pal1000 / mesa-dist-win Goto Github PK

I post the request as an "issue" 'cause it's the only way to contact pal1000 through github :-D

Man, can you do the same (windows binaries) for POCL now @1.0? https://github.com/pocl/pocl

I am getting illegal instruction when I trying to use opengl32.gll
From depends_x64
00:00:21.282: Second chance exception 0xC000001D (Illegal Instruction) occurred in "%DIR%\OPENGL32.DLL" at address 0x00007FF975EDB0E3 by thread 1.

Here I saw about AVX instruction set:
https://www.gamedev.net/forums/topic/648813-illegal-instruction-possible-causes/

How can I change the instruction set in opengl32.dll compilation? (maybe removing the AVX instruction set)

Hi,
I know isn't yet released but asking already in case I forget later.. mainly motivated on testing tesselation apps like Tessmark (requires env vars MESA_GL_VERSION_OVERRIDE=4.1 & MESA_GLSL_VERSION_OVERRIDE=410 to start_tessmark_windowed_1024x640)..

https://cgit.freedesktop.org/mesa/mesa/commit/?id=f09c466732e4a5b648d7503787777c926dd93c29

Add @donizettilorenzo to conversation as he showed interest in this topic as #5.

Recently Mesa3D landed clover state tracker support to llvmpipe driver which should provide an OpenCL CPU runtime similar to POCL that hopefully builds on Windows.

Roadmap

• Attempt clover build in standalone mode then test it with GPU Caps Viewer if builds successfully (fails: 1)
• Decide on a pre-built OpenCL ICD loader. Tested NVIDIA build and it worked just fine.
• Build in ICD mode then register and test it with GPU Caps Viewer if builds successfully.
• Update per app deployment tool to support OpenCL standalone install mode.
• Update system wide deployment tool to support OpenCL ICD install mode by complying with this spec.

Bonus

• Fake advertise CL GPU support on Microsoft Basic Display Adapter.

NV_depth_buffer_float: not supported
[06:30:28] Error/Warning in vertex shader:
[06:30:28] 0:1(10): error: GLSL 4.00 is not supported. Supported versions are: 1.10, 1.20, 1.30, 1.40, 1.00 ES, and 3.00 ES

can you help please with this error ?

Root cause: %LOCALAPPDATA%\Microsoft\WindowsApps is in PATH and %LOCALAPPDATA%\Microsoft\WindowsApps\python.exe is provided as a convenience shortcut to Python 3.x UWP version.

Sorry for writing an issue but i just wanted to thank you!!! This was very useful! I had a very hard time compiling Mesa.

Once more, Thank You!

@TheLastRar commented in #50.

With MinGW release and the hotfix my OpenGL application starts and works setting Gallium=zink
What hotfix did you do?

I'm trying to get the MinGW release to work, but my program just hangs (and then crashes a little later)
This happens even when I'm not settings Gallium environment variables.

Not an issue related to your installer, works fine, but I was wondering if you could take a look at this, not sure on how to contact you besides here.

https://www.opengl.org/discussion_boards/showthread.php/199760-Mesa3D-NVIDIA-CG-and-OpenGL-profiles-vp40-fp40

Cheers!

When trying to start with Far manager it reports "Can not execute ..., The operation was cancelled by the user. Any ideas for trblshoot?

I'm facing the unsupported Hardware acceleration error message that your great software responds to.
I installed Sketchup 2017 as a Chocolatey package.

2 remarks :

• This 2017 release of SketchUp is the last, installable, free version of SketchUp.
• the Chocolatey page mentions the Hardware requirement : it requires a 3D class video card with 1GB of memory or higher and hardware acceleration which my old Dell Latitude laptop does not have.

Here is the dedicated page
Once all the information required by mesa3d was filled in, I launched sketup.exe
But nothing happens anymore. So I sadly had to uninstall the package.

I don't know if the blocking is on the side of this old 2017 Sketchup version or if we I should rather study the specificity of the Chocolatey package.

Any idea ?

Makes it easier to integrate in automated pipelines like e.g. VCPKG and looks less "shady" than executing an exe. Developer mostly know archives and how to extract them.

Up until now I was maintaining build environment information in buildenvconf.md. This however has some drawbacks:

• it is a manual process;
• it only covers MSVC build, MSYS2 Mingw isn't covered.

To automate this process it is necessary to acquire version information for each dependency including optional ones and dump all this information to a file for MSVC build and another for Mingw build.

Common dependencies

• Windows;
• Resource Hacker;
• 7-zip
• Git for Windows
• Build commands and notes

MSYS2 build dependencies

• a dump of "pacman -Q" is enough.

MSVC build dependencies

• Visual Studio
• Windows SDK;
• Python and its required packages
• LLVM
• Ninja
• CMake
• winflexbison package
• flex
• Bison

I am trying to use mesa-dist-win on github CI runners. To do so I need to automate the deploy script.
I have found that in the systemwidedeploy.cmd script it tries to copy some files that do not exist in the package.
In the osmesa section, it has
@if "%deploychoice%"=="3" set osmesatype=gallium @if "%deploychoice%"=="4" set osmesatype=swrast @IF /I %PROCESSOR_ARCHITECTURE%==X86 copy "%mesaloc%\x86\osmesa-%osmesatype%\osmesa.dll" "%windir%\System32" @IF /I %PROCESSOR_ARCHITECTURE%==AMD64 copy "%mesaloc%\x86\osmesa-%osmesatype%\osmesa.dll" "%windir%\SysWOW64" @IF /I %PROCESSOR_ARCHITECTURE%==AMD64 copy "%mesaloc%\x64\osmesa-%osmesatype%\osmesa.dll" "%windir%\System32"

But in the package, the directories osmesa-galium and osmesa-swrast do not exist. If they did exist these lines would overwrite the osmesa.dll that had already been copied.

Using Mesa, I am unable to create an OpenGL context past version 3.3. Any later version gives the error: "WGL: Failed to create OpenGL context". Even though the normal driver supports OpenGL 4.3, using Mesa I can only use up to version 3.3. The OpenGL version is given as "OpenGL version 3.1 Mesa 19.0.2". Isn't Mesa 19.0.2 supposed to implement up to OpenGL 4.5?

Using glewinfo, it shows that OpenGL 4.1-4.5 are missing (although the extensions from those versions are not). However, I am still unable to create an OpenGL 4.6 context, which is marked as "OK".

Hi... I'm trying to get yuzu (switch emulator) works with mesa and I created a bat file with MESA_GL_VERSION_OVERRIDE=4.5COMPAT

Yuzu runs but it doesn't display anything, I can only hear the sound... I have black screen... Is this a bug? I have tried old version of mesa too, but it has the same problem D: Is there anyway to fix it?

Hi,
just forgot Zink got merged in Mesa some months ago and just curious if possible to build this driver also in your Mesa distribution.. as you know is a OpenGL 3.x driver that routes calls to underlying Vulkan driver on the system or that's the idea at least.. don't know if requires some specific Win patches or can build easily..
as said just a question, not a real need, and no pressure to integrate in case you aren't interested or it's currently unbuildable on Windows..
thanks..

Hello!

This is probably a user related issue (me) but I couldn't find another way to contact you and ask about the issue. I am running Windows 10 and I'm using AMD R9 280 GPU, the app I want to run is Cemu, which is an emulator of the Wii U.

Since AMD OpenGL drivers are not that good, I wanted to try the Mesa drivers and found this thread: https://www.opengl.org/discussion_boards/showthread.php/199121-Mesa3d-installing-for-Windows-10/page3

I downloaded the mesa-18.2.6-release-sfx.exe, ran it, then I ran the perappdeploy.cmd, and provided the info for the folder of cemu.exe. Now Cemu runs using mesa it seems (At least it says Generic GPU driver, instead of AMD GPU as it normally does), but the screen is completely black. I have tried a couple of different settings in the perappdeploy.cmd, but the result is the same. Have I missed something fundemental? Anything else that I might be doing wrong?

I am quite convinced that there is a large community out there that would very much appreciate it, if this solution would work.

Many thanks in advance!

Hello

I have OpenGL Version 2.1 and I want run Blender 2.80 that support OpenGL 3.3 minimum, so I downloaded Mesa 3d for windows and it requires ( libglpi.dll ) please I want it for Important ?

Thanks

I try to use your dll for offscreen rendering under Windows with .net core.

The following code works well under Linux, but fails under Windows: no exceptions, but glReadPixels returns buffer with zeros.

public static void Main(string[] args)
{
	int w = 800;
	int h = 600;
	int length = w * h * 4;
	byte[] buffer = new byte[length];
	IntPtr pnt = Marshal.AllocHGlobal(Marshal.SizeOf(buffer[0]) * length);

	var ctx = OSMesaCreateContext(GL.GL_RGBA, IntPtr.Zero);

	try
	{
		Marshal.Copy(buffer, 0, pnt, length);

		if (!OSMesaMakeCurrent(ctx, pnt, GL.GL_UNSIGNED_BYTE, 800, 600))
		{
			Console.WriteLine("OSMesaMakeCurrent failed!");
			return;
		}

		GL.glMatrixMode(GL.GL_PROJECTION);
		GL.glLoadIdentity();
		GL.glOrtho(0.0, 1.0f, 0.0, 1.0, 0.0, 1.0);
		GL.glMatrixMode(GL.GL_MODELVIEW);
		GL.glLoadIdentity();

		GL.glBegin(GL.GL_TRIANGLES);
		GL.glColor3f(1.0f, 0.0f, 0.0f); GL.glVertex2f(0.0f, 0.0f);
		GL.glColor3f(0.0f, 1.0f, 0.0f); GL.glVertex2f(0.5f, 1.0f);
		GL.glColor3f(0.0f, 0.0f, 1.0f); GL.glVertex2f(1.0f, 0.0f);
		GL.glEnd();

		GL.glReadPixels(0, 0, w, h, GL.GL_RGBA, GL.GL_UNSIGNED_BYTE, pnt);

		Marshal.Copy(pnt, buffer, 0, length);

		using (Image<Rgba32> image = Image.LoadPixelData<Rgba32>(buffer, w, h))
		{
			image.Save("test.bmp", new BmpEncoder());
		}
	}
	finally
	{
		OSMesaDestroyContext(ctx);
		Marshal.FreeHGlobal(pnt);				
	}

	Console.WriteLine("OK");
}

[DllImport("osmesa-gallium")]
public static extern IntPtr OSMesaCreateContext(uint format, IntPtr sharelist);
[DllImport("osmesa-gallium")]
public static extern IntPtr OSMesaCreateContextExt(uint format, int depthBits, int stencilBits, int accumBits, IntPtr sharelist);
[DllImport("osmesa-gallium")]
public static extern bool OSMesaMakeCurrent(IntPtr hdc, IntPtr buffer, uint type, int width, int height);
[DllImport("osmesa-gallium")]
public static extern void OSMesaDestroyContext(IntPtr hdc);

I put opengl32.dll and osmesa-gallium.dll into my bin folder.

I would be grateful if you could tell any suggestions what I could miss

Details are in the heading. Might worth a check... Different malware is reported in different versions. i.e. mesa3d-20.3.4-release reports Trojan:Script/Conteban.A!ml.

Installing your Mesa drivers in my Virtualbox VMs has greatly improved graphics compatibility. However, there's one application (Revit 2019) that won't use Mesa if Direct3D is not detected. It doesn't detect D3D, so it defaults to its own software renderer, but something strange happens because that renderer has become really slow (about 30 seconds per frame!!) after installing the Mesa drivers. I checked that all kinds of hardware acceleration are disabled for this VM. It runs Windows 7 64bit, latest service pack, and Virtualbox is 5.2.18.

Somehow, I have the feeling that if there was any Direct3D implementation that could be used on top of the Mesa drivers, then it would work fine... Is there any such implementation? Thanks!

I've installed mesa-18.1.0.600-1-sfx.exe using option 1 in 'systemwidedeploy.cmd' in VirtualBox 5.2.12 with guest Win10 Pro version 1803. If I try to start a command window using 'Run as Administrator', a error dialog pops up with "File System error (-1073741189)." Other attempts to perform actions requiring admin privileges result in the same error. I don't see this problem with VMware running a similar guest. I haven't found any explanation for what the error code means, nor any similar reports.

If this isn't the best forum to report this problem, please let me know. Thanks.

Hi!

Just a quick question: I'm reading that when you install these drivers, you cannot enable again the 3D acceleration from VirtualBox. That's fine for me (in fact I'm going to use mesa-dist-win because the 3D acceleration in VirtualBox is not good enough for the apps I use), but, in case I need to enable 3D acceleration again in the future, how should I proceed? Can I uninstall mesa-dist-win, enable 3D acceleration, and later reinstall mesa-dist-win again if I wish to use Mesa again?

Thanks!

MSYS2 Mingw is slightly faster than MSVC as it benefits from SSSE3 optimization via -march=core2 compiler flag which lacks equivalent in MSVC build and even further optimizations can be applied by changing march value to newer CPU generation. Historically @Alexpux didn't post regular updates to MSYS2 Mingw builds of Mesa so I feel I should tackle this as well. Reverse engineering his build doesn't appear to be very hard.

Tasks list

• Support locating and loading of MINGW64 and MINGW32 environments;
• Research necessary packages and patches and implement their management, use pre-built LLVM (info available here : https://github.com/Alexpux/MINGW-packages/tree/master/mingw-w64-mesa);
• Update Mesa3D build to to use toolchain=mingw option;
• Update distribution creation to support linux .a libraries;
• pass necessary compiler and linker flags.

Note: Deployment tools work unchanged.

Optional

• Self made LLVM build;
• swr build (needs upstream fixes).

Longer term

• Meson build support.

Hi,

On Windows 10, Windows Defender complains about finding 3 trojans inside the latest release package;

Can you check and clean if possible?

am using pre compiled modules and its slow any ideas how to fix this

Hi,
don't know if it's much to ask for, but anyway interested to try new LLVMpipe MSAA 4x support merged recently along with three new exts supported:
(mesamatrix)
2 days | llvmpipe: enable ARB_sample_shading
2 days | llvmpipe: enable GL_ARB_shader_texture_image_samples
2 days | llvmpipe: simple texture barrier implementation. (GL_ARB_texture_barrier)

Also few ours ago "wgl: add MSAA support" was merged fixing/adding that support to Windows..
https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4926

even Phoronix writed an article:
https://www.phoronix.com/scan.php?page=news_item&px=LLVMpipe-4x-MSAA-Bits

thanks anyway for your work!

Why the image presented to the output (monitor) seems like processed by a sharpening filter?
Mipmap LoD management issues?

PNG (screenshot taken via STAMP button, load the FULL RESOLUTION version ) -> https://ibb.co/XSTFsdG

No problem with LLVMPipe ! Output is regular!

Hi,

Would it be possible to make an installer that includes also the libs and the headers? I'm trying to port a unix program that uses the GLX API and need them while building that other program.

Thanks

I think being able to install via scripts would be a very useful feature. As it is now it is not possible to use the deploy scripts in provisioning systems. I am currently working with Github runners. As these are VMs created on-demand they need setting up at each run via scripts.
I did try and modify the systemwidedeploy.cmd to accept an argument (1 - 6) but it was getting very complicated bypassing the "pauses" and "goto deploy"
It would be great if you could add a script that accepts an argument instead of the menu based install.

Hello,
Any attempt to download recent msvc release files are blocked by antivirus claiming the Trojan:Win32/Cryptinject!ml
The mingw release seem to work fine

I can't seem to find spirv_to_dxil executable and/or libspirv_to_dxil in the release packages.

In the source tree it would be these build targets:
https://gitlab.freedesktop.org/mesa/mesa/-/blob/master/src/microsoft/spirv_to_dxil/meson.build

Do you build only drivers and this is intentional or can I possibly ask for these two to be added as well?

Please! 👍

https://www.phoronix.com/scan.php?page=news_item&px=Mesa-20.3-RC1-Released&utm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+Phoronix+%28Phoronix%29

To add Meson build support to the build script the following tasks must be completed:

Preparation phase

• Ensure everything is loaded and unloaded from PATH correctly;
• Split the build script into modules to handle growth triggered by this new feature support.

Python

• Recognize the Python version (py2/3 awareness);
• Extend py2/3 awareness with Python launcher support;
• Research what modules are needed for py3 / meson build and make the appropriate changes.

Mesa3D

• Support build with Meson and MsBuild if applicable;
• Support build with Meson and Ninja if applicable;
• LLVM Meson binary wrap support (required for llvmpipe and SWR);
• Retain current control over build targets or options.

Check new dependencies

• pkg-config and if needed libwinpthreads.

Finale

• Test for regressions and fix bugs.

Hi!

I was hoping to be able to run a program called Topaz Labs JPEG to RAW AI on a VMware server using software/CPU mode. It's a Qt5 application. The VMware server is running Windows Server 2008 R2.

I downloaded and installed the trial. When I try to run the program, it crashes instantly with this:

Faulting application name: Topaz JPEG to RAW AI.exe, version: 0.0.0.0, time stamp: 0x5c4f4ade
Faulting module name: unknown, version: 0.0.0.0, time stamp: 0x00000000
Exception code: 0xc0000005
Fault offset: 0x0000000000000000
Faulting process id: 0x16c0
Faulting application start time: 0x01d5e2c1e27805ff
Faulting application path: C:\Program Files\Topaz Labs\Topaz JPEG to RAW AI\Topaz JPEG to RAW AI.exe
Faulting module path: unknown
Report Id: 2167a7cf-4eb5-11ea-a2a8-0050560b8935

I can see in the installation directory there is an opengl32sw.dll file. Not sure if this is MESA based or what, but I tried renaming it to opengl32.dll. The program looks like it gets a little farther but it still crashes, this time with the following:

Faulting application name: Topaz JPEG to RAW AI.exe, version: 0.0.0.0, time stamp: 0x5c4f4ade
Faulting module name: Qt5Gui.dll, version: 5.12.0.0, time stamp: 0x5c0514b1
Exception code: 0xc0000005
Fault offset: 0x00000000002f1310
Faulting process id: 0x938
Faulting application start time: 0x01d5e2c2041f97ef
Faulting application path: C:\Program Files\Topaz Labs\Topaz JPEG to RAW AI\Topaz JPEG to RAW AI.exe
Faulting module path: C:\Program Files\Topaz Labs\Topaz JPEG to RAW AI\Qt5Gui.dll
Report Id: 42dcdf1f-4eb5-11ea-a2a8-0050560b8935

Finally, I tried copying in opengl32.dll from the latest 64 bit MESA mingw release into the application directory. This also didn't work; I get the same crash as above (0xc0000005 in Qt5Gui.dll).

(I also tried the MSVC release; it didn't have a different result.)

The program appears to have a Windows registry key for software mode:

[HKEY_CURRENT_USER\Software\Topaz Labs\Topaz JPEG to RAW AI\appMain]
"j2r_gpuOrCpu"=dword:00000001

(I assume 1 is for CPU; it's set to 0 by default.)

That also didn't have any effect.

Again, I did verify with Dependency Walker that the MESA opengl32.dll was loading from the application directory. It was.

From what I've researched online, both JPEG to RAW as well as Qt5 should support software mode. I have so far had a lot of success using software that uses OpenGL on a VMware server by using MESA.

Have you heard of anything like this before and know of a way to fix this?

My virtual machine is running Windows Server 2008 R2 (which is Windows 7 x64 based).

Currently the fileinfos are empty.
It would be nice to have at lest the mesa version somewhere in the fileinfos so i can identify what version of mesa i am currently using.

Hi,
just forgot to ask about your interest on this announcement of CL and GL over D3D12:
"https://www.collabora.com/news-and-blog/news-and-events/introducing-opencl-and-opengl-on-directx.html"
no hurry, just asking.. as right now isn't even merged on mesa master branch.. but that's the plan eventually.. seems right now implementation is changing fast : https://gitlab.freedesktop.org/kusma/mesa/-/merge_requests

really just opening issue so I get notified of eventual news on this .. :-)
thanks..

Hi,
now that one use case of CLonD3D12 and GLonD3D12 is clear:
https://devblogs.microsoft.com/directx/directx-heart-linux/
seems other use case Microsoft is interested is for Windows on ARM64 as this version doesn't have OpenGL/OpenCL support..
see:
https://gitlab.freedesktop.org/kusma/mesa/-/merge_requests/69
seems they are testing support for Photoshop OpenCL kernels on CLonD3D12 and this maybe useful jointly with also upcoming Windows ARM64 x64 emulation support..
anyway all that intro was to ask also how difficult would be to ship you mesa OpenGL software rasterizers for ARM64? SWR makes no sense but hope llvmpipe is Neon accelerated.. as Mesa supports Linux ARM64..
note that I haven't yet any Win ARM64 machine so just curious to ask..
thanks..

Phoronix found some difficulties to build AVX512 support so just asking? Perhaps fixed in 17.3 master branch?

Using the SWR drivers (GALLIUM_DRIVER=swr) results in a completely black screen. The code:

• Runs natively
• Works with llvmpipe and softpipe
• Reports no errors with glGetError()
• Does not contain function pointer for OpenGL functions which are NULL
• Has MESA_GL_VERSION_OVERRIDE and MESA_GLSL_VERION_OVERRIDE

Why do these not work?

Guys, any idea why virustotal has some alerts?
https://www.virustotal.com/gui/file/c170e5b02191528330b8dbf42e2a19aaab68ad4700872a2021896c385b6e85fb/detection

Hi, I'm trying to run Android Emulators on my Windows 10 PC which has an old GPU that doesn't support newer versions of OpenGL and was wondering if this driver can help with this. I understand that performance won't be great but I'm just trying to test some stuff out.

I get the follow error message when loading some games and was wondering if OpenGL ES 3.1 is supported, thanks a lot for any feedback in advance!

Hi,
seems Vallium was merged into Mesa last month:
https://www.phoronix.com/scan.php?page=news_item&px=Mesa-VALLIUM-Merged
currently it's in Mesa master (so should be in 20.3)..
don't know current status but might be very immature at this point..
anyway, any interest to also ship Vallium Vulkan CPU driver in your Win Mesa builds?

thanks..

Using "mesa3d-21.0.0-rc5-release-msvc.7z"

Gallium=llvmpipe -> perfectly working
Gallium=D3D12 -> perfectly working
Gallium=zink -> ?!?!?!?!

Recently read on Federico Dossena's page that he's busy lately and so I think WineD3D builds are not updated.
I'm not sure wined3d ver 6 released (it's rc6 on fdossena.org)
Nevertheless could you create a build of wined3d for windows?

P.S. sorry I didn't know how else to contact you..
P.P.S. Much thanks for both your swiftshader and mesa build!

After downloading mesa3d-19.1.3-release-mingw.exe-> unzipping -> Runningsystemwidedeploy.cmd and selecting 6. Update system-wide deployment. The drivers on my Azure Windows 10 vm don't appear to be updated. Running Android emulator I get the following message:

I was expecting the Microsoft Basic Render Driver to change?

Simple request: can we have the 20.1 RC1 builds to test it?

Thanks!

I just upgraded my pc with a new mobo, RAM and CPU. The CPU is Ryzen 5 2400G with Vega 11 onboard. Thing is, my monitor is a VGA D-sub, which is not supported by my mobo (Gigabyte GA-A320M-H), so when I install video drivers, the screen goes to noise after 2 seconds, then blacking it and it starts again. I tried to install Mesa3D instead and wasn't happy with the results and decided to remove it. Question is - how do i remove it?