eebrah has quit [Quit: nimehama! tupatane baadaye!]
mike_ has joined #arm-netbook
mike_ is now known as Guest75591
<rzk>
looks like I've fixed my ly-f1 and broke it in one time. now I have android instead of "init sdram fail" from boot0, but battery circuitry died, works only from AC charger
<rzk>
semi-fail reflow :D
mike_ has joined #arm-netbook
mike_ is now known as Guest66426
<rzk>
sitting with 300C heatgun in hand when its 30C in room looks like some sort of S&M.
hno has quit [Ping timeout: 265 seconds]
hno has joined #arm-netbook
gimli has joined #arm-netbook
rmull has joined #arm-netbook
<arete74>
1~
rellla has quit [Quit: ChatZilla 0.9.88.2 [Firefox 13.0.1/20120614114901]]
madmalkav has joined #arm-netbook
gimli_ has joined #arm-netbook
gimli_ has quit [Read error: Connection reset by peer]
<mnemoc>
hopefully this will help to wipe out the crappy arm9/arm11 tablets
newbie has joined #arm-netbook
newbie is now known as hp_
revident has joined #arm-netbook
j1nx has joined #arm-netbook
Quarx has quit []
<rmull>
This mailing list thread is... interesting...
<mnemoc>
freebirds' stuff?
<traeak>
mnemoc: for your kernel config i don't think you need ATA SFF support (it's a PATA backward compat feature) and you don't need Platform AHCI SATA suport, just the Softwinner one.
<traeak>
turning stuff off is good, no ?
<mnemoc>
traeak: a patch to improve sun4i_defconfig (generic linux) and sun4i_crane_defconfig (generic android) is highly welcomed
<traeak>
hehe
<mnemoc>
but tested changes only please :)
<traeak>
i'm still a bit perplexed as to why my sata throughput went down after it was very fast
<traeak>
about a month ago or so
<traeak>
i wonder if i did somehting bad by testing a bunch of drives i had handy (at that time)
<traeak>
regardless
<mnemoc>
tried jumping back to the latest .31 tag?
<traeak>
i did that whole exercise a few weeks back
<mnemoc>
to see if it's really a regression
<mnemoc>
so it's confirmed to be a regression?
<rm>
mnemoc, should I submit my wishlist to the config to github?
<traeak>
the proper answer is for me to dismantle my development laptop and test those drives again
<rm>
haven't actually tried yours yet
<rm>
but the one I started from was so annoying, dunno by whom, wiki doesn't even show page history
<traeak>
i hate the toshiba because they use a miniscule screw which drops into the case way too easily for hard drive rentsion
<mnemoc>
rm: prefer patches over wishlists.... but you can open a ticket for the wishlist too
<traeak>
retension
<traeak>
so anyone else here atall has an sata drive hooked up to their mele?
<mnemoc>
i don't anymore...
lupul has joined #arm-netbook
<mnemoc>
traeak: i'll try to find an spare 2.5 disc this weekend and look
<traeak>
thx
<traeak>
right now with hdparm i'm getting
<traeak>
~35MB/s cached and anywhere form 9.5 to 12.5MB/s disk reads
<traeak>
and i was getting dramatically higher before
<traeak>
not sure what drives the "cached reads" part, i think that's the controller
<rm>
btw, does anyone have a prebuilt armhf deb of 3.4 for the Mele?
<rzk>
finally, my ly-f1 works both linaro and android, time to fix ft5x touchscreen.
<rzk>
god bless chinese guys who invent cheap soldering stations.
<traeak>
mnemoc: thx....i found a spare 3.5" sata drive and tested, shows same throughput...the "Timing cached reads" the the interface speed part
<mnemoc>
traeak: and what do you get for the same disk in a different machine?
<traeak>
mnemoc: ugh, that's another pita problem :-p
<mnemoc>
:)
robws has quit [Ping timeout: 260 seconds]
<traeak>
mnemoc: we're in luck my machine here had a spare sata internal cable plugged in
<traeak>
drive gets 120MB/s
<traeak>
impressive
<traeak>
the sata interface speed at ~6GB/s
<traeak>
hmm....wonder what's on the drive (hehe)
<mnemoc>
not nice
<mnemoc>
no way to justify a 10% performance...
<traeak>
the interface speed is going to be grossly high on a xeon due to the dedicated interface speed
<traeak>
i do recall getting much higher. stupid me though, i didn't do a good job of keeping the old kernels around
<traeak>
meaning: i went on a cleaning spree ~a month ago or so
<mnemoc>
I get 14MB/s read on my uSD...
<specing>
So do I
<specing>
class 10
<traeak>
i just used some old class4 one i had sitting around
<traeak>
just switched from 2g to 4g one
<traeak>
7.7MB/s on each of them
DonkeyHotei has joined #arm-netbook
<specing>
hno: Have you written the gpio tool yet?
<hno>
no
<hno>
i should
<specing>
Would you like me to do it?
<specing>
Maybe it would be easier for me
<specing>
I have an ARM9 devkit here (not sure if armv7-a is backwards compatible wrt. PIO controllers) on which I know where the stuff is
<specing>
Oh hno
robws has joined #arm-netbook
<specing>
We would need to assure that script.bin -- assigned pins are never tested (unless you force it)
<j1nx>
./flashbench -a /dev/mmcblk0 --blocksize=1024
<j1nx>
align 2147483648 pre 534µs on 661µs post 546µs diff 121µs
<j1nx>
align 1073741824 pre 570µs on 719µs post 563µs diff 153µs
<j1nx>
align 536870912 pre 551µs on 700µs post 547µs diff 151µs
<j1nx>
align 268435456 pre 574µs on 708µs post 569µs diff 137µs
<j1nx>
align 134217728 pre 548µs on 676µs post 542µs diff 131µs
<j1nx>
align 67108864 pre 555µs on 681µs post 541µs diff 133µs
<j1nx>
align 33554432 pre 548µs on 691µs post 548µs diff 143µs
<j1nx>
align 16777216 pre 550µs on 680µs post 542µs diff 134µs
<j1nx>
align 8388608 pre 566µs on 707µs post 581µs diff 133µs
<j1nx>
align 4194304 pre 544µs on 558µs post 525µs diff 23.2µs
<j1nx>
align 2097152 pre 539µs on 551µs post 535µs diff 13.5µs
<j1nx>
align 1048576 pre 538µs on 548µs post 535µs diff 11.5µs
<j1nx>
align 524288 pre 537µs on 545µs post 535µs diff 8.74µs
<j1nx>
align 262144 pre 538µs on 548µs post 536µs diff 10.7µs
<mnemoc>
narf
<traeak>
uhoh
<traeak>
kick em!!!
<traeak>
hehe
<j1nx>
Do you see the big jump from 8 MB down to 4 MB (diff 133 us -> diff 23.2us)
<traeak>
mnemoc: great now you make me go learn how to kick someone :-p
<specing>
lol
<mnemoc>
traeak: why do you assume it was me? :)
<j1nx>
Above example shows a erasureblock size of 8Mb
nibb_ has quit [Read error: Connection reset by peer]
<traeak>
probalby worth having j1nx's stuff linked to somewhere on some wiki
nibb_ has joined #arm-netbook
<traeak>
that's another PITA, wiki farming
<j1nx>
So you will have the best performance on your SD if you align the start of your partition to the start of A erasure block
<hno>
j1nx, learn to use a pastebin please, fpaste.org is a good one.
<j1nx>
(I will document it when I am done with the EXT4 tests, but it is common knowledge)
<j1nx>
hno: sorry
<j1nx>
didn't think
<traeak>
j1nx: i know, but it's probably best to remind n00bs who have only run with 2.5/3.5" drives that they need to do some proper prep work on their bootable SD cards
<j1nx>
So if you have 8Mb rerasure blocks you calculate the start of your partion by: 8*1024^2/512 = 16384
<gimli>
j1nx: alsa ?
<j1nx>
Now this commonly used +16M you see everywhere for the is bigger then 16384
<hno>
j1nx, I wonder if not the tool is flawed. Unless the SD vendors are brain-damaged block size should be aligned to the default partition, which I guess in most is starting at sector 63.
<rm>
j1nx, there is no way there could be 8M erasure blocks
<j1nx>
gimli: Sry, not yet. Still fighting to have the rootfs as I want it to be, but keep getting back to the drawing board because of found flaws
<rm>
just no way
<rm>
on 2 SD cards the erase block is 64K on one, 128K on another
<rm>
this is readable from the card itself
<j1nx>
rm: erase blocks are commonly between 2 Mb and 16 Mb. Usually 4 Mb
<traeak>
mnemoc: gotcha, thx...i'm just wondering if i did something stupid and accidentally partly fried something is all
<j1nx>
rm: Pages, erase blocks and segments
<j1nx>
rm: Although it is usually possible to write single pages, the data cannot be overwritten without being erased first, and erasing is only possible in much larger units, typically between 128KB and 2MB
mike__ has quit [Remote host closed the connection]
mikey_w has quit [Remote host closed the connection]
<rm>
j1nx, it can if it has free erase block-sized areas, no?
Guest66426 has quit [Remote host closed the connection]
Guest75591 has quit [Remote host closed the connection]
<j1nx>
rm: read the lwn.net page. Good stuff there
robws has quit [Ping timeout: 244 seconds]
robws has joined #arm-netbook
<hno>
wonder if the SD standard have "discard". Probably it does.
<j1nx>
gimli: I am again rsync-ing a rootfs to the card. If I am happy I will take a look at ALSA
<traeak>
anyways i'm scratching my head at all the sd card stuff above, noticing that probably all my bootable cards need to be reformatted
revident has quit [Quit: Combustible lemons? Bah, I bring you weaponized asparagus!]
<traeak>
and repartitioned
<j1nx>
mnemoc: I guess this SD perfance stuff belongs more on your linux-sunxi wiki then it belongs on the rhombut wiki isn't it
<j1nx>
When I find some time I will add it
<mnemoc>
j1nx: :thumb-up: :)
<j1nx>
Although "noatime,nodiratime" will give the biggest boost
<j1nx>
linux boot parameter “elevator=noop” in u-boot will do fine as well
<traeak>
current SSDs i guess can rock and roll with sequential data streams
<mnemoc>
Timing buffered disk reads: 100 MB in 0.52 seconds = 194.11 MB/sec <--- cheap SSD on my amd e350 laptop
<traeak>
in fact i may have to get one of those for work stuff...set up a sata3 with some data to try to simulate a high throughput raid controller (that we can't afford)
<Turl>
mnemoc: how cheap?
<mnemoc>
~$80 for 120GB
<traeak>
they've droped a ton price wise
<mnemoc>
90 maybe... but it was less than 100
<traeak>
what brand/model? :-p
<mnemoc>
ocz vertex 2
<traeak>
okay...wasa looking at a vertex 4 for the simulation
<traeak>
price break still looks to be ~120GB for ssd's
<mnemoc>
I use deadline elevator... but i've never benchmarked them... buying an SSD was more of battery-life choice than performance
<Turl>
my battery is shot really
<Turl>
so should I buy one, it's for the performance
<mnemoc>
your is a desktop with builtin ups, not a laptop :p
<traeak>
heh
<traeak>
they also call those "boat anchors"
<hno>
Interesting. mSATA seems to take off for SSDs.
<hno>
A goal for next laptop is to have a reasonable SSD for system + a large HDD for various data not used all the time. mSATA might expand the available options there a bit.
<traeak>
or buy a laptop with an optical drive and replace it with an HDD caddy
<traeak>
that unfortunately forces you into 13.3", more likely 14"
<mnemoc>
13" is the max for a laptop imo
<mnemoc>
14" is already too large to carry
<mnemoc>
not that I carry mine a lot these days...
<hno>
I actually need an opticaldrive as well.
<hno>
well, maybe not if the HDD is large enought actually.
<xxiao>
mnemoc: watch out for your ocz ssd
<xxiao>
i assume you have a good backup strategy for that
<xxiao>
there is a known silicon bug that could render your SSD a sudden death and will never come back alive
<xxiao>
specifically for OCZ models
<traeak>
that's sweet
<xxiao>
i bought 8 of them and stress them for a SAN project a few months back
<mnemoc>
xxiao: I already RMAed it once :<
<traeak>
that's part of the PITA...trying to research what doesn't suck
<xxiao>
after 4 months of stress tests, 2 are totally dead
<Turl>
hno: look into hybrid disks too
<Turl>
hno: 'speed of a ssd' and huge storage
<traeak>
kind of like the old days...80GB western digital drives were the kiss of death
<Turl>
traeak: wait wha..
<Turl>
traeak: there's a 80GB WD @ home
<Turl>
been running for 10+ years :)
<traeak>
lucky you
<Turl>
on a P4 :P
<traeak>
we had a 50% fail rate on those :-p
<traeak>
athlonMP days
<traeak>
then the 120GB maxtors came out
<xxiao>
unless i need drive on the move, say in the car where mechanical just won't do, i use will SATA/SAS instead
<mnemoc>
xxiao: I was able to restore the data copying raw chunks of 4M each reconnecting a usb/sata adapter each time
<traeak>
those NEVER fail
<traeak>
anyways
<mnemoc>
xxiao: it halted short after reading 4M. reconnecting it did the magic
<xxiao>
mnemoc: in my case i have no data to recover, just reading/writing them under stress with various patterns of data
<traeak>
Turl: used to build 8 drive raid arrays with 3ware cards. find out really fast what's good and what's junk
<Turl>
:)
<xxiao>
once they're dead, i could never reuse them, include "reset" them based on some googling
<Turl>
traeak: I also have a maxtor 40?GB iirc with shitty sectors :<
<Turl>
from the P3 era
<traeak>
Turl: can't remmeber but the 40's may have been an older model
<mnemoc>
xxiao: I keep all important data in "the cloud", and the not important (music and stuff) in an external esata/usb raid1 device
<Turl>
that hard disk itself was a replacement for a smaller one that died btw
<Turl>
mnemoc: so you have your GBs of code on 'the cloud' ? :P
<xxiao>
talking about cloud, i'm thinking about use EC2 to do kernel compiling etc
<DonkeyHotei>
i remember some nasty failure of WD 160GB drives
<xxiao>
i have an account and really did not use it
<mnemoc>
Turl: normally my code is pretty small.... that awful linux-allwinner repo is the only exception
<xxiao>
how does it charge, by the time i run stuff there, or like a vps hosting plan charged monthly?
<Turl>
xxiao: unless you want to end bankrupt, I wouldn't :P
<DonkeyHotei>
and let's not forget the DeskStars
<traeak>
DonkeyHotei: WD were untouchable until the RE4 series came out
<traeak>
deathstars you mean
<DonkeyHotei>
yep
<mnemoc>
Turl: github has pretty decent plans for private repos :p
<xxiao>
Turl: i have to understand their pay scheme
<Turl>
seagate was pretty bad too wasn't it?
<traeak>
they ran too hot
<DonkeyHotei>
no, seagate's troubles were more recent
<Turl>
xxiao: usually renting a vps or dedicated server is cheaper
<xxiao>
traeak: right, SAS is also very hot, SSD is extremely cool
<DonkeyHotei>
mostly after the maxtor acquisition
<xxiao>
Turl: i have vps@linode, but not powerful enough to compile
<DonkeyHotei>
before they bought maxtor, they were top-notch
<Turl>
mnemoc: I pay like.. 1/3 of the cheapest github plan :)
* xxiao
may be able to leverage his linaro account
<Turl>
(the cheapest is like 9 bucks right?)
<ZaEarl>
hno, our next ultrathin will have an msata ssd plus a 1 tb hd.
<traeak>
i work somewhere else so i don't know aymore...been using samsung for spinning. they definitely run cool which seems to be a decent indicator of lifespan
<Turl>
I doubt github has a student discount of any sorts
<mnemoc>
Turl: I rent a real server with raid1 and have two fat real servers in the same DC from my employer.... but for small paid projects I prefer github instead of self-maintained stuff
<xxiao>
mnemoc: just curious, for paid projects why not bitbucket, which is what i use, it's good
j1nx has quit [Quit: Life without danger is a waste of oxygen]
<Turl>
mnemoc: for random snippets, ancient project, simple test and one shot code, I use dropbox
<Turl>
mnemoc: for free software I github
<mnemoc>
xxiao: bitbucket was hg only until short ago. and I'm more confortable with git
<xxiao>
mnemoc: i only use git with bitbucket
<Turl>
mnemoc: the rest is uni work and such, which I just set-url origin myserver:project.git and push through ssd
<Turl>
ssh*
<Turl>
mnemoc: bitbucket has free private repos now I think
<mnemoc>
:)
<Turl>
and they support git
<xxiao>
it's _very_ good
<mnemoc>
too lazy^Wbusy to consider relocation without a good reason
<Turl>
the only downside of github is that they go down at random :(
<hno>
ZaEarl, interesting, but I am not looking for a ultrathin. Need something that can boost noticeable CPU power and reasonable battery time.
<ZaEarl>
that's why I love git, etc. when our perforce server would go down, everyone was screwed. now it's not too bad when github is down.
<traeak>
hyow about a kickstarter project: "bug the chinese SOC vendors to open source the drivers that are supposed to be open source"
dfletcher has quit [Ping timeout: 245 seconds]
furan has joined #arm-netbook
<furan>
anybody tried any flash upgrades (soldering larger nand tsop) for the allwinner a10 devices?
dfletcher has joined #arm-netbook
<Turl>
the air remote is interesting
<mnemoc>
good night
<hno>
furan, not that I know of. But some have added a second NAND chip. Quite many A10 devices have space for a second NAND chip but only ship with one.
<traeak>
people with steady hands that can still see
<hno>
or reasonabe reflow equipment.
nibb_ has joined #arm-netbook
nibb__ has quit [Read error: Connection reset by peer]
DonkeyHotei has quit [Read error: Connection reset by peer]
DonkeyHotei has joined #arm-netbook
<CIA-122>
rhombus-tech: Jeff master * r58734d067506 /allwinner_a10/orders/kraln.mdwn:
<CIA-122>
rhombus-tech: Jeff master * re6348c8bfabc /allwinner_a10/orders/kraln.mdwn:
<CIA-122>
rhombus-tech: Courtney master * rfe3ad634085a /allwinner_a10/orders/fettsvett.mdwn: