Mesa 12.x?

[curaga]

FWIW, I agree with Phoronix. Modesetting is not a usable replacement.

[Juanito]

After some checking, I discovered that the previous Xorg-7.7-3d extension had been compiled for i686 after it failed to compile for i484.

This newer version does compile for i484, but then apparently fails to run.

My concern in using pentium-m would be that the i915 driver might run, but not i965 for example.

[curaga]

Pentium-m is above pentium3, so I don't think it would hinder i965 any way.

[Juanito]

To get the latest version of xf86-video-modesetting, it would be necessary to update xorg-server to the latest version, which is 1.18.4 at the time of writing this:

The xorg-server was updated to 1.18.4 

..let us know how the modesetting driver works.

[Juanito]

On my hardware the xorg-server doesn't seem to want to use modesetting...

Code: [Select]

[  1319.947] (==) AIGLX enabled
[  1319.947] (==) Matched intel as autoconfigured driver 0
[  1319.947] (==) Matched modesetting as autoconfigured driver 1
[  1319.947] (==) Matched fbdev as autoconfigured driver 2
[  1319.947] (==) Matched vesa as autoconfigured driver 3
[  1319.947] (==) Assigned the driver to the xf86ConfigLayout
[  1319.947] (II) LoadModule: "intel"
[  1319.948] (WW) Warning, couldn't open module intel
[  1319.948] (II) UnloadModule: "intel"
[  1319.948] (II) Unloading intel
[  1319.948] (EE) Failed to load module "intel" (module does not exist, 0)
[  1319.948] (II) LoadModule: "modesetting"
[  1319.948] (II) Loading /usr/local/lib/xorg/modules/drivers/modesetting_drv.so
[  1319.948] (II) Module modesetting: vendor="X.Org Foundation"
[  1319.948] compiled for 1.18.4, module version = 1.18.4
[  1319.948] Module class: X.Org Video Driver
[  1319.948] ABI class: X.Org Video Driver, version 20.0
[  1319.948] (II) LoadModule: "fbdev"
[  1319.949] (WW) Warning, couldn't open module fbdev
[  1319.949] (II) UnloadModule: "fbdev"
[  1319.949] (II) Unloading fbdev
[  1319.949] (EE) Failed to load module "fbdev" (module does not exist, 0)
[  1319.949] (II) LoadModule: "vesa"
[  1319.949] (II) Loading /usr/local/lib/xorg/modules/drivers/vesa_drv.so
[  1319.949] (II) Module vesa: vendor="X.Org Foundation"
[  1319.949] compiled for 1.18.0, module version = 2.3.4
[  1319.949] Module class: X.Org Video Driver
[  1319.949] ABI class: X.Org Video Driver, version 20.0
[  1319.949] (II) modesetting: Driver for Modesetting Kernel Drivers: kms
[  1319.949] (II) VESA: driver for VESA chipsets: vesa
[  1319.949] (--) using VT number 2

[  1319.952] (EE) open /dev/dri/card0: No such file or directory
[  1319.952] (WW) Falling back to old probe method for modesetting
[  1319.952] (EE) open /dev/dri/card0: No such file or directory
[  1319.952] (EE) Screen 0 deleted because of no matching config section.
[  1319.952] (II) UnloadModule: "modesetting"
[  1319.952] (II) Loading sub module "vbe"
[  1319.952] (II) LoadModule: "vbe"
[  1319.952] (II) Loading /usr/local/lib/xorg/modules/libvbe.so
[  1319.953] (II) Module vbe: vendor="X.Org Foundation"
[  1319.953] compiled for 1.18.4, module version = 1.1.0
[  1319.953] ABI class: X.Org Video Driver, version 20.0
[  1319.953] (II) Loading sub module "int10"
[  1319.953] (II) LoadModule: "int10"
[  1319.953] (II) Loading /usr/local/lib/xorg/modules/libint10.so
[  1319.955] (II) Module int10: vendor="X.Org Foundation"
[  1319.955] compiled for 1.18.4, module version = 1.0.0
[  1319.955] ABI class: X.Org Video Driver, version 20.0
[  1319.955] (II) VESA(0): initializing int10
[  1319.956] (II) VESA(0): Primary V_BIOS segment is: 0xc000
[  1319.956] (II) VESA(0): VESA BIOS detected

[Juanito]

So, it couldn't hurt to update the Tiny Core Linux repos with a fresh snapshot of xf86-video-intel, directly from git.

done for x86 and x86_64

[Juanito]

Xorg-7.7-3d recompiled for pentium-m and reposted.

I get this:

Code: [Select]

$ glxinfo | grep -i device
    Device: Mesa DRI Intel(R) Haswell Mobile x86/MMX/SSE2 (0xa16

[Juanito]

This forces modesetting to be used:

Code: [Select]

$ cat /usr/local/share/X11/xorg.conf.d/20-intel.conf
Section "Device"
    Identifier  "Intel Graphics"
    Driver      "modesetting"
    Option      "AccelMethod"    "glamor"
    Option      "DRI"            "3"
EndSection
..but dri3 is not enabled.

With this dri3 is enabled:

Code: [Select]

Section "Device"
   Identifier  "Intel Graphics"
   Driver      "intel"
   Option      "DRI" "3"
EndSection

[pq5190362]

FWIW, I agree with Phoronix. Modesetting is not a usable replacement.

Why not? Care to elaborate? Also, have you filed a bug for whatever issues you are experiencing with xf86-video-modesetting:

https://bugs.freedesktop.org/buglist.cgi?component=Driver%2Fmodesetting&order=changeddate%20DESC%2Cbug_status%2Cpriority%2Cassigned_to%2Cbug_id&product=xorg&query_format=advanced&resolution=---

?

 

..but dri3 is not enabled.

You should read the following thread:

https://bbs.archlinux.org/viewtopic.php?id=211792

In a nutshell:

xf86-video-modesetting does in fact use DRI3 by default, but Xorg.0.log does not seem to report it properly for whatever reason.

However, you should be able to check if DRI3 is enabled with the following command from a terminal:

Code: [Select]

LIBGL_DEBUG=verbose glxinfo | grep libgl
You should see:

libGL: Using DRI3 for screen 0

Do you?

 

[Juanito]

Yes, with the modesetting driver I get this:

Code: [Select]

$ LIBGL_DEBUG=verbose glxinfo | grep libgl
libGL: Can't open configuration file /home/tc/.drirc: No such file or directory.
libGL: pci id for fd 4: 8086:0a16, driver i965
libGL: OpenDriver: trying /usr/local/lib/dri/tls/i965_dri.so
libGL: OpenDriver: trying /usr/local/lib/dri/i965_dri.so
libGL: Can't open configuration file /home/tc/.drirc: No such file or directory.
libGL: Using DRI3 for screen 0

..seems odd that it doesn't appear in the xorg log though...

[curaga]

Why not? Care to elaborate?

Because it uses glamor, ie emulates 2d acceleration on OpenGL. This means that for anything with 2d functionality, it is likely slower, uses more memory, and requires lots of disk space for OpenGL support - this has been confirmed by phoronix benchmarks repeatedly.

By being a generic driver, it also cannot use any card-specific functionality, such as anti-tearing, proper XV acceleration, colorspace conversion, etc that have their own units.

[Juanito]

Benchmark scores

Integrated Graphics Chipset: Intel(R) HD Graphics 4400, corepure64 v7.2

Code: [Select]

    glmark2 2014.03
=======================================================
    OpenGL Information
    GL_VENDOR:     Intel Open Source Technology Center
    GL_RENDERER:   Mesa DRI Intel(R) Haswell Mobile
    GL_VERSION:    OpenGL ES 3.0 Mesa 12.0.1

[intel dri2]
glmark2 Score: 1890
glmark2-es2 Score: 1853

[intel dri3]
glmark2 Score: 1726
glmark2-es2 Score: 1722

[modesetting]
glmark2 Score: 1658
glmark2-es2 Score: 1646


[Misalf]

Hmm. Intel driver now loads successfully but I'm back at unusable 3D.
I guess I'd have to go back to 6.x to get 3D working.

[pq5190362]

Because it uses glamor, ie emulates 2d acceleration on OpenGL. This means that for anything with 2d functionality, it is likely slower, uses more memory, and requires lots of disk space for OpenGL support

That is a moot argument.

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

this has been confirmed by phoronix benchmarks repeatedly.

What you are saying is not true, but see for yourself:

https://www.phoronix.com/scan.php?page=news_item&px=GLAMOR-EXA-2D-Ubuntu-16.04
https://www.phoronix.com/scan.php?page=news_item&px=Radeon-Modesetting-DDX-Xenial
https://www.phoronix.com/scan.php?page=news_item&px=AMDGPU-DDX-Modesetting
https://www.phoronix.com/scan.php?page=news_item&px=Intel-DDX-May-Tests

 

Hmm. Intel driver now loads successfully but I'm back at unusable 3D.

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

I guess I'd have to go back to 6.x to get 3D working.

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

[Juanito]

..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?