ci: make ccache effective and observable, shallow checkout, add cache…

.github/workflows/cache_warmer.yml

@@ -0,0 +1,116 @@
+name: Cache Warmer
+
+# Primes the self-hosted runners' /tmp/ccache with the latest develop build
+# after every merge. Without this, the first PR runs after a large refactor
+# lands in develop find a cold/stale ccache and pay a full rebuild
+# (observed: Build step 2.7 min warm vs 37.5 min stale).
+#
+# IMPORTANT: keep the cmake configure flags below in sync with test.yml and
+# cuda.yml — ccache entries are keyed on the actual compiler flags, so the
+# warmer only helps if it compiles with the same configuration as the CI.
+
+on:
+  push:
+    branches:
+      - develop
+  workflow_dispatch:
+
+concurrency:
+  group: cache-warmer-${{ github.ref }}
+  cancel-in-progress: true
+
+defaults:
+  run:
+    shell: bash
+
+jobs:
+  warm-gnu:
+    name: Warm X64 ccache
+    runs-on: X64
+    if: github.repository_owner == 'deepmodeling'
+    container:
+      image: ghcr.io/deepmodeling/abacus-gnu
+      volumes:
+        - /tmp/ccache:/github/home/.ccache
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v7
+        with:
+          fetch-depth: 1
+          submodules: 'false'
+
+      - name: Take ownership of the workspace and update submodules
+        run: |
+          sudo chown -R $(whoami) .
+          git submodule update --init --recursive
+
+      - name: Install CI tools
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y gfortran ccache ninja-build
+          ccache --max-size=30G
+
+      - name: Install external tools from toolchain
+        run: |
+          cd toolchain
+          ./install_abacus_toolchain_new.sh --with-dftd4=install --dry-run -j8
+          ./scripts/stage4/install_stage4.sh
+          cd ..
+
+      # Mirrors the Configure flags of .github/workflows/test.yml.
+      - name: Configure & Build (prime ccache, no tests)
+        run: |
+          source toolchain/install/setup
+          cmake -B build -G Ninja \
+            -DBUILD_TESTING=ON \
+            -DCMAKE_EXPORT_COMPILE_COMMANDS=ON \
+            -DENABLE_MLALGO=ON \
+            -DENABLE_LIBXC=ON \
+            -DENABLE_LIBRI=ON \
+            -DENABLE_GOOGLEBENCH=ON \
+            -DENABLE_RAPIDJSON=ON \
+            -DENABLE_FLOAT_FFTW=ON \
+            -DENABLE_DFTD4=ON
+          cmake --build build -j8
+
+      - name: ccache statistics
+        run: ccache -s
+
+  warm-cuda:
+    name: Warm gpu ccache
+    runs-on: gpu
+    if: github.repository_owner == 'deepmodeling'
+    container:
+      image: ghcr.io/deepmodeling/abacus-cuda
+      volumes:
+        - /tmp/ccache:/github/home/.ccache
+      options: --gpus all
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v7
+        with:
+          submodules: recursive
+
+      - name: Install CI tools
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y ccache xz-utils ninja-build pkg-config
+          ccache --max-size=30G
+
+      - name: Install external tools from toolchain
+        run: |
+          cd toolchain
+          ./install_abacus_toolchain_new.sh --with-dftd4=install --dry-run -j8
+          ./scripts/stage4/install_stage4.sh
+          cd ..
+
+      # Mirrors the Configure & Build step of .github/workflows/cuda.yml.
+      - name: Configure & Build (prime ccache, no tests)
+        run: |
+          source toolchain/install/setup
+          rm -rf build
+          cmake -B build -G Ninja -DUSE_CUDA=ON -DBUILD_TESTING=ON -DENABLE_FLOAT_FFTW=ON
+          cmake --build build -j "$(nproc)"
+
+      - name: ccache statistics
+        run: ccache -s

.github/workflows/test.yml

@@ -24,7 +24,9 @@ jobs:
       - name: Checkout repository
         uses: actions/checkout@v7
         with:
-          fetch-depth: 0
+          # CMake only needs `git log -1` for the commit info; a shallow
+          # fetch is enough and saves ~10 min on cache-cold runners.
+          fetch-depth: 1
           # We will handle submodules manually after fixing ownership
           submodules: 'false'
 
@@ -47,6 +49,14 @@ jobs:
             xz-utils
           sudo pip install clang-format clang-tidy
 
+      - name: Setup ccache
+        run: |
+          # Default cache size (~10G) is too small for a full -O3 -g build
+          # plus daily develop churn; raise it and reset per-run counters.
+          ccache --max-size=30G
+          ccache --zero-stats
+          ccache -s
+
       - name: Install external tools from toolchain
         run: |
           cd toolchain
@@ -308,3 +318,7 @@ jobs:
           OMP_NUM_THREADS: '2'
         run: |
           ctest --test-dir build -V --timeout 1700 -E 'integrate_test|01_PW|02_NAO_Gamma|03_NAO_multik|04_FF|05_rtTDDFT|06_SDFT|07_OFDFT|08_EXX|09_DeePKS|10_others|11_PW_GPU|12_NAO_Gamma_GPU|13_NAO_multik_GPU|15_rtTDDFT_GPU|16_SDFT_GPU|17_DS_DFTU|MODULE_BASE|MODULE_IO|MODULE_HSOLVER|MODULE_CELL|MODULE_MD|MODULE_PSI|MODULE_ESTATE|MODULE_RI|MODULE_HAMILT|MODULE_PW|MODULE_LCAO|MODULE_AO|MODULE_NAO|MODULE_RELAX|MODULE_LR'
+
+      - name: ccache statistics
+        if: always()
+        run: ccache -s