robclark changed the topic of #aarch64-laptops to: Linux support for AArch64 Laptops (Chrome OS Trogdor Devices - Asus NovaGo TP370QL - HP Envy x2 - Lenovo Mixx 630 - Lenovo Yoga C630 - Lenovo ThinkPad X13s - and various other snapdragon laptops) - https://oftc.irclog.whitequark.org/aarch64-laptops
freekurt[m] has quit [Quit: Client limit exceeded: 20000]
<steev>
some day i'll do matrix again
<gabertron>
T14s update: got a bunch of help from konrad on the dts and my t14s is up and running now — will be working on cleaning up dts with konrad, but am able to run mainline mesa and display / gpu is working, WiFi works but seems flakey, usb detects stuff over the main typec port
<gabertron>
Booted off nvme in laptop presently — typically stuff like the audioreach doesn’t work yet but otherwise going well
<adamcstephens>
Lenovo definitely uses lvfs for some machines. For example I can get a number of firmware from lvfs for my x1 carbon gen7
<robclark>
afaict it is maybe (?) limited to machines that officially have a "ship with linux" option.. which isn't yet any aarch64 things (although they might be using lvfs for chromebook things?)
<robclark>
lenovo defn helped out with upstreaming stuff to l-f for x13s, so they seem to be friendly to linux on aarch64 even if it isn't quite ready for a "ships with linux" officially supported thing
ahoneybun[m] has quit [Quit: Client limit exceeded: 20000]
Segfault[m] has quit []
edrex[m] has quit []
<Nios34[m]>
I have an ES Yoga/Flex 5G. The modem disappeared on Windows after an update. Any idea?
<Nios34[m]>
In Device Manager, there used to be a Snapdragon SDX55 device and now it's gone
hightower3 has quit [Ping timeout: 480 seconds]
Segfault[m] has joined #aarch64-laptops
<Segfault[m]>
robclark: i did a bit more digging into the firefox issue, it seems like it uses GLES by default but they're uploading images in BGRA8 which mesa i think then converts to RGBA8
<Segfault[m]>
i can force it to upload textures as RGBA8 and swizzle them to BGRA8 later, that seems to fix the issue
<Segfault[m]>
well "fix", firefox still really struggles with stutters whenever it's uploading textures but that's a more general design issue on their side
<Segfault[m]>
it seems like they might've had a similar issue with some intel gpus a while ago and just added an override specific to them
jhovold has joined #aarch64-laptops
hightower2 has joined #aarch64-laptops
hightower2 has quit [Remote host closed the connection]
hightower2 has joined #aarch64-laptops
srinik has joined #aarch64-laptops
sporos11[m] has quit [Quit: Client limit exceeded: 20000]
<bluerise>
gabertron: does the T14s have anything fancy regarding selecting EL2 mode or DTS?
<bluerise>
in the BIOS?
smpl has joined #aarch64-laptops
neggles has quit [Quit: bye friends - ZNC - https://znc.in]
neggles has joined #aarch64-laptops
Dantheman825[m] has quit [Quit: Client limit exceeded: 20000]
agl7-x13s has joined #aarch64-laptops
agl7-x13s has quit []
Liso[m] has quit [Quit: Client limit exceeded: 20000]
Las[m] has quit [Quit: Client limit exceeded: 20000]
WirelesslyWired[m] has quit [Quit: Client limit exceeded: 20000]
ablu[m] has quit [Quit: Client limit exceeded: 20000]
agl7[m] has quit [Quit: Client limit exceeded: 20000]
aigotchi[m] has quit [Quit: Client limit exceeded: 20000]
Nei[m] has quit [Quit: Client limit exceeded: 20000]
akawolf[m] has quit [Quit: Client limit exceeded: 20000]
AlexMarty[m] has quit [Quit: Client limit exceeded: 20000]
anarchron has quit [Quit: Client limit exceeded: 20000]
Anton[m]1 has quit [Quit: Client limit exceeded: 20000]
ArtyomK[m] has quit [Quit: Client limit exceeded: 20000]
averyfreeman[m] has quit [Quit: Client limit exceeded: 20000]
baspar[m] has quit [Quit: Client limit exceeded: 20000]
bumble[m] has quit [Quit: Client limit exceeded: 20000]
pz[m]1 has quit [Quit: Client limit exceeded: 20000]
clover[m] has quit [Quit: Client limit exceeded: 20000]
QiuWenbo[m] has quit []
danielt has quit [Quit: Client limit exceeded: 20000]
davidebeatrici[m] has quit [Quit: Client limit exceeded: 20000]
emily[m] has quit [Quit: Client limit exceeded: 20000]
EnigmaCurry[m] has quit [Quit: Client limit exceeded: 20000]
enserzo[m] has quit [Quit: Client limit exceeded: 20000]
firlaev-hans-fiete[m] has quit [Quit: Client limit exceeded: 20000]
harvests[m] has quit [Quit: Client limit exceeded: 20000]
harvestz[m] has quit [Quit: Client limit exceeded: 20000]
HCG[m] has quit [Quit: Client limit exceeded: 20000]
hlr[m] has quit [Quit: Client limit exceeded: 20000]
szclsya[m] has quit [Quit: Client limit exceeded: 20000]
kazek[m] has quit [Quit: Client limit exceeded: 20000]
LoganLeland[m] has quit [Quit: Client limit exceeded: 20000]
LikeNeosMatrix[m] has quit [Quit: Client limit exceeded: 20000]
lollaritits[m] has quit [Quit: Client limit exceeded: 20000]
lun[m] has quit [Quit: Client limit exceeded: 20000]
malvi[m]1 has quit [Quit: Client limit exceeded: 20000]
matthew[m]123 has quit [Quit: Client limit exceeded: 20000]
mynery[m] has quit [Quit: Client limit exceeded: 20000]
NomadNaomie[m] has quit [Quit: Client limit exceeded: 20000]
nicholascw[m] has quit [Quit: Client limit exceeded: 20000]
Nick[m]123 has quit [Quit: Client limit exceeded: 20000]
Nao[m] has quit [Quit: Client limit exceeded: 20000]
nscnt[m] has quit [Quit: Client limit exceeded: 20000]
owc[m] has quit [Quit: Client limit exceeded: 20000]
patzek[m] has quit [Quit: Client limit exceeded: 20000]
PterKoczka[m] has quit [Quit: Client limit exceeded: 20000]
pine-clover[m] has quit [Quit: Client limit exceeded: 20000]
DocGalaxyBlock[m] has quit [Quit: Client limit exceeded: 20000]
quinine has quit [Quit: Client limit exceeded: 20000]
resuenehparg[m]1 has quit [Quit: Client limit exceeded: 20000]
sally[m]123 has quit [Quit: Client limit exceeded: 20000]
sally[m]1234 has quit [Quit: Client limit exceeded: 20000]
cenunix[m] has quit [Quit: Client limit exceeded: 20000]
shjim[m] has quit [Quit: Client limit exceeded: 20000]
Sobek[m] has quit [Quit: Client limit exceeded: 20000]
stefan-schmidt[m] has quit [Quit: Client limit exceeded: 20000]
steveej[m] has quit [Quit: Client limit exceeded: 20000]
Stirl[m] has quit [Quit: Client limit exceeded: 20000]
Sayatomoki[m] has quit [Quit: Client limit exceeded: 20000]
sz3m3k[m] has quit [Quit: Client limit exceeded: 20000]
M0133oracle[m] has quit [Quit: Client limit exceeded: 20000]
vadikas[m] has quit [Quit: Client limit exceeded: 20000]
wiizzard has quit [Quit: Client limit exceeded: 20000]
wiley[m] has quit [Quit: Client limit exceeded: 20000]
<colemickens>
adamcstephens: if you can go through this dts, and pull the `firmware-name` referenced files out, I can probably throw together a USB image over the weekend that you could test.
<colemickens>
Also, is there a cutting edge mesa branch that I should be looking at for using on the x1e devices?
ellyq has quit [Ping timeout: 480 seconds]
ellyq has joined #aarch64-laptops
colemickens_irc has joined #aarch64-laptops
ellyq has quit [Ping timeout: 480 seconds]
ellyq has joined #aarch64-laptops
srinik has quit [Remote host closed the connection]
srinik has joined #aarch64-laptops
ellyq has quit [Ping timeout: 480 seconds]
<Segfault[m]>
<robclark> "Segfault: yes, that is the issue..." <- fwiw if I use firefox's second upload mode it hits that memcpy path you're mentioning although it still seems kinda slow on the A618, I assume that's just down to memory bandwidth though?
srinik has quit [Ping timeout: 480 seconds]
valentine has quit [Quit: Client limit exceeded: 20000]
<robclark>
Segfault[m]: maybe something else going on w/ firefox? profile to see if cpu utilization is still high? I noticed that chromium had the same mis-swizzle issue (with gl+linux, didn't check gles) but was still fast/smooth
<Segfault[m]>
it's still high but that's just because on heavier websites it's often uploading new textures for almost every frame which is really slow
<Segfault[m]>
it's not totally maxed out and freezing for 100s of ms at a time once it's not falling back to convert_ubyte at least
<robclark>
I was looking at some anime fandom wiki page.. that seems like a good overly blinky page
<Segfault[m]>
reddit seems to be a good stress test too lol
<robclark>
Segfault[m]: maybe take an apitrace of loading/scrolling the page.. and then see if replaying the apitrace is faster? If so that would implicate something other than gpu.
<Segfault[m]>
nah in the trace every time it messes frames it's because it's uploading textures
<Segfault[m]>
though I will try an apitrace
<Segfault[m]>
the GPU seems fast enough, it just seems firefox isn't very good at keeping it fed
<robclark>
if ffox isn't double buffering the textures then uploading new data could stall on previous frame to complete.. although probably not, because normally it would be memcpy to staging buffer and then gpu blit to go linear->tiled+ubwc
<Segfault[m]>
ffox will upload any waiting textures before rendering a new frame
<Segfault[m]>
which yes causes stalls all the time if uploading textures takes more than a few ms
<robclark>
normally we use the gpu to copy updated pixels into the texture (rather than tiling or trying to UBWC compress on the CPU).. so normally it shouldn't stall
<robclark>
a perfetto trace would help make it clear.. but a bit of a learning curve if you've not used perfetto before
<Segfault[m]>
<robclark> "normally we use the gpu to..." <- oh yeah but I mean just the initial memcpy is already slow enough to cause ff to miss frames a lot of the time
<Segfault[m]>
at least on the 7c where memory bandwidth is low
<robclark>
it's not fast but I think it should be at least something like 20Gb/s .. so you should be able to afford to do at least ~340Mb worth of memcpy a frame ;-)
<Segfault[m]>
7c G2 is more like 12GB/s so once you account for time spent rendering (usually about 10ms) there's not a lot it can copy before missing a frame
<Segfault[m]>
tbf even on android with the original qcom drivers firefox really struggles on this soc, I'm not sure there's much freedreno could do to improve it
<robclark>
ideally unless ffox is doing something dumb, the rendering time should be pipelined with cpu setting up the next frame
<robclark>
and.. 10ms is a long time, even on sc7180.. I'd be curious what ffox is managing to do so badly
anonymix007[m] has quit [Quit: Client limit exceeded: 20000]
<robclark>
Segfault[m]: get a perfetto trace or send me an apitrace (from which I can get a perfetto trace)
\[m] has quit []
ellyq has joined #aarch64-laptops
pstef has quit [Read error: Connection reset by peer]