<div dir="ltr">Thank you for your suggestions!<div><br></div><div>After digging around, trying to find the mechanism (method) the Meson was actually using to find the dependencies, I was stuck, unable to determine what 'config-tool' meant. After reading the documentation, I concluded that I could manipulate 'cmake' or 'llvm-config' methods, something I understood. On my first attempt at a fix, I decided to just comment out the method altogether, using Meson's default 'auto' method:</div><div><br></div><div>dep_llvm = null_dep<br>with_llvm = false<br>if _llvm != 'false'<br> dep_llvm = dependency(<br> 'llvm',<br> version : _llvm_version,<br> modules : llvm_modules,<br> optional_modules : llvm_optional_modules,<br> required : (<br> with_amd_vk or with_gallium_radeonsi or with_gallium_swr or<br> with_gallium_opencl or _llvm == 'true'<br> ),<br> static : not _shared_llvm,<br># method : _llvm_method,<br> fallback : ['llvm', 'dep_llvm'],<br> )<br> with_llvm = dep_llvm.found()<br>endif<br><div><div dir="ltr" class="gmail_signature" data-smartmail="gmail_signature"><div dir="ltr"><div><br></div><div><br></div><div>Now, I am sure that there was very good reason to use 'config-tool', but I just don't know what it means. In the context of my system with my putting LLVM in my $HOME/opt folder, 'auto' seems to work just fine.</div><div><br></div><div>Right now I'm re-compiling LLVM with RTTI enabled because now it's complaining about that.</div><div><br></div><div>I'm happy to talk to developers that know more about Meson in order to create a PR/patch for this kind of thing. I'd also be happy to test other conditions (OSX, Windows, etc.) as long as I know what I'm looking for. </div><div><br></div><div>Anyway, thank you again! It takes a lot of discipline and commitment to offer help to newcomers on a mailing list.</div><div><br></div><div>Have a great week and stay healthy!</div><div><br></div><div>Andrew Jackman<br><a href="mailto:kd7nyq@gmail.com" target="_blank">kd7nyq@gmail.com</a></div></div></div></div><br></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Sun, Apr 26, 2020 at 6:56 PM Ken Moffat <<a href="mailto:zarniwhoop@ntlworld.com">zarniwhoop@ntlworld.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">On Sun, Apr 26, 2020 at 05:59:04PM +0300, Ruslan Kabatsayev wrote:<br>
> Hi,<br>
> <br>
> On Sun, 26 Apr 2020 at 07:36, Jackman <<a href="mailto:kd7nyq@gmail.com" target="_blank">kd7nyq@gmail.com</a>> wrote:<br>
> ><br>
> > Good evening! I hope you are all doing well.<br>
> ><br>
> > My goal is simply to get OpenCL 1.2+ working on my MXLinux (Debian Buster derivative) machine with an AMD GPU.<br>
> ><br>
[...]<br>
> ><br>
> > When I try to build Mesa, however, I am still getting this error:<br>
> ><br>
> > Message: libdrm 2.4.100 needed because amdgpu has the highest requirement<br>
> > Run-time dependency libdrm_intel found: YES 2.4.101<br>
> > Run-time dependency libdrm_amdgpu found: YES 2.4.101<br>
> > Run-time dependency libdrm_radeon found: YES 2.4.101<br>
> > Run-time dependency libdrm_nouveau found: YES 2.4.101<br>
> > Run-time dependency libdrm found: YES 2.4.101<br>
> > llvm-config found: YES (/home/jackman/opt/bin/llvm-config) 10.0.0<br>
> > Run-time dependency LLVM (modules: amdgpu, asmparser, bitreader, bitwriter, engine, ipo, mcdisassembler, mcjit, native, coroutines) found: NO (tried config-tool)<br>
> <br>
> This hints that LLVM is not fully functional. In particular, if you<br>
> look at the section for LLVM in BLFS[1], you'll find the following<br>
> cmake option is used there:<br>
> <br>
> -DLLVM_TARGETS_TO_BUILD="host;AMDGPU;BPF"<br>
> <br>
> You may want to enable these targets too. Then at least the amdgpu<br>
> module from the list should be found (not sure if the other modules<br>
> listed have also not been found).<br>
> <br>
<br>
Just to expand on that: in BLFS we added BPF for v4lutils. The<br>
AMDGPU target has been needed (for amdgpu, which the OP is trying to<br>
enable) for a long time. But we don't build any more than llvm,<br>
clang and compiler-rt so it is possible that one of the other parts<br>
of llvm might need another target.<br>
<br>
For mesa itself, we have some explanation of the options - maybe<br>
useful to reduce how much gets built. I think we removed the new<br>
iris target earlier, because of segfaults, or else we are going to<br>
(not relevant to amdgpu, nor to previous versions of mesa).<br>
<br>
OTOH, if someone is building all of it then perhaps the full set of<br>
drivers might be what they want - for me, building more than I need<br>
wastes time and space, but YMMV.<br>
<br>
ĸen<br>
-- <br>
He could send for Ptraci, his favourite handmaiden. She was special.<br>
Her singing always cheered him up. Life seemed so much brighter when<br>
she stopped. -- Pyramids<br>
_______________________________________________<br>
mesa-users mailing list<br>
<a href="mailto:mesa-users@lists.freedesktop.org" target="_blank">mesa-users@lists.freedesktop.org</a><br>
<a href="https://lists.freedesktop.org/mailman/listinfo/mesa-users" rel="noreferrer" target="_blank">https://lists.freedesktop.org/mailman/listinfo/mesa-users</a><br>
</blockquote></div>