08:45 <naobsd> oh I didn't notice there is libmali for X11 for RK3288 http://malideveloper.arm.com/resources/drivers/arm-mali-midgard-gpu-user-space-drivers/ r6p0

08:45 <naobsd> (and for fbdev too)

08:52 <naobsd> binaries for chromebook is still r5

08:52 <naobsd> and it seems r6p0 binary supports OpenCL

09:04 <naobsd> mmind00: is your mali r6p0 branch based on chromeos-3.14?

09:04 <naobsd> I thought kernel/userland release ver. need to be same... chromeos kernel uses r6 driver, but userland seems to be r5...

09:05 <naobsd> or I missed something?

09:05 <naobsd> btw

09:05 <naobsd> rockchip has r9 driver https://github.com/RockchipOpensourceCommunity/popmetal-kernel-3.14/commits/popmetal-ubuntu-mali_r9p0

09:05 <naobsd> (no idea about userland)

09:11 <naobsd> about mali400, rockchip has r4p0 driver https://github.com/RockchipOpensourceCommunity/px2-android-kernel-3.0/commit/77ddef751b8b5d7c35a9fe4d7f4d0c74009192ea

09:11 <naobsd> userland seems to be included in chipspark px2 ubuntu image

09:12 <naobsd> interesting part in ubuntu image is,

09:12 <naobsd> it has usr/lib/xorg/modules/drivers/malirga_drv.so

09:13 <naobsd> it seems to use /dev/rga

09:38 <naobsd> http://permalink.gmane.org/gmane.comp.lib.uclibc.buildroot/137714 buildroot for firefly? interesting, but, oh, somewhat-stable? :)

09:52 <c0d3z3r0> mmind00: some time ago you sent me that patch "net: ethernet: arc: don't create holes when allocating new skbs fails". is there any reason you didn't send it to the mailing list?

10:49 <mmind00> c0d3z3r0: most likely because I didn't know if it was correct :-)

10:50 <mmind00> naobsd: yep the kernel part is from chromeos ... and it seems someone pressured ARM to keep r5 and r6 stuff compatible

11:38 <naobsd> I see :)

12:35 <mmind00> naobsd: although when looking at the arm binary, 2mb on the chromeos side vs 21mb on the arm binary look suspicious ;-)

12:36 <mmind00> but in general the arm binary seems to run

12:39 <naobsd> mmind00: nice:) arm binary should have more function such as OpenCL

13:15 <naobsd> hmm... sometimes oom-killer does the work on firefly with mainline kernel

13:15 <naobsd> but there is no heavy process :(

14:10 <topi`> I wonder when we get the first boards based on the new RK3399...

14:10 <topi`> that's certainly a promising chip, and supports usb3.0 which is what's needed for connecting SATA disks without the slowness of usb2 bridge chips

14:11 <topi`> I wonder if it will be as expensive as the RK3288 initially was? I think I saw a figure of $40 for RK3288 but back in 2014

14:13 <topi`> the rk3368 has a slow memory interface (I just received a Geekbox and tested the memory perf), but it seems they have fixed that in RK3399

14:49 <naobsd> I think rk3368 target is not performance...

15:09 <naobsd> mmind00: opencl seems to be working with linux-next + mali related patches in somewhat-stable

16:00 <mmind00> naobsd: although that new driver version doesn't necessarily increase my opinion of the developers at the ARM graphics department ;-) ... I see hangs in the seemore.playcanvas.com webgl demo and strange new artifacts in glmark2

16:03 <naobsd> oh :(

16:05 <naobsd> I cannot try GLES yet... (no display/X11 system on my test env for now)

16:07 <naobsd> vdd_gpu: ramp_delay not set

16:07 <naobsd> hm

16:07 <naobsd> sleepy