From e51dcfd426d0bb475a3d12eed8d54a46b0f17444 Mon Sep 17 00:00:00 2001 From: danix Date: Tue, 31 Mar 2026 09:34:59 +0200 Subject: Added some of the packages I maintain on my personal system --- llama.cpp-vulkan/README | 22 +++++ llama.cpp-vulkan/doinst.sh | 27 ++++++ llama.cpp-vulkan/llama.cpp-vulkan.SlackBuild | 131 +++++++++++++++++++++++++++ llama.cpp-vulkan/llama.cpp-vulkan.info | 10 ++ llama.cpp-vulkan/rc.llama.cpp | 86 ++++++++++++++++++ llama.cpp-vulkan/slack-desc | 19 ++++ 6 files changed, 295 insertions(+) create mode 100644 llama.cpp-vulkan/README create mode 100644 llama.cpp-vulkan/doinst.sh create mode 100644 llama.cpp-vulkan/llama.cpp-vulkan.SlackBuild create mode 100644 llama.cpp-vulkan/llama.cpp-vulkan.info create mode 100644 llama.cpp-vulkan/rc.llama.cpp create mode 100644 llama.cpp-vulkan/slack-desc (limited to 'llama.cpp-vulkan') diff --git a/llama.cpp-vulkan/README b/llama.cpp-vulkan/README new file mode 100644 index 0000000..5509d44 --- /dev/null +++ b/llama.cpp-vulkan/README @@ -0,0 +1,22 @@ +llama.cpp + +LLM inference in C/C++ + +The main goal of llama.cpp is to enable LLM inference with minimal +setup and state-of-the-art performance on a wide range of hardware +locally and in the cloud. + + - Plain C/C++ implementation without any dependencies + - Apple silicon is a first-class citizen - optimized via ARM NEON, + Accelerate and Metal frameworks + - AVX, AVX2, AVX512 and AMX support for x86 architectures + - RVV, ZVFH, ZFH, ZICBOP and ZIHINTPAUSE support for RISC-V + architectures + - 1.5-bit, 2-bit, 3-bit, 4-bit, 5-bit, 6-bit, and 8-bit integer + quantization for faster inference and reduced memory use + - Custom CUDA kernels for running LLMs on NVIDIA GPUs (support for + AMD GPUs via HIP and Moore Threads GPUs via MUSA) + - Vulkan and SYCL backend support + - CPU+GPU hybrid inference to partially accelerate models larger than + the total VRAM capacity + diff --git a/llama.cpp-vulkan/doinst.sh b/llama.cpp-vulkan/doinst.sh new file mode 100644 index 0000000..0b91476 --- /dev/null +++ b/llama.cpp-vulkan/doinst.sh @@ -0,0 +1,27 @@ +config() { + NEW="$1" + OLD="$(dirname $NEW)/$(basename $NEW .new)" + # If there's no config file by that name, mv it over: + if [ ! -r $OLD ]; then + mv $NEW $OLD + elif [ "$(cat $OLD | md5sum)" = "$(cat $NEW | md5sum)" ]; then + # toss the redundant copy + rm $NEW + fi + # Otherwise, we leave the .new copy for the admin to consider... +} + +preserve_perms() { + NEW="$1" + OLD="$(dirname $NEW)/$(basename $NEW .new)" + if [ -e $OLD ]; then + cp -a $OLD ${NEW}.incoming + cat $NEW > ${NEW}.incoming + mv ${NEW}.incoming $NEW + fi + config $NEW +} + +preserve_perms etc/rc.d/rc.llama.cpp.new +config etc/rc.d/rc.llama.cpp.new + diff --git a/llama.cpp-vulkan/llama.cpp-vulkan.SlackBuild b/llama.cpp-vulkan/llama.cpp-vulkan.SlackBuild new file mode 100644 index 0000000..b78c515 --- /dev/null +++ b/llama.cpp-vulkan/llama.cpp-vulkan.SlackBuild @@ -0,0 +1,131 @@ +#!/bin/bash + +# Slackware build script for llama.cpp-vulkan + +# Copyright 2026 danix +# All rights reserved. +# +# Redistribution and use of this script, with or without modification, is +# permitted provided that the following conditions are met: +# +# 1. Redistributions of this script must retain the above copyright +# notice, this list of conditions and the following disclaimer. +# +# THIS SOFTWARE IS PROVIDED BY THE AUTHOR "AS IS" AND ANY EXPRESS OR IMPLIED +# WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF +# MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO +# EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, +# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; +# OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, +# WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR +# OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF +# ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +cd $(dirname $0) ; CWD=$(pwd) + +PRGNAM=llama.cpp-vulkan +SRCNAM=llama.cpp +VERSION=${VERSION:-b8581} +BUILD=${BUILD:-2} +TAG=${TAG:-_SBo} +PKGTYPE=${PKGTYPE:-tgz} + +if [ -z "$ARCH" ]; then + case "$( uname -m )" in + i?86) ARCH=i586 ;; + arm*) ARCH=arm ;; + *) ARCH=$( uname -m ) ;; + esac +fi + +if [ ! -z "${PRINT_PACKAGE_NAME}" ]; then + echo "$PRGNAM-$VERSION-$ARCH-$BUILD$TAG.$PKGTYPE" + exit 0 +fi + +TMP=${TMP:-/tmp/SBo} +PKG=$TMP/package-$PRGNAM +OUTPUT=${OUTPUT:-/tmp} + +if [ "$ARCH" = "i586" ]; then + SLKCFLAGS="-O2 -march=i586 -mtune=i686" + LIBDIRSUFFIX="" +elif [ "$ARCH" = "i686" ]; then + SLKCFLAGS="-O2 -march=i686 -mtune=i686" + LIBDIRSUFFIX="" +elif [ "$ARCH" = "x86_64" ]; then + SLKCFLAGS="-O2 -fPIC" + LIBDIRSUFFIX="64" +else + SLKCFLAGS="-O2" + LIBDIRSUFFIX="" +fi + +set -ex + +rm -rf $PKG +mkdir -p $TMP $PKG $OUTPUT +cd $TMP +rm -rf $SRCNAM-$VERSION +tar xvf $CWD/$SRCNAM-$VERSION.tar.gz +cd $SRCNAM-$VERSION +chown -R root:root . +find -L . \ + \( -perm 777 -o -perm 775 -o -perm 750 -o -perm 711 -o -perm 555 \ + -o -perm 511 \) -exec chmod 755 {} \; -o \ + \( -perm 666 -o -perm 664 -o -perm 640 -o -perm 600 -o -perm 444 \ + -o -perm 440 -o -perm 400 \) -exec chmod 644 {} \; + +mkdir -p build +cd build + cmake .. \ + -DCMAKE_C_FLAGS:STRING="$SLKCFLAGS" \ + -DCMAKE_CXX_FLAGS:STRING="$SLKCFLAGS" \ + -DCMAKE_INSTALL_PREFIX=/usr \ + -DLIB_SUFFIX=${LIBDIRSUFFIX} \ + -DMAN_INSTALL_DIR=/usr/man \ + -B build \ + -DCMAKE_BUILD_TYPE=Release \ + -DBUILD_SHARED_LIBS=ON \ + -DLLAMA_BUILD_TESTS=OFF \ + -DLLAMA_USE_SYSTEM_GGML=OFF \ + -DGGML_NATIVE=ON \ + -DGGML_ALL_WARNINGS=OFF \ + -DGGML_ALL_WARNINGS_3RD_PARTY=OFF \ + -DGGML_BUILD_EXAMPLES=OFF \ + -DGGML_BUILD_TESTS=OFF \ + -DGGML_LTO=ON \ + -DGGML_RPC=ON \ + -DGGML_VULKAN=ON \ + -DGGML_CUDA_FA_ALL_QUANTS=ON \ + -DLLAMA_BUILD_NUMBER="${VERSION#b}" \ + -Wno-dev + + cmake --build build -- + DESTDIR=$PKG cmake --install build +cd .. + +rm -f $PKG/{,usr/}lib${LIBDIRSUFFIX}/*.la + +find $PKG -print0 | xargs -0 file | grep -e "executable" -e "shared object" | grep ELF \ + | cut -f 1 -d : | xargs strip --strip-unneeded 2> /dev/null || true + +find $PKG/usr/man -type f -exec gzip -9 {} \; || true +for i in $( find $PKG/usr/man -type l ) ; do ln -s $( readlink $i ).gz $i.gz ; rm $i ; done || true + +mkdir -p $PKG/usr/doc/$PRGNAM-$VERSION +cp -a \ + AUTHORS CODEOWNERS CONTRIBUTING.md LICENSE README.md SECURITY.md media \ + $PKG/usr/doc/$PRGNAM-$VERSION +cat $CWD/$PRGNAM.SlackBuild > $PKG/usr/doc/$PRGNAM-$VERSION/$PRGNAM.SlackBuild + +mkdir -p $PKG/etc/rc.d +cat $CWD/rc.llama.cpp > $PKG/etc/rc.d/rc.llama.cpp.new + +mkdir -p $PKG/install +cat $CWD/slack-desc > $PKG/install/slack-desc +cat $CWD/doinst.sh > $PKG/install/doinst.sh + +cd $PKG +/sbin/makepkg -l y -c n $OUTPUT/$PRGNAM-$VERSION-$ARCH-$BUILD$TAG.$PKGTYPE diff --git a/llama.cpp-vulkan/llama.cpp-vulkan.info b/llama.cpp-vulkan/llama.cpp-vulkan.info new file mode 100644 index 0000000..1a5b023 --- /dev/null +++ b/llama.cpp-vulkan/llama.cpp-vulkan.info @@ -0,0 +1,10 @@ +PRGNAM="llama.cpp-vulkan" +VERSION="b8581" +HOMEPAGE="https://github.com/ggml-org/llama.cpp" +DOWNLOAD="https://github.com/ggml-org/llama.cpp/archive/b8581/llama.cpp-b8581.tar.gz" +MD5SUM="78bd502eca26be48813f50716e9b86fe" +DOWNLOAD_x86_64="" +MD5SUM_x86_64="" +REQUIRES="" +MAINTAINER="danix" +EMAIL="danix@danix.xyz" diff --git a/llama.cpp-vulkan/rc.llama.cpp b/llama.cpp-vulkan/rc.llama.cpp new file mode 100644 index 0000000..fab09d6 --- /dev/null +++ b/llama.cpp-vulkan/rc.llama.cpp @@ -0,0 +1,86 @@ +#!/bin/sh + +# Copyright 2026 danix +# All rights reserved. +# +# Redistribution and use of this script, with or without modification, is +# permitted provided that the following conditions are met: +# +# 1. Redistributions of this script must retain the above copyright +# notice, this list of conditions and the following disclaimer. +# +# THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED +# WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF +# MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO +# EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, +# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; +# OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, +# WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR +# OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF +# ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +# SERVER OPTIONS +# many options can be set as env variables. +# See https://github.com/ggml-org/llama.cpp/tree/master/tools/server + +RUNAS_USER= +RUNAS_GROUP=users +LLSRV=/usr/bin/llama-server +#MODEL_DIR= +MODEL= # which model to use +HOST=127.0.0.1 # ip to bind the server to +PORT=8181 # port for the server to listen to +CONTEXT= #size of context +TEMP=1.0 +TOP_K=20 +TOP_P=0.95 +MIN_P=0.00 +SLEEP_IDLE_SECONDS=500 # after how many seconds to unload the model from memory +REASONING_FORMAT=deepseek +GPU_LAYERS=all +SPLIT_MODE=none +NO_CONTEXT_SHIFT='--no-context-shift' +LOG_FILE=/var/log/llama-server/server.log +LLAMA_ARGS="--ctx-size $CONTEXT \ + --temp $TEMP \ + --top-k $TOP_K \ + --top-p $TOP_P \ + --min-p $MIN_P \ + --host $HOST --port $PORT \ + --reasoning-format $REASONING_FORMAT \ + -sm $SPLIT_MODE \ + $NO_CONTEXT_SHIFT \ + -ngl $GPU_LAYERS \ + --sleep-idle-seconds $SLEEP_IDLE_SECONDS \ + -m $MODEL" # alternatively one could set the model directory and pass --models-dir to the script + +case "$1" in + stop) + if /usr/bin/pgrep -f "$LLSRV" >/dev/null; then + echo "Stopping llama-server..." + killall llama-server 2>/dev/null + else + echo "llama-server is not running..." + exit 1 + fi + ;; + start) + echo "Starting llama-server..." + if [[ -d $(dirname $LOG_FILE) ]]; then + mkdir -p $(dirname $LOG_FILE) + touch $LOG_FILE + chown -R ${RUNAS_USER}:${RUNAS_GROUP} $(dirname $LOG_FILE) + fi + su $RUNAS_USER -c "$LLSRV $LLAMA_ARGS --log-file $LOG_FILE -lv 3 --log-timestamps" 1>/dev/null 2>&1 & + ;; + restart) + $0 stop + sleep 1 + $0 start + ;; + *) + echo "usage: $0 { start | stop | restart }" >&2 + exit 1 + ;; +esac diff --git a/llama.cpp-vulkan/slack-desc b/llama.cpp-vulkan/slack-desc new file mode 100644 index 0000000..273e15e --- /dev/null +++ b/llama.cpp-vulkan/slack-desc @@ -0,0 +1,19 @@ +# HOW TO EDIT THIS FILE: +# The "handy ruler" below makes it easier to edit a package description. +# Line up the first '|' above the ':' following the base package name, and +# the '|' on the right side marks the last column you can put a character in. +# You must make exactly 11 lines for the formatting to be correct. It's also +# customary to leave one space after the ':' except on otherwise blank lines. + + |-----handy-ruler------------------------------------------------------| +llama.cpp-vulkan: llama.cpp-vulkan (LLM inference in C/C++) +llama.cpp-vulkan: +llama.cpp-vulkan: Port of Facebook's LLaMA model in C/C++ with Vulkan GPU optimizations +llama.cpp-vulkan: +llama.cpp-vulkan: The main goal of llama.cpp is to enable LLM inference with minimal +llama.cpp-vulkan: setup and state-of-the-art performance on a wide range of hardware +llama.cpp-vulkan: locally and in the cloud. +llama.cpp-vulkan: +llama.cpp-vulkan: Home: https://github.com/ggml-org/llama.cpp +llama.cpp-vulkan: +llama.cpp-vulkan: -- cgit v1.2.3