<cosm>
but I have an idea why it behaves differently between A8 and A9
<cosm>
on A8 NEON registers had high latency, whereas they've fixed that in A9
<cosm>
so on A8 you generally need long, sparse (in regards to access to a single register) loops
<cosm>
A8 SoC also generally tend to have much lower memory bandwidth
<cosm>
*SoCs
<ssvb>
no, it's just because NEON can bypass L2 cache on Cortex-A8 and do direct streaming from/to memory, with up to 12 outstanding requests in the flight
<ssvb>
"5.3 Non-blocking NEON loads" and "NMB, NEON (data) miss buffer" are explained there
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
Guest28017 has quit [Ping timeout: 276 seconds]
w00tc0d3 has joined #linux-sunxi
w00tc0d3 is now known as Guest8361
seere_ has joined #linux-sunxi
seere has quit [Ping timeout: 252 seconds]
wingrime has joined #linux-sunxi
TheSeven has quit [Ping timeout: 245 seconds]
TheSeven has joined #linux-sunxi
tomboy65 has quit [Ping timeout: 272 seconds]
[GuS] has quit [Ping timeout: 240 seconds]
tomboy65 has joined #linux-sunxi
Gerwin_J has quit [Quit: Gerwin_J]
slact has joined #linux-sunxi
<slact>
greetings
slact has quit [Ping timeout: 252 seconds]
slact has joined #linux-sunxi
pseudomind has joined #linux-sunxi
pseudomind has quit [Client Quit]
pseudomind has joined #linux-sunxi
pseudomind has quit [Quit: pseudomind]
wingrime has quit [Quit: Lost terminal]
kuldeepdhaka has quit [Ping timeout: 265 seconds]
slact has quit [Ping timeout: 252 seconds]
pseudomind has joined #linux-sunxi
cosm has quit [Ping timeout: 276 seconds]
cosm has joined #linux-sunxi
kuldeepdhaka has joined #linux-sunxi
slact has joined #linux-sunxi
kuldeepdhaka has quit [Max SendQ exceeded]
<slact>
can I use FEL mode to read/write NAND?
<slact>
I really want to run patwood's a20 branch of the kernel on an a23 tablet, but don't want to brink it before I can re-frash it or have boot from mmc
<slact>
bah, brick and re-flash
kuldeepdhaka has joined #linux-sunxi
<wens>
I don't think anyone that owns an a23 device here has tried it
<slact>
i don't mind being the first
<slact>
so do you know if I could use fel tools to write NAND?
<slact>
I can read it from rooted android if need be
<libv>
slact: so you have found working serial on the inet 86dz?
<wens>
you'll have to look at allwinner's tools, phoenixsuite or something
<slact>
ok, so linux-sunxi FEL tools can't be used to write to NAND?
<wens>
not that I know of
<slact>
libv: Unfortunately not Rx. I tried all the test points on the front and back of the board, all the mmc pins, and some other random ones -- no luck.
<slact>
so boot0 doesn't expose NAND as ram at some address or anything?
<libv>
slact: still, you should be able to fix up most of that page.
<slact>
I certainly will, libv.
<libv>
slact: do you still have the original android installed?
<wens>
slact: NAND is accessed using it's controller, which gives you _raw_ access to the flash
<slact>
(As you might've notice, i've already begun)
<wens>
allwinner has some libnand blob over that
<slact>
libv: yes
<slact>
wens: that'll work, too. I can read it raw, write it to a cubieboard, and read it from there?
pseudomind has quit [Quit: pseudomind]
<slact>
(cubie2)
<wens>
raw doesn't even include wear leveling, so it's risky I think
<slact>
"fuck it, we'll do it live"
<wens>
while it's impossible to brick the device, you do risk breaking your NAND
<slact>
How could I break the whole chip?
<slact>
I understand I much burn a sector or two at worst, but the driver will wear-level around that eventually, no?
<slact>
*might burn
<wens>
I meant the default setup
<slact>
(and anyway, I am willing to proceed)
<wens>
anyway, give it a try, and document it :)
<slact>
...but how?...
<slact>
it sounds like raw NAND is exposed at some address during FEL mode?
<slact>
or will i need to write some FEL code to turn on the controller and talk to it?
<wens>
the controller is exposed. you have to send it commands to read/write the NAND
<wens>
the latter
<slact>
got it, i'll check how much work this'll take
<slact>
(unless maybe you can point me to some allwinner tool that can do this at least for the a20?)
<libv>
slact: when you wander into android, can you get the identification strings out?
<libv>
slact: before you erase your nand and install something else
<slact>
writing them up now
<libv>
ok, thanks
<slact>
libv: done
<slact>
i'll add a few more pictures later
sehraf has joined #linux-sunxi
ganbold_ has quit [Ping timeout: 264 seconds]
ganbold_ has joined #linux-sunxi
<slact>
now with pictures!
FreezingCold has quit [Ping timeout: 265 seconds]
deasy has joined #linux-sunxi
vertex_ has quit [Quit: Page closed]
FreezingCold has joined #linux-sunxi
<slact>
ok, i'm all photoed out.
<libv>
slact: what about those two pads in the middle of the back of the board?
<slact>
you mean the scruffy metal pads?
<libv>
slact: no, the two round ones :)
<libv>
they seem to be near the soc, but they might not be enabled in the fex
<slact>
yeah, i tried them all
<slact>
using the sophisticated method of punching buttons on the console while dragging a wire around the pads, i don't see Rx on the back
<slact>
Is there any other way to look, or any known chip pin it could be on?
<libv>
wens also had difficulty enabling RX on a23, but eventually did succeed
<slact>
yeah, i think he was able to dump fex info, so he could edit that stuff. i'm still stuck at boot0.
<libv>
you should be able to get the fex and the dram info from android
<slact>
there's a "magic.bin" at nanda, which doesn't parse with bin2fex
<slact>
and some bitmaps, that's all
<libv>
:(
<slact>
can I load a kernel to ram in FEL, then load it via a normal boot?
netforhack has joined #linux-sunxi
<netforhack>
Hello
<slact>
or add it to the correct nand partition, and change a uEnv parameter via fel?
<slact>
(hello)
<netforhack>
can i use kernel & u-boot from fedora 20 image on another rootfs ?
<netforhack>
anyone alive ?
<ccaione>
netforhack: yes
bonbons has joined #linux-sunxi
libcg has joined #linux-sunxi
<netforhack>
does mainline have full support a10s ?
cajg_ has joined #linux-sunxi
netforhack has quit [Quit: Leaving]
cajg has quit [Ping timeout: 240 seconds]
<ccaione>
not yet
npcomp has quit [Ping timeout: 240 seconds]
npcomp has joined #linux-sunxi
<mnemoc>
moin
ganbold_ has quit [Ping timeout: 264 seconds]
<ccaione>
moin
<mnemoc>
moin ccaione
ganbold_ has joined #linux-sunxi
cajg_ is now known as cajg
bertrik has joined #linux-sunxi
bertrik has quit [Changing host]
bertrik has joined #linux-sunxi
Fusing3 has joined #linux-sunxi
JohnDoe_71Rus has quit [Quit: KVIrc Insomnia 4.0.2, revision: 402, sources date: 20100627, built on: 2010-08-02 17:44:33 UTC 402 http://www.kvirc.net/]
ganbold_ has quit [Ping timeout: 264 seconds]
bonbons has quit [Remote host closed the connection]
bonbons has joined #linux-sunxi
ganbold_ has joined #linux-sunxi
bonbons has quit [Read error: Connection reset by peer]
bonbons has joined #linux-sunxi
Net147 has joined #linux-sunxi
mmarker1 has quit [Read error: Connection reset by peer]
mmarker has joined #linux-sunxi
\\Mr_C\\ has quit [Quit: .]
popolon has joined #linux-sunxi
popolon has joined #linux-sunxi
kuldeepdhaka has quit [Excess Flood]
popolon has quit [Quit: Quitte]
kuldeepdhaka has joined #linux-sunxi
kuldeepdhaka has quit [Ping timeout: 240 seconds]
lioka has quit [Ping timeout: 240 seconds]
Gerwin_J has joined #linux-sunxi
kuldeepdhaka has joined #linux-sunxi
kuldeepdhaka_ has joined #linux-sunxi
nicksydney has quit [Remote host closed the connection]
kuldeepdhaka has quit [Ping timeout: 252 seconds]
nicksydney has joined #linux-sunxi
kuldeepdhaka_ has quit [Max SendQ exceeded]
Skaag has quit [Ping timeout: 240 seconds]
jlj has joined #linux-sunxi
bonbons has quit [Quit: Leaving]
<jlj>
Is the 3.10/experimental sunxi branch complete or are major sections missing (like USB, sound or such)?
<jlj>
as compared to 3.4
bonbons has joined #linux-sunxi
bgal has joined #linux-sunxi
<mnemoc>
jlj: ignore that 3.10 branch
bonbons has quit [Ping timeout: 245 seconds]
<mnemoc>
jlj: it's "on hold"/abandoned atm
<jlj>
mnemoc: ah, okay. So it's 3.4 that's the way to go until mainline support is ready?
bonbons has joined #linux-sunxi
<mnemoc>
jlj: we might need an in-between stable, 3.10 or maybe 3.14... but 3.10/experimental is nothing more than an old experiment atm
<jlj>
mnemoc: Okay, then I know the status, thanks
paulk-aldrin has joined #linux-sunxi
Seppoz has joined #linux-sunxi
<wens>
what happened to the "devices" section on the wiki main page?
FreezingCold has quit [Ping timeout: 240 seconds]
leviathanch has joined #linux-sunxi
jlj has quit [Ping timeout: 252 seconds]
leviathanch has quit [Ping timeout: 240 seconds]
tomboy65 has quit [Ping timeout: 272 seconds]
tomboy65 has joined #linux-sunxi
k3rN has joined #linux-sunxi
smotocel69 has joined #linux-sunxi
jlj has joined #linux-sunxi
Net147 has quit [Quit: HydraIRC -> http://www.hydrairc.com <- The professional IRC Client :D]
jlj has quit [Ping timeout: 255 seconds]
<cosm>
ssvb (or anyone else): do you get any noticeable improvements from using hardware cursors in xf86-video-fbturbuo?
<cosm>
I'm experimenting with adding some hardware acceleration for RK chips and the first thing I've tried was hardware cursor support
<cosm>
and in my totally improper benchmark for wiggling the mouse around as far as I can, I see no difference in load
<cosm>
*as fast
wingrime has joined #linux-sunxi
<cosm>
I see around 35% cpu utilization (on a RK3188 at 1.6GHz) from both X and the application under the cursor, presumeably handling events
<cosm>
I can't see much difference in cursor update rate under high system load either
<ssvb>
cosm: the hardware cursor is useful because it can hover over the hardware overlay
<cosm>
yes, but do you get any tangible benefits, as in lower system load or lower latency?
<ssvb>
cosm: that's not a performance enhancement, but mostly one of the XV and DRI2 overlay enablers
<cosm>
oh, right
<cosm>
that makes sense
jlj has joined #linux-sunxi
ddc has joined #linux-sunxi
<ssvb>
cosm: also some performance improvement probably can be observed if you don't wiggle the cursor, but just have it hovering at the same position over some quickly animated window
<ssvb>
cosm: with the software cursor, you are going to see some flicker, because the cursor gets removed, window gets updated and the cursor gets added again over and over again
<ssvb>
cosm: and with the hardware cursor, nothing needs to be done
ddc has quit [Ping timeout: 252 seconds]
<cosm>
ssvb: you're right, I can see some difference when hovering over the animated area in gtkperf
<cosm>
but point taken, it's more about overlays
<cosm>
thanks
<ssvb>
you are welcome
<cosm>
ssvb: so are you willing to accept rockchip patches in fbturbo?
<cosm>
ssvb: uhm, yeah, those patches are a bit of a PITA to deal with since they break hard if everything isn't just right and it's somewhat difficult to follow what's going on in commits called 'update' which change 50K lines of code :)
<cosm>
I guess that stuff works on RK3066 but I've abandoned my attempt to port it to rk3188
jlj has quit [Ping timeout: 276 seconds]
jlj has joined #linux-sunxi
xinj has quit [Ping timeout: 264 seconds]
k3rN has quit [Remote host closed the connection]
k3rN has joined #linux-sunxi
k3rN has quit [Excess Flood]
k3rN has joined #linux-sunxi
k3rN has quit [Excess Flood]
smotocel69 has quit [Quit: Leaving]
k3rN has joined #linux-sunxi
k3rN has quit [Excess Flood]
k3rN has joined #linux-sunxi
jlj has quit [Quit: leaving]
bgal has quit [Ping timeout: 265 seconds]
bonbons has quit [Quit: Leaving]
bonbons has joined #linux-sunxi
jinzo has joined #linux-sunxi
bonbons has quit [Read error: Connection reset by peer]
Skaag has joined #linux-sunxi
paulk-aldrin has quit [Remote host closed the connection]
bonbons has joined #linux-sunxi
bonbons has quit [Quit: Leaving]
ecelis has quit [Ping timeout: 276 seconds]
bonbons has joined #linux-sunxi
Black_Horseman has quit [Quit: Zwi se logou mou!!!]
<mnemoc>
libv: sorry, no more messing with the http server. all setup done. I should have been transparent (SIGHUP-only) but had to do a real restart twice :\
<ssvb>
libv: have you never observed this screen shaking glitch on a10 before?
<Turl>
ssvb: wow, awesome :D
wingrime has quit [Ping timeout: 252 seconds]
lioka has joined #linux-sunxi
daemon has joined #linux-sunxi
pacopad has quit [Remote host closed the connection]
<ArdaXi>
hi guys, according to http://linux-sunxi.org/Cedrus sunxi-vdpau should support OSD, but mpv fails silently with VDPAU_OSD=1 and /dev/g2d chmodded to 666
libcg has quit [Quit: bye bye]
<Turl>
ArdaXi: using fbturbo?
<ArdaXi>
yes
k3rN has joined #linux-sunxi
k3rN has quit [Excess Flood]
<ArdaXi>
I've built the latest sunxi-vdpau, linked mpv against that
k3rN has joined #linux-sunxi
k3rN has quit [Excess Flood]
k3rN has joined #linux-sunxi
k3rN has quit [Excess Flood]
k3rN has joined #linux-sunxi
k3rN has quit [Excess Flood]
k3rN has joined #linux-sunxi
k3rN has quit [Excess Flood]
k3rN has joined #linux-sunxi
ganbold_ has quit [Remote host closed the connection]
tomob has left #linux-sunxi [#linux-sunxi]
physis has quit [Ping timeout: 240 seconds]
Black_Horseman has joined #linux-sunxi
Black_Horseman has quit [Changing host]
Black_Horseman has joined #linux-sunxi
physis has joined #linux-sunxi
nicksydney has quit [Remote host closed the connection]
maksimlin has joined #linux-sunxi
nicksydney has joined #linux-sunxi
leviathanch has joined #linux-sunxi
lkcl has quit [Ping timeout: 240 seconds]
<rz2k>
ArdaXi: came at day time at EU and find jemk - he is around sometimes.