check if reboot fixes it, then report back here.

/Estel

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

After update: about 10-12% on idle at 250MHz (dsp profile) and 30% when playing flac by media player widget (reading by conky)

EDIT- On ideal at 500MHz - 6% idle, 19-21% when playing music

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

Continuation of:
http://talk.maemo.org/showpost.php?p...postcount=1007

After killing pulseaudio manually (from xterm), hildon-home started to use 99% of CPU power, @900mhz. Strange, as I don't have any "multimedia" wikdgets (hardly any widgets at all...). Killing hildon-home Bring everything back to normal - no more oddities.
---

pulseaudio CPU usage, during playback of .ogg q=6 file, is ~7% @ 500 mhz. mafw-dbus-wrapper adds another 3-4%, and openmediaplayer throws 2-3% in. Total CPU usage during playback of said file, with screen on and conky, is 23-30% - would be 5-7% less with device locked and no Conky.

Looks just about OK, if not right, if not "great!" ;)

/Estel

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

pulseaudio-module-bluetooth - conflicts pulseaudio << 0.9.14-2
pulseaudio-esound-compat - conflicts pulseaudio << 0.9.14-2
pulseaudio - conflicts pulseaudio-module-nokia-voice << 0.9.15.0-42

Is that correct? All other pulseaudio-related packages upgraded fine.

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

Both pulseaudio versions do take about 14-16% CPU of 600Mhz when playing "Closer to where you want me" (standard supplied song)

-no difference between old and updated Pulsaudio in CPU usage
-memory footprint: 1,88 % for old and 1,63% for new
-Pulseaudio goes to sleep after quit

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

the correct way is:

stop pulseaudio
start pulseaudio

Though I recommend a reboot, just in case

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

Are you sure you are on PR1.3? As that looks really strange.

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

Exactly same errors here, but fapman let me upgrade them all, and everything works fine. I'm definitely on PR1.3. I remember this error, seen it before (can't recall why I've reinstalled pulseaudio bits), and it ws always harmless.

Control file mess-up by Nokia?

/Estel

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

2 more packages in -thumb - calendar-backend and EDS. Will wait for a couple of days and if there are no issues, will issue a new -mp

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

Installed, no issues so far. What the hell is EDS responsible for, so could do more thorough testing (calendar tested in every strange way I was able to try)?
---

Different thing - lately (= few weeks, maybe longer), I have observed, that many times in idle, xorg uses slightly more % of CPU time, than it used to. May it be similar thing as with thumb packaging of pulseaudio?

/Estel

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

http://wiki.maemo.org/Documentation/...on_Data_Server


Hmm, it might be because I recompiled clutter with -O3 -ffast-math -ftree-vectorize. The same flags that was causing pulseaudio to use more CPU than usual :D. Though for a correct measurement you need to have your CPU frequency locked ;)

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

do we actually need esound-compat for anything? apt-get seems to be fine with deleting it (no packages depend on it), but with dependencies you never know.

does any program actually use esd?

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

No idea, but it seems stock -pr depends on it, so...

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

AFAIK it has been a long-standing "bug" ("upstream") that pulseaudio included the esound compatibility as a dependency. Maybe 10 years ago it made sense. I spotted this package yesterday in my Ubuntu 12.04 and purged it with no adverse effects.

I have just done that on my N900. Seems OK. I don't want to start a discussion on what CSSU should do and what it should not do, but I personally think that it's about time to stop caring about one-to-one compatibility with Nokia PR/SSU, so CSSU-team could test/check if this dependency is actually needed, and if not, remove it.

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

Updated to the new packages.
Something should test?

Anyway I havnt notise anything strange yet.

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

IIRC i read somewhere its for network sound in XDMCP sessions...:confused:

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

should not be required. obviously with XDMCP you'd need a way of getting "network transparency", but pulseaudio does that.

AFAIK no programs use esd anymore. The only rdependency of pulseaudio-esound-compat was libesd0 (in extras/testing/devel), which, for all I care, is not used by anything (I don't even have libesd0 installed).

If at some point some program would use esd (why?) then it will pull libesd0, which in turn will pull the pulseaudio-esound-compat, so no problemo here :)

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

Tested it with locked frequency, and still it seems to use a little more CPU than it should (although, hard to be sure, as difference isn't big - 1-2% - and xorg doesn't use much in idle, anyway).

Do you think it's worthwhile to recompile it for cssu-thumb, like pulseaudio? Also what are supposed benefits of parameter mentioned by you (-O3 -ffast-math -ftree-vectorize)? Maybe increased CPU usage is worth it, if we gain something else?

Sorry, no idea about those compiling parameters here, so question may be silly :)

/Estel

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

Had anyone the experience that alarms didn't sound after upgrading PA packages? Even if i had rebooted 1x after update, not one of my test alarms gave a sound, not on 'silent' nor on 'Normal' profile. Only after the second reboot everything went back to normal.
(Btw: IIRC this happened to me at least 3-4 times before...)

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

updated to latest released files. any thing to check or any changes from previous ones ?

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

New version (21.2011.38-1Tmaemo5.1+thumb4) is out, check the first post for changes

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

Just updated. All good. Thanks freemangordon.

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

Those are optimization parameters. O3 is very aggresive, and requires user-engaging-brain before using (you may want to search for info on pointer aliasing). fast-math will make some floating point operations faster, at the cost of losing strict IEEE floating point compliance (no biggie for most purposes, but again, user needs to know what he's doing). tree-vectorize will attempt to use parallelizing instructions (sse2, neon, etc.)

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

updated to latest version. im facing a problem when i run apt-get update. CSSU-thumb repo updates without any problem. error log attached.

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

There is no error, it is a warning about an expired key, which might be explained by the disinterest of nokia in the N900 repository.

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

Actually, in case of cssu-thumb, it's also matter of merlin's repos being not signed "officially" by certificate acceptable by N900, usually (haven't checked attached logs).

This can be fixed, by adding it to trusted zone manually - command somewhere in this thread.

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

latest update seems to be working fine here. only problem is it failed to automatically reboot on install again. showed "nokia" white screen then turned off.

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

There is no "real error" in your log, those keys are expiried, se here

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

Actually when you look at the warning you'll realize, that it is not an unauthorized key, but rather an expired key (on the nokia repos), the repos on my server (if signed, not all are) are signed by a key which is added to apt-key by the newest community-ssu-enabler, thus residing in the "thrusted zone"

never give answers without checking the logs ;)

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

That's why I have mentioned not checking logs in the first place, but point taken, anyway :)

BTW, where I have missed official info about adding your key to apt-key? Not that there are any trust issues (none of them), but messing with suer's trust zone is something that should be written "bold red" in changelogs, methinks, no matter what.

/Estel

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

latest cssu thumb update working great without any problem yet. N900 is now more fluid and responsive.

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

The latest CSSU update introduced several CSSU repositories besides Community SSU (testing) which I had installed previously - just Community SSU (I guess it is CSSU stable); Community SSU (devel); Community SSU (thumb).

So - instead of following the instructions in first post of this thread - can I just enable Community SSU (thumb) repository in HAM and upgrade my system?

Maybe it is good to include corresponding info also to the first post... :)

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

no, you need to follow the instructions.

Re: [ANNOUNCE] CSSU-thumb thread - stable Thumb2 on N900
 

Well, I have to say I've experienced some weird issues with a PolarCell battery I got from "wannsee-electronic" on eBay.

At first it worked as fine as the OEM battery, but then my phone started disconnecting my 2G/3G connection. Was running Power Kernel and Thumb and at first thought it was the culprit.

This mostly happened when I unlocked the phone and I started typing, few characters made through ssh before it "lost" the network/sim. Tried two different sim-cards, pushing the SIM down with piece of paper etc, nothing seemed to work. Found out that overclocking or underclocking seemed to help, it seemed to be worst at stock speed. It also cut the connection at times when I left the phone with keyboard open and pressed some key on it (and thus activating the keyboard backlight, with the screen still on).

After a while the phone started locking, crashing, and finally it wouldn't even charge or boot most of the time. All this was fixed when I replaced the FAULTY BATTERY. So I'm not sure about the power regulator-theory (or then it also is bit faulty on... faulty batteries :p), but it really seemed to be with some voltage irregularity caused by the battery.