Mesa 12.x?

[pq5190362]

..but both the tests you link to for intel ddx versus modesetting and my own testing with glmark2 show that intel ddx is faster than modesetting, no?

xf86-video-modesetting might be a bit slower in 2D performance than xf86-video-intel according to the benchmarks, but by how much?

Do you notice a difference?

I have to admit I am using xf86-video-intel @ SNA @ DRI3 @ TearFree myself, but I am not a fanboy of xf86-video-intel and still like the idea of a unified DDX driver (xf86-video-modesetting).

I mainly am using xf86-video-intel because it has the TearFree option, which works quite well (although not quite perfect).

With xf86-video-modesetting there is tearing. Admittedly though, that apparently could be fixed by using a compositor (which I currently do not): https://whirm.eu/posts/fix-for-xorgs-modesetting-driver-tearing/ .

[Misalf]

Have you tried xf86-video-modesetting? Does 3D work with xf86-video-modesetting?

Via an Xorg config?

Code: [Select]

Section "Device"
Identifier  "Card0"
Driver      "modesetting"
Option      "AccelMethod"  "glamor"
Option      "DRI" "2"
EndSection

I tried DRI 1, 2, and 3. Results in:

Code: [Select]

[ 16549.418] (WW) glamor requires at least 128 instructions (64 reported)
[ 16549.418] (EE) modeset(0): Failed to initialize glamor at ScreenInit() time.
[ 16549.418] (EE)
Fatal server error:
[ 16549.418] (EE) AddScreen/ScreenInit failed for driver 0
[ 16549.418] (EE)
[ 16549.418] (EE)
Please consult the The X.Org Foundation support
at http://wiki.x.org
 for help.
[ 16549.418] (EE) Please also check the log file at "/var/log/Xorg.0.log" for additional information.
[ 16549.418] (EE)
[ 16549.434] (EE) Server terminated with error (1). Closing log file.


Maybe it's the kernel (i.e. the DRM driver) which is causing your issue?

Well, something is broken since 7.0 with intel i195.

[pq5190362]

Via an Xorg config?

You could probably also uninstall all xf86-video-* drivers, which should result in xf86-video-modesetting being used (it's built into xorg-server).

I tried DRI 1, 2, and 3.

You can not configure what type of DRI is being used with xf86-video-modesetting. xf86-video-modesetting does not have an option for configuring what type of DRI is being used.

Just type the following command into a terminal and see for yourself:

Code: [Select]

man modesetting
As you can see from the manpage, there is no such option. xf86-video-modesetting uses DRI3.

Results in:

[...]

Well, something is broken since 7.0 with intel i195.

Hmm, no idea, maybe Juanito and curaga can help.

[pq5190362]

Results in:

Looks like your hardware does not support glamor (which is being used by xf86-video-modesetting), see:

https://lists.freedesktop.org/archives/xorg/2016-July/058152.html

[Juanito]

Well, something is broken since 7.0 with intel i195.

I believe it's the drm pointer error, which I presume is fixed in later kernel versions, but I have not seen a patch for our kernel version.

On another note, I compiled the intel vulkan driver, but have not yet figured out how to test it...

[curaga]

That is a moot argument.

xf86-video-ati and xf86-video-amdgpu are also using glamor, just like xf86-video-modesetting .

If you don't care about disk, RAM, and performance, why not use Ubuntu?

-ati certainly does not default to glamor, it defaults to EXA. -amdgpu doesn't matter, because the hw supported by it does not have any 2d functionality.


Also, your own Intel link showed the Intel driver is faster. Cough.

[pq5190362]

-ati certainly does not default to glamor, it defaults to EXA.

You certainly should type this into a terminal:

Code: [Select]

man radeon
and then read this:

Option "AccelMethod" "string"
              Chooses  between  available  acceleration  architectures.  Valid
              values are EXA (for pre-TAHITI GPUs) and  glamor  (for  R300  or
              higher). The default is glamor as of TAHITI, otherwise EXA.

 

[curaga]

I'm aware. Tahiti is one of the last GPUs it supports. The huge majority defaults to EXA.

[pq5190362]

In some cases glamor is much faster than EXA on AMD hardware, and xf86-video-modesetting @ glamor is even faster than xf86-video-ati @ glamor:

https://www.phoronix.com/scan.php?page=news_item&px=GLAMOR-EXA-2D-Ubuntu-16.04

 

[Juanito]

seems the intel vulkan driver is not quite there yet:

Code: [Select]

$ VK_ICD_FILENAMES=/usr/local/share/vulkan/icd.d/intel_icd.json vulkaninfo
WARNING: Haswell Vulkan support is incomplete
gen7_pipeline.c:213: FINISHME: VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO
...

===========
VULKAN INFO
===========

Vulkan API Version: 1.0.21

INFO: [loader] Code 0 : Found manifest file /usr/local/share/vulkan/icd.d/intel_icd.json, version "1.0.0"

Instance Extensions:
====================
Instance Extensions count = 3
VK_KHR_surface                      : extension revision 25
VK_KHR_xcb_surface                  : extension revision  5
VK_EXT_debug_report                 : extension revision  3


Layers: count = 0
=======
Presentable Surface formats:
============================
GPU id       : 0 (Intel(R) Haswell Mobile)
Surface type : VK_KHR_xcb_surface
Format count = 1
B8G8R8A8_SRGB


Device Properties and Extensions :
==================================
GPU0
VkPhysicalDeviceProperties:
===========================
apiVersion     = 0x400005  (1.0.5)
driverVersion  = 1 (0x1)
vendorID       = 0x8086
deviceID       = 0x0a16
deviceType     = INTEGRATED_GPU
deviceName     = Intel(R) Haswell Mobile
...
..and if I try to run any of the test images, except smoketest:

Code: [Select]

$ VK_ICD_FILENAMES=/usr/local/share/vulkan/icd.d/intel_icd.json tri
WARNING: Haswell Vulkan support is incomplete
gen7_pipeline.c:213: FINISHME: VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO
...
Segmentation fault

[pq5190362]

I'm aware. Tahiti is one of the last GPUs it supports. The huge majority defaults to EXA.

Will no longer be true with xf86-video-ati 7.9.0, see:

https://cgit.freedesktop.org/xorg/driver/xf86-video-ati/commit/?id=f11531c99fcd6473f58b4d10efaf3efd84304d8e