<maemish_>
Epiphany browser starts quickly but after it opens and you see the page it crashes.
<maemish_>
This is seen in xterm after that
<freemangordon>
get a backtrace
rafael2k has joined #maemo-leste
rafael2k has quit [Quit: Leaving]
rafael2k has joined #maemo-leste
Twig has quit [Remote host closed the connection]
rafael2k_ has joined #maemo-leste
xmn has quit [Ping timeout: 252 seconds]
rafael2k has quit [Ping timeout: 255 seconds]
rafael2k_ is now known as rafael2k
<Wizzup>
freemangordon: could easily be out of ram, too
<Wizzup>
trying to browse the modern web on a n900 is like using a 2004-2005 celeron (which is actually way more powerful) to try and browse the modern web
<Wizzup>
maybe some super simple thing like dillo can work, but it won't support all the new stuff
<sicelo>
yeah. realistically, people should mostly just stop using/obtaining N900 if browsing is a main requirement. there's not a lot that we can do about the web getting bloated, and that trend won't stop
Danct12 has joined #maemo-leste
nohit has quit [Ping timeout: 255 seconds]
nohit has joined #maemo-leste
Danct12 has quit [Remote host closed the connection]
<n900>
but still awesome as an irc client! :)
<Wizzup>
:p
uvos__ has joined #maemo-leste
<uvos__>
Wizzup: ping
<uvos__>
maemish_: sicelo: yeah even the mapphones are pushing it when used to browse modern webpages
<uvos__>
im suprised qtwebrowser manages to dsiplay any page of any complexity on n900
<uvos__>
ff just ooms immidatly so its better than that :)
<freemangordon>
right
<maemish_>
But Maemo 5 with Opera Mobile, Fennec or Micro-b with nginx loads just fine without problems. So it is not just the question of bloated web.
<freemangordon>
maemish_: right, but none of the 'modern' browsers is optimized for 256 MB of ram
<freemangordon>
also, which exactly page microb or opera have no issue loading? besides tmo that is
<freemangordon>
also, do you have swap enabled? also, make sure to disable ramzswap
<maemish_>
I have been using Opera just fine for browsing. Or Fennec. And Micro-b with nginx works suprisingly well.
<freemangordon>
for some reason it is enabled on n900 making things even worse
<freemangordon>
ok, but lets have one site that loads ok in fremantle and use it as a benchmark
<freemangordon>
"fine for browsing" is too general
<maemish_>
I have not done any tweaks for swap. Thought there is swap by default.
<freemangordon>
nope, iirc
<freemangordon>
besides ramzswap and that shall be disabled
<freemangordon>
so make sure to disable ramzswap and enable emmc swap
<freemangordon>
and try again
<freemangordon>
but don;t expect miracles, FF is FUBAR after v78
<maemish_>
Is there any info how to? Like to an idiot level?
<freemangordon>
give me the output of "swapon -s"
<maemish_>
Not having device at hand. At work. Will do in the evening. Thanks for bothering to help.
<freemangordon>
ok, I will be a bit busy in the evening, but in general you need swapon/swapoff commands
<freemangordon>
and you have to identify emmc swap partition
<freemangordon>
(s)fdisk -l
<maemish_>
Have been active on the forum helping others without using N900 for a year. Lot to remember.
<maemish_>
Your info made this idiot to say "Huh?"
<freemangordon>
hmm?
<freemangordon>
why is that
<maemish_>
But thanks. Gonna see if I can with that info and with help from my friend go forward. It is no problem.
<freemangordon>
I mean - maybe ask google how to list enabled swap etc
<freemangordon>
ok
<maemish_>
Thanks.
<sixwheeledbeast>
`blkid -t TYPE=swap` ?
* freemangordon
just learned something new :)
<sixwheeledbeast>
so performance is better if swap is on the device?
<freemangordon>
for leste, yes
<sixwheeledbeast>
ah
<freemangordon>
like for fremantle swap has to be on uSD ;)
<sixwheeledbeast>
that's why i wondered.
<freemangordon>
sixwheeledbeast: on fremantle /opt is on emmc
<uvos__>
idk if thats true
<freemangordon>
uvos__: what is "thats"?
<uvos__>
for d4 a modern sdcard is mutch faster with random writes
<uvos__>
than its emmc
<freemangordon>
not for n900
<uvos__>
and seqential is mutch faster too
<uvos__>
if you enable 100mhz hack
<freemangordon>
we talk n900 here
<uvos__>
n900 can do this hack too
<freemangordon>
but, swap on emc means 2 io channels
<uvos__>
freemangordon: i would not be so sure, i have a sdcard that can saturate the bus at 50Mhz with random writes
<freemangordon>
one for root and another for swap
<uvos__>
ie 25MB/s
<freemangordon>
on n900 emmc does more than 15MB/s
<freemangordon>
iirc
<uvos__>
well thats al lot slower than a good application class sdcard
<Wizzup>
uvos__: pong
<freemangordon>
combined with separate channel gives better performance in general
<freemangordon>
swap io should not wait for rottfs io
<uvos__>
freemangordon: sure maybe, but its not that clear cut as you make it out to be
<freemangordon>
*rootfs
<freemangordon>
also, you talk about some hacks and special cards
<uvos__>
freemangordon: no hacks
<uvos__>
freemangordon: for faster random
<uvos__>
but yes you need a "special" ie modern application class sdcard
<freemangordon>
uSD on n900 is 2 lines, IIRC
<freemangordon>
emmc is 4
<uvos__>
right
<uvos__>
but this dosent matter if emmc cant saturate its bus
<freemangordon>
sure
<freemangordon>
but emmc on n900 is *very* fast
<uvos__>
Wizzup: just wonderning what you wanted
<uvos__>
freemangordon: 15 years ago, sure
<freemangordon>
can;t remember the benchmarks from back then
<maemish_>
You guys can't even believe how lucky I feel to follow this conversation.
<Wizzup>
uvos__: just a sec
<maemish_>
Peak behind the scenes to esoteric info.
<freemangordon>
uvos__: unfortunately google does not give any results for n900 emmc benchmark
<uvos__>
freemangordon: ok well we should disable zswap and add a script that asks the user where to place swap maybe
<freemangordon>
but I can assure you that copy over usb mass storage is done with 16MB/s
<freemangordon>
right
<freemangordon>
zswap on n900 is very bad idea
<uvos__>
thats ~seqential then and really slow,
<bencoh>
16MB/s is very decent for a 2009 device
<uvos__>
compeared to a modern sdcard
<freemangordon>
not really sequential
<sixwheeledbeast>
someone on tmo claimed 13000Kb/s using flasher3.5 output ?
<uvos__>
shoudl be fairly seqentuail, depending on allocations on the card ofc
<bencoh>
I'd say it counts as sequential as well yeah
<freemangordon>
sixwheeledbeast: this is not good benchmark
<freemangordon>
sixwheeledbeast: I remember running benchmarks locally on the device back then
<freemangordon>
but neither can remember nor I can find the results :)
<freemangordon>
uvos__: still, my gut feeling tells me that a separate io channel for swap will be better than even the fastest uSD card, as it is not only the raw speed thet matters
<sixwheeledbeast>
i'd like to stick to my cycling swap defragging method if required, so just interested in performance for leste having all on the card.
<freemangordon>
maybe provide so benchmarks if possible
<freemangordon>
*some
<sixwheeledbeast>
it does make sense bandwidth wise i agree
<uvos__>
freemangordon: that might be true, is deffinatly true when the sdcard is otherwise loaded, but idk how this presents itself in regular usage
<freemangordon>
IIRC omap3 runs its mmc bus on max 54MHz, right?
<uvos__>
something to play around with
<uvos__>
freemangordon: yes, scard spec requires 1.8v trancievers if you want to do more
<uvos__>
which we dont have
<uvos__>
*sdcard
<freemangordon>
so, theoretical bus bandwith for uSD is 54/4 = 13.5 MB/s, no?
<uvos__>
bus with should be 4 and 8 not 2 and 4
<uvos__>
iirc, def true for d4
<freemangordon>
umm, have to check it, but I think on n900 it is 2 and 4
<uvos__>
i thin 2 isent valid for sdhc
<freemangordon>
does anyone have n900 boot dmesg log around?
<freemangordon>
kernel tells us the bus width
<freemangordon>
"mmc1: switch to bus width 1 failed"
<bencoh>
maybe, lemme check
<bencoh>
hmm, nope, sorry. are you looking for fremantle or leste btw?
<freemangordon>
well, we are looking for upstream kernel :)
<bencoh>
ah, nevermind then
<freemangordon>
joerg: do you remember uSD vs eMMC bus widhts on n900?
<freemangordon>
uvos__: right, according to schematic we have MMC_DATA0 to MMC_DATA3 for MMC1 and MMC_DATA0 to MMC_DATA7 for MMC2
<freemangordon>
so uSD is 4 bits
<bencoh>
mmc1 has 4, mmc2 has 8
<freemangordon>
:nod:
<bencoh>
looks like we're looking at the same pdf :]
<freemangordon>
joerg: nevermind
<freemangordon>
bencoh: yeah
<freemangordon>
uvos__: so, max theoretical transfer is 27MB/s
<uvos__>
freemangordon: right so any modern uhs sdcard can massivley outperform emmc
<uvos__>
seqentally
<uvos__>
and A class ones with do mutch faster random
<freemangordon>
I am still not convinced, because those 16MB/s are over USB
<freemangordon>
if we bench eMMC on the device, I will not be surprised if we see like 25MB/s
<freemangordon>
maybe I can do that soon
<freemangordon>
remember back then in HDD times? all the recommendations were to have swap on a separate HDD (and on a separate IDE channel), even if that HDD is slower than rootfs one?
<uvos__>
sure
<freemangordon>
the same here
<uvos__>
but on d4 emmc can do about 40MB/s
<uvos__>
sequentally
<freemangordon>
d4 is another beer
<uvos__>
and like 10 random
<freemangordon>
HW wise those are incomparable
* sixwheeledbeast
still uses spinning disks with two swap partitions on another disk...
<uvos__>
so if n900 is same ballpark the increased random speed helps more than io contention hurts
<freemangordon>
uvos__: afaik eMMC on n900 has very high random speed compared to uSD cards
<freemangordon>
everything but 256 MB of rootfs is on eMMC
<uvos__>
i have a sdcard that can saturate random writes on d4 at 100MHz bus speed
<freemangordon>
wow
<uvos__>
ie about 45MB/s
<uvos__>
its not that special even
<uvos__>
just a A2 card
<freemangordon>
may I have a model? I am struggling to find a good uSD card here, no matter the price
<freemangordon>
*the model
<uvos__>
sure
<uvos__>
but i dont have it on me rn
<freemangordon>
no hurry
<uvos__>
ill post it later
<freemangordon>
ok
rafael2k_ has joined #maemo-leste
rafael2k has quit [Ping timeout: 252 seconds]
rafael2k__ has joined #maemo-leste
rafael2k_ has quit [Ping timeout: 255 seconds]
uvos__ has quit [Ping timeout: 255 seconds]
maemish_ has quit [Quit: Connection closed for inactivity]
<joerg>
>>mmc1 has 4, mmc2 has 8<< just watch out for mmcs getting renamed, at least in fremantle
<joerg>
they swap places/names/numbers
maemish_ has joined #maemo-leste
<joerg>
iirc max speed on uSD in N900 always maxed out at 12.5MB/s, in line with 25MHz@4bit ?
<joerg>
there are faster uSD interfaces but... prolly not for OMP34xx
<joerg>
OMAP*
uvos__ has joined #maemo-leste
<uvos__>
joerg: should be 50MHz (max speed of standart sdhc @ 5V) @4bit
<uvos__>
practicly about 22MB/s or so with overhead
<joerg>
yeah, but... maybe a signal integrity issue? I don't know
<joerg>
anyway the 12some MB/s I recall quite reliably
<joerg>
it's what mc etc shown me when I copied huge amounts of data to/from uSD
<joerg>
possibly for write though, which must be even slower than read
<uvos__>
sure your usd card must be able to saturate the bus ofc
* joerg
idly wonders what uSD if any is in Iron900
<joerg>
hmmm, mount doesn't suggest there's ANY uSD in my IroN900