mirrors/guru
1
0 Fork 0
mirror of https://github.com/gentoo-mirror/guru.git synced 2026-10-31 05:40:31 -04:00
Code Issues Packages Projects Releases Wiki Activity
38,506 Commits 1 Branch 0 Tags
d7d229efb7986e0bc94e052d4277e7796bd99ebe
Commit Graph
45 Commits
Author SHA1 Message Date
Gilles Dartiguelongue e55fc4cf6e sci-misc/llama-cpp: media-libs/shaderc is only needed for ggml-vulkan 
Signed-off-by: Gilles Dartiguelongue <gilles.dartiguelongue@esiee.org>
 2026-09-18 19:20:18 +02:00
Gilles Dartiguelongue ad11a53782 sci-misc/llama-cpp: match upstream ROCm wmma support removal 
Refs: https://github.com/ggml-org/llama.cpp/pull/26046
Signed-off-by: Gilles Dartiguelongue <gilles.dartiguelongue@esiee.org>
 2026-09-18 19:20:18 +02:00
Gilles Dartiguelongue dcde740596 sci-misc/llama-cpp: add 0.4.1 
Signed-off-by: Gilles Dartiguelongue <gilles.dartiguelongue@esiee.org>
 2026-09-18 19:20:17 +02:00
Gilles Dartiguelongue 15942baf01 sci-misc/llama-cpp: allow building tests 
Signed-off-by: Gilles Dartiguelongue <gilles.dartiguelongue@esiee.org>
 2026-09-18 19:20:17 +02:00
Gilles Dartiguelongue 8c3f71e8a2 sci-misc/llama-cpp: drop unnecessary src_install override 
ggml-rpc-server is installed via tools.

Signed-off-by: Gilles Dartiguelongue <gilles.dartiguelongue@esiee.org>
 2026-09-18 19:20:17 +02:00
Gilles Dartiguelongue a1d271a41d sci-misc/llama-cpp: set ROCm to minimum required version 
Have not seen a reason to set a newer version so follow upstream written requirement.

Signed-off-by: Gilles Dartiguelongue <gilles.dartiguelongue@esiee.org>
 2026-09-18 19:20:17 +02:00
Gilles Dartiguelongue 27feb9b491 sci-misc/llama-cpp: fix hipcc usage deprecation warning 
Signed-off-by: Gilles Dartiguelongue <gilles.dartiguelongue@esiee.org>
 2026-09-18 19:20:17 +02:00
Gilles Dartiguelongue 726b1b8c45 sci-misc/llama-cpp: Cosmetic changes for GGML handling 
Add some explicit comments and separations to GGML related variables
to ease synchronization with other projects embedding GGML
such as whisper.cpp, stable-diffusion.cpp, etc.

Signed-off-by: Gilles Dartiguelongue <gilles.dartiguelongue@esiee.org>
 2026-09-18 19:20:17 +02:00
Gilles Dartiguelongue d889498f57 sci-misc/llama-cpp: improve ggml collision handling 
Instruct CMake to install llama.cpp libraries in a libdir subfolder and
add RPATH allowing it to load its libraries and ggml copy.
Do not drop includes as llama can be a dependency of whisper.cpp or
other packages.

Signed-off-by: Gilles Dartiguelongue <gilles.dartiguelongue@esiee.org>
 2026-09-18 19:20:17 +02:00
Gilles Dartiguelongue e78d5a8e4d sci-misc/llama-cpp: sort IUSE 
Complaint from pkgcruft.

Signed-off-by: Gilles Dartiguelongue <gilles.dartiguelongue@esiee.org>
 2026-09-18 19:20:16 +02:00
Gilles Dartiguelongue d7af112aa2 sci-misc/llama-cpp: add 0.4.0_p10948 
Switch from legacy -DAMDGPU_TARGETS to -DGPU_TARGETS.

Signed-off-by: Gilles Dartiguelongue <gilles.dartiguelongue@esiee.org>
 2026-09-18 19:20:13 +02:00
Paul Zander 8b951481b6 sci-misc/llama-cpp: do not use ROCM_USEDEP 
ROCM_USEDEP should not be used for flag propagation anymore, since every
other core ROCm package requires AMDGPU_TARGETS to be set globally
(otherwise dependency solver tree explodes in size).

Fixes: 4b0e0ad9e6
See-also: 61d7e4718f
Signed-off-by: Paul Zander <negril.nx+gentoo@gmail.com>
 2026-09-15 16:41:36 +02:00
Gilles Dartiguelongue ae36817fe3 sci-misc/llama-cpp: drop unnecessary CMake variable 
Was used with the blas patch in commit d233181e96.

Closes: https://bugs.gentoo.org/975598
Signed-off-by: Gilles Dartiguelongue <gilles.dartiguelongue@esiee.org>
 2026-09-13 23:10:53 +02:00
Gilles Dartiguelongue c9f900361a sci-misc/llama-cpp: download and install web UI assets 
Restore web UI feature until upstream completes rework on its
build artifacts so they embed the assets.

For release ebuilds, read the nightly-tag.txt attached to the tag on
GitHub in order to set the relevant LLAMA_UI_VERSION. For live ebuilds,
pick the latest "nightly" tag when updating. This is not optimal but
spares nodejs[npm] dependency and build time as the UI itself does not
change much between releases.

Bug: https://github.com/ggml-org/llama.cpp/issues/27316
Closes: https://bugs.gentoo.org/979245
Refs: https://github.com/ggml-org/llama.cpp/blob/master/tools/ui/README.md
Signed-off-by: Gilles Dartiguelongue <gilles.dartiguelongue@esiee.org>
 2026-09-10 17:42:38 +02:00
Gilles Dartiguelongue cbf3610dec sci-misc/llama-cpp: fix insecure runpath QA error 
* QA Notice: The following files contain insecure RUNPATHs
 *  Please file a bug about this at https://bugs.gentoo.org/
 *  with the maintainer of the package.
 *   /var/tmp/portage/sci-misc/llama-cpp-0.3.0/image/usr/bin/ggml-rpc-servern    RPATH: ::::::::::::::::::::
 *

Closes: https://bugs.gentoo.org/975597
Signed-off-by: Gilles Dartiguelongue <gilles.dartiguelongue@esiee.org>
 2026-09-10 17:42:38 +02:00
Gilles Dartiguelongue e8be7b944e sci-misc/llama-cpp: import ggml management from ::gentoo 
Import CPU flag management from ggml ebuild in ::gentoo.
Main difference is avx516bw is not added as it is an optional block of
avx512l and only avx512l appears to be used by GGML.
It avoids automagically enabling when only of the option is enabled.

Signed-off-by: Gilles Dartiguelongue <gilles.dartiguelongue@esiee.org>
 2026-09-10 17:42:37 +02:00
Gilles Dartiguelongue 7d7f03f375 sci-misc/llama-cpp: sync live ebuild 
Various QA improvements: Use || die in remove instruction. Switch GGML_NATIVE to OFF to better match cmake documentation. Drop obsolete LLAMA_BUILD_WEBUI option. Add missing ROCM_USEDEP.

Signed-off-by: Gilles Dartiguelongue <gilles.dartiguelongue@esiee.org>
 2026-09-10 17:42:37 +02:00
Gilles Dartiguelongue 01ba8f9f2f sci-misc/llama-cpp: better manage dev/release build 
Allow managing both releases and pre-releases.
- Follow upstream instructions on setting LLAMA_BUILD_IS_DEV.
- LLAMA_BUILD_COMMIT is only set in live ebuild to avoid maintenance
  headaches with releases.
- LLAMA_BUILD_NUMBER is set to tag value or 9999 for live ebuild.

Refs: https://github.com/ggml-org/llama.cpp/blob/master/docs/release.md#building-a-release
Signed-off-by: Gilles Dartiguelongue <gilles.dartiguelongue@esiee.org>
 2026-09-10 17:42:37 +02:00
Craig Andrews ad27c53326 sci-misc/llama-cpp: rcp-server was renamed to ggml-rpc-server 
See:
https://github.com/ggml-org/llama.cpp/commit/c299a92c38b6de6a1139617652b66081828648db

Signed-off-by: Craig Andrews <candrews@gentoo.org>
 2026-06-29 13:10:12 -04:00
Paul Zander 46865dde7c Revert "sci-misc/llama-cpp: introduce webui USE flag" 
This reverts commit 27205a89af.

The checksums are unstable. This simply doesn't work in this manner.

Signed-off-by: Paul Zander <negril.nx+gentoo@gmail.com>
 2026-05-22 21:01:26 +02:00
Paul Zander 4b4650d059 Revert "sci-misc/llama-cpp: introduce lto USE flag" 
This reverts commit 91f8575242.

We do not want `USE=lto` .
https://wiki.gentoo.org/wiki/LTO#Exceptions_.28USE.3D.22lto.22.29

There is no need to have a `USE=lto` here. All it does is enable
`CMAKE_INTERPROCEDURAL_OPTIMIZATION`, which only passes a potentially
bad version of `-flto`.

Signed-off-by: Paul Zander <negril.nx+gentoo@gmail.com>
 2026-05-22 20:48:14 +02:00
Michael Egger 91f8575242 sci-misc/llama-cpp: introduce lto USE flag 
Signed-off-by: Michael Egger <egger.m@protonmail.com>
 2026-05-22 20:29:20 +02:00
Michael Egger 27205a89af sci-misc/llama-cpp: introduce webui USE flag 
Signed-off-by: Michael Egger <egger.m@protonmail.com>
 2026-05-22 20:29:20 +02:00
Craig Andrews 427e19afde sci-misc/llama-cpp: disable building the webui 
Signed-off-by: Craig Andrews <candrews@gentoo.org>
 2026-05-14 14:08:34 -04:00
Craig Andrews 6a097c56d1 sci-misc/llama-cpp: depend on dev-util/spirv-headers 
See: https://github.com/ggml-org/llama.cpp/pull/21572

Signed-off-by: Craig Andrews <candrews@gentoo.org>
 2026-04-14 10:54:38 -04:00
Craig Andrews 98a0200bd2 sci-misc/llama-cpp: add openssl use flag 
Addresses automagic dependency on dev-libs/openssl

Signed-off-by: Craig Andrews <candrews@gentoo.org>
 2026-03-30 15:18:10 -04:00
Rok Faith 71642b949e sci-misc/llama-cpp: update 9999: add USE="examples wmma" 
Signed-off-by: Rok Faith <rok@faith.si>
Part-of: https://github.com/gentoo/guru/pull/420
Closes: https://github.com/gentoo/guru/pull/420
Signed-off-by: Paul Zander <negril.nx+gentoo@gmail.com>
 2026-01-26 15:10:11 +01:00
Rok Faith 04ca4460ca sci-misc/llama-cpp: update 9999: change USE hip -> rocm 
Signed-off-by: Rok Faith <rok@faith.si>
Part-of: https://github.com/gentoo/guru/pull/420
Signed-off-by: Paul Zander <negril.nx+gentoo@gmail.com>
 2026-01-26 15:10:11 +01:00
Sergey Alirzaev b2cf10b408 sci-misc/llama-cpp: + flexiblas backend 
Signed-off-by: Sergey Alirzaev <l29ah@riseup.net>
 2025-11-10 03:24:29 +01:00
Sergey Alirzaev 0d5b89bb35 sci-misc/llama-cpp: fixup bdep 
Signed-off-by: Sergey Alirzaev <l29ah@riseup.net>
 2025-11-08 03:40:54 +01:00
Sergey Alirzaev 3733474f9c sci-misc/llama-cpp: handle automagic openmp dep 
enabled by default as otherwise results in severe performance hit
see https://github.com/ggml-org/llama.cpp/issues/12171

Signed-off-by: Sergey Alirzaev <l29ah@riseup.net>
 2025-08-10 02:35:22 +02:00
Sergey Alirzaev bf6fdb43eb sci-misc/llama-cpp: + opencl backend, bump 
Signed-off-by: Sergey Alirzaev <l29ah@riseup.net>
 2025-07-28 00:59:07 +02:00
Paul Zander 451623c534 sci-misc/llama-cpp: set CUDAHOSTCXX via cuda.eclass 
Closes: https://bugs.gentoo.org/960549
Signed-off-by: Paul Zander <negril.nx+gentoo@gmail.com>
 2025-07-22 20:25:30 +02:00
Sergey Alirzaev bd53f5db4f sci-misc/llama-cpp: fix hip/rocm, remove old 
Signed-off-by: Sergey Alirzaev <l29ah@riseup.net>
 2025-07-10 02:16:05 +02:00
Sergey Alirzaev 1ac2399f36 sci-misc/llama-cpp: sync live 
Signed-off-by: Sergey Alirzaev <l29ah@riseup.net>
 2025-07-04 04:17:10 +02:00
Sergey Alirzaev 36a39a0a80 sci-misc/llama-cpp: avoid clashing with whisper.cpp 
Signed-off-by: Sergey Alirzaev <l29ah@riseup.net>
 2025-07-03 21:32:59 +02:00
Alexey Korepanov d233181e96 sci-misc/llama-cpp: support cuda, update upstream url 
Signed-off-by: Alexey Korepanov <kaikaikai@yandex.ru>
 2025-06-16 21:01:09 +08:00
Sergey Alirzaev 1b3ace659d sci-misc/llama-cpp: enable rpc 
Signed-off-by: Sergey Alirzaev <l29ah@riseup.net>
 2025-05-17 14:57:34 +02:00
Sergey Alirzaev e01288b799 sci-misc/llama-cpp: Add ROCm 6.3 HIP support 
Closes: https://bugs.gentoo.org/949856
Signed-off-by: Sergey Alirzaev <l29ah@riseup.net>
 2025-02-20 15:57:54 +01:00
Alexey Korepanov a48cf7afbb sci-misc/llama-cpp: support OpenBLAS or BLIS backends 
A significant speed up in some workflows when running on a CPU

Signed-off-by: Alexey Korepanov <kaikaikai@yandex.ru>
 2025-02-09 11:55:04 +00:00
Alexey Korepanov fb105c916a sci-misc/llama-cpp: add dependencies on curl and numpy 
Signed-off-by: Alexey Korepanov <kaikaikai@yandex.ru>
 2025-02-08 16:49:15 +00:00
Sergey Alirzaev b228272c6c sci-misc/llama-cpp: strip march=native 
Closes: https://bugs.gentoo.org/949130
Signed-off-by: Sergey Alirzaev <l29ah@riseup.net>
 2025-02-01 12:19:51 +01:00
Sergey Alirzaev ab2a79180f sci-misc/llama-cpp: new package, add 0_pre4576, 9999 
Signed-off-by: Sergey Alirzaev <l29ah@riseup.net>
 2025-01-29 01:11:11 +01:00
Takuya Wakazono db9e23974f Revert "sci-misc/llama-cpp: new package, add 0_pre4564, 9999" 
This reverts commit e933540939.

This ebuild has too many issues.
- missing manifest
- use of deprecated eclass
- broken dependency
- undescribed USE flags
- visible keyword for vcs version

Please get advice on a GitHub PR or IRC before committing

Signed-off-by: Takuya Wakazono <pastalian46@gmail.com>
 2025-01-28 00:10:33 +09:00
Sergey Alirzaev e933540939 sci-misc/llama-cpp: new package, add 0_pre4564, 9999 
Signed-off-by: Sergey Alirzaev <l29ah@riseup.net>
 2025-01-27 12:52:32 +01:00