LanDi has quit [Remote host closed the connection]
dlezcano has quit [Ping timeout: 258 seconds]
dlezcano has joined #linux-exynos
dlezcano has quit [Ping timeout: 272 seconds]
dlezcano has joined #linux-exynos
ssvb has quit [Remote host closed the connection]
ssvb has joined #linux-exynos
afaerber has quit [Quit: Verlassend]
afaerber has joined #linux-exynos
vishp has joined #linux-exynos
vishp has joined #linux-exynos
circ-user-VymHm has joined #linux-exynos
circ-user-VymHm is now known as vinipetrucci
afaerber has quit [Quit: Verlassend]
<vinipetrucci>
hi all, i have a quick question. does anyone here have successfully boot a Linux kernel that works with HMP (Heterogeneous Multi-Processing) mode allowing task migration across all 8 cores on Samsumg Chromebook 2 (Exynos 5 Octa 5420?). Any thoughts, pointers? I'm following this tutorial here: https://wiki.linaro.org/WorkingGroups/PowerManagement/Doc/Chromebook2_boot and would like to know if this is a good direction.
<Wizzup_>
I've tried it some time ago, but didn't get it to work yet
<vinipetrucci>
what steps have you tried? can you share your status?
<javier__>
vinipetrucci: it is working for me when using linux-next and exynos_defconfig but disabling the big.LITTLE switcher (CONFIG_BL_SWITCHER)
<javier__>
vinipetrucci: but I heard is not stable on exynos5420 as it is in exynos5422. I didn't have any issues but didn't stress it a lot either
<vinipetrucci>
nice! i see
<javier__>
vinipetrucci: what issues are you having and more importantly, why do you want this?
<Wizzup_>
Well, the why is obvious. Power doesn't matter much and you can use more cores
<vinipetrucci>
i want to do research on hetcmp scheduling with real-time constraints
<vinipetrucci>
right now, i just want to have some pointers/directions that worked previously
<vinipetrucci>
to build/boot the kernel etc
<javier__>
Wizzup_: for doing research or hacking on the scheduler makes sense but if not I don't think so, you have no guarantees that performance will be better
<vinipetrucci>
better than homocmp?
<vinipetrucci>
or previous switcher scheduler?
<vinipetrucci>
performance, yes. but how about power/energy?
<javier__>
vinipetrucci: I mean, the scheduler thinks that is a SMP system so depending on your workloads it may even worst to to have all cores
<Wizzup_>
javier__: you've said this before, but I'm quite certain that using all the cores simultaneously is faster :)
<javier__>
Wizzup_: for which workload? :)
<Wizzup_>
make -j9
<vinipetrucci>
javier__: which scheduler are you talking about? default CFS scheduler or Linaro's HMP mode?
<javier__>
vinipetrucci: default CFS scheduler
<vinipetrucci>
javier__: yes, definitely. CFS is not hetero-aware...
<javier__>
nod, that's my point
<Wizzup_>
which is a valid one, but not always
<Wizzup_>
as a gentoo user... :)
<vinipetrucci>
javier__: can you describe in more detail how did you make it work with the chromebook 2? what kind of LInux does it run? ChromimOS, Ubuntu?
<javier__>
vinipetrucci: debian jessie but it shouldn't really care what user-space you have
<javier__>
vinipetrucci: you just need to enable CONFIG_EXYNOS5420_MCPM (which exynos_defconfig already does) and disable the b.L switcher
<vinipetrucci>
javier__: cool
<vinipetrucci>
javier__: and for boot loading did you that non-verified u-boot strategy?
<javier__>
vinipetrucci: yes, I'm doing chained booting using the nv-uboot
<javier__>
and just loading a kernel image + dtb instead of the signed FIT required by the verified boot
<vinipetrucci>
javier__: do you have any pointers on that chained booting thing? sorry, first time i heard about this
* Wizzup_
should retry mainline today or tomorrow
<Wizzup_>
wifi working now?
<Wizzup_>
vinipetrucci: it's on the wiki I think
<javier__>
vinipetrucci: I think linaro folks have a step by step how to
<vinipetrucci>
javier__: nice! i was looking at that link this morning, thanks so much.. i will try that right now
<javier__>
vinipetrucci: you are welcome
<javier__>
Wizzup_: most things are working on mainline, if you have a peach pi then display is working, for peach pit you need some in-flight patches
<Wizzup_>
I always forget which one is pi and pit, but I have the 5800
<Wizzup_>
I think that is pi
<javier__>
yes, the touchpad is not working on pi though
<javier__>
and is working on peach pit (5420)
<Wizzup_>
oh
<javier__>
yeah, there are some support missing on the atmel driver...
<Wizzup_>
so I need linux-next rather than linux right?
<javier__>
Wizzup_: yes
<Wizzup_>
ack
<Wizzup_>
will test that in a bit :)
<javier__>
Wizzup_: ok, you need to enable all CONFIG_*_MAX77802 options that got disabled by mistake
<javier__>
I'm about to test a patch for that
<Wizzup_>
alright
<javier__>
s/test/post
afaerber has joined #linux-exynos
<si1v3r>
I'm looking at the samsung chromebook 2. Is the 13 or the 11 better supported?
_whitelogger has joined #linux-exynos
<sjoerd>
should be pretty similar support wise
<si1v3r>
So the touchpad is working on XE503C12?
<si1v3r>
peach pit?
<sjoerd>
I think those patches landed now
<sjoerd>
peach pi (13") can be made to work as well but needs one or two extra patches
<javier__>
si1v3r: what sjoerd said, peach pi needs a patch for the atmel driver but the DTS support is linux-next
<si1v3r>
Ok, so it's possible to run 100% unmodified upstream?
<si1v3r>
with next?
<javier__>
si1v3r: yes, most peripherals should be working (modulo wifi I think, or at least I haven't tested it)
<javier__>
iirc there is some support missing in the regulator core but that is only when running at the maximum frequency I think
<si1v3r>
That's not too bad. If I can run faster than the dual 1.7 is will be fine. I really like the idea of double the ram.
<si1v3r>
Man someone needs to sell a consumer device that's just a web browser and a terminal. heh
<si1v3r>
I'd buy like 3 of em.
<sjoerd>
javier__: I think the max feq. is something you can't hit on peach pi but pit is fine in that respect
<javier__>
sjoerd: you can't because the operating points are not defined in the DTS or?
<sjoerd>
javier__: Because you need to boost the voltage when going from 1.8 to 2.1 ghz i believe
<sjoerd>
the operating point also isn't in the dts
<javier__>
sjoerd: right, that's what I was referring with the support in the regulator core, the downstream ChromeOS kernel has the voltage-locking thingy but was not accepted upstream
<javier__>
sjoerd: probably that's why the operating point is not in the dts...
<sjoerd>
Oh for the wifi, yes sorry
<sjoerd>
javier__: I believe the operating point is not in the dts as the system boots with a too low voltage to hit 2.1 so you need to scale
<javier__>
I see, makes sense
<sjoerd>
could be wrong though
liquidAcid has joined #linux-exynos
tilmann has quit [Ping timeout: 272 seconds]
tilmann has joined #linux-exynos
tilmann has quit [Client Quit]
liquidAcid has quit [Quit: Leaving]
vishp has quit [Quit: This computer has gone to sleep]