freemangordon | KotCzarny: zram is compcache with a new name, right? | 13:49 |
---|---|---|
KotCzarny | compcache i think compressses pages in mem | 13:51 |
freemangordon | compcache creates block device as well | 13:51 |
M4rtinK | on the other hand we have had some good experience with zRAM (previously called compcache) | 00:30 |
---|
vi__ | freemangordon: hey, FMG. how are things? Did you ever get anywhere with compcache? | 18:31 |
---|
LjL-Alps | 1) is it typically worth to make some adjustments to the default Maemo 5 swap parameters, and if so, are the parameters suggested by Swappolube any good? 2) does compcache at least tend to not make matter worse? | 01:55 |
---|---|---|
LjL-Alps | is it typically worth making changes to swappiness and other parameters (and if so, are the ones proposed by Swappolube any good)? and is compcache worth a shot? | 14:46 |
iDont | divVerent: setting up compcache might buy you enough memory for temporarily disabling your regular swap, i.e. compcache/swapoff/cryptsetup/swapon/(optional)swapoff_compcache | 19:44 |
---|
cehteh | what would help a lot would be more ram or compcache/zram at least | 23:00 |
---|---|---|
cehteh | but nokias kernel changes for swapping introduce some bugs in compcache | 23:00 |
freemangordon | no, compcache, google for it | 12:31 |
---|
vi_ | freemangordon: is there any chance you could enable compcache statistics, at least for some of these testing versions? | 16:28 |
---|
freemangordon | vi_: compcache won't make it :(. No matter what I do, compression/decompression eats all of the cpu. Or I am too stupid and can't think of a better algo | 10:28 |
---|---|---|
freemangordon | vi_: sorry, didn't see your queries, I was tweaking compcache for the last couple of days and didn't look at IRC | 10:40 |
freemangordon | zeq1: for mtd or compcache? | 12:10 |
---|---|---|
freemangordon | ShadowJK: Nokia algo in swap is like spaghetti, I would prefer to reimplement a better algo in compcache | 15:20 |
freemangordon | vi__: do you still use compcache? | 19:25 |
---|
vi_ | dont look at me, I use compcache! | 15:18 |
---|
kerio | (re:compcache) | 07:07 |
---|
user_ | compcache rulez. | 23:42 |
---|
kerio | together with compcache, too - which is stupid | 15:47 |
---|
kerio | hm, swap on uSD + compcache makes things a little bit slower | 00:43 |
---|
kerio | thumb+compcache is pretty great | 16:14 |
---|
vi_ | runs fine, even with 96mb compcache. | 18:26 |
---|
Timmy | can i enable compcache on a bfs kernel? | 21:41 |
---|
vi__ | https://code.google.com/p/compcache/wiki/CompilingAndUsingNew | 13:17 |
---|
Hurrian | japh, compcache and 805MHz fixes most lack-of-ram problems | 13:21 |
---|---|---|
freemangordon | Raimu: you don't have enough free RAM man, I was getting those while I was experimenting with compcache. remove it. it is not needed wit cssu-thumb ;) | 19:21 |
vi__ | Raimu: are you using compcache? | 19:22 |
vi__ | Raimu: Do not change the min_free_kb value when you start compcache. | 19:25 |
freemangordon | unfortunately at this point compcache brings only problems | 19:27 |
Raimu | freemangordon: Hey, is there a reason the compcache baked into KP doesn't give out the stats when rzc asks for them? | 19:28 |
Estel_ | freemangordon, could it be related to compcache? http://talk.maemo.org/showpost.php?p=1225181&postcount=266 | 01:35 |
---|---|---|
Raimu | vi__: Hey -- I was to ask, are you still using compcache on n900? What size/params did you settle on? | 19:09 |
[DarkGUNMAN] | vi_ you been meddling much with compcache? | 12:12 |
---|---|---|
* vi_ would happily take a CPU hit for higher compcache compression. | 12:18 | |
vi_ | AFAI there is somthing broken with backing swap in our compcache. | 13:33 |
vi__ | by how much does n900 compcache compress data? | 23:42 |
---|
freemangordon | even an average TtF of 100000 hours means 700 years according to a document vi_ posted somewhere (compcache thread?) | 17:16 |
---|
vi__ | compcache | 00:40 |
---|---|---|
Estel_ | vi__, as You remember re freemangordon, your compcache setup wasn;t working properly even for opera mini in background and something | 00:41 |
Estel_ | and it failed as soon as *real* swapping wa smvoed out of compcache | 00:47 |
Estel_ | the main problem is that compcache is Your primary swapo | 00:47 |
Estel_ | old, unused data, sits in compcache... | 00:47 |
vi__ | Estel_: did you try compcache yet? | 20:38 |
---|---|---|
Estel_ | in compcache thread on tmo, Your own thread, freemangordon :P | 20:42 |
vi_ | I must add that this crash only happens with those wacky compcache settings. | 20:10 |
---|---|---|
vi_ | ok, I will create a non dsp overclocking profile and test it. I should add, this ONLY happens with compcache module loaded | 20:31 |
freemangordon | vi_, I see, could be a nug in compcache too. ok | 20:32 |
vi_ | freemangordon: COMPCACHE RULEZ | 15:39 |
---|---|---|
Hurrian | side note: tried out your kernel tunables for compcache, it's working really nice ;) | 15:57 |
freemangordon | vi_, 128MB for compcache is not a good idea, I was just able to bring my device to its knees . sygic, ovi maps, microb opening FB 2-3 more light applications and it entered swap trashing inferno | 16:13 |
ShadowJK | I found compcache with backing swap is brutal, because the access pattern seems to be optimized for making flash go slow | 16:24 |
freemangordon | vi_ agree, but by lowering compcache size by half, you will prevent that | 16:34 |
vi_ | Raimu: compcache was always going to be bikg, we just never figured the right settings to use it. | 18:30 |
* luke-jr is pretty sure he's used compcache on N810 and/or N900 | 18:31 |
cehteh | a shame that compcache/zram is broken too :/ | 00:26 |
---|
freemangordon | because what I saw in kernel block driver during my times with compcache is that when there are n pages of swap to be written, the driver searches it's mapping table for continuous space of at least n clusters | 19:03 |
---|
Hurrian|hurr | with swap on sd, and the right swap settings for compcache, you'll get max [strike]placebo[/strike]responsiveness | 18:09 |
---|
Hurrian | on a side note, i tried using compcache again, with 120512KB disk size | 01:50 |
---|---|---|
Hurrian | i'm guessing only certain apps should be allowed to swap out into compcache | 01:52 |
ShadowJK | How does compcache work when it gets full? Does it work like another level or LRU between RAM and real swap? | 02:14 |
user___ | who here uses n900 compcache | 00:00 |
---|---|---|
vi_ | tramp: anyone who has a little experiance of compcache on the b900. | 00:05 |
infobot | vi_ meant: tramp: anyone who has a little experiance of compcache on the n900. | 00:05 |
vi_ | Sicelo: do you use compcache on the n900 | 00:21 |
vi__ | Just got compcache on the go... | 19:00 |
---|---|---|
vi__ | robbiethe1st: apologies, it is calld 'compcache-tools' | 19:44 |
cehteh | more ram and compcache ftw :P | 10:54 |
---|
jurop88 | anybody with hints on ramzez/compcache on N900? | 19:07 |
---|---|---|
DocScrutinizer | jurop88: maybe some user will pop up here and tell about their efort to implement compcache etc into powerkernel | 19:14 |
DocScrutinizer | Shapeshifter: jacekowski cehteh: compcache et al that is | 19:17 |
cehteh | compcache would be vastly more useful, put your efforts into that | 14:18 |
---|---|---|
Hurrian | compcache is available as a patch for 2.6.28, making me think that its one of the things just disabled until tested thoroughly | 14:20 |
cehteh | compcache can be externally compiled and loaded, i just didnt tried yet | 14:21 |
cehteh | make compcache instance of lets say 300MB .. will eat at max approx 150MB ram, leaves 100MB uncompressed ram ...100+300 .. net result is about 400MB | 14:26 |
cehteh | someome should get compcache running and packaged | 10:39 |
---|
Hurrian | Pali , how do I get compcache working in power48? | 13:52 |
---|
vi__ | what about compcache? | 22:23 |
---|
Hurrian | robbiethe1st : ramzswap/compcache + kill allocating task on OOM should be enough for a mobile device | 11:04 |
---|---|---|
Hurrian | gurus, the N9/50 uses compcache/ramz right? | 11:05 |
DocScrutinizer | Hurrian: (N9/50 uses compcache/ramz) tell me how to check | 11:09 |
[DarkGUNMAN] | would be nice to see. Might run better if compcache gets sorted out for power48 kernel | 13:16 |
---|---|---|
[DarkGUNMAN] | Pali is it true you are trying to get compcache to work wth the next verson of kernel-power? | 14:06 |
vi___ | MohammadAG: will compcache make all my dreams come true? | 22:01 |
cehteh | i'd like to have compcache/zram on the n900 ... | 23:09 |
---|---|---|
cehteh | zakkkkm: on what? zram/compcache? guess not | 23:10 |
zakkkkm | cehteh, yeah. zram/compcache | 23:11 |
ShadowJK | Rumours of compcache used on N900+1 by default | 06:13 |
---|
yacc | Actually, I wonder if somebody that compiles the kernel could add compcache, ... | 04:07 |
---|---|---|
yacc | jacekowski, the really cool part about compcache is that it's so much faster than normal hdd based swap that OOM killer starts almost immediatly when memory fills up :) | 04:20 |
alanon_maemo | I remember on the n800 the 1 thing that actually made a real difference was ramzswap / compcache | 18:45 |
---|---|---|
alanon_maemo | do any other mobile os's use ramzswap / compcache ? | 19:02 |
cehteh | please package compcache for the n900 :) | 01:40 |
---|---|---|
iDont | Compcache for the N900 would be great, maybe another time :P | 01:42 |
cehteh | jacekowski: compcache is blazing fast, it works on slower processors | 01:43 |
cehteh | the compcache page has some results | 01:44 |
cehteh | still no official compcache for the n900 .. :/ | 01:35 |
---|
tobis87 | Hi, Anyone else noticed that Pr1.3 changed the way swap works? I noticed a lot of "zero pages" with rzscontrol from compcache. But I have no idea, what this actually does: http://pastebin.com/iJYM4sR5 | 14:26 |
---|
yacc | Wondering => does maemo support compcache? | 00:25 |
---|
100 matches in 3875 log files with 2456684 lines (8.4 seconds).
Generated by irclogsearch.py 2.15.1 by Marius Gedminas - find it at mg.pov.lt!