Drawing errors with tightvnc

[bsdphk]

FreeBSD CURRENT r333667 arm64 on Thinkpad T480

[ 323.338] (--) PCI:*(0:0:2:0) 8086:5917:17aa:225d rev 7, Mem @ 0xf0000000/16777216, 0xe0000000/268435456, I/O @ 0x0000e000/64, BIOS @ 0x????????/65536

drm-next-kmod-4.11.g20180505_1

tightvnc-1.3.10_4

See this picture:

http://phk.freebsd.dk/misc/drawerror.png

The horizontal black bars should not be there, and their position will vary with each repainting of the local window, for instance if you iconify it.

[johalun]

Thanks for reporting! Can you please try the stable release, pkg install drm-stable-kmod and see if that works better.

[bsdphk]

Will try when I have a chance, may take some days.

[wulf7]

I am observing similar artifacts while running xfreerdp and drm-next-kmod:

Installing drm-stable-kmod solves the problem.

12.0-CURRENT r333766 amd64 on Acer Aspire S7
[ 9.979] (--) PCI:*(0:0:2:0) 8086:0166:1025:0746 rev 9, Mem @ 0x50000000/4194304, 0x80000000/268435456, I/O @ 0x00002000/64, BIOS @ 0x????????/65536

[johalun]

@wulf7 Thanks for the report! Will setup a xfreerdp environment and investigate. Can you see these artifacts on any other application? Are you using intel or modesetting X driver?

[wulf7]

I`m using intel driver. Modesetting driver with drm-next-kmod does not suffer from these artifacts in my setup. I do not see artifacts on other applications (mostly xfce-bundled)

[johalun]

@wulf7 The Intel driver is more or less deprecated. I recommend modesetting unless there's a strong reason not to use it. To use that, uninstall xf86-video-intel and remove any Driver "intel" entry from xorg.conf. At any rate, we're working on a new release now and will hopefully work out these renderbugs we are seeing.

[wulf7]

I recommend modesetting unless there's a strong reason not to use it.

Thank you for suggestion. The thing I am most concerned about is a power consumption as my laptop has a battery of very limited capacity (3900mAh by design, 3000mAh now, decayed). That is why I prefer to run native 2d acceleration rather than mapped to OpenGL one.
I wonder if anybody compared UXA vs SNA vs GLAMOR in terms of battery life?

[valpackett]

I doubt it's going to be significant. Does Intel even have actual dedicated 2D hardware like tiny embedded GPUs?

And how much are the 2D operations accelerated by SNA or GLAMOR even used these days? Modern clients do their own rendering on CPU or with GL/Vulkan. And compositors use GL…

Anyway, you can measure power usage with Intel pcm tools. The command is something like pcm.x 1.

[wulf7]

I doubt it's going to be significant.

Maybe it is not significant but it is at least measurable: http://cynic.cc/blog/posts/sna_acceleration_vs_uxa/

And compositors use GL…

I use xfwm4 from ports as compositor. It does not support GL yet

Anyway, you can measure power usage with Intel pcm tools. The command is something like pcm.x 1.

Interesting tool. Thanks for pointing it. Usually I use discharge current sensor integrated into battery so I have to disconnect power cord to measure power usage. Unfortunately, I am too lazy to write some deterministic 2d graphics tests to make measurements after.

Anyway, SNA is not working for me due to I915_GEM_EXECBUFFER2 ioctl failures so it is time to give modesetting driver a chance.

[wulf7]

Looks like 48dbaae fixed xfreerdp for me (drm 4.16)

[johalun]

Cool! There's one more commit which fixes the same issue but in a different place so if you're building 4.16 from source, pull the latest version. Pkgs will be updated today.

[zeising]

the ports tree was updated with the latest snapshots earlier today.