Quoting: SinaCutieAMD RX 460 was the last card I used.

That's strange, since Polaris is quite well supported. I didn't have any major issues with RX 480. What exactly was the problem with your card?

View this comment - View article - View full comments

I updated the wiki, and put the build script here: https://gist.github.com/shmerl/f4e5f76871239158cf083e37c5da56f4 [External Link]

View this comment - View article - View full comments

Example of how you can do it (assuming you put your mesa in /opt/mesa-${branch})

mesa_run.sh

#!/bin/bash

export mesa_dir=${mesa_dir:-"mesa-master"}

if [[ "$1" == "32" ]]; then
   arch_dir="x86"
   vk_arch="i686"
   shift
else
   arch_dir="x86_64"
   vk_arch="x86_64"
fi

mesa_env="LD_LIBRARY_PATH=/opt/${mesa_dir}/${arch_dir}:$LD_LIBRARY_PATH LIBGL_DRIVERS_PATH=/opt/${mesa_dir}/${arch_dir} EGL_DRIVERS_PATH=/opt/${mesa_dir}/${arch_dir} VK_ICD_FILENAMES=/opt/${mesa_dir}/${arch_dir}/radeon_icd.${vk_arch}.json"

echo "Mesa env:"
echo "${mesa_env}"
echo "==========================================="

export ${mesa_env}
$@

View this comment - View article - View full comments

Quoting: slaapliedjeSo are you just launching with an LD_PRELOAD? Or do you change the .desktop file? I was going to ask about steam command parameters, but I don't think you use it due to DRM, right?

I'm using LD_LIBRARY_PATH and a bunch of variables needed for OpenGL and Vulkan custom locations. See in the linked wiki page. I'll update it soon and will also put my common scripts somewhere so you can see examples.

I.e. I made a script mesa_run.sh which launches its parameter with custom mesa, and I simply use that when I want to run something using that custom Mesa, rather default system one. You surely can use the same idea in desktop launchers and other scripts.

View this comment - View article - View full comments

Quoting: natis1For starters, Gentoo includes git versions of all packages in its repos so no need to git clone, ./configure, make, make install anything manually. Arch does something similar with its aur.

More importantly though, when you build a package from source using a git repository as the source on Arch that package is tracked by your package manager and can be updated along with the rest of your system (pacaur -Syu --devel). As far as I know, Debian and variants lets you build a package from source, but doesn't automatically keep it up to date. If you install Mesa git but never update it it will quickly fall behind the fixed release version.

I see. But all those methods install Mesa as a replacement for stable one. That's not the goal for me in this case. I like to have experimental / master Mesa alongside stable one, and only use it on demand when needed. So in such case building it and placing it in custom location works best.

Here is an example how to do it: https://www.gamingonlinux.com/wiki/Building_Mesa_from_source

View this comment - View article - View full comments

Actually correction, the refresh is going to be 12 nm. 7 nm will be used already in Navi, so that's in 2019.

View this comment - View article - View full comments

Quoting: slaapliedjeWell, the open source driver went through many names / iterations. radeon, radeonhd, radeonsi, radv, etc. Nvidia has had nv, nouveau.

I think you are mixing up kernel driver (radeon / amdgpu), and API implementations (radeonsi/r600 for OpenGL, radv/amdvlk for Vulkan and etc.). Having different names is quite reasonable, if they have actually different code for different hardware, or simply different implementations.

View this comment - View article - View full comments

Quoting: natis1I personally disagree. It's much harder to build Mesa from source on a distro like Ubuntu or openSUSE than Arch or Gentoo.

How exactly is it harder? Check how official distro packages configure and build Mesa, and use that as a starting point.

And in this case, I don't need PPAs and binary click installers, because I don't want to replace system Mesa with a manually built one, especially if it's a master branch. I need to use them in parallel.

View this comment - View article - View full comments

Quoting: slaapliedjeAlso, even with the open source stuff, you have to try and get latest Mesa on your system, and that's generally a pretty major thing to do on many of the binary distributions (not so bad on something like Arch.)

It's only as major as you are not yet familiar with building Mesa from source. Once you learn that, it's noting major, just compilation time. One thing many are missing that there is no need to replace system Mesa to use the latest one for running some games. You can use system packaged one for everything else, and use your custom built one for launching games that need the newest features. That's exactly what I do, so I don't need any package repositories for that and can build Mesa any time I feel like doing it.

Sure, building stuff from source might sound intimidating, but it's just some learning curve, nothing insurmountable.

View this comment - View article - View full comments

Quoting: mrdeathjrSadly amd dont have opengl 4.6 for now

That's incorrect. It has these:

GL_ARB_indirect_parameters
GL_ARB_pipeline_statistics_query
GL_ARB_polygon_offset_clamp
GL_ARB_shader_atomic_counter_ops
GL_ARB_shader_draw_parameters
GL_ARB_shader_group_vote
GL_ARB_texture_filter_anisotropic
GL_ARB_transform_feedback_overflow_query
GL_KHR_no_error

What's missing are these:

GL_ARB_gl_spirv
GL_ARB_spirv_extensions

Is Cemu using these two? I really doubt it does.

View this comment - View article - View full comments