TheSeven has quit [Remote host closed the connection]
TheSeven has joined #linux-sunxi
derethor has quit [Quit: Leaving]
<Turl>
[hawk]: can you send a patch to the list? :) a copy of the fex section would be good for reference too
_BJFreeman has joined #linux-sunxi
BJfreeman is now known as Guest56939
_BJFreeman is now known as BJfreeman
Guest56939 has quit [Ping timeout: 252 seconds]
<[hawk]>
Turl: sure... tomorrow! In the meantime I also fixe the bma250 :-)
BJfreeman has quit [Ping timeout: 252 seconds]
<Turl>
[hawk]: ok, thx :)
\\Mr_C\\ has quit []
tzafrir has joined #linux-sunxi
TheSeven has quit [Disconnected by services]
[7] has joined #linux-sunxi
\\Mr_C\\ has joined #linux-sunxi
_BJFreeman has joined #linux-sunxi
rz2k has joined #linux-sunxi
_BJFreeman is now known as BJfreeman\
BJfreeman\ is now known as BJfreeman
tzafrir has quit [Ping timeout: 252 seconds]
Tsvetan2 is now known as Tsvetan
_BJFreeman has joined #linux-sunxi
BJfreeman has quit [Ping timeout: 252 seconds]
_BJFreeman is now known as BJfreeman
Soru has quit [Ping timeout: 255 seconds]
Soru has joined #linux-sunxi
<oliv3r>
x86 ouch
<oliv3r>
wait
n01 has joined #linux-sunxi
n01 has quit [Client Quit]
n01 has joined #linux-sunxi
<oliv3r>
i have no clue how that got compiled to x86
<oliv3r>
but new one is: a31-meminfo-static: ELF 32-bit LSB executable, ARM, EABI5 version 1 (SYSV), statically linked, for GNU/Ld
eebrah|away is now known as eebrah
n01 has quit [Quit: Changing server]
n01 has joined #linux-sunxi
<oliv3r>
where is mnemoc; how bad is he?
n01 has quit [Ping timeout: 246 seconds]
BJfreeman has quit [Quit: had a good time]
Tsvetan has quit [Ping timeout: 246 seconds]
<oliv3r>
i think very strongly, that the DRAM controller is not allwinner IP; but some other IP
<oliv3r>
i don't think AW has many docs
<oliv3r>
and the way it is setup ... just messy
<oliv3r>
thoughts?
rellla has joined #linux-sunxi
rellla has quit [Ping timeout: 248 seconds]
rellla has joined #linux-sunxi
tkoskine has quit [Ping timeout: 268 seconds]
tkoskine has joined #linux-sunxi
<rellla>
hramrach_: i'll check that later. i'm busy with my newborn these days ;) on windows it plays fine with vlc at least. mediainfo says: http://pastebin.com/a0ZyFSf0
<oliv3r>
winwhat?
<oliv3r>
you heretic
<rellla>
oliv3r: i forgot, that this channel is called "linux"-sunxi ;) you're right. bad bad windows!
<oliv3r>
you dissapoint me rellla! :(
<oliv3r>
you may leave your geek-card at the door
Soru_ has joined #linux-sunxi
Soru has quit [Ping timeout: 264 seconds]
rellla2 has joined #linux-sunxi
vicenteH has joined #linux-sunxi
mturquette has quit [Ping timeout: 256 seconds]
mturquette has joined #linux-sunxi
<oliv3r>
what is an 'mbus'
<oliv3r>
i thought it was some random databus (i2c/can etc)
paulk-desktop has joined #linux-sunxi
<mripard_>
oliv3r: reading marvell patches? :)
<oliv3r>
mripard_: not at all; why?
<oliv3r>
mripard_: sun[4567]i all enable the GPS clock (wtf why?) and the 'mbus' clocks
<oliv3r>
which is even more strange because in the a10 manual, mbus clocks aren't even listed
<oliv3r>
then there's the ACE clock, which sun5i needs (only)
<oliv3r>
but I don't quite understand yet what the ACE bus exactly is or more importantly, how it relates to the SDRAM clock
<oliv3r>
Turl: Turl Turl
<oliv3r>
Turl: http://sprunge.us/VQeH this is your clock debug paste, why is ahb_sdram not enabled and running at 168 MHz?
<oliv3r>
Turl: pll5_ddr is running at 480 MHz but also not enabled
<oliv3r>
or does the AHB have it's own connection to the sdram? (for dma of the various drivers I expect)
<oliv3r>
that would explain the ahb_sdram being null; but not pll5_ddr being 0 :)
notmart has joined #linux-sunxi
<oliv3r>
the counts that is
<oliv3r>
hno: who wrote the gross of dram.c of the SPL? it's almost identical to the a* stuff from allwinner-boot/boot0/drv/init_dram
<oliv3r>
hno: once i've sent my 2 naming patches (almost done with 2nd set) and we accept those, we have to talk about some cleanup; some things are silly now (partially because we don't know certain things and just guess things) so with some testing we should be able to make things better
<rz2k>
oliv3r: ace might refer to 'advanced crypto engine'
<ssvb>
for me it just prints "aw_clkevt_init: sun7i_timer0_clockevent mult 6442450, max_delta_ns 170000, min_delta_ns 1000, cpumask 0xc04607e8, irq 54" to the serial console and reboots
<rz2k>
try disabling both 8250 and 'winners serial driver' in character devices
<rz2k>
i have olinuxino that doesnt work with anything serial related in kernel
<ssvb>
rz2k: does not seem to make any difference, I'll probably try to downclock dram just in case
<oliv3r>
ssvb: the cubie2 only has 240 MHz memory bandwith, won't that affect performance even more?
<ssvb>
cortex-a8 has non-pipelined vfp, that's a huge performance problem
<oliv3r>
so we boo at a20 and pray for a40? :)
<rz2k>
that x3 is not everything you can get, i bet i goofed up something in gcc settings
<techn_>
sw may have real bad bugs
<ssvb>
even arm11 has pipelined vfp and can beat cortex-a8 in some floating point benchmarks :)
<ssvb>
it's good that cortex-a7 has fixed this performance regression
<ssvb>
must be good for scalar floating point workloads
<rz2k>
weird thing is
<rz2k>
openssl is slower
<rz2k>
than a10
jemk has quit [Quit: bye]
<rz2k>
probably needs recompilation with -mcpu=cortex-a7 for real performance
<ssvb>
neon simd is roughly twice worse per clock compared to cortex-a8
<rz2k>
also about caches, what cahce controller a7 has?
<rz2k>
because standard lx20 is not compiled in kernel in 3.3 kernels
<rz2k>
(i tried enabling it, but it was broken, so probably it is not lx20)
<ssvb>
for integer code a7 has very limited dual issue support, only instructions with immediate operands and branches can dual issue
<rm>
ssvb, this says 512KB L2 cache, maybe you should edit this if you know it's wrong http://linux-sunxi.org/A20
<ssvb>
a8 can dual issue most instructions, as long as there are no data dependencies
<rz2k>
classic ARM, fix one thing, break another? :)
<ssvb>
the "peak" performance (on hand optimized code) is almost twice worse per cycle, but on the average quality compiler generated code the difference might be not so big
<oliv3r>
though with 2 cores and a mp2 hammering the memory bus; some extra cache would be good
<ssvb>
Raspbery Pi is perfectly happy with 128K of high latency external L2 cache!
piyushverma has joined #linux-sunxi
<techn_>
ssvb: which display resolution you are using?
<techn_>
is a20 still suffering same issue with 1080p?
<ssvb>
it seems to be hardcoded to 1280x720 50hz in the cubieboard2 image from nand
<ssvb>
and I still can't boot my own kernel :(
lkcl has quit [Read error: Operation timed out]
rellla has quit [Remote host closed the connection]
<ssvb>
hmm, "Data is only allocated to the L2 cache when evicted from the L1 memory system, not when first fetched from the system" seems to imply that Cortex-A7 uses exclusive L1 / L2 caches
<ssvb>
so in fact it may be 32K + 32K + 256K total for two cores
<oliv3r>
i'd feel bad for you; but i don't even have a cubie2 yet :(
<oliv3r>
it's stuck somewhere in the mail
<ssvb>
oliv3r: I have "liberated" mine from the customs office yesterday :)
<oliv3r>
i don't even know where mine is
<oliv3r>
all i know, some one ate it :(
<oliv3r>
of course i broke things again in my own build :S
tinti has joined #linux-sunxi
<naobsd>
ssvb: why you can't boot own kernel on cb2?
<ssvb>
naobsd: something is not right, it prints a single line to the serial console and reboots
<ssvb>
oliv3r: using a phone charger, maybe it's indeed not a very good idea
<naobsd>
ssvb: strange, it can boot on my cb2
<oliv3r>
ssvb: if i power the olinxino over usb-otg; it won't boot more then that
<oliv3r>
if i power it through a 12V adapter (5V is a little to light for it) -> no go
<oliv3r>
let me grab a 1 amp USB adapter, that has to be enough
<naobsd>
I'm using PC's USB port, probably it's 500mA... it can boot if nothing is connected to USB/SATA on cb2
<ssvb>
naobsd: what kind of fex file did you use? could you please add some detailed instructions to the wiki (with all the repositories, commmit id's, etc.)?
<ssvb>
naobsd: but I'll try to test some other PSU first
<oliv3r>
ssvb: 1 amp USB adapter works
<naobsd>
I see, which wiki page
<oliv3r>
svb 500 mA isn't enough to power it, maybe 700 mA will be enough
<leowt>
already created an issue to see if we can get a community just like linux-sunxi for rk
<oliv3r>
note to self (and maybe hno) SUN7I wants dlccr[i] as 0xf << 14; 0x4 << 14 (sun45) seems to work too
<leowt>
btw, what can you ppl tell about a20? it has the same heating issues as a10?
<oliv3r>
what heating issue does a10 have?
<leowt>
oliv3r: full freq
<leowt>
crashes & etc
<leowt>
a13 doesnt
<oliv3r>
what freq you running it at?
<leowt>
1000
<oliv3r>
must be your board specifically
<oliv3r>
i run my tablet at 1000+
<leowt>
if i do a kernel compile
<leowt>
on a10
<leowt>
with 1000+
<oliv3r>
ah well that is extremly stressfull yeah :)
<leowt>
it crashes
<oliv3r>
it might be that 1GHz is just too much for it
<oliv3r>
how do you know its a heating issue?
<leowt>
well, the kernel compile thing is just to test that kind of stress
<leowt>
because it is
<leowt>
=P
<leowt>
~90 degrees last read
<leowt>
celsium
<oliv3r>
90 is hot, but not bad
<leowt>
external thermo coupler
<oliv3r>
if you cool the chip
<leowt>
yes but last read
<oliv3r>
does it work then?
<leowt>
yep
<leowt>
i have a alu cooler just on top of it
<leowt>
with an thermal pad
<oliv3r>
and it works then?
<leowt>
with a13 its cool all the time
<leowt>
no cooling
<leowt>
ok, it is not a good example
<oliv3r>
well a13 is sun5i
<leowt>
because of it apackaging
<leowt>
and etc
<leowt>
but, i have other socs
<leowt>
bga, and of course nobody wants a possible crashing soc
<leowt>
so i was asking what about a20
<leowt>
does it need a cooler too?
<oliv3r>
well you can only farily test it if everybody stress tests there a10
<oliv3r>
like a cubieboard
<oliv3r>
if all cubies fail at 1 GHz kernel compiles
<oliv3r>
then 1 GHz is too much
<oliv3r>
but icidentally; A20 runs at 960 MHz 'max'
<leowt>
why is that?
<oliv3r>
because they over estimated the A10 :p
<oliv3r>
'need' is all relative
<oliv3r>
nothing 'needs' a cooler
<oliv3r>
all need a cooler
<oliv3r>
you can downclock it a bit and then it doesn't need one
<oliv3r>
you can overclock it and then it needs one
<leowt>
that is all relative
<leowt>
xD
<oliv3r>
and down/over is really just what the manufacture decides is 'normal operating freq'
<naobsd>
leowt: making community for rk sounds good. long time I could find very few people who can play with source
<leowt>
naobsd: all i have from rk is a 8168, so im glad to help. I am stuck in the kernel compile, the soc is new, supposedly the sources for rk8188 are the same for this one
<leowt>
already submitted a patch to rkflashtools for 8168, i have everything
<leowt>
found uart on the board too
<leowt>
oliv3r: yeah but it is not cool to say the normal operating freq is 1000mhz while in reality it is not like that
<leowt>
if you get what i mean
<naobsd>
sorry, rk thing is little off-topic here
<leowt>
naobsd: yeah, true.
<oliv3r>
leowt: they probably never stress tested it that high that hard
<oliv3r>
so they tried this, tried that 'oh runs at 1200 MHz' Good 1200 MHz is max freq
<leowt>
haha
<oliv3r>
and quickly it turned out that 1100 is really the max
<oliv3r>
and appearantly it's not even 1100 under very heavy stress
<leowt>
good quality cert btw
<leowt>
XD
<leowt>
wharever
<oliv3r>
well how would you determine max freq :)
<leowt>
next week i will recieve an a20 and an a31 tablet
<oliv3r>
'try and see' :)
<leowt>
ill post info in the wiki
_BJFreeman has joined #linux-sunxi
_BJFreeman is now known as BJfreeman
<oliv3r>
hno: have you tested a20 patches with cubie1? I haven't checked yet when/why; but right now cubie1 detect 0 ? memory and errors out; could be I broke something though
<oliv3r>
i'll check that later
<oliv3r>
no time now :S
<oliv3r>
sun7i obviously works
<drachensun>
oliv3r: Ran the new a31 meminfo static, got this