From a7444d2ed89d8271fa182f67d25c8e414b7a25a7 Mon Sep 17 00:00:00 2001
From: xlliu-scitix <xlliu@scitix.ai>
Date: Sun, 21 Dec 2025 03:12:56 +0000
Subject: [PATCH 1/3] chore(scripts): add install scripts

---
 scripts/env.sh          |  6 ++++
 scripts/install_sihpc   | 66 +++++++++++++++++++++++++++++++++++++++++
 scripts/uninstall_sihpc | 25 ++++++++++++++++
 3 files changed, 97 insertions(+)
 create mode 100644 scripts/env.sh
 create mode 100644 scripts/install_sihpc
 create mode 100644 scripts/uninstall_sihpc

diff --git a/scripts/env.sh b/scripts/env.sh
new file mode 100644
index 0000000..5e2610b
--- /dev/null
+++ b/scripts/env.sh
@@ -0,0 +1,6 @@
+#!/bin/bash
+export SIHPC_HOME=/usr/local/sihpc
+export PATH=$SIHPC_HOME/bin:$PATH
+export LD_LIBRARY_PATH=$SIHPC_HOME/lib:/usr/local/cuda/lib64:$LD_LIBRARY_PATH
+export OMPI_MCA_opal_prefix=$SIHPC_HOME
+export OPAL_PREFIX=$SIHPC_HOME
\ No newline at end of file
diff --git a/scripts/install_sihpc b/scripts/install_sihpc
new file mode 100644
index 0000000..01e370a
--- /dev/null
+++ b/scripts/install_sihpc
@@ -0,0 +1,66 @@
+#!/bin/bash
+set -euo pipefail
+
+PREFIX="/usr/local/sihpc"
+LIBCONF="/etc/ld.so.conf.d/sihpc.conf"
+PROFILE_SH="/etc/profile.d/sihpc.sh"
+
+echo "Installing SiHPC runtime to: $PREFIX"
+
+if [ ! -d "$PREFIX" ]; then
+    mkdir -p "$PREFIX"
+fi
+cp -r ./* "$PREFIX/"
+echo "Files installed to $PREFIX"
+
+if [ ! -f "$LIBCONF" ]; then
+    echo "$PREFIX/lib" > "$LIBCONF"
+    [ -d "$PREFIX/lib64" ] && echo "$PREFIX/lib64" >> "$LIBCONF"
+    echo "Added $LIBCONF"
+else
+    if ! grep -q "$PREFIX/lib" "$LIBCONF"; then
+        echo "$PREFIX/lib" >> "$LIBCONF"
+    fi
+    if [ -d "$PREFIX/lib64" ] && ! grep -q "$PREFIX/lib64" "$LIBCONF"; then
+        echo "$PREFIX/lib64" >> "$LIBCONF"
+    fi
+    echo "Updated existing $LIBCONF"
+fi
+
+ldconfig
+echo "ldconfig updated"
+
+if [ ! -f "$PROFILE_SH" ]; then
+    cat <<EOF > "$PROFILE_SH"
+# Auto-generated by SiHPC installer
+if [ -f $PREFIX/env.sh ]; then
+    source $PREFIX/env.sh
+fi
+EOF
+    chmod +x "$PROFILE_SH"
+    echo "Added $PROFILE_SH"
+else
+    echo "$PROFILE_SH already exists, skipping."
+fi
+
+for f in /etc/bash.bashrc /etc/bashrc; do
+    if [ -f "$f" ]; then
+        if ! grep -q "$PREFIX/env.sh" "$f"; then
+            {
+                echo ""
+                echo "# SiHPC environment"
+                echo "if [ -f $PREFIX/env.sh ]; then"
+                echo "    source $PREFIX/env.sh"
+                echo "fi"
+            } >> "$f"
+            echo "Added SiHPC source to $f"
+        fi
+    fi
+done
+
+echo
+echo "SiHPC installation completed successfully!"
+echo "Installed to: $PREFIX"
+echo "Library config: $LIBCONF"
+echo "Auto env setup: $PROFILE_SH"
+echo "Run 'source $PREFIX/env.sh' now to activate current shell."
\ No newline at end of file
diff --git a/scripts/uninstall_sihpc b/scripts/uninstall_sihpc
new file mode 100644
index 0000000..20cabe2
--- /dev/null
+++ b/scripts/uninstall_sihpc
@@ -0,0 +1,25 @@
+#!/bin/bash
+# sihpc-uninstaller.sh
+
+set -e
+
+SIHPC_ROOT="/usr/local/sihpc"
+
+if [ ! -d "$SIHPC_ROOT" ]; then
+    echo "sihpc install dir $SIHPC_ROOT not exist."
+    exit 1
+fi
+
+echo "=============================="
+echo " uninstall sihpc"
+echo " install dir: $SIHPC_ROOT"
+echo "=============================="
+
+echo "deleting $SIHPC_ROOT ..."
+rm -rf "$SIHPC_ROOT"
+
+echo "please check shell config（~/.bashrc, ~/.zshrc etc.），"
+echo "remove $SIHPC_ROOT/bin from  PATH"
+
+echo "sihpc unintall done！"
+exit 0
\ No newline at end of file

From 5eec99dc086d31ff99e62c2a88ea52dbe05cbbd5 Mon Sep 17 00:00:00 2001
From: xlliu-scitix <xlliu@scitix.ai>
Date: Sun, 21 Dec 2025 15:12:18 +0800
Subject: [PATCH 2/3] add dockerfile and ci workflow (#1)

---
 .github/workflows/pre-check.yml        |  29 +++++
 .github/workflows/release.yml          |  37 ++++++
 docker/Dockerfile.cuda12.x.ubuntu20.04 | 153 +++++++++++++++++++++++++
 3 files changed, 219 insertions(+)
 create mode 100644 .github/workflows/pre-check.yml
 create mode 100644 .github/workflows/release.yml
 create mode 100644 docker/Dockerfile.cuda12.x.ubuntu20.04

diff --git a/.github/workflows/pre-check.yml b/.github/workflows/pre-check.yml
new file mode 100644
index 0000000..f293c61
--- /dev/null
+++ b/.github/workflows/pre-check.yml
@@ -0,0 +1,29 @@
+on:
+  pull_request:
+  workflow_dispatch:
+
+jobs:
+  build-only:
+    runs-on: ubuntu-22.04
+    steps:
+      - uses: actions/checkout@v4
+      - uses: docker/setup-buildx-action@v3
+      - name: Free disk space
+        run: |
+          sudo rm -rf /usr/share/dotnet
+          sudo rm -rf /usr/local/lib/android
+          sudo rm -rf /opt/ghc
+          sudo rm -rf /opt/hostedtoolcache
+          sudo docker system prune -af || true
+          df -h
+      - name: Build run package
+        run: |
+          docker buildx build \
+            -f docker/Dockerfile.cuda12.x.ubuntu20.04 \
+            --platform linux/amd64 \
+            --target package \
+            --output type=local,dest=dist \
+            .
+      - name: List artifacts
+        run: |
+          ls -lh dist
\ No newline at end of file
diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
new file mode 100644
index 0000000..86b42f8
--- /dev/null
+++ b/.github/workflows/release.yml
@@ -0,0 +1,37 @@
+name: Release run installer
+
+on:
+  push:
+    tags:
+      - "v*"
+
+permissions:
+  contents: write
+
+jobs:
+  build-release:
+    runs-on: ubuntu-22.04
+
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Setup Docker Buildx
+        uses: docker/setup-buildx-action@v3
+
+      - name: Build .run installer (Docker)
+        run: |
+          docker buildx build \
+            -f docker/Dockerfile.cuda12.x.ubuntu20.04 \
+            --platform linux/amd64 \
+            --build-arg BUILD_DATE=$(date +%Y%m%d) \
+            --output type=local,dest=dist \
+            .
+
+      - name: List artifacts
+        run: ls -lh dist
+
+      - name: Upload to GitHub Release
+        uses: softprops/action-gh-release@v2
+        with:
+          files: |
+            dist/*.run
diff --git a/docker/Dockerfile.cuda12.x.ubuntu20.04 b/docker/Dockerfile.cuda12.x.ubuntu20.04
new file mode 100644
index 0000000..efe1a62
--- /dev/null
+++ b/docker/Dockerfile.cuda12.x.ubuntu20.04
@@ -0,0 +1,153 @@
+###########################
+# Build-time configuration
+###########################
+
+# Base OS and CUDA versions
+ARG UBUNTU_VERSION=20.04
+ARG CUDA_VERSION=12.8.1
+ARG CUDART_VERSION=12.8.90
+ARG CUDART_MAJOR_VERSION=12
+
+# NCCL versions
+ARG NCCL_PACKAGE_VERSION=2.27.7-1+cuda12.4
+ARG NCCL_SO_VERSION=2.27.7
+
+# OpenMPI versions
+# - MPI_VERSION: full OpenMPI version
+# - MPI_SERIES: major.minor series used in download URL
+ARG MPI_VERSION=4.1.8
+ARG MPI_SERIES=4.1
+
+# Build date (override at build time)
+ARG BUILD_DATE=20251221
+
+###########################
+# Build Stage
+###########################
+FROM nvidia/cuda:${CUDA_VERSION}-devel-ubuntu20.04 AS build
+
+# Re-declare build args for this stage (values are inherited)
+ARG UBUNTU_VERSION
+ARG CUDA_VERSION
+ARG CUDART_VERSION
+ARG NCCL_PACKAGE_VERSION
+ARG NCCL_SO_VERSION
+ARG MPI_VERSION
+ARG MPI_SERIES
+ARG BUILD_DATE
+
+ENV DEBIAN_FRONTEND=noninteractive
+WORKDIR /workspace
+
+# -------------------------
+# 1. Base build dependencies
+# -------------------------
+RUN mv /etc/apt/sources.list.d/cuda*.list /tmp/disabled-cuda.list || true && \
+    apt-get -o Acquire::http::No-Cache=true update && \
+    apt-get install -y --no-install-recommends \
+        build-essential gcc g++ curl git wget ca-certificates \
+        make automake autoconf libtool pkg-config \
+        python3 python3-pip gzip xz-utils && \
+    rm -rf /var/lib/apt/lists/*
+
+# -------------------------
+# 2. Install CUDA keyring and restore NVIDIA repository
+# -------------------------
+RUN wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-keyring_1.1-1_all.deb && \
+    dpkg -i cuda-keyring_1.1-1_all.deb && \
+    apt-get update
+
+# -------------------------
+# 3. Install NCCL (pinned version)
+# -------------------------
+RUN apt-mark unhold libnccl2 libnccl-dev || true && \
+    apt-get install -y --no-install-recommends \
+    libnccl2=${NCCL_PACKAGE_VERSION} \
+    libnccl-dev=${NCCL_PACKAGE_VERSION} && \
+    apt-mark hold libnccl2 libnccl-dev && \
+    ldconfig && \
+    rm -rf /var/lib/apt/lists/*
+
+# -------------------------
+# 4. Build OpenMPI from source
+# -------------------------
+RUN wget https://download.open-mpi.org/release/open-mpi/v${MPI_SERIES}/openmpi-${MPI_VERSION}.tar.gz && \
+    tar zxvf openmpi-${MPI_VERSION}.tar.gz && \
+    cd openmpi-${MPI_VERSION} && \
+    ./configure --prefix=/usr/local/sihpc --with-cuda=/usr/local/cuda && \
+    make -j$(nproc) && make install && \
+    rm -rf /workspace/openmpi-${MPI_VERSION} /workspace/openmpi-${MPI_VERSION}.tar.gz
+
+# -------------------------
+# 5. Build nccl-tests
+# -------------------------
+RUN git clone https://github.com/scitix/nccl-tests.git -b sicl && \
+    cd nccl-tests && \
+    make MPI=1 MPI_HOME=/usr/local/sihpc && \
+    mkdir -p /usr/local/sihpc/libexec/nccl-tests && \
+    cp -rf build/*_perf /usr/local/sihpc/libexec/nccl-tests/ && \
+    cp scripts/nccl_perf /usr/local/sihpc/bin/nccl_perf && \
+    cp scripts/nccl_test /usr/local/sihpc/libexec/nccl-tests/nccl_test && \
+    cp scripts/env.sh /usr/local/sihpc/env.sh && \
+    cp scripts/install_sihpc /usr/local/sihpc/bin/install_sihpc && \
+    cp scripts/uninstall_sihpc /usr/local/sihpc/bin/uninstall_sihpc && \
+    rm -rf /workspace/nccl-tests
+
+# -------------------------
+# 6. Collect runtime libraries (strict selection)
+# -------------------------
+RUN set -eux && \
+    mkdir -p /usr/local/sihpc/lib && \
+    cp /usr/local/cuda/lib64/libcudart* /usr/local/sihpc/lib/ && \
+    cp /usr/lib/x86_64-linux-gnu/libnccl.so* /usr/local/sihpc/lib/
+    # cp /lib/x86_64-linux-gnu/libltdl.so.7.3.1 /usr/local/sihpc/lib/ && \
+    # cp /usr/lib/x86_64-linux-gnu/libhwloc.so* /usr/local/sihpc/lib/ && \
+    # cp /usr/lib/x86_64-linux-gnu/libevent_core* /usr/local/sihpc/lib/ && \
+    # cp /usr/lib/x86_64-linux-gnu/libevent_pthreads* /usr/local/sihpc/lib/
+
+# -------------------------
+# 7. Fix library symlinks
+# -------------------------
+RUN cd /usr/local/sihpc/lib && \
+    rm -f libcudart.so libcudart.so.12 && \
+    ln -sf libnccl.so.2.27.7 libnccl.so.2 && \
+    ln -sf libnccl.so.2 libnccl.so && \
+    ln -sf libcudart.so.12.8.90 libcudart.so.12 && \
+    ln -sf libcudart.so.12 libcudart.so
+    # rm -f libevent_core-2.1.so.7 && \
+    # ln -sf libhwloc.so.15.1.0 libhwloc.so.15 && \
+    # ln -sf libhwloc.so.15.1.0 libhwloc.so && \
+    # ln -sf libevent_core-2.1.so.7.0.0 libevent_core-2.1.so.7 && \
+    # ln -sf libevent_core-2.1.so.7 libevent_core-2.1.so && \
+    # ln -sf libevent_pthreads-2.1.so.7.0.0 libevent_pthreads-2.1.so.7 && \
+    # ln -sf libevent_pthreads-2.1.so.7 libevent_pthreads-2.1.so && \
+    # ln -sf libltdl.so.7.3.1 libltdl.so.7 && \
+    # ln -sf libltdl.so.7 libltdl.so
+
+###########################
+# Package Stage
+###########################
+FROM ubuntu:20.04 AS package
+
+# Re-declare args for this stage (values are inherited)
+ARG UBUNTU_VERSION
+ARG NCCL_PACKAGE_VERSION
+ARG MPI_VERSION
+ARG BUILD_DATE
+
+# Expose versions/date as environment variables for runtime shell expansion
+ENV NCCL_PACKAGE_VERSION=${NCCL_PACKAGE_VERSION} \
+    MPI_VERSION=${MPI_VERSION} \
+    BUILD_DATE=${BUILD_DATE}
+
+COPY --from=build /usr/local/sihpc /usr/local/sihpc
+
+WORKDIR /
+RUN apt-get update && apt-get install -y --no-install-recommends makeself && \
+    chmod +x /usr/local/sihpc/bin/install_sihpc && \
+    SAFE_NCCL_PKG=$(printf '%s\n' "${NCCL_PACKAGE_VERSION}" | tr '+' '-') && \
+    PACKAGE_FILENAME="sicl-nccl${SAFE_NCCL_PKG}-ompi${MPI_VERSION}-ubuntu${UBUNTU_VERSION}-${BUILD_DATE}.run" && \
+    makeself --gzip /usr/local/sihpc \
+    "${PACKAGE_FILENAME}" \
+    "SiHPC MPI + NCCL + NCCL-tests Portable Installer" \
+    ./bin/install_sihpc

From 42c1025d7dcb58fc64eb11bd09f73530bdfd3891 Mon Sep 17 00:00:00 2001
From: xlliu-scitix <xlliu@scitix.ai>
Date: Sun, 21 Dec 2025 07:18:22 +0000
Subject: [PATCH 3/3] add free disk space in release.yml ci

---
 .github/workflows/release.yml | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
index 86b42f8..6cfd95d 100644
--- a/.github/workflows/release.yml
+++ b/.github/workflows/release.yml
@@ -18,6 +18,15 @@ jobs:
       - name: Setup Docker Buildx
         uses: docker/setup-buildx-action@v3
 
+      - name: Free disk space
+        run: |
+          sudo rm -rf /usr/share/dotnet
+          sudo rm -rf /usr/local/lib/android
+          sudo rm -rf /opt/ghc
+          sudo rm -rf /opt/hostedtoolcache
+          sudo docker system prune -af || true
+          df -h
+
       - name: Build .run installer (Docker)
         run: |
           docker buildx build \