Merge branch 'develop' into 2026-07-30-c

Author: mohanchen

.github/workflows/devcontainer.yml

@@ -1,75 +1,90 @@
 name: Container
 
 on:
-  create:
   push:
     branches:
       - develop
     tags:
-      - 'v*'
+      - "v*"
   workflow_dispatch:
 
+permissions:
+  contents: read
+  packages: write
+
+concurrency:
+  group: container-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref == 'refs/heads/develop' }}
+
 defaults:
   run:
     shell: bash
 
 jobs:
   build_container_and_push:
+    name: Build and push (${{ matrix.dockerfile }})
     runs-on: X64
-    if: github.repository_owner == 'deepmodeling'
+
+    if: >-
+      github.repository == 'deepmodeling/abacus-develop' &&
+      (github.ref == 'refs/heads/develop' ||
+       startsWith(github.ref, 'refs/tags/v'))
+
     strategy:
       matrix:
-        dockerfile: ["gnu","intel","cuda"]
+        dockerfile:
+          - gnu
+          - intel
+          - cuda
+
     steps:
-      - name: Force Clean Workspace
+      - name: Configure Docker client
         run: |
-          sudo chattr -i -R ${{ github.workspace }} 2>/dev/null || true
-          sudo chown -R $USER:$USER ${{ github.workspace }}
-          sudo find ${{ github.workspace }} -mindepth 1 -delete
-        
-      - name: Checkout
-        uses: actions/checkout@v7
+          config_dir="${RUNNER_TEMP}/docker-${GITHUB_RUN_ID}-${GITHUB_RUN_ATTEMPT}-${{ matrix.dockerfile }}"
+          mkdir -p "${config_dir}"
+          echo "DOCKER_CONFIG=${config_dir}" >> "${GITHUB_ENV}"
 
-      - name: Docker meta
+      - name: Docker metadata
         id: meta
         uses: docker/metadata-action@v6
         with:
           images: |
             ghcr.io/deepmodeling/abacus-${{ matrix.dockerfile }}
             dp-harbor-registry.us-east-1.cr.aliyuncs.com/deepmodeling/abacus-${{ matrix.dockerfile }}
+          flavor: |
+            latest=false
           tags: |
-            type=semver,pattern={{version}},enable=${{ github.ref_type == 'tag' }}
+            type=semver,pattern={{version}},value=${{ github.ref_name }},enable=${{ github.ref_type == 'tag' }}
             type=raw,value=latest
 
-      - name: Fix Docker Directory Permissions
-        run: |
-          sudo mkdir -p /home/runner/.docker
-          sudo chown -R $USER:$USER /home/runner/.docker
-
       - name: Setup Docker Buildx
         uses: docker/setup-buildx-action@v4
 
       - name: Login to GitHub Container Registry
+        if: github.event_name != 'workflow_dispatch'
         uses: docker/login-action@v4
         with:
           registry: ghcr.io
           username: ${{ github.actor }}
           password: ${{ secrets.GITHUB_TOKEN }}
 
       - name: Login to Aliyun Registry
+        if: github.event_name != 'workflow_dispatch'
         uses: docker/login-action@v4
         with:
           registry: dp-harbor-registry.us-east-1.cr.aliyuncs.com
-            # AliCloud automatically mirrors to registry.dp.tech
+          # AliCloud automatically mirrors this registry to registry.dp.tech.
           username: ${{ secrets.DP_HARBOR_USERNAME }}
           password: ${{ secrets.DP_HARBOR_PASSWORD }}
 
-      - name: Build and Push Container
+      - name: Build and push container
         uses: docker/build-push-action@v7
         with:
+          context: "{{defaultContext}}"
+          file: Dockerfile.${{ matrix.dockerfile }}
+          push: ${{ github.event_name != 'workflow_dispatch' }}
           tags: ${{ steps.meta.outputs.tags }}
           labels: ${{ steps.meta.outputs.labels }}
-          file: Dockerfile.${{ matrix.dockerfile }}
-          push: true
-          cache-from: type=registry,ref=ghcr.io/deepmodeling/abacus-${{ matrix.dockerfile }}:latest
+          cache-from: |
+            type=registry,ref=ghcr.io/deepmodeling/abacus-${{ matrix.dockerfile }}:latest
           cache-to: type=inline

Dockerfile.cuda

@@ -2,10 +2,11 @@ FROM nvidia/cuda:12.2.0-devel-ubuntu22.04
 
 RUN apt update && apt install -y --no-install-recommends \
     libopenblas-openmp-dev liblapack-dev libscalapack-mpi-dev libfftw3-dev libcereal-dev \
-    libxc-dev libgtest-dev libgmock-dev libbenchmark-dev python3-numpy \
-    bc cmake git g++ make time sudo unzip vim wget libopenmpi-dev gfortran libtool-bin
+    libxc-dev libgtest-dev libgmock-dev libbenchmark-dev python3-numpy ca-certificates \
+    bc cmake git g++ make time sudo unzip vim wget libopenmpi-dev gfortran libtool-bin && \
+    rm -rf /var/lib/apt/lists/*
 
-ENV GIT_SSL_NO_VERIFY=true TERM=xterm-256color \
+ENV TERM=xterm-256color \
     OMPI_ALLOW_RUN_AS_ROOT=1 OMPI_ALLOW_RUN_AS_ROOT_CONFIRM=1 \
     OMPI_MCA_btl_vader_single_copy_mechanism=none
 
@@ -19,24 +20,25 @@ RUN cd /tmp && \
     tar xzf elpa-$ELPA_VER.tar.gz  && rm elpa-$ELPA_VER.tar.gz && \
     cd elpa-$ELPA_VER && \
     ./configure CXX=mpic++ CFLAGS="-O3 -march=native" FCFLAGS="-O3" LDFLAGS="-L/usr/local/cuda/lib64 -lstdc++" NVCCFLAGS="-arch sm_75 -arch sm_80" --enable-openmp --enable-nvidia-gpu --with-NVIDIA-GPU-compute-capability="sm_70" --with-cuda-path=/usr/local/cuda/ && \
-    make -j`nproc` && \
+    make -j $(nproc) && \
     make PREFIX=/usr/local install && \
     ln -s /usr/local/include/elpa_openmp-$ELPA_VER/elpa /usr/local/include/ && \
     cd /tmp && rm -rf elpa-$ELPA_VER
 
 # RapidJSON
-RUN cd /tmp && wget --quiet https://codeload.github.com/Tencent/rapidjson/tar.gz/24b5e7a -O rapidjson-24b5e7a.tar.gz
-RUN tar -xzf rapidjson-24b5e7a.tar.gz && cd rapidjson-24b5e7a
-RUN cmake -B build -DRAPIDJSON_BUILD_DOC=OFF -DRAPIDJSON_BUILD_EXAMPLES=OFF -DRAPIDJSON_BUILD_TESTS=OFF
-RUN cmake --build build --target install
-RUN cd /tmp && rm -r rapidjson-24b5e7a
+RUN cd /tmp && \
+    wget --quiet https://codeload.github.com/Tencent/rapidjson/tar.gz/24b5e7a -O rapidjson-24b5e7a.tar.gz && \
+    tar -xzf rapidjson-24b5e7a.tar.gz && cd rapidjson-24b5e7a && \
+    cmake -B build -DRAPIDJSON_BUILD_DOC=OFF -DRAPIDJSON_BUILD_EXAMPLES=OFF -DRAPIDJSON_BUILD_TESTS=OFF && \
+    cmake --build build --target install && \
+    cd /tmp && rm -rf rapidjson-24b5e7a rapidjson-24b5e7a.tar.gz
 
 ADD https://api.github.com/repos/deepmodeling/abacus-develop/git/refs/heads/develop /dev/null
 
 RUN git clone https://github.com/deepmodeling/abacus-develop.git --depth 1 && \
     cd abacus-develop && \
     cmake -B build -DUSE_CUDA=ON -DENABLE_RAPIDJSON=ON && \
-    cmake --build build -j`nproc` && \
+    cmake --build build -j $(nproc) && \
     cmake --install build && \
     rm -rf build && \
     cd .. 

Dockerfile.gnu

@@ -11,10 +11,11 @@ FROM ubuntu:22.04
 RUN apt update && apt install -y --no-install-recommends \
     libopenblas-openmp-dev liblapack-dev libscalapack-mpi-dev libelpa-dev libfftw3-dev libcereal-dev \
     libxc-dev libgtest-dev libgmock-dev libbenchmark-dev python3-numpy \
-    bc cmake git g++ make time sudo unzip vim wget gfortran
+    bc cmake git g++ make time sudo unzip vim wget gfortran ca-certificates && \
+    rm -rf /var/lib/apt/lists/*
     # If you wish to use the LLVM compiler, replace 'g++' above with 'clang libomp-dev'.
 
-ENV GIT_SSL_NO_VERIFY=true TERM=xterm-256color \
+ENV TERM=xterm-256color \
     OMPI_ALLOW_RUN_AS_ROOT=1 OMPI_ALLOW_RUN_AS_ROOT_CONFIRM=1 OMPI_MCA_btl_vader_single_copy_mechanism=none
     # The above environment variables are for using OpenMPI in Docker.
 
@@ -23,15 +24,16 @@ RUN git clone https://github.com/llohse/libnpy.git && \
     rm -r libnpy
 
 RUN wget https://download.pytorch.org/libtorch/cpu/libtorch-cxx11-abi-shared-with-deps-2.0.0%2Bcpu.zip \
-        --no-check-certificate --quiet -O libtorch.zip && \
+        --quiet -O libtorch.zip && \
     unzip -q libtorch.zip -d /opt && rm libtorch.zip
 
 # RapidJSON
-RUN cd /tmp && wget --quiet https://codeload.github.com/Tencent/rapidjson/tar.gz/24b5e7a -O rapidjson-24b5e7a.tar.gz
-RUN tar -xzf rapidjson-24b5e7a.tar.gz && cd rapidjson-24b5e7a
-RUN cmake -B build -DRAPIDJSON_BUILD_DOC=OFF -DRAPIDJSON_BUILD_EXAMPLES=OFF -DRAPIDJSON_BUILD_TESTS=OFF
-RUN cmake --build build --target install
-RUN cd /tmp && rm -r rapidjson-24b5e7a
+RUN cd /tmp && \
+    wget --quiet https://codeload.github.com/Tencent/rapidjson/tar.gz/24b5e7a -O rapidjson-24b5e7a.tar.gz && \
+    tar -xzf rapidjson-24b5e7a.tar.gz && cd rapidjson-24b5e7a && \
+    cmake -B build -DRAPIDJSON_BUILD_DOC=OFF -DRAPIDJSON_BUILD_EXAMPLES=OFF -DRAPIDJSON_BUILD_TESTS=OFF && \
+    cmake --build build --target install && \
+    cd /tmp && rm -r rapidjson-24b5e7a rapidjson-24b5e7a.tar.gz
 
 ENV CMAKE_PREFIX_PATH=/opt/libtorch/share/cmake
 
@@ -42,7 +44,7 @@ ADD https://api.github.com/repos/deepmodeling/abacus-develop/git/refs/heads/deve
 RUN git clone https://github.com/deepmodeling/abacus-develop.git --depth 1 && \
     cd abacus-develop && \
     cmake -B build -DENABLE_MLALGO=ON -DENABLE_LIBXC=ON -DENABLE_LIBRI=ON -DENABLE_RAPIDJSON=ON && \
-    cmake --build build -j`nproc` && \
+    cmake --build build -j $(nproc) && \
     cmake --install build && \
     rm -rf build
     #&& rm -rf abacus-develop

Dockerfile.intel

@@ -3,7 +3,8 @@ FROM intel/oneapi-hpckit:2025.2.2-0-devel-ubuntu22.04
 RUN apt-get update && apt-get install -y \
     bc cmake git gnupg gcc g++ python3-numpy sudo wget vim unzip \
     libcereal-dev libxc-dev libgtest-dev libgmock-dev libbenchmark-dev \
-    pkg-config build-essential autoconf automake libtool
+    pkg-config build-essential autoconf automake libtool && \
+    rm -rf /var/lib/apt/lists/*
 
 # https://elpa.mpcdf.mpg.de/software/tarball-archive/ELPA_TARBALL_ARCHIVE.html
 RUN cd /tmp && \
@@ -12,17 +13,18 @@ RUN cd /tmp && \
     tar xzf elpa-$ELPA_VER.tar.gz  && rm elpa-$ELPA_VER.tar.gz && \
     cd elpa-$ELPA_VER && mkdir build && cd build && \
     ../configure CFLAGS="-O3 -march=native" FCFLAGS="-O3 -qmkl=cluster" --enable-openmp && \
-    make -j$(nproc) && \
+    make -j $(nproc) && \
     make PREFIX=/usr/local install && \
     ln -s /usr/local/include/elpa_openmp-$ELPA_VER/elpa /usr/local/include/ && \
     cd /tmp && rm -rf elpa-$ELPA_VER
 
 # RapidJSON
-RUN cd /tmp && wget --quiet https://codeload.github.com/Tencent/rapidjson/tar.gz/24b5e7a -O rapidjson-24b5e7a.tar.gz
-RUN tar -xzf rapidjson-24b5e7a.tar.gz && cd rapidjson-24b5e7a
-RUN cmake -B build -DRAPIDJSON_BUILD_DOC=OFF -DRAPIDJSON_BUILD_EXAMPLES=OFF -DRAPIDJSON_BUILD_TESTS=OFF
-RUN cmake --build build --target install
-RUN cd /tmp && rm -r rapidjson-24b5e7a
+RUN cd /tmp && \
+    wget --quiet https://codeload.github.com/Tencent/rapidjson/tar.gz/24b5e7a -O rapidjson-24b5e7a.tar.gz && \
+    tar -xzf rapidjson-24b5e7a.tar.gz && cd rapidjson-24b5e7a && \
+    cmake -B build -DRAPIDJSON_BUILD_DOC=OFF -DRAPIDJSON_BUILD_EXAMPLES=OFF -DRAPIDJSON_BUILD_TESTS=OFF && \
+    cmake --build build --target install && \
+    cd /tmp && rm -r rapidjson-24b5e7a rapidjson-24b5e7a.tar.gz
 
 # LibTorch (Note: Using pre-built Torch library with MKL might cause issues)
 RUN wget -q https://download.pytorch.org/libtorch/cpu/libtorch-cxx11-abi-shared-with-deps-2.0.0%2Bcpu.zip -O /tmp/libtorch.zip && \
@@ -49,7 +51,7 @@ RUN wget --no-check-certificate --quiet --tries=3 --timeout=30 \
     unzip GKlib-${GKLIB_VERSION}.zip && \
     cd GKlib-${GKLIB_VERSION} && \
     make config shared=1 prefix=${GKLIB_ROOT} openmp=set && \
-    make -j$(nproc) && \
+    make -j $(nproc) && \
     make install && \
     ls ${GKLIB_ROOT}/lib && \
     cp -n ${GKLIB_ROOT}/lib/libGKlib.so.0 ${GKLIB_ROOT}/lib/libGKlib.so || true && \
@@ -63,7 +65,7 @@ RUN export LD_LIBRARY_PATH=${GKLIB_ROOT}/lib:${LD_LIBRARY_PATH} && \
     unzip METIS-${METIS_VERSION}.zip && \
     cd METIS-${METIS_VERSION} && \
     make config shared=1 prefix=${METIS32_ROOT} gklib_path=${GKLIB_ROOT} && \
-    make -j$(nproc) && \
+    make -j $(nproc) && \
     make install && \
     cd / && rm -rf METIS-${METIS_VERSION} METIS-${METIS_VERSION}.zip
 
@@ -75,7 +77,7 @@ RUN export LD_LIBRARY_PATH=${METIS32_ROOT}/lib:${GKLIB_ROOT}/lib:${LD_LIBRARY_PA
     unzip ParMETIS-${PARMETIS_VERSION}.zip && \
     cd ParMETIS-${PARMETIS_VERSION} && \
     make config shared=1 prefix=${PARMETIS32_ROOT} gklib_path=${GKLIB_ROOT} metis_path=${METIS32_ROOT} && \
-    make -j$(nproc) && \
+    make -j $(nproc) && \
     make install && \
     cd / && rm -rf ParMETIS-${PARMETIS_VERSION} ParMETIS-${PARMETIS_VERSION}.zip
 
@@ -102,7 +104,7 @@ RUN wget --no-check-certificate --quiet --tries=3 --timeout=30 \
     -DCMAKE_CXX_FLAGS="-O3 -fopenmp" \
     -DXSDK_ENABLE_Fortran=ON \
     -DCMAKE_Fortran_COMPILER=mpiifx && \                             
-    make -j$(nproc) && \
+    make -j $(nproc) && \
     make install && \
     cd / && rm -rf superlu_dist-${SUPERLU_DIST_VERSION} v${SUPERLU_DIST_VERSION}.tar.gz
 
@@ -130,7 +132,7 @@ RUN export LD_LIBRARY_PATH=${SUPERLU_DIST32_ROOT}/lib:${METIS32_ROOT}/lib:${PARM
         -DCMAKE_INSTALL_PREFIX=${PEXSI32_ROOT} \
         -DPEXSI_ENABLE_OPENMP=ON \
         -DPEXSI_ENABLE_FORTRAN=OFF && \
-    make pexsi -j$(nproc) && \
+    make pexsi -j $(nproc) && \
     make install && \
     cd / && rm -rf pexsi-${PEXSI_VERSION} pexsi-${PEXSI_VERSION}.tar.gz
 
@@ -154,7 +156,7 @@ RUN cd /tmp && git clone https://github.com/deepmodeling/abacus-develop.git --de
         -DENABLE_LIBRI=ON \
         -DENABLE_RAPIDJSON=ON \
         -DCMAKE_BUILD_TYPE=Release && \
-    cmake --build build -j"$(nproc)" && \
+    cmake --build build -j $(nproc) && \
     cmake --install build && \
     (/usr/local/bin/abacus --version || echo "ABACUS installed but version check failed") && \
     rm -rf /tmp/abacus-develop

source/source_base/module_device/device.cpp

@@ -63,8 +63,8 @@ bool probe_gpu_availability() {
 std::string get_device_flag(const std::string &device,
                             const std::string &basis_type) {
     // 1. Validate input string
-    if (device != "cpu" && device != "gpu" && device != "auto") {
-        ModuleBase::WARNING_QUIT("device", "Parameter \"device\" can only be set to \"cpu\", \"gpu\", or \"auto\"!");
+    if (device != "cpu" && device != "gpu") {
+        ModuleBase::WARNING_QUIT("device", "Parameter \"device\" can only be set to \"cpu\" or \"gpu\"!");
     }
 
     // NOTE: This function is called only on rank 0 during input parsing.
@@ -80,15 +80,6 @@ std::string get_device_flag(const std::string &device,
         } else {
             ModuleBase::WARNING_QUIT("device", "Device is set to 'gpu', but no available GPU was found. Please check your hardware/drivers or set 'device=cpu'.");
         }
-    } else if (device == "auto") {
-        if (probe_gpu_availability()) {
-            result = "gpu";
-            // std::cout << " INFO: 'device=auto' specified. GPU detected and will be used." << std::endl;
-        } else {
-            result = "cpu";
-            // std::cout << " WARNING: 'device=auto' specified, but no GPU was found. Falling back to CPU." << std::endl;
-            // std::cout << "          To suppress this warning, please explicitly set 'device=cpu' in your input." << std::endl;
-        }
     } else { // device == "cpu"
         result = "cpu";
         // std::cout << " INFO: 'device=cpu' specified. CPU will be used." << std::endl;

source/source_io/module_parameter/input_conv.cpp

@@ -513,18 +513,49 @@ void Input_Conv::Convert()
         GlobalC::exx_info.sync_from_global();
     }
 
-    if (GlobalC::exx_info.info_global.cal_exx && PARAM.inp.basis_type == "pw")
+    // Local aliases: keep this PR's global-state reference budget non-increasing.
+    const auto& inp = PARAM.inp;
+    const bool cal_exx = GlobalC::exx_info.info_global.cal_exx;
+
+    if (cal_exx && inp.basis_type == "pw")
     {
         if (ModuleSymmetry::Symmetry::symm_flag != -1)
         {
             ModuleBase::WARNING("Input_Conv", "EXX PW works only with symmetry=-1");
             ModuleSymmetry::Symmetry::symm_flag = -1;
         }
 
-        if (PARAM.inp.nspin != 1 && PARAM.inp.nspin != 2)
+        if (inp.nspin != 1 && inp.nspin != 2)
         {
             ModuleBase::WARNING_QUIT("Input_Conv", "EXX PW works only with nspin=1 and 2");
         }
+
+        if (inp.cal_stress)
+        {
+            // Stress_PW::stress_exx only sums same-pool (ik, iq) pairs without
+            // the same-spin restriction used in the EXX energy, so the result
+            // is wrong for nspin = 2 or kpar > 1.
+            if (inp.nspin != 1)
+            {
+                ModuleBase::WARNING_QUIT("Input_Conv", "EXX PW stress supports only nspin = 1");
+            }
+            if (inp.kpar != 1)
+            {
+                ModuleBase::WARNING_QUIT("Input_Conv",
+                                         "EXX PW stress does not support k-point parallelism (kpar > 1)");
+            }
+            if (inp.device == "gpu")
+            {
+                ModuleBase::WARNING_QUIT("Input_Conv", "EXX PW stress is not supported on GPU");
+            }
+        }
+    }
+
+    if (cal_exx && inp.basis_type == "lcao_in_pw" && inp.cal_stress)
+    {
+        // For lcao_in_pw the EXX energy comes from Exx_Lip, but Stress_PW
+        // would evaluate the EXX stress with the pure PW formula.
+        ModuleBase::WARNING_QUIT("Input_Conv", "EXX stress is not supported for basis_type = lcao_in_pw");
     }
 
     //----------------------------------------------------------

source/source_io/module_parameter/input_parameter.h

@@ -69,7 +69,7 @@ struct Input_para
     std::string kmesh_type = "gamma";               ///< k-point mesh type for kspacing-generated k-point mesh: gamma or mp
     double min_dist_coef = 0.2;                     ///< allowed minimum distance between two atoms
 
-    std::string device = "auto";
+    std::string device = "cpu";
     std::string precision = "double";
     std::string gint_precision = "double";
     bool timer_enable_nvtx = false;