Intel's Unreleased 3.0 X.Org Driver Gets More Fixes For DRI3/Present

[darkbasic]

glamor used to be slow, and since I was among the first (forcedly) early adopters I also was the #1 hater. Fortunately since the last year it is fast enough for everyday's use and I don't hate it anymore. I'm glad to see I'm not the only one who had to resort to modesetting because of xf86-video-intel bugs.

@everyone saying glamor is still slow:
You don't have the slightest idea of what *SLOW* means because you never experienced seconds lags when typying something in Konsole. Glamor was that slow initially.

[Sloth]

Recently got rid of xf86-video-intel. No regrets

This. Same here. Switched to modesetting. Vastly more stable and everything I use is faster (gtkperf by ~30%.) Haswell mobile and Desktop.

[pq1930562]

The amdgpu (and radeon for SI+ hw) DDX's are basically identical to modesetting -- they also use GLAMOR, and by extension are just as slow.

And how about xf86-video-nouveau?

Why do xf86-video-intel, xf86-video-ati, xf86-video-amdgpu and xf86-video-nouveau still exist and are still maintained when xf86-video-modesetting would work as well for all of them?

[pq1930562]

xf86-video-amdgpu contains amdgpu kernel specific code so i doubt that xf86-video-modesetting works.

Hmmm, but couldn't the guys from AMD (i.e. bridgman , agd5f and so) include this code in xf86-video-modesetting as well?

To me it sounds like a good idea to have one unified DDX driver (xf86-video-modesetting) which is being maintained and improved by everyone instead of everyone working on a different DDX driver (i.e. xf86-video-intel, xf86-video-ati, xf86-video-amdgpu, xf86-video-nouveau, xf86-video-openchrome and so on).

[darkbasic]

Xorg is going to die anyway, so old DDX are not that important anymore.

[haagch]

"This is a non-accelerated driver,"

But when using glamor it's still opengl accelerated.

You also linked to the separate xf86-video-modesetting that has been deprecated since 2014 when it was included in mainline xorg server.

[pq1930562]

Did you read the man page, video-modesetting is non accelerated graphics.

No, I had not read it, since you edited it in afterwards and it's on the previous page.

And as already mentioned by haagch , it is accelerated through glamor and you apparently also linked to an old version of xf86-video-modesetting.

Besides, what is your argument here? I was referring to the post from imirkin where he said that xf86-video-modesetting would be as fast as xf86-video-amdgpu and as fast as xf86-video-ati, since xf86-video-modesetting is using glamor for acceleration just like xf86-video-amdgpu and xf86-video-ati.

[pq1930562]

Look from your package manger, you will notice several other xorg drivers than for amdgpu,nouveau or intel. More better, look at source code of those xorg drivers, maybe you then figure why implementing everthing to one driver is not a good idea. They have not mixed those drivers in the kernel and for a good reason, modularity is they key in sw development.

Then why is xf86-video-modesetting better on Intel hardware than xf86-video-intel?

[darkbasic]

If you mean X11, xorg.org has opposite opinion. Read more, then you know more. What gamers do not need is 140 usd usb audio modules, so try figure contexts.

"X.Org Server (xorg for friends) is the free and open source implementation of the display server for the X Window System stewarded by the X.Org Foundation."
What I do not really need is someone who annoys me. Go playing with your Debian Testing, hurry before the next freeze will force you to use 6 months old software.

Michael how am I supposed to put someone in ignore list in vBulletin?

[Masush5]

On my old Sandy Bridge laptop modesetting is much more reliable than video-intel which causes X to crash or lock-up randomly regardless of using DRI 2/3 or UXA/SNA.
Regarding the performance of glamor i found that oddly enough using the video-ati ddx with glamor acceleration gives me ~40% less performance in gtkperf than using modesetting (which uses glamor as well) or EXA acceleration. Modesetting and EXA perf is nearly identical. Not sure what the reason for this might be, i thought the glamor codepaths in video-ati and modesetting to be very similar.
This is on a HD 5850.