15:47 <libv> yay. demo for fosdem is done.

15:47 <Wizzup> \o/

15:47 <Wizzup> How do you manage external display?

15:48 <libv> i currently hope for the best

15:48 <libv> namely that the 2011 projector in the classroom in building K does 720p and accepts the mode

15:49 <libv> because this hdmi->vga scaler thing doesn't

15:49 <libv> it works nicely with the cubietruck though, but not with the exynos original kernel

15:50 <libv> and installing a kernel with a fully working mali kernel driver is just too much of a risk so near to the conference

15:50 <libv> worst case scenario, there's a repeat of 2012, but with an 11" display as opposed to a 7" :)

15:53 <Wizzup> :)

16:12 <daniels> libv: fwiw, the current source from malideveloper applies and works just fine, although mainline exynos_drm doesn't support fencing, so it may or may not be totally unsynchronised

16:13 <libv> daniels: ok, good to know for the future

16:13 <libv> right now, it is mapping fbdev, and just memcpying it over

16:14 <libv> i tried using kms bo/planes and that's just a straight panic

16:14 <libv> try stopping X with hdmi attached, and another straight panic

16:14 <daniels> shrug, planes wfm both on 3.8 and mainline

16:14 <libv> all 3.4.0

16:14 <daniels> 3.4?!

16:15 <libv> it was a working setup, with a halfway working kernel, X and mali

16:15 <libv> which made me unbelievably happy

16:15 <libv> when i started doing lima, i had an android that didn't even have LD_PRELOAD

16:15 <daniels> well yeah, the only two working kernel trees i know of are chromeos-3.8, and ~3.19

16:16 <daniels> those all have working kernel + x + mali, although 3.19 retains a number of corner-case failures unless you take padovan's branch to convert it to the atomic helpers

16:16 <daniels> if you're using a random branch of a 2.5-year old kernel, i wouldn't really be expecting too much success

16:16 <libv> hah

16:17 <libv> https://github.com/hrw/chromebook-linux doesn't even exist anymore

16:17 <daniels> https://chromium.googlesource.com/chromiumos/third_party/kernel/+log/chromeos-3.8/

16:17 <daniels> that's what's shipped on snow

16:17 <daniels> (and pit/pi)

16:18 <libv> daniels: i have to spend tomorrow printing schedules and stuff, so no way am i going to change this halfway working setup

16:18 <libv> i'm not as comfortable as when i showed q3a on mali utgard, but i am pretty happy with the result i have right now

16:19 <libv> q3a was working like half jan back, and i had 2 weeks left to see whether i could make it as fast as the binary

16:19 <libv> here, i am glad that it works :)

16:20 <libv> as midgard really is completely different

16:20 <libv> it's obviously related, but i was only able to use the texture swizzle

16:20 <libv> s/use/re-use/

16:21 <libv> anyway, more on that on sunday

16:21 <amitk> daniels: the only thing that keeps me even interested in the trying to get mali working is the memory of how much worse it was with imagination to rev kernel versions :)

16:21 <daniels> amitk: :D

16:22 <amitk> and from all accounts, nothing has changed with them

16:25 <daniels> except that they've ~stopped shipping gpus

16:27 <libv> who, img or arm?

16:29 <libv> img is doing very well, they are just desperate to be a complete ARM competitor. if samsung gets that rumoured gpu out, arm is not doing that well anymore.

16:31 <WarheadsSE> You mean in terms of pvr vs mali libv ?

16:31 <libv> yes, as i am not sure what daniels would've meant otherwise

16:32 <WarheadsSE> just being sure

16:32 <WarheadsSE> sadly, i can never make it to fosdem or embedded world, etc.

16:32 <libv> mali t series ships in... exynos, rockchip.

16:32 <libv> and ...

16:32 <WarheadsSE> Allwinner

16:34 <libv> nope, they ship either mali-400 or powervr

16:35 <libv> not even mali-450

16:36 <daniels> yes, with 4xx having completely taken over from sgx, and no-one licensing rgx

16:38 <WarheadsSE> libv: ah, my confusion. I thought that was a "

16:38 <WarheadsSE> "t" family