maxwelld | i am always using custom xorg.conf for the lapdock. | 00:02 |
---|---|---|
maxwelld | not sure there can be a better way. | 00:02 |
maxwelld | i remove it when i restart and want just to use d4. then i copy it and restart if i want to use the lapdock. | 00:02 |
sicelo | maxwelld: no need. i just built sphone without the forced rotation. https://pasteboard.co/NsDBRG9PAJKa.png | 00:13 |
maxwelld | i think my issue is different. without the xorg.conf xorg is able to run x on d4. but it is not able to run x on motorola lapdock | 00:21 |
maxwelld | so i use my custom xorg.conf | 00:21 |
maxwelld | but i have to do that manually every time. but i am ok with that probably? | 00:21 |
sicelo | arno11, and finally ... i've tested the calls on N900 with Leste. this is super awesome really :-) | 00:58 |
freemangordon | tmlind: do you have any idea where to look for usleep_range() latency? I enabled ftrace function tracer but don't know how to analyze the results. BTW, it turned out we have the same issue on d4. | 07:35 |
tmlind | freemangordon: ok, sorry not sure what might be broken at this point, will try to find my old notes on timer tests so maybe it can be narrowed down a bit | 07:37 |
tmlind | seems short usleep_range() can take much longer than specified? | 07:38 |
freemangordon | on average 400us | 07:38 |
freemangordon | according to cyclictest | 07:38 |
freemangordon | setting pm_qos_resume_latency_us makes no difference | 07:39 |
freemangordon | increasing the priority however does | 07:39 |
freemangordon | cyclictest -p 80 -m -i 100 results in 40us average | 07:40 |
freemangordon | on d4 | 07:40 |
freemangordon | so maybe it is just a matter of scheduler priority, dunno | 07:43 |
tmlind | sounds like if priority affects it | 08:10 |
freemangordon | mhm | 08:10 |
freemangordon | but calling cpu_latency_qos_add_request() does nothing | 08:10 |
freemangordon | OTOH, returning -EAGAIN from pwm_ir_suspend seems to make things better | 08:11 |
freemangordon | so I am starting to think this is idle related, like CPU enter too deep idle state which needs lots of time to resume from | 08:12 |
freemangordon | *enetrs | 08:12 |
freemangordon | argh | 08:12 |
freemangordon | *enters | 08:12 |
tmlind | ok possible | 08:17 |
arno11 | sicelo: cool :) | 09:01 |
tmlind | seems like reloading the d4 modem modules including phy-mapphone-mdm6600 every two weeks works around the modem weirdness, uptime again close to 23 days, this still the old v6.1 based kernel | 09:29 |
tmlind | so maybe some modem firmware bug triggering some bug on linux | 09:29 |
Wizzup | tmlind: nice digging | 10:01 |
Wizzup | I wonder if this is also the case for turning it in offline mode or not | 10:01 |
tmlind | not sure what goes wrong, but i've had uptimes past two weeks by doing modem reload | 10:27 |
Wizzup | ok | 10:28 |
Wizzup | I do sometimes put my d4 into offline mode, which is why I asked | 10:29 |
Wizzup | this doesn't unload the modules but I suppose it puts the modem in some other mode at least | 10:29 |
tmlind | ah ok yeah no idea if offline mode is enough to avoid the issue | 10:29 |
tmlind | i've noticed that the qmi data connection at some point can run out of connection id's so maybe that's related | 10:30 |
Wizzup | hmm | 10:31 |
arno11 | sicelo: callerId and landscape mode work very well with your patches :) | 12:19 |
arno11 | gprs0 issue is also solved :) thx | 12:19 |
sicelo | arno11: yay! good to know | 12:33 |
Wizzup | sicelo: any prs you want me to look at? | 12:35 |
sicelo | Wizzup: https://github.com/maemo-leste/libicd-network-ofono/pull/1 and https://github.com/maemo-leste/libicd-network-ipv4/pull/7 | 12:36 |
sicelo | those fix the gprs issue that arno was referring to just now | 12:36 |
Wizzup | ok | 12:37 |
Wizzup | i'm planning to take a holiday for ~7 days or so (from work), so I will have more time for these things | 12:39 |
Wizzup | let me look at these now | 12:39 |
Wizzup | for iproute2, do we not need the netmask at all? | 12:39 |
Wizzup | isn't that passed by icd2 itself? | 12:39 |
Wizzup | doesn't look like it is | 12:40 |
sicelo | we do. it's in $subnet | 12:40 |
Wizzup | oh, right. | 12:42 |
Wizzup | ty for pointing that out | 12:43 |
Wizzup | sicelo: do we want to explicitly depend on iproute2? | 12:44 |
sicelo | that might be good, yes | 12:44 |
sicelo | but something already depends on it ... not sure what. i didn't have to install it manually | 12:44 |
sicelo | maybe something in devuan | 12:44 |
sicelo | but yeah, explicitly depending on it might be useful, in case that 'mysterious thing' disappears or changes :-) | 12:45 |
sicelo | the sphone patch for callerId is at https://paste.debian.net/plain/1292456 | 12:49 |
Wizzup | ok, let me fix up debian/control | 12:51 |
Wizzup | sicelo: what is the typo being fixed? | 13:38 |
Wizzup | here https://github.com/maemo-leste/libicd-network-ofono/pull/1/files | 13:38 |
sicelo | easier to see it from this view, https://github.com/maemo-leste/libicd-network-ofono/pull/1/commits/944850b6444feb511dc97652d9987af994fb8ecc | 13:41 |
sicelo | i made it a separate commit in case it turns out unnecessary. looks to me like the decision to use dns[0] or dns[1] was meant to be made depending on availability or non-availability of dns[1], not dns[0] | 13:43 |
Wizzup | ty | 13:45 |
Wizzup | good eye! | 13:46 |
Wizzup | I'll get this going now. | 13:46 |
arno11 | Wizzup: sicelo: i finally successfully built a pcsx-rearmed package for Leste and neon optimized :) it works full speed + sound on N900 :D | 17:37 |
arno11 | should work on D4 with far better rendering with few modifs | 17:42 |
Wizzup | great | 17:47 |
Wizzup | what modifs? | 17:47 |
arno11 | for example the key mapping is maybe different | 17:48 |
arno11 | (hardcoded) | 17:48 |
arno11 | and maybe we should use different cflags for optimization | 17:49 |
arno11 | and for better rendering it is just a question of settings | 17:50 |
arno11 | (opengl rendering with scalex2 + filters for example) | 17:51 |
arno11 | but when you see how it runs on n900, could be awesome on D4 really | 17:52 |
arno11 | pcsx burns 'only' 300mA with pulseaudio on n900 | 17:55 |
Wizzup | right, ideally we'd have one build for all devices, but maybe some config files | 17:58 |
Wizzup | in any case, cool | 17:59 |
Wizzup | ! | 17:59 |
arno11 | yes cool :) i think i just need to check few things about d4 and update config file | 18:00 |
arno11 | anyway maybe it works well with n900 optimization | 18:02 |
arno11 | i'll update my repo and then maybe you can try on your D4 ? | 18:03 |
Wizzup | sounds good, I'll need some rom to try it on, I haven't done much with the ps recently :D | 18:04 |
arno11 | ok, and btw the HLE bios integrated in the emu works well | 18:06 |
arno11 | so you just need a .bin game ;) | 18:07 |
bencoh | arno11: assuming you can use fremantle on n900, you might want to benchmark pcsx-rearmed there and compare | 18:16 |
arno11 | yes :) with no real benchmark, it already seems similar in term of perf | 18:17 |
arno11 | definitely better in term of power cons. | 18:18 |
arno11 | and really don't know why | 18:18 |
arno11 | on leste cpu usage is around 50-60% @25 fps in Gran Turismo for example | 18:23 |
arno11 | brb testing with fremantle | 18:24 |
arno11 | surprisingly it works better on leste @ the same freq (850MHz): similar cpu usage at around 25 fps but smoother on leste with better sound | 18:58 |
arno11 | of course with hildon compo deactivated | 19:00 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!