That depends on what is your target. If you're running qemu on a x86 host, it's better to not do software emulation and build for qemux86_64 as well, and then run qemu with kvm, so it executes directl

You might want to try enabling gallium-llvm only for the target mesa: PACKAGECONFIG:append:x86-64:pn-mesa = ' gallium-llvm' Alex

You need to build mesa, not mesa-native. And please show exact changes you made. Alex

Try latest poky master? Alex

The following works for me. I get tmp/work/core2-64-poky-linux/mesa/2_22.2.0-r0/packages-split/mesa-vulkan-drivers/usr/lib/libvulkan_intel.so tmp/work/core2-64-poky-linux/mesa/2_22.2.0-r0/packages-spl

I think for software vulkan you need to pass in -Dvulkan-drivers=swrast when building mesa for the target. Alex

I would first check whether any of that work in the branch has meanwhile landed upstream. The post is from 2021. Have you tried the software Vulkan in the guest? Alex

You need to clarify what you want to do exactly. Why can’t you use the existing recipe? Alex

Do keep in mind that PARALLEL_MAKE can and should be set per recipe, so you can make-limit only the worst items. Alex <Oliver.Westermann@...> wrote:

Anything written in C++ tends to consume 1-2 Gb of ram per compiler process. If that lands you in OOM, you probably should limit that with PARALLEL_MAKE:pn-nodejs, but otherwise that is the sad realit

TL;DR version =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D If you want to get everything needed to run a yocto build: - clone the bootstrap repository (ask your distribution maintainer where = it is): $ gi

vulkan-samples recipe is simply using the standard cmake installation routine from cmake.bbclass, so all questions should be aimed to upstream. And upstream doesn't want you to run installation at all

I finally got a moment to look at this. The binary with the samples is being built (you can find it in $WORKDIR/build/app/bin/x86_64/vulkan_samples), but is not installed. So if you want to get it pac

Next you need to check if you have the mesa-vulkan-drives package in deploy-* in the same mesa builddir, and does it contain the needed driver file? Alex

You need to check package-split/ to find out why. Alex

This is not logs from mesa, this is logs from your image recipe. You need to start with mesa. Alex

You need to go to mesa's build directlory and inspect what is getting built, with which options, and how it is installed and packaged. All the needed bits are there, but it does take a bit of 'yocto s

What often works is running it under strace - you might be able to see the one key syscall that causes the failure. Otherwise you might have to go through the heavier gdb step through. Alex

It helps if you say what you did precisely and show the complaint. Alex

Don’t bother with virgl yet. Start with the pure software implementation in Mesa. Alex