Re: BFS for the power kernel
 

what about the backported compat-wireless patches for injection form lxp ? Can they also be included ?

If updated regularly AND the whole kernel ecosystem is kept running, with injection, scheduling, f-cam, etc AND providing indeed a faster user experience, then this has the potential of getting (a lot) better than the original power kernel. I would be very tempted.

Thanks for all the effort so far, both you and titan (for providing the power series in the first place) !

Re: BFS for the power kernel
 

I've built the modules for the BFS kernel, but I'm not sure lxp would be happy with me sharing them..

Re: BFS for the power kernel
 

lxp has offered the patches to titan for inclusion in the 'standard' power series (there was no response, as far as I know). Maybe power47 will include it and it's due any day, we (I) don't know. In that case it would come to the derived power-bfs kernel automatically, after you've done your stuff anyway.

Anyway, I guess it's worth a question to lxp ... I donated to get the drivers, as I feel people should get some reward if they put time and effort into it (and need the money), but not smart enough myself to compile it all together with an existing power kernel and keep it all working :)

Re: BFS for the power kernel
 

Yeah, lxp helped me a lot to get them to compile - it was straightforward once I knew the 'trick', but without that, I'd still be trying now :D I'm sure he wouldn't mind; I will send him an email today. However, they are currently built against the v42 of the BFS kernel, so I really need to update my source first!

Re: BFS for the power kernel
 

mmmm based on power46, that would mean all PR1.3 improvements, h-e-n patches, injection patches and bfs (with build in scheduletool) and probably lots of stuff I forgot... nice. Don't forget f-cam drivers, make it 'eyes-closed' installable and you have a winner.

There might be an updated h-e-n out this month (as Joerg was hinting at). Hopefully it's only gui/frontend stuff. If extra kernel mods are needed, then work might have to be redone ...

Also there are kernel patches being developed as part of the improving responsiveness thread, to replace the m-locker library solution.

All these things should ideally come together somewhere and properly performance/stability tested :)

Thanks for getting the already known integration done.

Re: BFS for the power kernel
 

Hello Tigerite mind sharing with me your compiled kernel? Eversince I'm always having trouble compiling bfs to the kernel power and now i really want it badly on lxp's kernel. Anyways I also donated anyways to lxp so I don't think he would mind sharing it to another donator? :p

Re: BFS for the power kernel
 

Of course I'll gladly share the compiled kernel, but lxp's modules that were part of his download don't work with it - it's the modules that need to be compiled against the kernel that's the tricky part. Anyway hopefully I will get chance today to patch my v42 source to the latest v46 and then rebuild the modules against that, I still need to email lxp to make sure he's okay with it first though.

Re: BFS for the power kernel
 

Thanks! Just hit me a PM :)
Well based on lxp's license on his injection package it should be ok. That's the main reason why he included patches and sources on the first place.

Re: BFS for the power kernel
 

What about the wonder-patch? I think the wonder-patch wins over bfs and bfs is made from a noob and what is this for a name? brain **** scheduler?

Re: BFS for the power kernel
 

Are you seriously calling Con Kolivas a noob? He isn't, read his Wikipedia page he started developing CPU schedulers for the linux kernel in 2004.

Re: BFS for the power kernel
 

I'm not interested in vanilla kernel-power and, if I'm thinking about the same patch as you do, the patch won't work for kernel-bfs (as BFS doesn't support cgroups). I even doubt whether the patch will help kernel-power because of the way Nokia configured Maemo.

All the features from kernel-power are also to be found in kernel-bfs (which include those you've named with the exception of the injection drivers, which aren't part of kernel-power either). F-cam drivers can be found on the garage page of kernel-bfs as well.

As of now, kernel-bfs is a niche project; those who actively want to use it most likely will use it, but the vast majority sticks to the vanilla kernel(-power).

Please read up on Con Kolivas. Also, use whatever you think suits you the best :)

Re: BFS for the power kernel
 

5% improvement in pulseaudio isn't that impressive. also it's called stock kernel not vanilla. but it sound interesting the brain **** scheduler :p.

Re: BFS for the power kernel
 

If you can do better, you're welcome :)

Re: BFS for the power kernel
 

dude, what is the difference between the wonder-patch and bfs? and about con he is a dr. med. he is not even a dr. math or dr. informatic. it is his hobby!

Re: BFS for the power kernel
 

I'm not a kernel expert, so I don't know :p (do you?)

I just think that is not so nice to ctiticise other's work only because you wanted something else. If you want wonder-patch so much, I think you should try to apply it to power kernel by yourself instead of keep asking...

Re: BFS for the power kernel
 

:)

yes i want it very much also bfs cannot hurt so plz patch kernel-power with it. it would be awesome. also there is compcache a nice patch too. I want it too.

Re: BFS for the power kernel
 

please stop asking, if you read how it works it turns out that it's impossible to put on our phone. we don't have multiple ttys. you can get the same effect by modifying the cgroups on the stock kernel. go play with that instead.

here, look at this fail: http://forums.internettablettalk.com...ad.php?t=65933

Re: BFS for the power kernel
 

What's your point? The maintainer of the linux kernel scheduler (Ingo Molnar) used the ideas (albeit not the code) from his earlier work to write the CFS scheduler for the kernel. And it seemed to me the reasons that Con's actual work wasn't accepted into the kernel were political more than anything else. Just because he didn't study it at Uni doesn't mean he is clueless in this field.

Re: BFS for the power kernel
 

My point is I'm new to this device I have already ported my tune-up script from my router and other linux box and I have just compiled read-ahead so that I can modify this parameter of my device and it is feeling great and I believe it is more respnsive but I met here a lot of ppl who think of themself higher then the rest but when I asked they don't answer to me or doesn't take me serious. Also with my new read-ahead it is feeling great and I think I don't need to move swapspace to microsd!!

Re: BFS for the power kernel
 

Yeah, I get you're trying to promote what you have done, but half your posts come across like spam, barely related and sometimes not at all to the topic being discussed. And that is hardly a valid reason to denigrate Con Kolivas's work, and imply he doesn't really know what he is doing.

P.S. Your sig is overly large and comes across as obnoxious, can you try to tidy it and make it smaller.

Re: BFS for the power kernel
 

That's not true! Every post from me is related to coding and tune-up! For example I'm also looking for the fleury and hierholzer algo which is in part a key algo for augmentend reality. Also I don't think my sig is too big. Not at the moment!

Re: BFS for the power kernel
 

But they aren't related to this topic. If you want to talk about them, then it is probably better to create your own thread about them. And your sig is bigger than the text of most of your posts, that counts as too big IMO. This has gone quite offtopic, so I'm not going to discuss it further in this thread.

Re: BFS for the power kernel
 

That isn't true! If you know about augmented reality then u will know about fleury and hierholzer algo as tree and also u will know about breadt-first and depth-first alog also u will know that ingo's scheduler is just a red-black tree with some tuneable paramter how he search the tree! If I'm right a scheduler can be build from a quadtree, too. A quadtree offers other benifits but is usually used in distributing algo power to multi-cpu situations.

Re: BFS for the power kernel
 

Okay, so I got the go-ahead from lxp (some days ago) to post the modules freely. Sorry for the delay, I've just been so busy lately. The modules do require a slightly modified kernel, as the rx51power_defconfig requires CONFIG_CFG80211=m - by default it's set to y, perhaps this could be changed for -bfs6? I've also added two patches to the default kernel: one which iDont himself recommended to me (but was unsure whether it was draining more battery, so didn't include it in bfs5), to do with anti I/O stalling - for details go here; the other I found myself, which makes the nandsim module work on the N900 - without it, there isn't enough memory to simulate 256MB of nand (and run anything on the system) and there's no way of creating a backing file. Details of that one can be found here (neither patch needed modifying to apply to BFS so I left them exactly as found on the pages linked).

I have tested the debs myself on my N900 and lxp's modules against the new kernel - everything works for me, at least, but of course, YMMV..

http://ifile.it/uh697nz/kernel-bfs_2...bfs5_armel.deb
http://ifile.it/mws48k1/kernel-bfs-m...bfs5_armel.deb
http://ifile.it/93xaj2f/kernel-bfs-h...bfs5_armel.deb
http://ifile.it/wm3rck2/linux-kernel...bfs5_armel.deb
http://ifile.it/c07nkrv/kernel-bfs-b...bfs5_armel.deb
http://ifile.it/we6jrc0/kernel-bfs-f...bfs5_armel.deb
http://ifile.it/si1jpk7/lxp-modules_2.6.28-bfs5.tar.bz2

Re: BFS for the power kernel
 

Thanks a lot! I've very good result in speed increase with this switch
echo 1 > /proc/sys/kernel/sched_compat_yield but then i get random reboots. Can anybody confirm this please?

Re: BFS for the power kernel
 

I'd gladly test the setting but I don't have a sched_compat_yield in my /proc/sys/kernel and so I can't write to it :confused:

Re: BFS for the power kernel
 

Please stop cluttering this thread. Sched_compat_yield is a tunable for CFS and therefore completely irrelevant to this thread; it doesn't exist in kernel-bfs.

Your last five posts are unrelated to (kernel-)bfs. If you want to discuss things like augmented reality or scheduler models, I recommend starting a new thread. This isn't meant offensive, just pointing it out.

Will be done ;)

That pretty much sums it up. Also, as Adam said himself, the patch needs some testing. I can push it to the kernel-bfs tree, but won't enable it by default for that reason. I didn't experience any problems while using it, although I had the feeling of accelerated battery drain (hard to determine though) as you said. Quantitative tests is what we need :)

I'll send you a PM regarding the nandsim patch later today.

Re: BFS for the power kernel
 

any chance of getting an tiny "multiboot" item/enabler DEB as well for this kernel ? Then it could be an (almost) eyes-closed install :)

Also, do we know upfront how much root space the total of these DEBs will take ? Just as a precaution ...

Thanks for the effort so far !!

Re: BFS for the power kernel
 

The bootimg provides the flasher-capable kernel. It shouldn't be difficult to drop and create a new .ITEM and append -bfs5 to the loader.

Re: BFS for the power kernel
 

Assuming that you already have multiboot and the latest version of this kernel installed:

• install the bootimg
• open xterm
• become root (root or sudo gainroot)
• cd /boot/multiboot
• cp zImage-2.6.28-bfs5 vmlinuz-2.6.28-bfs5
• cd /etc/multiboot.d
• vi BFS.item (or leafpad BFS.item) -> insert the following text

ITEM_NAME="Maemo BFS"
ITEM_KERNEL="2.6.28-bfs5"
ITEM_MODULES=ext3

• save and close
• reboot leaving keyboard open
• select Maemo BFS
• enjoy :D

Re: BFS for the power kernel
 

Isn't he asking for a .deb? Should be done easy!

Re: BFS for the power kernel
 

-_-

Isn't it easy? You have just to copy&paste...what is the difficult part?

Re: BFS for the power kernel
 

I don't have multiboot installed, yet! Also I'm used to edit grub files!

Re: BFS for the power kernel
 

-_-'

Do you want me to teach you how to install and configure it? Maybe I could make you a drawing...

http://talk.maemo.org/msearch.php

If you don't want to use multiboot, then I can't help you, sorry!

Re: BFS for the power kernel
 

No! He was asking!

Re: BFS for the power kernel
 

He thanked me, so I think I solved his problem.

Re: BFS for the power kernel
 

I don't know, but are u really serious? U want me to learn from u? U show me your drawing? this powersearch?

are u up for kidtalk?

Re: BFS for the power kernel
 

This project looks great, installed it on my N900
No lagging here, with GL_HWSync on and schedtool tweak applied.

Havent really had time to do crazy things with 30+ apps. :)

Anyways, can you compile kernel-power-settings (nope, not overclocking, I just like to go to xlv for power savings.)
Pretty please?

Re: BFS for the power kernel
 

I'm glad to hear that :)

Anyway, look a few posts back or click this direct link for kernel-power-settings.

Do note that HAM will try to update it to the official kernel-power-settings, pulling in kernel-power (so do not update it ;))

Re: BFS for the power kernel
 

Thanks!
Time to see what this new kernel could do.