mnemoc changed the topic of #arm-netbook to: EOMA: Embedded Open Modular Architecture - Don't ask to ask. Just ask! - http://elinux.org/Embedded_Open_Modular_Architecture/EOMA-68 - ML arm-netbook@lists.phcomp.co.uk - Logs http://ibot.rikers.org/%23arm-netbook or http://irclog.whitequark.org/arm-netbook/ - http://rhombus-tech.net/
<hramrach> does u-boot not work with hardfloat?
voronaam has quit [Quit: Leaving.]
<aholler_> no, it doesn't use floats
<hramrach> it does not use floats so does not work when compiled with float support?
<hramrach> sounds odd
<aholler_> double negotation
Jef91|AFK is now known as Jef91
<hramrach> the weird part is that adding float support to the compiler breaks program that does not use float
<hramrach> but perhaps it has manually crafted assembly routines which would have to be updated to the new abi
<aholler_> 1. it doesn't make sense to compile it float support. 2. something (irqs or similiar) might fail because they don't care and don't save the registers
<aholler_> s/ it / with /
<ibot> aholler_ meant: 1. with doesn't make sense to compile it float support. 2. something (irqs or similiar) might fail because they don't care and don't save the registers
<aholler_> ibot: learn fuzzy ;)
<hramrach> it makes sense to compile it with the default compiler
<hramrach> and if the registers are unused it shall not fail if not saved
<hramrach> but the abi is different
<aholler_> which abi?
<hramrach> not sure. you can't mix softfloat and hardfload
<hramrach> probably different calling convention
<hramrach> or something
<aholler_> it isn't userland. it doesn't use floats, no float-abi in u-boot
<hramrach> then why it would fail with hf?
<aholler_> who said that?
<aholler_> read what you asked and what I've answered
Jef91 has quit [Ping timeout: 248 seconds]
<aholler_> it might fail, but should not
<hramrach> oh well
<aholler_> anyway it doesn't make sense and default u-boot make disables floats at all
<hramrach> the page recommends softload compiler
<rz2k> it is not recomended to use neon/vfp in u-boot
<aholler_> as said, you might run into problems if the compiler would use some float-stuff to optimize something
<aholler_> but I don't think it does that if you don't call it with -mfpu=foo or similiar
<aholler_> and the makefiles from u-boot don't do that
<hramrach> there is no man page for gcc -.-
<aholler_> just build u-boot and don't care about your compiler. it does the right thing
<aholler_> at least if you don't use a compiler like 4.5 with bugs
<hramrach> 4.5 has bugs?
Jef91 has joined #arm-netbook
<hramrach> arm-linux-gnueabihf-gcc (Debian 4.5.3-12) 4.5.3
<aholler_> no, never ;) but if unpatched, volatile doesn't work with 4.5.0 and 4.5.1
<hramrach> awesome
<hramrach> is that normal that kernel fails to build when you enable random options?
<hramrach> ipv6 nf and nf6 would probably apply in this case
<hramrach> net/built-in.o: In function `tcp_is_local6': linux-2.6/net/ipv4/tcp.c:3366: undefined reference to `rt6_lookup'
<aholler_> git grep rt6_lookup, enable missing options, try again.
<aholler_> such happens
<hramrach> git has builtin grep?
<hramrach> cool
Jef91 is now known as Jef91|AFK
<hramrach> the function is in ipv6/route.c and does not seem to be ifdefed
<hramrach> how do I tell what selects that file?
<aholler_> look at the Makefile
<aholler_> or do an educated guess ;)
<hramrach> it's the default :s
<hramrach> thanks for suggestions
<hramrach> but can't fid what broke it
<hramrach> just building with ipv6 does not work
<aholler_> do you have ipv6 as a module? make it static
<hramrach> yes, have it as module
<hramrach> it used to work
<aholler_> seems not to be the case with your random config options
stefanro has quit [Ping timeout: 255 seconds]
<hramrach> yes, as builtin it works
<hramrach> thanks
uwe__ has joined #arm-netbook
uwe_ has quit [Ping timeout: 255 seconds]
stefanro has joined #arm-netbook
traeak has quit [Remote host closed the connection]
RITRedbeard has joined #arm-netbook
nemik_ has joined #arm-netbook
nemik has quit [Ping timeout: 260 seconds]
lerc has joined #arm-netbook
Jef91|AFK is now known as Jef91
rz2k has quit []
kaspter has joined #arm-netbook
soooga has joined #arm-netbook
Jef91 has quit [Quit: Instantbird 1.3 -- http://www.instantbird.com]
gsilvis has quit [Read error: Connection reset by peer]
gsilvis has joined #arm-netbook
rz2k has joined #arm-netbook
<bsdfox_> are people using sunxi-3.0 or 3.4?
aholler has joined #arm-netbook
aholler_ has quit [Ping timeout: 248 seconds]
servili007 has quit [Read error: Connection reset by peer]
cheng has joined #arm-netbook
ganbold_ has joined #arm-netbook
Quarx has joined #arm-netbook
rz2k has quit []
gimli has joined #arm-netbook
cheng has quit [Quit: Leaving]
ZaEarl has quit [Ping timeout: 276 seconds]
kaspter has quit [Remote host closed the connection]
Yaku has joined #arm-netbook
eebrah has joined #arm-netbook
<provel_> ssvb: hi! Are the drivers working actually? I can't find a way to make them work with mali framebuffer ones
eebrah has quit [Ping timeout: 246 seconds]
soooga has quit [Quit: Leaving]
rellla has joined #arm-netbook
<provel_> anoyone else noticed that samba is very slow? seems it can't go over 500KB/s
<provel_> also ethtool gives error checking wakeonlan (expected) and linkstatus (less expected...)
<provel_> ok.... eth0 seems set to 10Mbit half duplex.... this explains something.... but how I fix it?
<provel_> (emele a2000)
<provel_> still not that fast, samba now 2MB/sec
<rellla> network performance should be tested (with iperf etc.) if realistic speed can be reached, even if ethtool tells you 100MBs/FD.
<rellla> i'm not for sure at my environment, because my switch/infrastructure seems to be slow at all.
<provel_> TY
<provel_> changed emac to 19 ... works... now I'll try to test
<provel_> no, speed still looks less than 2 MB
<provel_> at least with samba
<provel_> I'll try to completely remove the emac_power line
<provel_> no way... 2MB/s max with samba
<rellla> did you try another benchmarking tool, e.g. iperf?
<rellla> 2MB/s at both directions?
<techn_> how about ftp?
<provel_> [ 3] 0.0-10.1 sec 28.1 MBytes 23.4 Mbits/sec
<provel_> iperf result seems better
<provel_> this as iperf client
<techn_> provel_: whats your freq?
<provel_> similar as server : [ 3] 0.0-10.1 sec 25.5 MBytes 21.3 Mbits/sec
<provel_> my freq?
<techn_> cpu freq
<provel_> how I see? I have ondemand
<provel_> mele a2000
<techn_> cat /proc/cpuinfo
<provel_> mele a2000mips
<provel_> up to 1008 bogomips
<provel_> no freq there
<provel_> bogomips from 59 to 1008
<techn_> oh.. try this: http://linux-sunxi.org/Cpufreq
<provel_> hold a sec... I was with wireless pc link
<buZz> swap ondemand for performance :)
<provel_> with wired it's [ 3] 0.0-10.0 sec 49.1 MBytes 41.0 Mbits/sec
<provel_> done ..... [ 3] 0.0-10.0 sec 58.4 MBytes 48.8 Mbits/sec
<provel_> ethtool says full duplex
<rellla> uuaaahhh
<provel_> keeping it more times sequantially I got [ 3] 0.0-10.0 sec 60.5 MBytes 50.6 Mbits/sec
<rellla> techn_: thanks, i missed changing to performance on my second mele. now iperf says:
<rellla> [ 4] 0.0-10.1 sec 107 MBytes 89.5 Mbits/sec
<rellla> seems to have it done so far
<provel_> rellla.... what you inserted in emac_power line in the fex ? nothing or 19?
<rellla> provel_: not sure anymore. i'm "refexing" my bins. wait a minute.
<rellla> according to my script.bin's filesize it should be without the power entry. on mele a2000 and a100.
<rellla> iperf server: mele a100
<rellla> iperf client: mele a2000 -> 89MB/s
<rellla> iperf server: mele a2000
<rellla> iperf client: mele a100 -> 49MB/s
<rellla> hm..?
<rellla> is this normal behaviour or a problem with my switch: http://pastebin.com/v81aHiBb
<rellla> or an issue with my meles?
jquip has joined #arm-netbook
jquip has quit [Ping timeout: 255 seconds]
<provel_> dunno, I have just one line in iperf
<rellla> started with flag "-d". bidirectional test
<provel_> ok, I check
<provel_> Whew
<provel_> mine really different
<provel_> gets faster in opposite direction...
<provel_> started on PC
<provel_> started on mele it's still faster but in first test instead than second (client)
<provel_> looks like mele as client is much faster for me, 86/90 MBs
<provel_> *Mb/s, sorry
<rellla> so to get reliable results, i should start with a minimal wired network infratructure.... i think, there could be some problems with my wiring.
<rellla> looks like 90MB/s is quite good
<provel_> yes but only with mele as client testing bidir
<provel_> hum, also in no-bidir
<provel_> mele as server maxspeed=55Mbs
<provel_> now I'll recheck mele as samba server
<rellla> i'm no network expert. i also have problems with i386 at work atm. GBit-Net, and a poor performance to my linux machine. seems to result on a bad interface/driver. windows to windows machine performance is good. windows -> samba is VERY bad at least in one direction... <- to be done in the future ;-(
<provel_> maxspd = 5.9MBs, avg = 4,5 MBs ca
<rellla> so it's a samba problem ?!
<provel_> no, 55Mbs is slightly more than 6 MBs
<provel_> so it's decent
<provel_> samba it's decent.... eth0 is still about half speed
<rellla> oh. Mb /MB, sorry.
hg_5 has joined #arm-netbook
<provel_> np.... however this way it's for sure way better than initail issue... maybe this change should be reflected in sunxi-bsp fex file for mele a1000
<rellla> it should. and be commented on issue report at github. and to give a hint to governor settings.
<hramrach> provel_: if you achieve half speed of interface with sambe it's cool
<hramrach> try netperf or something
<hramrach> oh, you did
<hramrach> so try if htere are some parameters that can be tweaked with miitool or ethtool or what
<hramrach> but when the etherenet card is cheap don't expect miracles
<hramrach> server should be easy, though
<rellla> hramrach: ethernet card is cheap, as a10 is cheap ;-)
<hramrach> I can achieve 400Mbit on Realtek Gbit card
<hramrach> you get the idea
<hramrach> but the problem is receivin there
<hramrach> probably not good enough interrupt mitigation
<hramrach> I don't really see how sending can go wrong if the card is not total fail but there might be slight ahrdware or driver flaw and you get issues
<provel_> hramrach: it's half speed when transmitting
<provel_> hramrach: and 85-90% speed when receiving
<provel_> Gbit cards are different and have different issues
<provel_> cable for example is not a real issue with Fast Eth.
<provel_> but can be a limitation with GbEth
<provel_> the slower part is mele sending.... likely due to some driver issue
<provel_> mele recv is 85-90% speed which is good for a little box
<provel_> in any case samba 3 seems working fine... samba 4 still to test
<hramrach> I can test speed when I get the board booting
<hramrach> need to find a uSD card
penguin42 has joined #arm-netbook
<jukivili> does vendor driver work with wifi in mk802?
<penguin42> ooh, my mk809 has arrived
<orly_owl> 9? a10?
<penguin42> 809 - rk3066
<orly_owl> oh, meh
<rellla> provel_: ~90Mb/s from a100 to my dockstar and vice versa
<rellla> bidirectional test 90/50 Mb/s
<rellla> but only ~50Mb/s from a2000 to dockstar
<rellla> bidirectional test 70/4 Mb/s
<rellla> seems that something goes wrong with a2000. i need some time to check this out....
<penguin42> orly_owl: Meh - a10 is a solved problem - boring!
<orly_owl> no it's not, lima isnt fully functional yet
<penguin42> true
<penguin42> right, I think I need an infinite number of extra USB peripherals now for it
<orly_owl> as long as you have the money
<penguin42> orly_owl: Well, it was only £40 which for a dual core A9 was too tempting
<provel_> rellla: ok so it's a2000 issue
<provel_> rellla: maybe something physically different?
<rellla> maybe another fex-setting is needed...
<rellla> i'll open it and check board revisions
<provel_> android had a fex/bin file?
<provel_> ok I found the bin in nanda
<provel_> the only difference with script.bin is that in android line is not removed, it's "emac_power =" without a value
<provel_> hum... the only emac difference.... all the file is quite different
<provel_> hum... I think fex were uptated
<provel_> -machine = "A10-EVB-V1.1"
<provel_> +machine = "A10-EVB-V1.2"
<provel_> there's differences in plls, card_burn_para, uart, jtag/dram, mali clkdiv, again uart, ctp_screen_max_y, lcd
<provel_> usb, ir
<provel_> android fex is completely different here....
<provel_> I try to give a shot with android fex....
<provel_> nop, seems no good
<provel_> sigh I was hoping to fix the issue of usb drive moving devices
rsalveti has quit [Ping timeout: 255 seconds]
<provel_> anyone else noticed that when inserting a second drive the first one moves from sda to sdc (maybe related to the fact first one has 2 partitions) ?
<provel_> likely it's the same without the entry or just without the value
rsalveti has joined #arm-netbook
<mnemoc> mripard: what's the base branch for your gpio patches?
<mnemoc> mripard: missed the v2 of pinctrl-sunxi so I'm kind of lost
eFfeM has joined #arm-netbook
rellla has quit [Quit: rellla]
<mripard> mnemoc: 2s, i push it on my github
<mripard> done, branches sunxi-pinctrl and sunxi-gpio on https://github.com/mripard/linux
<techn_> mnemoc: last commit killed my wlan :/
<techn_> ERR: unknown usbc_no(2)
<techn_> L90 usb_hcd_servers.c
<jukivili> unification patch also changed Kconfig.. did you enable new sunxi-ehci options?
<techn_> jukivili: hmm.. I think I should be using same kconfig as in repo
<techn_> depconfig I mean
<jukivili> hm..
<techn_> jukivili: which config you mean. .If I'll check it from sysfs
<jukivili> do you have CONFIG_USB_SUNXI_EHCI=y, CNFIG_USB_SUNXI_OHCI=y and CONFIG_USB_SUNXI_COMMON=y?
<techn_> EHCI, OHCI and COMMON are y
<provel_> which branch is that? I haven't at all in stage/...3.4
<techn_> I'm using 3.0
<techn_> usb keyboard works ok
<techn_> lsusb shows two 2.0 root hub.. one 1.1 root hub and logitech keyboard
<techn_> and keyboard is connected to that 1.1 root hub :/
<provel_> which device is that? mele?
<techn_> mini-x
<jukivili> hm.. I have one modification in mk802.fex .. 'usb_host_init_state = 1' for [usbc2]
<jukivili> I think those lines need to be restored in usb_hcd_servers.c
<techn_> couple secs.. I'll try usb_host_init_state = 1
<techn_> jukivili: that did the trick
<slapin_nb> hno: ping
<techn_> jukivili: but I think thats just workaround :/
<mnemoc> mripard: thank you
tTTttt has quit [Quit: leaving]
Pulser has joined #arm-netbook
<jukivili> techn_: true, correct fix would be to remove 'if (sw_ehci->host_init_state == 0)' check in ehci-sunxi.c:sw_ehci_hcd_probe
<jukivili> maybe.. depends if we want to start ignoring host_init_state config and always enable host controllers
<techn_> previously it ignored that? since it helped if I explicitly defined it on
<jukivili> seems that previously if it was 0, sunxi usb-manager thing to decided whatever host controller should be enabled or not. Now if it's 0, it stays so and controller stays dead.
<ssvb> provel_: "Are the drivers working actually? I can't find a way to make them work with mali framebuffer ones" - hi, which drivers are you asking about?
<provel_> ssvb x86 ... fbdev
<provel_> I tried using them... maybe I should change something insinde xorg config
<provel_> or maybe them are requiring r3p1 framebuffer ones?
<ssvb> provel_: you mentioned "mali framebuffer ones", do you have non-x11 mali blobs installed?
<provel_> yes
<provel_> should I install x11 ones?
<provel_> but with thos I have xf86-video-mali
<penguin42> hmm nice set of binaries on this mk809 install - it's got things like tcpdump,strace,nc and oprofield - is that normal to find in an android image?
<ssvb> depends on what you want to get, if you want x11 integration working, then you need DRI2 enabled x11 mali blobs
<provel_> ssvb : ok.... but how I ensure your ones are used instead than xf86-video-mali ? Them just overwrite files?
<provel_> I'll try to remove them from /usr/lib/xorg/modules/drivers ..
<ssvb> provel_: compile and install xf86-vide-fbdev in the same way as xf86-video-mali, and edit /etc/X11/xorg.conf : https://gist.github.com/4462216
<ssvb> provel_: but if you are interested in OpenGL ES acceleration, xf86-video-mali is still faster at the moment (partially because of shadow framebuffer, partially because of not wrapping pixmaps)
<ssvb> provel_: currently the only advantage of xf86-video-fbdev is non-crippled 2D performance
<provel_> oook
* penguin42 wonders what the offensive_work_normalizer does
<penguin42> word!
<provel_> hum... no way.... even with that config doesn't starts
<provel_> lots of this error:
<provel_> (EE) FBDEV(0): FBIOPUTCMAP: Invalid argument
<ssvb> ignore this error
<ssvb> it does not matter
<ssvb> anything else?
<provel_> then no error report
<provel_> no EE at all other than that
<ssvb> then it should work, how do you know that it does not?
<provel_> screen clears and I get one line of console input
<provel_> but then doesn't srart the process i started with xinit
ganbold__ has joined #arm-netbook
<ssvb> can you pastebin your /var/log/Xorg.0.log ?
<provel_> ok
<provel_> cat /var/log/Xorg.0.log
<provel_> ops
<provel_> loads submodule shadow ... maybe I have to explicitly disable instead than comment out?
ganbold_ has quit [Ping timeout: 252 seconds]
<provel_> no, same behaviour
<bsdfox_> if I make sun4i_defconfig then make one simple change (compile in softwinner sata support) when I go to make uImage modules it is clearing my config-- scripts/kconfig/conf --silentoldconfig Kconfig
<provel_> maybe it's beacause I'm not on r3p1
<bsdfox_> then making me reconfigure from the command line
<ssvb> provel_: can you run Xorg in gdb? I only tested tested mali blobs r3p0
<ssvb> provel_: and 1.12 version of xorg-server, but not sure if this is relevant
<provel_> ssvb: will need some time
<provel_> launching gdb startx will suffice?
<ssvb> try just "gdb Xorg"
<provel_> ok
hg_5_ has joined #arm-netbook
<provel_> doh ... I need some moments.... but how I take logs after "run" ?
hg_5 has quit [Ping timeout: 255 seconds]
<provel_> ssvb: ok it did same thing also inside gdb
<provel_> still no other errors in Xorg*log
<ssvb> provel_: did something happen? it either works, or terminates (segfaults, quits, etc.)
<provel_> no same as before
<provel_> a little complain about plymouth ply-terminal and nothing else
<provel_> I suppose it's because its r2p4
<ssvb> provel_: if it works, then try to run some X application, for example "DISPLAY=:0 xterm"
<provel_> no doesn't works...... it's beacause r2p4 or because DISPLAY:=1
<ssvb> provel_: "it does not work" is not a great bugreport, I have no clue what is happening in your system
<provel_> ok.... r3p0 will work with 3.0 kernel?
<ssvb> does it terminate? yes/no?
<provel_> no seems it freezes without displaying anything
<provel_> oh... yes no process Xorg
<provel_> installing r3p0
<ssvb> provel_: http://linux-sunxi.org/Mali400 suggests r3p0, how did you end up with r2p4?
<provel_> same behaviuor..... maybe I was already on r3p0... just followed http://linux-sunxi.org/Mali400 DRI2 drivers section
<provel_> ok I'll stop for now.... not useful for anyone
<ssvb> provel_: can you also check the original http://cgit.freedesktop.org/xorg/driver/xf86-video-fbdev/ ?
<ssvb> provel_: ok, thanks, there are surely more useful things for both of us to do :)
hg_5_ has quit [Quit: ChatZilla 0.9.89-rdmsoft [XULRunner 1.9.0.17/2009122204]]
hg_5 has joined #arm-netbook
<bsdfox_> ssvb, what do you run right now?
eFfeM has quit [Quit: Leaving.]
<ssvb> bsdfox_: many different things
<bsdfox_> lets say I haven't touched my mele since early november and just compiled a new 3.0.52 kernel. would you run fbdev? does cedarx work at all yet?
<ssvb> I never tried cedarx myself
<bsdfox_> ssvb, which mali driver ar eyou using?
<ssvb> r3p0
<Soravux> Anybody has a recommended compilation toolchain for armv71hf? I tried the linaro one but the 2012-12 version segfaults on my setup and the 2012-11 version requires me to add "-lrt -lpthread" to the LIBS in the Markefile of sunxi-tools (but compiles successfully afterwards).
hg_5 has quit [Quit: ChatZilla 0.9.89-rdmsoft [XULRunner 1.9.0.17/2009122204]]
<bsdfox_> h36sa@linaro-ubuntu-desktop:~$ DISPLAY=":0" es2gears
<bsdfox_> libEGL warning: failed to create a pipe screen for Mali DRI2
<bsdfox_> libEGL warning: DRI2: failed to open Mali DRI2 (search paths /usr/lib/arm-linux-gnueabihf/dri)
ZaEarl has joined #arm-netbook
<bsdfox_> I'm guessing that's indicating that 3d acceleration isn't setup properly?
hg_5 has joined #arm-netbook
<provel_> Soravux: I used linaro server with sunxi-bsp, and added thing only when needed
<provel_> semms decent even if it isn't rocket fast
<provel_> toolchain is as per wiki
<provel_> bsdfox_: yes.... are you using r3p1 with kernel 3.0? you should use r3p0
<bsdfox_> provel_, I'm using r3p0
<bsdfox_> on 3.0.52
<provel_> wai a sec, Il'' give you my scrit... are you on linaro ?
<bsdfox_> yeah
<bsdfox_> I'm getting lots of UMP<1>: No handler for IOCTL...
<provel_> copy as rebmali..... create dir malibuild in home and launch as sudo rebmali x11
<provel_> will do all what's in the wiki
<bsdfox_> yeah that's what I've already run
<provel_> works 10%% here
<provel_> did you copied the conf in xorg drivers?
<bsdfox_> yeah
<provel_> here works like a charm
<bsdfox_> what kernel are you on
<provel_> 3.0.52
<bsdfox_> can you pastebin your dmesg?
<provel_> it's not-so-fast cause of the memcpy, but worked out-of-the box
<provel_> with my script
<provel_> (which miss only the conf copy)
<penguin42> grr hacking on this mk809 would be easier if the dozy amazon vendor I ordered a uSD card off had sent the right one
<penguin42> what format are the image files that get loaded by Android boot loaders? I can see this rk3066 image has an ANDROID! at the start
<bsdfox_> provel_, can you pastebin your dmesg?
Quarx has quit []
<bsdfox_> https://gist.github.com/4462732 my es2_info looks wrong
<provel_> bsdfox my WHOLE dmesg?
<bsdfox_> provel_, yeah
<provel_> there you are http://pastebin.com/RswqH4Gm
<provel_> I have no issues with ump
<provel_> my fex/bin file is sane and almost all taken from sunxi-bsp
<bsdfox_> provel_, does your es2_info look better than mine?
<provel_> no es2_info... which package should I install?
<hramrach> es2_info is packaged in mesa-utils-extra or such in Ubuntu
<bsdfox_> ahh your mali clk_div is 3 too
<provel_> ? your?
<provel_> my android script.bin specify div 4
<provel_> but in linux I'm using div 3
<provel_> seems working
<bsdfox_> where do you specify that?
<bsdfox_> also are you on rev c?
<hramrach> mesa-utils-extra: /usr/bin/es2_info
<bsdfox_> provel_, does your EGL_VERSION show EGL_VERSION: 1.4 Linux-r3p0-04rel0 ?
<provel_> installing
<bsdfox_> mine just says = 1.4 (DRI2)
<provel_> bsdfox_ : fex is the ascii file to creatre script.bin with bootloader
<provel_> and sunxi tools
<bsdfox_> I'm using my script.bin from android
<provel_> es2_info wont work as can't open display ?
<provel_> no good
<provel_> script.bin from android is no good for linux
<bsdfox_> provel_, try it from X :)
<hramrach> that or select a driver that does not use X
<provel_> he needs X testing... I'll try xinit after reboot
<provel_> with xinit I lose output
<bsdfox_> where did you get DRI2? libdri2-1 linaro or did you build from source?
<provel_> linaro I think
<provel_> no way.... will not work... however your issue is script.bin 90% sure
<provel_> my mele won't even start with android script.bin
<bsdfox_> provel_, your es2_info won't work?
<provel_> I can't get output as I use plain xorg without ubuntu
<provel_> redirection gave nothing
<provel_> xterm don't worked
<provel_> I use xorg only with retroarch +gles
<provel_> there works
<provel_> (sorry my xorg is linaro-server + install xorg)
<provel_> (I don't wanted all the gnome fuff)
<hramrach> eglut supports both kms and x11 but I don't see how they are selected
<bsdfox_> do you have any *mali* in /usr/lib/arm-linux-gnueabihf/dri?
<provel_> not there
<hramrach> you probably want es1_info
<hramrach> which is not packaged
<bsdfox_> I think I might just be missing the mali dri binary
<provel_> there I have only libglsl.so
<provel_> I haven't there
<hramrach> git://anongit.freedesktop.org/git/mesa/demos
mentis has joined #arm-netbook
<provel_> use my script and install x11
<provel_> have to go, bye
provel_ has quit [Quit: leaving]
<bsdfox_> libEGL warning: DRI2: failed to open Mali DRI2 (search paths /usr/lib/arm-linux-gnueabihf/dri)
<bsdfox_> hramrach, do you know where linaro would store that by default?
<hramrach> I have no idea about linaro, sorry
<hramrach> I will possibly have when I am finished with the ARM board but for now I am looking how to make a bootable SD card
lerc has quit [Read error: Connection reset by peer]
mentis has quit [Remote host closed the connection]
<penguin42> well this mk809/rk3066 seems ok, there are instructions out there how to get it to boot a kernel image off the recovery part that seem to work (although not found the instructions how to build the image file yet) and it gives a text console on hdmi so I'm fairly happy
<jinzo> hmz, anyone has any idea what's up with rellla's xbmc?
<jinzo> /empatzero's?
rz2k has joined #arm-netbook
<hramrach> I can't build es2_info because libEGL ahs undefined reference to wl_display_connect
<hramrach> it's awesome that the bianry runs under these circumstances :>
<bsdfox_> I think I'm missing something fundamental
<bsdfox_> hramrach, do you get those errors about failed to open Mali DRI2?
<ssvb> hramrach: be sure to use the compiled libUMP.so and not the libUMP.so blob bundled with the driver
<ssvb> hramrach: and (most of) the linking issues should be solved
<bsdfox_> can anyone pastebin a working Xorg.0.log?
piezo has joined #arm-netbook
rsalveti has quit [Ping timeout: 276 seconds]
rsalveti has joined #arm-netbook
<hramrach> ssvb: I am trying to build the mesa utils on pc
<hramrach> and I am sure that compiling Mesa without Wayland support would resolve these issues
rsalveti_ has joined #arm-netbook
<hramrach> btw that blog about uxa finally delivering faster accelerated 2d than software is interesting
<hramrach> it's not much new that fglrx performance is abysmal and that exa performance leaves something to be desired
<hramrach> but the question you have to ask is this
<hramrach> how long does it take to render the trace in number fo display frames? ho many display frames the trace contains? how long did it take the application to produce the trace?
rsalveti has quit [Ping timeout: 272 seconds]
rsalveti_ is now known as rsalveti
<hramrach> In my experience exa tends to be good enough because if something takes long to render it typically takes long to produce the render commands as well
<hramrach> there are exceptions, though
<hramrach> not everything is aceclerated and when you hit a software fallback with exa the performance is abysmal
<hramrach> also there are some limitations. for me rendering a screenful or few of log in an xterm-like application is instanteous but give it enough screenfuls and eventually the rendering is so slow you can read pretty much everything that scrolls past
<hramrach> there is no gradual slowdown. either it renders instantly or some limit is hit and it renders slowly
lkcl has quit [Read error: Operation timed out]
tinti_ has quit [Ping timeout: 248 seconds]
<ssvb> hramrach: for xterm scrolling you can try to add "+j" option to the command line (This option indicates that xterm should not do jump scrolling), otherwise it can't be really used for benchmarking
netchip has joined #arm-netbook
<netchip> hey
<hramrach> ssvb: I don't use it for benchmarking. I only observe that in practice there is some limit on the amount of scrolling exa is willing to accelerate
<netchip> is this channel for Allwinner chips?
<hramrach> also
<netchip> nice
<netchip> How are the Allwinner A10 tablets running on Android 4.2?
<netchip> Fluid, or laggy?
<hramrach> I guess this is more for tablets running Linux
* specing slaps netchip
<specing> since when does hardware run on software?
<netchip> specing, sorry sorry =/ I was too lzay to change it when I had written it
<netchip> hramrach, Android is Linux...
<specing> no
<hramrach> GNU/Linux if you want to be pedantic
<penguin42> hramrach: No
<specing> android is stripped down, horibly-javatized Linux
<specing> and there is no GNU there
<netchip> lolwut
<netchip> I am running bash on my ANdroid
<netchip> and also Linux kernel
<netchip> lol
XenGi is now known as XenGi_
<ZaEarl> android does use a linux kernel
<penguin42> hramrach: Android has finally made there a point of saying GNU/Linux as opposed to Android/Linux
<jelly-home> there are two kinds of people, those that think Linux is the OS, and those that think Linux is any OS with a Linux kernel. They're ALL WRONG AND SHOULD NOT DISCUSS THIS
<hramrach> indeed
<netchip> it's a kernel
<ZaEarl> but Android does not use the GNU C library like most typical "Linux" distros.
<netchip> they modified ir
<netchip> it*
<hramrach> and there is no GNOME on Android
<netchip> I can port
<netchip> everyone can
<hramrach> where G in GNOME stands for GNU iirc
<netchip> you needs lots of time
<netchip> but everyoen can
<hramrach> no, you can't port GNOME to Android
<netchip> no?
<netchip> someone ported glibc to Android
<netchip> IIRC
<ZaEarl> GNOME won't run on Bionic
<netchip> chroot...
<hramrach> you could make it but then it will no longer be GNOME and Android
<netchip> anyway
<netchip> it doesn't matter
<penguin42> indeed
<netchip> anyway, let's get back to my original question, changed - specially for specing -, Does Android run fluid on an Allwinner A10 chip?
<ZaEarl> yup
<barqux> yeah, its ok
<netchip> Android 4.2*
<ZaEarl> yes
<netchip> bit like N7, but slower?
<netchip> nvm
<netchip> that's a dumb comparisation
<netchip> because an Allwinner A10 tablet is 90 euro
<bsdfox_> https://gist.github.com/4463371 (EE) MALI(0): mali_drm_open_master DRM OPEN (fd: 0x7)
<netchip> and a N7 250 euro
<techn_> bsdfox_: chmod /deb/tmp
<techn_> .. /dev/ump :)
<hramrach> that does not necessarily mean N7 performs better
<bsdfox_> techn_, crwxrwxrwx 1 root root 245, 0 Jan 5 11:49 /dev/ump
<netchip> hramrach, I can't image a singecore outperforms a tweaked quad core...
<netchip> imagine*
<penguin42> netchip: The A10 is a single core Cortex A8 running at 1GHz, in the end that's going to be a heck of a lot slower than a N7
<techn_> oh.. run x as root?
ganbold___ has joined #arm-netbook
<netchip> I need something with proper uboot sources, kernel code with git history, and unlocked bootloader
<netchip> well
<hramrach> it depends on the benchamrk
<penguin42> no
<netchip> and UART, Ubuntu etc
<penguin42> netchip: Well that is the nice thing about A10 the sources are out there etc
<netchip> penguin42, Does it have UART?
<hramrach> don't know what quad core N7 has, and Hz means nothing unless it's the same CPU and bus divider
<netchip> hramrach, tegra 3
<penguin42> netchip: Yes
<hramrach> well, then it may well perform very poorly
<bsdfox_> techn_, same error root or user
<penguin42> hramrach: In the end most cortex's out there are faster clock/clock than the A8 in an A10
<hramrach> because nVidia are notorious for bad drivers
<netchip> remember it's Google
<netchip> Google won't accept bad drivers
<netchip> in fact, I think Google has access to those source code
<netchip> codes*
ganbold__ has quit [Ping timeout: 276 seconds]
<hramrach> then there is good chance it will perform better at least in some ways
<techn_> bsdfox_: I checked my logs and I get same EE
<hramrach> but you should note that A8 can perform two NEON instructions every clock while most of those multicores can do only 1
lerc has joined #arm-netbook
<netchip> hramrach, I think 90 euro for an Allwinner tablet is very nice. Performance/price seen
<netchip> and the fact you can run everything on it
<techn_> it seems that it wont block gles :/
<penguin42> netchip: Watch out whether the tablet has a uart you can get to
<bsdfox_> techn_, ok. do you get lots of these? UMP<1>: No handler for IOCTL. cmd: 0x400c9007, arg: 0xbebbb468
<netchip> penguin42, I think those chinese tablets of under 100 euro don't have invested in closing the UART ports ;)
<techn_> bsdfox_: which mali libs you are using.. and which kernel driver?
<bsdfox_> r3p0 3.0.52
<penguin42> netchip: Well they won't have invested in putting them somewhere convenient, so hopefully you'll find a good place you can get to them on the board
<hramrach> netchip: yes, 90 eur is nice for a tablet
<barqux> hramrach: could you mention some applications where neon or simd makes miracles happen?
<hramrach> video playback?
<barqux> mooore moooore
<techn_> bsdfox_: sounds like you have some mitchmatch .. try finding libUMP.so .. if you have it twice somewhere
<penguin42> you can use neon for things like memcpy's and string ops, which are occasionally useful
<penguin42> and A8 does have I think a wider bus to one of the caches for Neon but I forget which
<ssvb> bsdfox_: do you have mali enabled in the kernel config? and also CONFIG_FB_SUNXI_UMP=y?
<bsdfox_> CONFIG_SUNXI_MALI_RESERVED_MEM=y
<bsdfox_> CONFIG_DRM_MALI=m
<bsdfox_> CONFIG_MALI400_DEBUG=y
<bsdfox_> CONFIG_MALI=m
<bsdfox_> CONFIG_MALI400=m
<hramrach> barqux: in my experience video playback is *the* application on tablets that requires any amount of power
<bsdfox_> CONFIG_FB_SUNXI_UMP=y
<techn_> ldd /usr/bin/glmark2-es2 ? is that pointing to correct files?
<bsdfox_> techn_, I don't think it is
<bsdfox_> libEGL.so.1 => /usr/lib/arm-linux-gnueabihf/mesa-egl/libEGL.so.1 (0x4002e000)
<bsdfox_> libGLESv2.so.2 => /usr/lib/arm-linux-gnueabihf/mesa-egl/libGLESv2.so.2 (0x4001b000)
<bsdfox_> also it appears I have a /lib/libUMP.so and a /usr/lib/libUMP.so
<barqux> hramrach: yes that makes sense
<netchip> does an Allwinner A10 plays 720p?
<techn_> netchip: yes.. but depend filetype
<netchip> you mean bitrate?
<hramrach> supposedly it plays 1080p too
<hramrach> more like that the codec is one for which accelerated decoding is supported
<ZaEarl> A10 can play 40mbps 1080p streams smoothly
<barqux> penguin42: yes, simd for memcpy and strcmps etc etc is useful but I guess it's more common in multimedia
<bsdfox_> techn_, renaming /lib/libUMP.so seems to have fixed that IOCTL error
<bsdfox_> I still don't have accelerated 3d as it's using the mesa libs
<penguin42> barqux: Hey the problem with memcpy etc is it only pays off for very large copies generally
<bsdfox_> hmm I've got a bunch of crap in /lib I don't want
<hramrach> possibly video frames? :p
<penguin42> hramrach: Perfect :-)
<netchip> On the A10 you can play Angry Birds play fluid on a tv screen with a resolution of 1080p :O
<barqux> penguin42: really? shouldn't performance be noticeable even for smaller amounts of data if you're memory aligning it?
* penguin42 needs a collection of random USB crap; my dad knicked my usb hubs and I don't have a spare USB keyboard/mouse
<penguin42> barqux: memcpy is easier because you know the length to start with; string stuff is much harder because it might be a small string that's not aligned
<penguin42> barqux: So you can start off with a 'is it big, is it aligned, ok go for Neon else do it easy' - but actually things like ldrd/strd are pretty fast as well, so as long as you're quick enough to keep up with the memory bus the Neon doesn't help
<penguin42> barqux: And neon supposedly eats more power and can be dynamically turned off per-process, so people worrying about whether it's worth turning on for something like a memcpy
<datagutt> netchip
<datagutt> please don't start kanging cm for allwinner
<netchip> nah
<netchip> my plan is to contribute it to CM
<datagutt> code is working on cm for mk802iii
<bsdfox_> techn_, should I just update the symlinks to point elsewhere?
<datagutt> cody*
<netchip> datagutt, mk802iii is?
<datagutt> a usb-sized device you connect to your tv
<datagutt> so you get android no your tv
<datagutt> Oh
<netchip> isn't that a Mediatek?
<datagutt> that does not use allwinner
<datagutt> the mk802+ and mk802ii does though
<datagutt> netchip: + and ii uses allwinner
<datagutt> xplod is working on the + version
<netchip> ok
<netchip> but I want a tablet
<datagutt> Ah
<netchip> not an usb-sized thingy for on your TV
<netchip> because I go buy an ODROID X2 for that ;-)
<barqux> penguin42: yes, sounds valid. It's a pain if power consumption increases on these devices. Sometimes it makes you miss the previous generation of cellphones. easily a week without recharge!
<datagutt> netchip: buying exynos-powered device to dev for? are you insane?
<netchip> datagutt, rmcc is insane?! He bought an ODROUD U2
<netchip> I don't think he is
<netchip> are there any hwcomposer sources for the Allwinner A10?
<netchip> that's important for >Android4.0
<bsdfox_> ok I'm getting closer.. symbol lookup error: /usr/lib/libGLESv2.so.2: undefined symbol: XextCreateExtension
<ssvb> bsdfox_: now this is easy, you are definitely using wrong libUMP.so, the compiled one should have proper libXext dependency
<bsdfox_> h36sa@linaro-ubuntu-desktop:~$ ldd /usr/lib/libGLESv2.so.2
<bsdfox_> libUMP.so => /usr/lib/libUMP.so (0x4025d000)
<bsdfox_> h36sa@linaro-ubuntu-desktop:~$ ls -la /usr/lib/libUMP.so
<bsdfox_> -rw-r--r-- 1 root root 30301 Jan 5 12:32 /usr/lib/libUMP.so
<bsdfox_> I will recompile it now that I cleaned up some libs
<bsdfox_> yay that looks like it worked
<bsdfox_> about 90fps es2gears
eebrah has joined #arm-netbook
<hramrach> netchip: I don't get what are these exynos powered devices for
<hramrach> the A10 does the job for 'get android in your TV' and 'get android in your tablet'
<hramrach> exynos may be close to 'get ARM on your desktop' but not quite. some SATA/PCIe would be helpful as well as more RAM
<netchip> well
<netchip> Did you ever see an Nexus 10?
<netchip> It's DAMN quick
<hramrach> you may possibly get more RAM than those 2G you get on ODROID but no board exists afaik that has more
<hramrach> I did not see Nexus
<netchip> I highly recommend you check youtube
<hramrach> dos not work for me
<netchip> what do you mean?
<penguin42> the 10 is reputed to have a wonderful display
<hramrach> ENOFLASH
<netchip> hramrach, Ah, you live in a country where they block the Internet?
<hramrach> wonderful display is something you will not find on the A10 tablets, yes
<netchip> penguin42, yeah
<netchip> it's also a great audio chip
<netchip> and it's 80% open source
<netchip> only wifi and gfx not (and bootloader)
<hramrach> then again it will be quite dirty most of the time so I wonder how wonderful it has to be to make a difference even then
<bsdfox_> I've got mali clkdiv=4 in my script.bin/fex (using the one for android on my mele a2000). Is it as simple as changing the div to 3 ti improve my performance?
<hramrach> netchip: plus half of the time I get some stupid message taht the video is unavailable even if I install flash
<hramrach> so cba doing that
<netchip> =/
<hramrach> when youtube makes it hard to view their videos I am not going to bend over backwards to view them. they usually aren't that awesome
eebrah has quit [Quit: ChatZilla 0.9.89 [Firefox 17.0.1/20121128204232]]
<hramrach> netchip: btw the countries where they do not block internet are very few and very obscure
<hramrach> they do block internet in the US, the EU, China
<netchip> hramrach, Yeah, they do here, in the EU, but nobody sees it. They call you paranoid if you say it's blocked.
<netchip> but they don't do DPI
<hramrach> I did see a blocked site. It was postered all over it. am I seeing things?
<hramrach> they do DPI
<hramrach> to protect MPAA IP
<hramrach> or maybe their DPI is actually pointing a pencil in the Yellow Pages and going after the person so selected
<hramrach> because they also went after person whi did not have a computer or was not alive
rzk|2 has joined #arm-netbook
rz2k has quit [Ping timeout: 252 seconds]
rzk|2 has quit [Read error: Connection reset by peer]
<bsdfox_> wow changing mali clkdiv from 4 to 3 improved my glmark2 score from 25 to 44
<rm> hramrach, you can't view the Gangnam Style video in Germany: http://ompldr.org/vZ3loaA/2012-12-18T133009Z-gema.png
<hramrach> yes, stuff like that
<hramrach> plus sometimes it's just broken I guess
lkcl has joined #arm-netbook
servili007 has joined #arm-netbook
XenGi_ is now known as XenGi
mtndew420 is now known as mtndew`
netchip has quit [Ping timeout: 248 seconds]
mtndew` has quit [Changing host]
mtndew` has joined #arm-netbook
rz2k has joined #arm-netbook
mtndew` has quit [Quit: sigh]
Yaku has quit []
<barqux> hey I get some sort of graphical artefacts in my console when using fbcon on mk802ii, sunxi 3.0.52 kernel
<barqux> the current line kinda turns to rubbish for a while and when I continue typing it fixes itself
<barqux> does anyone recognize that?
<techn_> barqux: known problem.. feel free to fix it :)
<barqux> oh, ok!
<barqux> techn_: thing is, I'm new to messing with the kernel. Any hint to where one might start looking at this issue?
<hramrach> barqux: what kinds of artifacts?
<hramrach> there should be interface for drawing on the framebuffer directly form userspace so you can try that and see what artifacts you get with differnt patterns and guess what might cause that
<barqux> hramrach: it's as if random stripes and/or pixels just turn black. Nothing major really
<hramrach> always the same or varies over time?
<barqux> it varies as far as I can tell
<hramrach> hard to catch then I guess
<barqux> if I leave it, it changes once in a while
<hramrach> but still can try what drawing directly does
<barqux> yah, can look into it
<jinzo> dual core MIPS's from Ingenic finally, they're a bit behind their schedule, but still an interesting offer if they'll have the right price
<penguin42> jinzo: That price would have to be pretty damn low to be competitive against all the cheap A9's that are now around
<hramrach> or they would have to include some killer feature those don't have
<penguin42> nod
<jinzo> they went preatty low with the previous one
<jinzo> (first one under 100USD, ICS, etc...)
merbzt has joined #arm-netbook