tmlind | freemangordon: nice yeah it's smoother now :) the tearing issue is still there and the ants are still there though, maybe they are both related | 06:40 |
---|---|---|
tmlind | hmm is the tearing issue even more visible now? | 06:41 |
tmlind | freemangordon: yeah ok not looping waiting for blits to complete makes sense :) | 06:49 |
tmlind | freemangordon: maybe also test what happens if you revert "drm/omap: Fix drm_handle_vblank() handling for command mode panels"? without that wl does not get updated, but it may not be needed for xorg at all | 07:05 |
tmlind | actually, maybe in omap_crtc_framedone_irq() try moving the if (omap_state->manually_updated) { } part to the end after wake_up(&omap_crtc->pending_wait)? | 07:20 |
tmlind | seems like the vblank interrupt should be generated only after framedone has completed, then the simulated vblank interrupt triggers the redraw of the command mode display | 07:21 |
tmlind | here's a patch to try, untested so far: http://muru.com/linux/d4/vblank-after-framedone.patch | 07:51 |
tmlind | not seeing a difference with my testcase, might be worth testing with xorg though | 08:21 |
tmlind | looks like .cpu_cache_flags = OMAP_BO_UNCACHED makes no difference for me | 08:22 |
freemangordon | tmlind: ok | 09:18 |
freemangordon | goign tot try | 09:18 |
tmlind | yeah worth trying to try to narrow the remaining issues down | 09:21 |
freemangordon | I think tearing is the only one remaining in omapdrm | 09:21 |
freemangordon | it looks to me the only way we have tearing is to start frame upload while another one is in progress | 09:23 |
freemangordon | not sure how to verify though | 09:24 |
tmlind | well if the command modem vblank simulation forces refresh the wrong time it could cause tearing | 09:26 |
freemangordon | yes, that's what I mean | 09:26 |
tmlind | so maybe test first with the vblank-after-framedone.patch applied, then without that and the revert to get rid of the vblank patch | 09:27 |
freemangordon | yep, rebooting with the new kernel | 09:27 |
tmlind | ok | 09:27 |
tmlind | i mean second test without vblank-after-framedone.patch and the vblank patch reverted | 09:27 |
freemangordon | mhm | 09:28 |
tmlind | ok | 09:28 |
freemangordon | vblank patch is not upstream, right? | 09:28 |
tmlind | nope | 09:30 |
freemangordon | still tears, lemme try with that reverted | 09:31 |
tmlind | ok | 09:33 |
freemangordon | hmm | 09:40 |
freemangordon | still tears | 09:40 |
freemangordon | but that patch is not needed :) | 09:40 |
tmlind | any issues with terminal not updating? | 09:40 |
freemangordon | in h-d? | 09:40 |
tmlind | in h-d droid4 lcd with no hdmi cable connected | 09:40 |
freemangordon | or without compositing manager? | 09:40 |
tmlind | well not sure.. | 09:41 |
freemangordon | ok, xterm is fulscreen | 09:41 |
freemangordon | and updates fine | 09:41 |
freemangordon | hmm, rotation transition is way faster now | 09:41 |
tmlind | ok might as well revert that one too for m-l, seems this is is only needed for wl | 09:42 |
freemangordon | mhm | 09:42 |
freemangordon | hmm, maybe I shall do page fliip instead of 'dirty fb" | 09:45 |
freemangordon | it could be that "dirty fb" requests immediate redraw | 09:45 |
freemangordon | glmark2 Score: 84 | 09:50 |
freemangordon | I think this revert affected the performance | 09:50 |
tmlind | nice | 09:52 |
freemangordon | but I am not sure reverting it is the right thing to do | 09:52 |
tmlind | yeah sounds like some variant is still needed | 09:53 |
freemangordon | anyway, have to stop for few hours, will continue later on | 09:53 |
freemangordon | ttyl | 09:53 |
tmlind | ttyl | 09:53 |
uvos | new sphone incomming: | 20:35 |
uvos | supports typing + on keypad via longpress | 20:35 |
uvos | all the ui bits are now modules so you can unload the dialer, the call ui, the sms ui etc speratly and replace them with something else if you like | 20:36 |
uvos | sphone now allows you to choose at runtime between using a glib mainloop and needing only gtk/glib or using a glib/qt mainloop that allows qt and gtk windows at the same time | 20:37 |
uvos | so replacing the ui modules one at a time with qt ones is an option if someone is so inclined | 20:38 |
uvos | + various fixes | 20:39 |
freemangordon | sounds good | 21:19 |
Wizzup | ok, I have SDWire | 22:05 |
Wizzup | and some dc-dc power module | 22:06 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!