1. Forum
  2. Usenet
  3. LINUX.GENTOO.USER

  • Re: [gentoo-user] New SPIRV flag for LLVM

    From Eli Schwartz@21:1/5 to gevisz on Sun Jul 13 19:30:01 2025
    This is an OpenPGP/MIME signed message (RFC 4880 and 3156) --------------dvoItMoe0GwHEB4L0EqxgYHe
    Content-Type: text/plain; charset=UTF-8
    Content-Transfer-Encoding: quoted-printable

    On 7/13/25 9:33 AM, gevisz wrote:
    Updating my system today, I have found that two LLVM
    packages would be recompiled with new LLVM_TARGETS="SPIRV”.

    Googling LLVM_TARGETS SPIRV Gentoo gives almost
    no information except for this (in Russian): https://www.linux.org.ru/forum/general/16891018
    where people complain about LLVM_TARGETS, SPRV
    and Gentoo in general.


    Can't help you with that, but: https://bugs.gentoo.org/953517


    I started re-compilation of these two packages but ChatGPT
    says that I do not need SPIRV LLVM TARGET for my
    old Radeon HD 4770 video card.


    ... oh well, if you want to believe every ludicrous lie told by ChatGPT,
    go right ahead without me.


    Was it so difficult to issue a news about this new LLVM target
    explaining who and why needs it?


    We could issue 700 news articles a day about every possible commit in
    all of Gentoo, but *why*? It's a USE flag. The default change is more convenient for various reasons, and has little value in removing. It
    costs nothing other than time to recompile.

    Since when does that merit "STOP AND READ BEFORE UPDATING"? Feel free to
    skip it, for now.


    --
    Eli Schwartz

    --------------dvoItMoe0GwHEB4L0EqxgYHe--

    -----BEGIN PGP SIGNATURE-----

    wnsEABYIACMWIQTnFNnmK0TPZHnXm3qEp9ErcA0vVwUCaHPq7AUDAAAAAAAKCRCEp9ErcA0vV0jg AQDOp4CtFQIMHFREFMXxlEPXWNszFxEtJcrTBbrF9YP80QD/cnme3wkaWFrjzP1LcFpWFI95gESX XNqbJc2X/MRnSA0=
    =6IEQ
    -----END PGP SIGNATURE-----

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Eli Schwartz@21:1/5 to gevisz on Thu Jul 17 22:30:01 2025
    This is an OpenPGP/MIME signed message (RFC 4880 and 3156) --------------050YRaKOgqOEij5MjCwYmxTT
    Content-Type: text/plain; charset=UTF-8
    Content-Transfer-Encoding: quoted-printable

    On 7/17/25 3:54 PM, gevisz wrote:

    We could issue 700 news articles a day about every possible commit in
    all of Gentoo, but *why*? It's a USE flag.

    It is not a usual USE flag. It is a USE flag that does not appear in
    $ equery uses llvm-core/llvm
    * Found these USE flags for llvm-core/llvm-20.1.7:
    U I
    - - abi_x86_32 : 32-bit (x86) libraries
    + + binutils-plugin : Build the binutils plugin
    - - debug : Enable extra debug codepaths, like asserts
    and extra output. If you want to get meaningful backtraces see https://wiki.gentoo.org/wiki/Project:Quality_Assurance/Backtraces
    - - doc : Build and install the HTML documentation
    and regenerate the man pages
    - - exegesis : Enable performance counter support for llvm-exegesis tool that can be used to measure host machine
    instruction characteristics
    - - libedit : Use the libedit library (replacement for readline)
    + + libffi : Enable support for Foreign Function
    Interface library
    - - llvm_targets_ARC : ARC (Argonaut RISC Core) embedded CPU
    target [EXPERIMENTAL]
    - - llvm_targets_CSKY : C-SKY CPU target [EXPERIMENTAL]
    - - llvm_targets_DirectX : DirectX target [EXPERIMENTAL]
    - - llvm_targets_M68k : Motorola 68000 target [EXPERIMENTAL]
    - - llvm_targets_Xtensa : Tensilica Xtensa (ESP32) CPU target [EXPERIMENTAL]
    - - test : Enable dependencies and/or preparations
    necessary to run tests (usually controlled by FEATURES=test but can be toggled independently)
    - - verify-sig : Verify upstream signatures on distfiles
    + + xml : Add support for XML files
    - - z3 : Enable support for sci-mathematics/z3
    constraint solver
    + + zstd : Enable support for ZSTD compression

    I think that making a news about USE flags
    that is not reported by equery uses command
    make no more than one news per year. :)


    I see no logical connection between the two. equery uses --force-masked
    shows it, and many more.

    I promise you, far far far more than once per year does a
    package.use.force get added to some package in the tree.



    The default change is more
    convenient for various reasons, and has little value in removing. It
    costs nothing other than time to recompile.

    I have a 20-year-old computer with a Phenom processor.
    So, time to recompile such a heavy package as LLVM does matter for me.


    I empathize greatly.

    As a former user of a binary distro I would not be a Gentoo user at all,
    let alone a Developer, if Gentoo didn't have binary packages. It was
    *the* reason I switched, and quickly became a Developer.

    For a computer that old, I think you would benefit a lot from optional
    binary support:

    https://www.gentoo.org/news/2023/12/29/Gentoo-binary.html

    It supports USE flags perfectly -- only installing binaries that match
    your USE, automatically building from source otherwise.

    You can do it per package, even:

    emerge --sync
    emerge -puDU @world

    # packages which are a big burden to compile, listed in -puDU @world
    emerge -1auU --getbinpkg llvm-core/llvm gcc <...>

    emerge -auDU @world



    --
    Eli Schwartz

    --------------050YRaKOgqOEij5MjCwYmxTT--

    -----BEGIN PGP SIGNATURE-----

    wnsEABYIACMWIQTnFNnmK0TPZHnXm3qEp9ErcA0vVwUCaHla8gUDAAAAAAAKCRCEp9ErcA0vVyXi AP4uYZfGTXKuTkt3YBF6H6ynF4sktDaBfVggYRPxrARAoQD/YwYB3AlMAWi7GSs5/KqffCML8fBy SLC9b5lse1xL0gA=
    =d2Qi
    -----END PGP SIGNATURE-----

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)