alyssa changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard & Bifrost - Logs https://freenode.irclog.whitequark.org/panfrost - <daniels> avoiding X is a huge feature
stikonas_ has quit [Remote host closed the connection]
yann has quit [Ping timeout: 256 seconds]
vstehle has quit [Ping timeout: 240 seconds]
cwabbott has quit [Quit: cwabbott]
cwabbott has joined #panfrost
jernej has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
jernej has joined #panfrost
nerdboy has quit [Ping timeout: 240 seconds]
davidlt has joined #panfrost
vstehle has joined #panfrost
buzzmarshall has quit [Remote host closed the connection]
raster has joined #panfrost
yann has joined #panfrost
Elpaulo has quit [Quit: Elpaulo]
stikonas has joined #panfrost
hrw has joined #panfrost
<hrw> morning
<tomeu> hrw: hi there
<hrw> does someone uses panfrost on rockpro64 board?
<hrw> http://paste.debian.net/1146528/ shows how it fails for me on 5.6.6 and 5.7-rc4 kernel (same way on both) and I wonder where the problem is
<hrw> Fedora kernels
<icecream95> hrw: Is the simpleondemand module in your initramfs?
<hrw> icecream95: have to check
<hrw> icecream95: drivers/devfreq/governor_simpleondemand.ko you mean?
<hrw> icecream95: it is not present
<hrw> looks like nothing depends on it so it was omitted
<hrw> panfrost depends on drm and gpu-sched. gpu-sched does not depend on anything
<icecream95> hrw: panfrost depends on it at runtime when it calls devm_devfreq_add_device
<hrw> icecream95: at runtime. while initramfs scripts check 'modinfo' output iirc. will look closer
<hrw> thanks for hints
<icecream95> This is arguably a kernel bug, and panfrost should explicitly depend on having it so it gets picked up by modinfo
<hrw> Kconfig says it depends on. hm.
hrw has quit [Remote host closed the connection]
hrw has joined #panfrost
kaspter has quit [Remote host closed the connection]
kaspter has joined #panfrost
icecream95 has quit [Quit: leaving]
stikonas has quit [Read error: Connection reset by peer]
stikonas has joined #panfrost
raster has quit [Quit: Gettin' stinky!]
raster has joined #panfrost
chewitt has joined #panfrost
indy has quit [Quit: ZNC - http://znc.sourceforge.net]
<hrw> is t860 capable of >fullhd? or is it limitation of rockpro64 board?
<hrw> I have it connected to hdmi port of my 3440x1440 display
<urjaman> well that's a funky-ass resolution
<alyssa> hrw: Kevin is natively 2400x1600 fwiw
<urjaman> but yeah it's not a thing of the T860, but of the display output hardware, and the clock configuration of the chipset, etc
<alyssa> ^^
<alyssa> Performance degrades on high resolutions but there shouldn't be architectural limits at 3440px
<hrw> thanks
<hrw> I will move board later to 24" fhd screen but now it is more handy to have it connected here
indy has joined #panfrost
<hrw> urjaman: better to have 34" 3440x1440 than 2x 27" 2560x1440 :D
<urjaman> the first guess i'd have is that maybe the kernel doesnt know how to pull off the exact HDMI clock required for that display at that resolution
<urjaman> i havent looked at the RK3399 code for that tho
<alyssa> IIRC there's a limit of 2560 (or 2600?) pixels somewhere in the display stack on rk3399, but don't know if that affects HDMI
<urjaman> (but the RK3288 needs a bunch of patches for the HDMI to support anything but like the usual 1080p displays which happen to have a clock that's an exact divider from the default PLL frequency)
<robmur01> yup, the upstream HDMI driver has a really limited whitelist of pixel clocks for TV resolutions and steadfastly refuses to allow anything else
<urjaman> oh yeah atleast on RK3288 one of the VOP's can do upto 2560ish (i forget what exactly) and the other can do upto 4K ...
<urjaman> dunno if the kernel knows to pick the correct VOP because even that code was real bad the last time i checked
<robmur01> if I keep using my board through my lovely 1920x1200 monitor in awkwardly-stretched 1920x1080 it'll eventually irritate me enough to start hacking in more rates from downstream
<urjaman> but if you're lucky and it picks the bigger one for the first display it finds (which is what happened for my T760... only that the first display was the eDP LCD for which the little is the better choice)
<urjaman> i'm still carrying essentially my slightly modified version of what chromeos did for the RK3288 in my kernel tree
<urjaman> so i can run my 1920x1200 2443BW just fine, tho these days almost never do
<hrw> looks like I found another issue with board ;D it has DP on usb-c but without DT info probably
<alyssa> Ah, yes,
<alyssa> USB-C <--> DP <--> HDMI <--> DVI
<alyssa> It's a marvel I can see anything at all.
<hrw> I have one of those C->C+A+HDMI dongles from some conference
<robmur01> Unfortunately, if it's not a Chromebook, then mainline says "No USB-C for you!"
<Lyude> hdmi and dvi are basically the same, just one has audio
<Lyude> (and sometimes ethernet)
<urjaman> and HDMI can clock faster
<Lyude> urjaman: mhm, dp doesn't need to though :)
<robmur01> the tcphy driver needs converting to the new kernel typec API, since apparently only Chromebooks are allowed to use extcon...
* alyssa hugs Kevin
<urjaman> (though technically i dont see why you couldnt run single-link DVI as fast as HDMI, except that they didnt want people to make DVI screens like that)
<hrw> robmur01: I see
<alyssa> robmur01: can I just say that the fact that midgard requires software pack/unpack in blend shaders is just insanity on top of insanity
<robmur01> alyssa: Welcome to Computers. Please leave your sanity in the 1980s as you enter.
gcl has quit [Quit: leaving]
<alyssa> robmur01: I'm trying to decide between links2, lynx, and elinks, for the most modern experience. Thoughts?
<robmur01> Netscape 4.7 will forever be the best
<hrw> alyssa: w3m!
<hrw> robmur01: which chromebook has working usb-c?
<alyssa> hrw: Hm.. what about dillo though?
<robmur01> dunno, I would have assumed all of them - the main thing is that there it's all handled by the cros-ec thingy, as opposed to boards like mine that use a discrete type C port controller chip
<hrw> alyssa: dillo requires gtk. w3m can run on bare text console
<alyssa> fltk, but true :)
<hrw> omg. forgot about fltk...
raster- has joined #panfrost
raster has quit [Disconnected by services]
raster- has quit [Client Quit]
raster has joined #panfrost
<daniels> hrw: chromebooks have been all-type-C-all-the-time for quite a few years now ...
<daniels> robmur01: tbf x86 does something very similar, except instead of talking to cros-ec across spi, you talk to ACPI
<alyssa> Ah, yes, the spec that mandates interpreting dynamic code in kernelspace to shutdown your computer.
<daniels> the future is nwo
<daniels> also, now.
<robmur01> the NWO is now
<daniels> intel are gradually exposing more of how you should speak to Type-C, but at least the original implementations just handled everything hidden away from the OS: the first XPS I got about 5 years ago (Sky Lake) dealt with power/DP/etc switching in magic firmware, and Linux never had any visibility or control over it
<alyssa> the now is nwo
* alyssa hides in the past
<robmur01> I actually quite like the concept of ACPI in terms of hiding platform-specific gunk
<robmur01> the tables I dumped from this machine total ~480KB; there's an upstream DTB which comes in at ~38KB. However you look at it that's a lot of stuff that Linux has to know about and have code to deal with, and Windows gets to breezily ignore...
davidlt has quit [Ping timeout: 256 seconds]
cwabbott has quit [Ping timeout: 272 seconds]
kinkinkijkin has quit [Remote host closed the connection]
cwabbott has joined #panfrost
<daniels> alyssa: i was always more of a WWF guy tbh
jolan has quit [Quit: leaving]
<alyssa> daniels: WWF?
<daniels> alyssa: nWo was a World Championship Wrestling faction, back when that was a thing ... I was more of a World Wrestling Federation kid
<alyssa> ah
raster has quit [Quit: Gettin' stinky!]
icecream95 has joined #panfrost
<mmind00> robmur01: actually chromebooks also shouldn't use extcon ... and the chromeos-people did actually work at making the cros-ec play with the type-c framework
<mmind00> (for typec)
<mmind00> there have been a lot of patches landed recently, but I'm not sure how far that has progressed
<robmur01> mmind00: yeah, I know it's only because of legacy, since extcon is the worst non-abstraction :)
<robmur01> are said patches in mainline or CrOS?
<mmind00> robmur01: mainline
<mmind00> though I don't really know how usable that is for rk3399 so far
cowsay has quit [Read error: Connection reset by peer]
cowsay has joined #panfrost
nerdboy has joined #panfrost
robmur01 has quit [Ping timeout: 260 seconds]
robmur01 has joined #panfrost