freemangordon | ok, seems I lack the theory how a tear-free, zero-copy xorg driver shall be implemented, any hints appreciated | 07:33 |
---|---|---|
Wizzup | well tearing is prevalent on many devices, so I am not sure if that's a big deal atm | 07:58 |
freemangordon | but, if I am going to do a driver from scratch, why not take it into consideration? | 07:59 |
Wizzup | that's true, I think there are a few X drivers that have tearless operation | 08:03 |
Wizzup | one of the tearfree modsetting bug reports links to radeon drivers that solve this | 08:03 |
Wizzup | and there's als the modesetting patch that makes modesetting tearless on non-rotated displays | 08:03 |
freemangordon | hmm, seems xorg pvr_drv source code was available in the 4x versions of graphics sdk | 08:03 |
freemangordon | but not in 5x, for some reason | 08:04 |
Wizzup | this entire bug report might be interesting to read: https://gitlab.freedesktop.org/xorg/xserver/-/issues/244 | 08:04 |
freemangordon | my browser is too old for gitlab :( | 08:05 |
freemangordon | (ff 66) | 08:05 |
Wizzup | tearfree on amdgpu also has a big perf cost btw | 08:05 |
freemangordon | myba they copy the front buffer on start of every operation | 08:06 |
freemangordon | *maybe | 08:06 |
freemangordon | I mean - I am thinking on it for the last couple of days | 08:06 |
freemangordon | and still have no clear vision on how shall changes be copied between buffers :) | 08:07 |
Wizzup | the maybe check out this branch, but keep in mind that it doesn't reportedly work for everyone https://gitlab.freedesktop.org/vliaskov/xserver/-/commits/tearfree-1.20.9 | 08:08 |
Wizzup | but tearing being problematic is one of the reasons folks wanted to move to wayland | 08:13 |
Wizzup | I wonder if this is not something we want to tacke later | 08:13 |
Wizzup | tackle* | 08:13 |
Wizzup | this https://unix.stackexchange.com/questions/182801/why-is-video-tearing-such-a-problem-in-linux suggests to read https://magcius.github.io/xplain/article/index.html | 08:15 |
Wizzup | "After reading this (wonderful explanation) of how an Xorg server works I realized that X server paints window updates directly to the memory card at any random time unless you use a compositing enabled window manager." and below on the stackexchange link | 08:15 |
Wizzup | but yeah I don't have answers just pointers | 08:15 |
Wizzup | maybe you already read all this | 08:15 |
Wizzup | I'd be fine with tearing in place, if it's not artifacting and just tearing | 08:16 |
Wizzup | often with xorg is also seems to require special window manager workaround as well | 08:16 |
freemangordon | Wizzup: well, we are compositing (h-d) | 08:16 |
freemangordon | but, on d4 I see tearing when h-d scrolls | 08:17 |
Wizzup | yes, but kwim with compositing still has problems in certain vsync modes it seems for example | 08:17 |
freemangordon | so something is not quite right | 08:17 |
Wizzup | kwin* | 08:17 |
Wizzup | it could be related to opendrm, or when/how the x server syncs, how the exact page flip is timed, etc | 08:17 |
Wizzup | I think it'd be perhaps better to fix any hangs, get this stuff going, and then look at it again a bit later, but that's up to you of course | 08:18 |
Wizzup | as in I don't think this will be simple at all | 08:18 |
freemangordon | sure it is not | 08:18 |
freemangordon | the point is that I don;t see any point in fixing MS | 08:18 |
Wizzup | in any case I might read this since it seems good https://magcius.github.io/xplain/article/index.html | 08:18 |
Wizzup | well others might fix tearing in MS, there are patches for some of it | 08:19 |
freemangordon | it is not about the tearing | 08:19 |
Wizzup | oh | 08:19 |
freemangordon | I just don;t see it fit for our usecase | 08:19 |
freemangordon | rotation being one of the major issues | 08:19 |
freemangordon | olso, glamor is too slow even on d4, forget about n900 | 08:20 |
Wizzup | mhm | 08:20 |
freemangordon | my replacement is semi-useful, but it still does texture copy with a shader | 08:20 |
freemangordon | maybe this cannot be avoided, because of the compositing | 08:21 |
freemangordon | but in any case, I first want to see if I can get pvr_drv working | 08:21 |
freemangordon | because if this works, we get HW accelerated EXA for free | 08:22 |
freemangordon | and I will have to just add dri3/PRESENT on top, IIUC | 08:22 |
Wizzup | ok | 08:24 |
freemangordon | yeah :D | 08:25 |
freemangordon | XOrg-sources.tar.gz | 08:25 |
Wizzup | ? | 08:26 |
freemangordon | pvr_drv should be there, hopefully | 08:26 |
freemangordon | hmm, cannot find it | 08:32 |
freemangordon | yeah, not available, will have to RE it :( | 08:46 |
Wizzup | is it large? | 09:02 |
freemangordon | no, but still | 09:04 |
freemangordon | ~50k | 09:04 |
freemangordon | but I have one with DWARF info, so not much of an issue | 09:04 |
Wizzup | ok | 09:04 |
freemangordon | my concern is if this will help, as API has changed | 09:05 |
freemangordon | so, I will have to start porting it to 1.17 only after I have that REed | 09:06 |
Wizzup | yeah | 09:06 |
freemangordon | without even knowing if it is goingt o work | 09:06 |
freemangordon | but yeah | 09:06 |
Wizzup | uvos: hm call hangup didn't work for me and it kept vibrating after call was hung up on remote side | 12:21 |
mighty17[m] | hey is there any omap device with wm1811/wm8994 audio? | 15:09 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!