| freemangordon | uvos: do you have an idea how to test compositing performance? | 18:21 |
|---|---|---|
| Wizzup | I think one of the x perf tools has a test for it | 18:21 |
| freemangordon | thanks | 18:22 |
| Wizzup | x11perf -compwinwin500 | 18:22 |
| freemangordon | I don;t know what this is measuring, but even if I disable compositing in the driver, it makes no difference in the score | 18:35 |
| Wizzup | maybe it's not a good test then | 18:36 |
| buZz | freemangordon: maybe run compiz? :D does that still exist? | 18:38 |
| buZz | or https://wiki.archlinux.org/title/xcompmgr | 18:39 |
| freemangordon | anyway, it is not compositing that slows down scrolling | 18:39 |
| Wizzup | well that's not an easy measurable 'test' | 18:39 |
| Wizzup | and h-d does compositing also | 18:39 |
| buZz | right | 18:39 |
| freemangordon | yes, I think what h-d does is the issue | 18:39 |
| buZz | ah, maybe using wrong opengl calls or something? | 18:40 |
| freemangordon | because with CheckComposite() etc returning TRUE and doing nothing I see no measurable performance increase when h-d is running | 18:40 |
| freemangordon | no, most -probably TFP is very slow | 18:41 |
| freemangordon | or... dunno | 18:41 |
| buZz | maybe need to try with a compositing-disabled window manager? | 18:41 |
| buZz | like fluxbox or something | 18:41 |
| freemangordon | well, if I stop h-d it is enough | 18:41 |
| freemangordon | fps increases by factor of 2-3 | 18:41 |
| freemangordon | I was just hoping it is xorg compositing that is the issue | 18:42 |
| freemangordon | maybe I should have done that before wasting a week on REing the blob :( | 18:49 |
| buZz | hehe cool, i now have 1080p output on hdmi , and indeed until you kill/restart h-d maemo is a bit confused | 20:44 |
| buZz | hmm, https://www.ti.com/lit/ml/swpt034b/swpt034b.pdf?ts=1665176547841 <- that says the 'display controller' has 'multi-pipelines' and 'multioutput' https://i.imgur.com/GiOYKvO.png | 23:05 |
| uvos | buZz: yes so? | 23:08 |
| uvos | freemangordon: so what dose compositing accelration in xorg do exactly | 23:09 |
| buZz | uvos: i'm trying to find if hdmi -and- lcd on could work | 23:09 |
| uvos | sure ofc | 23:09 |
| uvos | not just in theory it works in fine in practicce | 23:10 |
| buZz | because when i enable hdmi, the lcd turns off :P | 23:10 |
| uvos | just hildon dosent know what to do with it | 23:10 |
| uvos | no its just black beacuse hildon freaks out | 23:10 |
| uvos | it works fine with plain x | 23:10 |
| buZz | ah, maybe i shouldnt add --primary? | 23:10 |
| uvos | dosent help | 23:10 |
| uvos | hildon dosent know about that | 23:10 |
| uvos | this is use other window manager teretory | 23:11 |
| buZz | hmhm | 23:11 |
| uvos | freemangordon: while compositing x dosent really render anything, i though it was something about accelerating the offscreen buffers but idk. | 23:11 |
| buZz | but with just mirroring? | 23:11 |
| uvos | cant do that either | 23:12 |
| buZz | n900 can :P | 23:12 |
| uvos | since the dsi pannel can do one resolution, and one reslution only | 23:12 |
| buZz | yeah thats fine | 23:12 |
| uvos | and its unlikey the external montior can do the same resolution | 23:12 |
| uvos | so its unlikely to ever work | 23:12 |
| uvos | the n900 works entirely differently | 23:12 |
| uvos | it dosent have 2 display outputs at all | 23:12 |
| buZz | you can display a 960x560 buffer on a 1280x720 screen though :P | 23:13 |
| uvos | its just a dumb converter chip that drives the tv out | 23:13 |
| uvos | buZz: the mode has to line up | 23:13 |
| uvos | often the montiors will accept other scann rates thatn edid tells xorg | 23:13 |
| uvos | but you have to add a custom modline for that | 23:13 |
| uvos | never tried it | 23:13 |
| uvos | but that could work | 23:13 |
| buZz | hmhm | 23:15 |
| uvos | so the way that would work | 23:15 |
| buZz | well one of my hdmi 'tvs' is just a capture unit ;) | 23:15 |
| uvos | is use cvt to generate a modeline | 23:15 |
| uvos | then add the modeline to the crtc using xrandr | 23:15 |
| uvos | and activate it | 23:15 |
| uvos | then you can mirror the internal display to that | 23:16 |
| uvos | and kill hildon | 23:16 |
| buZz | hmm :) | 23:16 |
| uvos | should work if the external display accepts the scann rate (not guarenteed) | 23:16 |
| uvos | the other thing that might work | 23:17 |
| uvos | is to use a native resolution on the external display and add a xorg (xrandr --output --transfrom <matrix>) transformation matrix to the output to match the effective resolution to the one of the internal display | 23:19 |
| uvos | but that will only work if the ddx supports transformations | 23:19 |
| uvos | i think only glamor dose that | 23:19 |
| uvos | so no nvm | 23:19 |
| buZz | hehe laaame; https://i.imgur.com/YQkLDKb.png | 23:44 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!
