thanks to ncnn ❤

.astylerc

@@ -0,0 +1,35 @@
+# astyle -n -r "benchmark/*.h,*.cpp" "src/*.h,*.cpp" "tests/*.h,*.cpp" "tools/*.h,*.cpp" "examples/*.h,*.cpp"
+
+# brace style
+--style=allman
+
+# tab
+--attach-namespaces
+--attach-extern-c
+--attach-closing-while
+
+# indentation
+--indent-preproc-define
+--indent-col1-comments
+--min-conditional-indent=0
+--max-continuation-indent=120
+
+# padding
+--pad-oper
+--pad-comma
+--pad-header
+--align-pointer=type
+--align-reference=type
+
+# formatting
+--break-closing-braces
+--attach-return-type
+--attach-return-type-decl
+--keep-one-line-blocks
+--keep-one-line-statements
+--convert-tabs
+--max-code-length=200
+--mode=c
+
+# other
+--lineend=linux

.ci/linux-x64-cpu-gcc.yml

@@ -0,0 +1,119 @@
+name: linux-x64-cpu-gcc
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.ci/linux-x64-cpu-gcc.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'tests/**'
+    - 'tools/**'
+    - '!tools/pnnx/**'
+    - 'examples/**'
+  mr:
+    target-branches: [master]
+    paths:
+    - '.ci/linux-x64-cpu-gcc.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'tests/**'
+    - 'tools/**'
+    - '!tools/pnnx/**'
+    - 'examples/**'
+concurrency:
+  group: linux-x64-cpu-gcc-${{ ci.head_ref }}
+
+jobs:
+  linux-gcc:
+    name: linux-gcc
+    strategy:
+      matrix:
+        include:
+          - { SSE2: 'OFF', AVX: 'OFF', AVX2: 'OFF', AVX512: 'OFF' }
+          - { SSE2: 'ON',  AVX: 'OFF', AVX2: 'OFF', AVX512: 'OFF' }
+          - { SSE2: 'ON',  AVX: 'ON',  AVX2: 'OFF', AVX512: 'OFF' }
+          - { SSE2: 'ON',  AVX: 'ON',  AVX2: 'ON',  AVX512: 'OFF' }
+          - { SSE2: 'ON',  AVX: 'ON',  AVX2: 'ON',  AVX512: 'ON'  }
+
+    runs-on:
+      pool-name: docker
+      container:
+        image: bkci/ci:ubuntu
+    steps:
+    - name: checkout
+      checkout: self
+      with:
+        strategy: FRESH_CHECKOUT
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: install-deps
+      run: |
+        apt-get update
+        apt-get install -y libprotobuf-dev protobuf-compiler libopencv-dev
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DNCNN_SSE2=${{matrix.SSE2}} -DNCNN_AVX=${{matrix.AVX}} -DNCNN_AVX2=${{matrix.AVX2}} -DNCNN_AVX512=${{matrix.AVX512}} -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j $(nproc)
+    - name: test
+      run: cd build && ctest --output-on-failure -j $(nproc)
+    - name: build-shared
+      run: |
+        mkdir build-shared && cd build-shared
+        cmake -DNCNN_SSE2=${{matrix.SSE2}} -DNCNN_AVX=${{matrix.AVX}} -DNCNN_AVX2=${{matrix.AVX2}} -DNCNN_AVX512=${{matrix.AVX512}} -DNCNN_SHARED_LIB=ON ..
+        cmake --build . -j $(nproc)
+    - name: build-noint8
+      run: |
+        mkdir build-noint8 && cd build-noint8
+        cmake -DNCNN_SSE2=${{matrix.SSE2}} -DNCNN_AVX=${{matrix.AVX}} -DNCNN_AVX2=${{matrix.AVX2}} -DNCNN_AVX512=${{matrix.AVX512}} -DNCNN_INT8=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j $(nproc)
+    - name: test-noint8
+      run: cd build-noint8 && ctest --output-on-failure -j $(nproc)
+
+  linux-gcc-cpp03-nostdio-nostring-simplestl:
+    runs-on:
+      pool-name: docker
+      container:
+        image: bkci/ci:ubuntu
+    steps:
+    - name: checkout
+      checkout: self
+      with:
+        strategy: FRESH_CHECKOUT
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: build-nostdio
+      run: |
+        mkdir build-nostdio && cd build-nostdio
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/host.gcc-c++03.toolchain.cmake -DNCNN_BUILD_TESTS=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j $(nproc)
+    - name: test-nostdio
+      run: cd build-nostdio && ctest --output-on-failure -j $(nproc)
+    - name: build-nostdio-nostring
+      run: |
+        mkdir build-nostdio-nostring && cd build-nostdio-nostring
+        cmake -DNCNN_STDIO=OFF -DNCNN_STRING=OFF -DNCNN_BUILD_TESTS=OFF -DNCNN_BUILD_BENCHMARK=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j $(nproc)
+    - name: build-simplestl
+      run: |
+        mkdir build-simplestl && cd build-simplestl
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/host-c.gcc.toolchain.cmake -DNCNN_STDIO=ON -DNCNN_STRING=ON -DNCNN_SIMPLESTL=ON -DNCNN_BUILD_TESTS=ON -DNCNN_BUILD_BENCHMARK=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j $(nproc)
+    - name: test-simplestl
+      run: cd build-simplestl && ctest --output-on-failure -j $(nproc)
+    - name: build-simplestl-simpleomp
+      run: |
+        mkdir build-simplestl-simpleomp && cd build-simplestl-simpleomp
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/host-c.gcc.toolchain.cmake -DNCNN_STDIO=ON -DNCNN_STRING=ON -DNCNN_SIMPLESTL=ON -DNCNN_SIMPLEOMP=ON -DNCNN_BUILD_TESTS=ON -DNCNN_BUILD_BENCHMARK=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j $(nproc)
+    - name: test-simplestl-simpleomp
+      run: cd build-simplestl-simpleomp && ctest --output-on-failure -j $(nproc)

.ci/pnnx.yml

@@ -0,0 +1,125 @@
+name: pnnx
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.ci/pnnx.yml'
+    - 'tools/pnnx/**'
+    - '!tools/pnnx/README.md'
+  mr:
+    target-branches: [master]
+    paths:
+    - '.ci/pnnx.yml'
+    - 'tools/pnnx/**'
+    - '!tools/pnnx/README.md'
+concurrency:
+  group: pnnx-${{ ci.head_ref }}
+
+jobs:
+  ubuntu:
+    strategy:
+      matrix:
+        include:
+          - torch-version: 1.8.1
+            torchvision-version: 0.9.1
+            torchvision-cache-key: '0_9_1'
+
+          - torch-version: 1.9.1
+            torchvision-version: 0.10.1
+            torchvision-cache-key: '0_10_1'
+
+          - torch-version: 1.10.0
+            torchvision-version: 0.11.1
+            torchvision-cache-key: '0_11_1'
+
+          - torch-version: 1.11.0
+            torchvision-version: 0.12.0
+            torchvision-cache-key: '0_12_0'
+
+          - torch-version: 1.12.0
+            torchvision-version: 0.13.0
+            torchvision-cache-key: '0_13_0'
+
+          - torch-version: 1.13.0
+            torchvision-version: 0.14.0
+            torchvision-cache-key: '0_14_0'
+
+          - torch-version: 2.0.0
+            torchvision-version: 0.15.1
+            torchvision-cache-key: '0_15_1'
+
+    runs-on:
+      pool-name: docker
+      container:
+        image: bkci/ci:ubuntu
+    steps:
+    - name: checkout
+      checkout: self
+      with:
+        strategy: FRESH_CHECKOUT
+        enableGitLfs: false
+
+    - name: install-deps
+      run: |
+        apt-get update
+        apt-get install -y python3-pip libjpeg-dev libpng-dev libprotobuf-dev protobuf-compiler
+        python3 -m pip install --upgrade pip
+        apt-get remove -y python3-setuptools
+        pip3 install -U setuptools
+        pip3 install -U pytest wheel twine requests einops
+
+    - name: setup pytorch
+      run: |
+        export PYTHONUSERBASE=${{ci.workspace}}/torch-${{matrix.torch-version}}
+        pip3 install --user torch==${{matrix.torch-version}}+cpu torchvision==${{matrix.torchvision-version}}+cpu -f https://download.pytorch.org/whl/torch_stable.html
+
+    - name: cache-torchvision
+      id: cache-torchvision
+      uses: cache@1.*
+      with:
+        cachePaths: torchvision-${{matrix.torchvision-version}}-install
+        cacheKey: torchvision-${{matrix.torchvision-cache-key}}-linux-install-20211228
+    - name: checkout-torchvision
+      if: steps.cache-torchvision.outputs.cacheHit != 'true'
+      checkout: https://github.com/pytorch/vision.git
+      with:
+        pullType: TAG
+        refName: v${{matrix.torchvision-version}}
+        localPath: vision
+        enableSubmodule: false
+        enableGitLfs: false
+    - name: torchvision
+      if: steps.cache-torchvision.outputs.cacheHit != 'true'
+      run: |
+        cd vision
+        mkdir -p build; cd build
+        cmake -DCMAKE_INSTALL_PREFIX=${{ci.workspace}}/torchvision-${{matrix.torchvision-version}}-install -DTorch_DIR=${{ci.workspace}}/torch-${{matrix.torch-version}}/lib/python3.9/site-packages/torch/share/cmake/Torch -DCMAKE_BUILD_TYPE=Release ..
+        cmake --build . -j $(nproc)
+        cmake --build . --target install
+
+    - name: build-ncnn
+      run: |
+        export PYTHONUSERBASE=${{ci.workspace}}/torch-${{matrix.torch-version}}
+        mkdir build && cd build
+        cmake -DCMAKE_BUILD_TYPE=Release -DNCNN_PYTHON=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j $(nproc)
+        cd ..
+        export CMAKE_BUILD_PARALLEL_LEVEL=$(nproc)
+        pip3 install --user .
+
+    - name: build-pnnx
+      run: |
+        export PYTHONUSERBASE=${{ci.workspace}}/torch-${{matrix.torch-version}}
+        cd tools/pnnx
+        mkdir build && cd build
+        cmake -DCMAKE_BUILD_TYPE=Release -DTorchVision_INSTALL_DIR=${{ci.workspace}}/torchvision-${{matrix.torchvision-version}}-install ..
+        cmake --build . -j 16
+
+    - name: test
+      run: |
+        export PYTHONUSERBASE=${{ci.workspace}}/torch-${{matrix.torch-version}}
+        export OMP_NUM_THREADS=1
+        export MKL_NUM_THREADS=1
+        export MKL_ENABLE_INSTRUCTIONS=SSE4_2
+        cd tools/pnnx
+        cd build && ctest --output-on-failure -j 16

.ci/test-coverage.yml

@@ -0,0 +1,910 @@
+name: test-coverage
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.ci/test-coverage.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/**'
+    - 'tests/**'
+    - 'toolchains/**'
+  mr:
+    target-branches: [master]
+    paths:
+    - '.ci/test-coverage.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/**'
+    - 'tests/**'
+    - 'toolchains/**'
+concurrency:
+  group: test-coverage-${{ ci.head_ref }}
+
+jobs:
+  linux-gcc-gpu:
+    name: linux-gcc-gpu
+    runs-on:
+      pool-name: docker
+      container:
+        image: bkci/ci:ubuntu
+    steps:
+    - name: checkout
+      checkout: self
+      with:
+        strategy: FRESH_CHECKOUT
+        enableGitLfs: false
+
+    - name: install-deps
+      run: |
+        apt-get update
+        apt-get install -y lcov libvulkan-dev
+        curl https://uploader.codecov.io/verification.gpg | gpg --no-default-keyring --keyring trustedkeys.gpg --import
+        curl -Os https://uploader.codecov.io/latest/linux/codecov
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM.sig
+        gpgv codecov.SHA256SUM.sig codecov.SHA256SUM
+        shasum -a 256 -c codecov.SHA256SUM
+        chmod +x codecov
+
+    - name: cache-swiftshader
+      id: cache-swiftshader
+      uses: cache@1.*
+      with:
+        cachePaths: swiftshader-install
+        cacheKey: swiftshader-linux-install-20230420-1
+
+    - name: checkout-swiftshader
+      if: steps.cache-swiftshader.outputs.cacheHit != 'true'
+      checkout: https://github.com/google/swiftshader.git
+      with:
+        pullType: COMMIT_ID
+        refName: dd55e592406dc0bae219df11adec6363840aff4a
+        localPath: swiftshader
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: swiftshader
+      if: steps.cache-swiftshader.outputs.cacheHit != 'true'
+      run: |
+        cd swiftshader
+        git -c submodule."third_party/git-hooks".update=none submodule update --init --recursive
+        mkdir -p build; cd build
+        cmake -DCMAKE_INSTALL_PREFIX=install -DSWIFTSHADER_BUILD_PVR=FALSE -DSWIFTSHADER_BUILD_TESTS=FALSE -DSWIFTSHADER_ENABLE_ASTC=FALSE -DSWIFTSHADER_WARNINGS_AS_ERRORS=FALSE -DREACTOR_BACKEND=Subzero -DREACTOR_DEFAULT_OPT_LEVEL=Default -DCMAKE_BUILD_TYPE=Release ..
+        cmake --build . -j $(nproc)
+        mkdir ${{ci.workspace}}/swiftshader-install
+        cp Linux/* ${{ci.workspace}}/swiftshader-install
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_AVX2=ON -DNCNN_AVX512=OFF -DNCNN_XOP=OFF -DNCNN_OPENMP=OFF -DNCNN_VULKAN=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j $(nproc)
+    - name: test
+      run: |
+        printf "[Processor]\nThreadCount=4\n" > build/tests/SwiftShader.ini
+        export VK_ICD_FILENAMES="${{ci.workspace}}/swiftshader-install/vk_swiftshader_icd.json"
+        cd build && ctest --output-on-failure -j 4
+    - name: lcov-collect
+      run: |
+        cd build
+        lcov -d ./src -c -o lcov.info
+        lcov -r lcov.info '/usr/*' -o lcov.info
+        lcov -r lcov.info '*/build/*' -o lcov.info
+        lcov -r lcov.info '*/glslang/*' -o lcov.info
+        lcov --list lcov.info
+    - name: codecov
+      run: ./codecov -t ${{settings.CODECOV_TOKEN.access_token}} -C ${{ ci.sha }} -B ${{ ci.head_ref }} -f build/lcov.info
+
+  linux-gcc-gpu-lavapipe:
+    name: linux-gcc-gpu-lavapipe
+    runs-on:
+      pool-name: docker
+      container:
+        image: bkci/ci:ubuntu
+    steps:
+    - name: checkout
+      checkout: self
+      with:
+        strategy: FRESH_CHECKOUT
+        enableGitLfs: false
+
+    - name: install-deps
+      run: |
+        apt-get update
+        apt-get install -y lcov libvulkan-dev libxcb-shm0
+        curl https://uploader.codecov.io/verification.gpg | gpg --no-default-keyring --keyring trustedkeys.gpg --import
+        curl -Os https://uploader.codecov.io/latest/linux/codecov
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM.sig
+        gpgv codecov.SHA256SUM.sig codecov.SHA256SUM
+        shasum -a 256 -c codecov.SHA256SUM
+        chmod +x codecov
+
+    - name: cache-lavapipe
+      id: cache-lavapipe
+      uses: cache@1.*
+      with:
+        cachePaths: lavapipe-install
+        cacheKey: lavapipe-linux-install-20211127-4
+
+    - name: checkout-lavapipe
+      if: steps.cache-lavapipe.outputs.cacheHit != 'true'
+      checkout: https://github.com/mesa3d/mesa.git
+      with:
+        pullType: COMMIT_ID
+        refName: cd39180cfab20734744b379b085cc3b5c2cecd3a
+        localPath: mesa
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: lavapipe
+      if: steps.cache-lavapipe.outputs.cacheHit != 'true'
+      run: |
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye main' | tee -a /etc/apt/sources.list
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye-updates main' | tee -a /etc/apt/sources.list
+        apt-get update
+        apt-get build-dep -y mesa
+        mkdir -p "${{ci.workspace}}/lavapipe-install"
+        cd mesa
+        mkdir build
+        cd build
+        meson -Dprefix="${{ci.workspace}}/lavapipe-install" -Dbuildtype=release -Db_lto=true -Db_ndebug=true -Dplatforms="x11" -Ddri3=enabled -Ddri-drivers="" -Dgallium-drivers=swrast -Dgallium-vdpau=disabled -Dgallium-xvmc=disabled -Dgallium-omx=disabled -Dgallium-va=disabled -Dgallium-xa=disabled -Dgallium-opencl=disabled -Dopencl-native=false -Dvulkan-drivers=swrast -Dshader-cache=disabled -Dgles1=disabled -Dgles2=disabled -Dopengl=false -Dgbm=disabled -Dglx=disabled -Degl=disabled -Dllvm=enabled -Dvalgrind=disabled -Dlibunwind=disabled -Dlmsensors=disabled ..
+        ninja -j$(nproc)
+        ninja install
+        find ${{ci.workspace}}/lavapipe-install
+        cat ${{ci.workspace}}/lavapipe-install/share/vulkan/icd.d/lvp_icd.x86_64.json
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_AVX2=ON -DNCNN_AVX512=OFF -DNCNN_XOP=OFF -DNCNN_OPENMP=OFF -DNCNN_VULKAN=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j $(nproc)
+    - name: test
+      run: |
+        export LP_NUM_THREADS=4
+        export VK_ICD_FILENAMES="${{ci.workspace}}/lavapipe-install/share/vulkan/icd.d/lvp_icd.x86_64.json"
+        cd build
+        ctest --output-on-failure -j 4
+    - name: lcov-collect
+      run: |
+        cd build
+        lcov -d ./src -c -o lcov.info
+        lcov -r lcov.info '/usr/*' -o lcov.info
+        lcov -r lcov.info '*/build/*' -o lcov.info
+        lcov -r lcov.info '*/glslang/*' -o lcov.info
+        lcov --list lcov.info
+    - name: codecov
+      run: ./codecov -t ${{settings.CODECOV_TOKEN.access_token}} -C ${{ ci.sha }} -B ${{ ci.head_ref }} -f build/lcov.info
+
+  linux-gcc-x64:
+    name: linux-gcc-x64
+    strategy:
+      matrix:
+        # openmp: ['OFF', 'ON']
+        include:
+          - { SSE2: 'OFF', AVX: 'OFF', XOP: 'OFF', F16C: 'OFF', FMA: 'OFF', AVX2: 'OFF', AVX512: 'OFF', AVX512VNNI: 'OFF', AVXVNNI: 'OFF', AVX512BF16: 'OFF', AVX512FP16: 'OFF'}
+          - { SSE2: 'ON',  AVX: 'OFF', XOP: 'OFF', F16C: 'OFF', FMA: 'OFF', AVX2: 'OFF', AVX512: 'OFF', AVX512VNNI: 'OFF', AVXVNNI: 'OFF', AVX512BF16: 'OFF', AVX512FP16: 'OFF'}
+          - { SSE2: 'ON',  AVX: 'ON',  XOP: 'OFF', F16C: 'OFF', FMA: 'OFF', AVX2: 'OFF', AVX512: 'OFF', AVX512VNNI: 'OFF', AVXVNNI: 'OFF', AVX512BF16: 'OFF', AVX512FP16: 'OFF'}
+          - { SSE2: 'ON',  AVX: 'ON',  XOP: 'OFF', F16C: 'ON',  FMA: 'ON',  AVX2: 'ON',  AVX512: 'OFF', AVX512VNNI: 'OFF', AVXVNNI: 'OFF', AVX512BF16: 'OFF', AVX512FP16: 'OFF'}
+          - { SSE2: 'ON',  AVX: 'ON',  XOP: 'OFF', F16C: 'ON',  FMA: 'ON',  AVX2: 'ON',  AVX512: 'ON',  AVX512VNNI: 'ON',  AVXVNNI: 'OFF', AVX512BF16: 'OFF', AVX512FP16: 'OFF'}
+
+    runs-on:
+      pool-name: docker
+      container:
+        image: bkci/ci:ubuntu
+    steps:
+    - name: checkout
+      checkout: self
+      with:
+        strategy: FRESH_CHECKOUT
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: install-deps
+      run: |
+        apt-get update
+        apt-get install -y lcov
+        curl https://uploader.codecov.io/verification.gpg | gpg --no-default-keyring --keyring trustedkeys.gpg --import
+        curl -Os https://uploader.codecov.io/latest/linux/codecov
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM.sig
+        gpgv codecov.SHA256SUM.sig codecov.SHA256SUM
+        shasum -a 256 -c codecov.SHA256SUM
+        chmod +x codecov
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_OPENMP=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON \
+            -DNCNN_SSE2=${{matrix.SSE2}} \
+            -DNCNN_AVX=${{matrix.AVX}} \
+            -DNCNN_XOP=${{matrix.XOP}} \
+            -DNCNN_F16C=${{matrix.F16C}} \
+            -DNCNN_FMA=${{matrix.FMA}} \
+            -DNCNN_AVX2=${{matrix.AVX2}} \
+            -DNCNN_AVX512=${{matrix.AVX512}} \
+            -DNCNN_AVXVNNI=${{matrix.AVXVNNI}} \
+            -DNCNN_AVX512VNNI=${{matrix.AVX512VNNI}} \
+            -DNCNN_AVX512BF16=${{matrix.AVX512BF16}} \
+            -DNCNN_AVX512FP16=${{matrix.AVX512FP16}} \
+            ..
+        cmake --build . -j $(nproc)
+    - name: test
+      run: cd build && ctest --output-on-failure -j $(nproc)
+    - name: lcov-collect
+      run: |
+        cd build
+        lcov -d ./src -c -o lcov.info
+        lcov -r lcov.info '/usr/*' -o lcov.info
+        lcov -r lcov.info '*/build/*' -o lcov.info
+        lcov --list lcov.info
+    - name: codecov
+      run: ./codecov -t ${{settings.CODECOV_TOKEN.access_token}} -C ${{ ci.sha }} -B ${{ ci.head_ref }} -f build/lcov.info
+
+  linux-gcc-arm:
+    name: linux-gcc-arm
+    strategy:
+      matrix:
+        # openmp: ['OFF', 'ON']
+        include:
+          - { GNU_INLINE_ASM: 'ON'}
+          - { GNU_INLINE_ASM: 'OFF'}
+
+    runs-on:
+      pool-name: docker
+      container:
+        image: bkci/ci:ubuntu
+    steps:
+    - name: checkout
+      checkout: self
+      with:
+        strategy: FRESH_CHECKOUT
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: install-deps
+      run: |
+        apt-get update
+        apt-get install -y lcov g++-arm-linux-gnueabi g++-arm-linux-gnueabihf libcapstone4 libglib2.0-0
+        curl https://uploader.codecov.io/verification.gpg | gpg --no-default-keyring --keyring trustedkeys.gpg --import
+        curl -Os https://uploader.codecov.io/latest/linux/codecov
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM.sig
+        gpgv codecov.SHA256SUM.sig codecov.SHA256SUM
+        shasum -a 256 -c codecov.SHA256SUM
+        chmod +x codecov
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: cache@1.*
+      with:
+        cachePaths: qemu-install
+        cacheKey: qemu-arm-install-20220831-1
+
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cacheHit != 'true'
+      checkout: https://github.com/qemu/qemu.git
+      with:
+        pullType: COMMIT_ID
+        refName: 621da7789083b80d6f1ff1c0fb499334007b4f51
+        localPath: qemu
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: qemu
+      if: steps.cache-qemu.outputs.cacheHit != 'true'
+      run: |
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye main' | tee -a /etc/apt/sources.list
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye-updates main' | tee -a /etc/apt/sources.list
+        apt-get update
+        apt-get build-dep -y qemu
+        cd qemu
+        ./configure --prefix=${{ci.workspace}}/qemu-install --target-list=arm-linux-user --disable-system
+        make -j$(nproc)
+        make install
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/arm-linux-gnueabi.toolchain.cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_GNU_INLINE_ASM=${{matrix.GNU_INLINE_ASM}} -DNCNN_VFPV4=ON -DNCNN_ARM82=OFF -DNCNN_OPENMP=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j $(nproc)
+    - name: test
+      run: |
+        export PATH=${{ci.workspace}}/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-arm TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/arm-linux-gnueabi" ctest --output-on-failure -j $(nproc)
+    - name: lcov-collect
+      run: |
+        cd build
+        lcov -d ./src -c -o lcov.info
+        lcov -r lcov.info '/usr/*' -o lcov.info
+        lcov -r lcov.info '*/build/*' -o lcov.info
+        lcov --list lcov.info
+
+    - name: build-armhf-vfpv3-d16
+      run: |
+        mkdir build-armhf-vfpv3-d16 && cd build-armhf-vfpv3-d16
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/arm-linux-gnueabihf-vfpv3-d16.toolchain.cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_GNU_INLINE_ASM=${{matrix.GNU_INLINE_ASM}} -DNCNN_VFPV4=OFF -DNCNN_ARM82=OFF -DNCNN_OPENMP=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j $(nproc)
+    - name: test-armhf-vfpv3-d16
+      run: |
+        export PATH=${{ci.workspace}}/qemu-install/bin:$PATH
+        cd build-armhf-vfpv3-d16
+        TESTS_EXECUTABLE_LOADER=qemu-arm TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/arm-linux-gnueabihf" ctest --output-on-failure -j $(nproc)
+    - name: lcov-collect-armhf-vfpv3-d16
+      run: |
+        cd build-armhf-vfpv3-d16
+        lcov -d ./src -c -o lcov.info
+        lcov -r lcov.info '/usr/*' -o lcov.info
+        lcov -r lcov.info '*/build-armhf-vfpv3-d16/*' -o lcov.info
+        lcov --list lcov.info
+
+    - name: codecov
+      run: |
+        ./codecov -t ${{settings.CODECOV_TOKEN.access_token}} -C ${{ ci.sha }} -B ${{ ci.head_ref }} -f build/lcov.info
+        ./codecov -t ${{settings.CODECOV_TOKEN.access_token}} -C ${{ ci.sha }} -B ${{ ci.head_ref }} -f build-armhf-vfpv3-d16/lcov.info
+
+  linux-gcc-aarch64:
+    name: linux-gcc-aarch64
+    strategy:
+      matrix:
+        # openmp: ['OFF', 'ON']
+        include:
+          - { GNU_INLINE_ASM: 'ON', ARM82: 'OFF', ARM82DOT: 'OFF', ARM82FP16FML: 'OFF', ARM84BF16: 'OFF', ARM84I8MM: 'OFF', ARM86SVE: 'OFF'}
+          - { GNU_INLINE_ASM: 'ON', ARM82: 'ON',  ARM82DOT: 'OFF', ARM82FP16FML: 'OFF', ARM84BF16: 'OFF', ARM84I8MM: 'OFF', ARM86SVE: 'OFF'}
+          - { GNU_INLINE_ASM: 'ON', ARM82: 'ON',  ARM82DOT: 'ON',  ARM82FP16FML: 'ON',  ARM84BF16: 'OFF', ARM84I8MM: 'OFF', ARM86SVE: 'OFF'}
+          - { GNU_INLINE_ASM: 'ON', ARM82: 'ON',  ARM82DOT: 'ON',  ARM82FP16FML: 'ON',  ARM84BF16: 'ON',  ARM84I8MM: 'ON',  ARM86SVE: 'OFF'}
+          - { GNU_INLINE_ASM: 'OFF', ARM82: 'ON',  ARM82DOT: 'ON',  ARM82FP16FML: 'ON',  ARM84BF16: 'ON',  ARM84I8MM: 'ON',  ARM86SVE: 'OFF'}
+
+    runs-on:
+      pool-name: docker
+      container:
+        image: bkci/ci:ubuntu
+    steps:
+    - name: checkout
+      checkout: self
+      with:
+        strategy: FRESH_CHECKOUT
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: install-deps
+      run: |
+        apt-get update
+        apt-get install -y lcov g++-aarch64-linux-gnu libcapstone4 libglib2.0-0
+        curl https://uploader.codecov.io/verification.gpg | gpg --no-default-keyring --keyring trustedkeys.gpg --import
+        curl -Os https://uploader.codecov.io/latest/linux/codecov
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM.sig
+        gpgv codecov.SHA256SUM.sig codecov.SHA256SUM
+        shasum -a 256 -c codecov.SHA256SUM
+        chmod +x codecov
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: cache@1.*
+      with:
+        cachePaths: qemu-install
+        cacheKey: qemu-aarch64-install-20220831-1
+
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cacheHit != 'true'
+      checkout: https://github.com/qemu/qemu.git
+      with:
+        pullType: COMMIT_ID
+        refName: 621da7789083b80d6f1ff1c0fb499334007b4f51
+        localPath: qemu
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: qemu
+      if: steps.cache-qemu.outputs.cacheHit != 'true'
+      run: |
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye main' | tee -a /etc/apt/sources.list
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye-updates main' | tee -a /etc/apt/sources.list
+        apt-get update
+        apt-get build-dep -y qemu
+        cd qemu
+        ./configure --prefix=${{ci.workspace}}/qemu-install --target-list=aarch64-linux-user --disable-system
+        make -j$(nproc)
+        make install
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/aarch64-linux-gnu.toolchain.cmake \
+            -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_OPENMP=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON \
+            -DNCNN_GNU_INLINE_ASM=${{matrix.GNU_INLINE_ASM}} \
+            -DNCNN_ARM82=${{matrix.ARM82}} \
+            -DNCNN_ARM82DOT=${{matrix.ARM82DOT}} \
+            -DNCNN_ARM82FP16FML=${{matrix.ARM82FP16FML}} \
+            -DNCNN_ARM84BF16=${{matrix.ARM84BF16}} \
+            -DNCNN_ARM84I8MM=${{matrix.ARM84I8MM}} \
+            ..
+        cmake --build . -j $(nproc)
+    - name: test
+      run: |
+        export PATH=${{ci.workspace}}/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-aarch64 TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/aarch64-linux-gnu" ctest --output-on-failure -j $(nproc)
+    - name: lcov-collect
+      run: |
+        cd build
+        lcov -d ./src -c -o lcov.info
+        lcov -r lcov.info '/usr/*' -o lcov.info
+        lcov -r lcov.info '*/build/*' -o lcov.info
+        lcov --list lcov.info
+    - name: codecov
+      run: ./codecov -t ${{settings.CODECOV_TOKEN.access_token}} -C ${{ ci.sha }} -B ${{ ci.head_ref }} -f build/lcov.info
+
+  linux-gcc-mipsisa32r6el:
+    name: linux-gcc-mipsisa32r6el
+    strategy:
+      matrix:
+        OPENMP: ['OFF', 'ON']
+
+    runs-on:
+      pool-name: docker
+      container:
+        image: bkci/ci:ubuntu
+    steps:
+    - name: checkout
+      checkout: self
+      with:
+        strategy: FRESH_CHECKOUT
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: install-deps
+      run: |
+        apt-get update
+        apt-get install -y lcov g++-mipsisa32r6el-linux-gnu libcapstone4 libglib2.0-0
+        curl https://uploader.codecov.io/verification.gpg | gpg --no-default-keyring --keyring trustedkeys.gpg --import
+        curl -Os https://uploader.codecov.io/latest/linux/codecov
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM.sig
+        gpgv codecov.SHA256SUM.sig codecov.SHA256SUM
+        shasum -a 256 -c codecov.SHA256SUM
+        chmod +x codecov
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: cache@1.*
+      with:
+        cachePaths: qemu-install
+        cacheKey: qemu-mipsel-install-20220831-1
+
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cacheHit != 'true'
+      checkout: https://github.com/qemu/qemu.git
+      with:
+        pullType: COMMIT_ID
+        refName: 621da7789083b80d6f1ff1c0fb499334007b4f51
+        localPath: qemu
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: qemu
+      if: steps.cache-qemu.outputs.cacheHit != 'true'
+      run: |
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye main' | tee -a /etc/apt/sources.list
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye-updates main' | tee -a /etc/apt/sources.list
+        apt-get update
+        apt-get build-dep -y qemu
+        cd qemu
+        ./configure --prefix=${{ci.workspace}}/qemu-install --target-list=mipsel-linux-user --disable-system
+        make -j$(nproc)
+        make install
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/mipsisa32r6el-linux-gnu.toolchain.cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_MSA=OFF -DNCNN_MMI=OFF -DNCNN_OPENMP=${{matrix.OPENMP}} -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j $(nproc)
+    - name: test
+      run: |
+        export PATH=${{ci.workspace}}/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-mipsel TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/mipsisa32r6el-linux-gnu" ctest --output-on-failure -j $(nproc)
+    - name: lcov-collect
+      run: |
+        cd build
+        lcov -d ./src -c -o lcov.info
+        lcov -r lcov.info '/usr/*' -o lcov.info
+        lcov -r lcov.info '*/build/*' -o lcov.info
+        lcov --list lcov.info
+    - name: codecov
+      run: ./codecov -t ${{settings.CODECOV_TOKEN.access_token}} -C ${{ ci.sha }} -B ${{ ci.head_ref }} -f build/lcov.info
+
+  linux-gcc-mipsisa64r6el:
+    name: linux-gcc-mipsisa64r6el
+    strategy:
+      matrix:
+        OPENMP: ['OFF', 'ON']
+
+    runs-on:
+      pool-name: docker
+      container:
+        image: bkci/ci:ubuntu
+    steps:
+    - name: checkout
+      checkout: self
+      with:
+        strategy: FRESH_CHECKOUT
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: install-deps
+      run: |
+        apt-get update
+        apt-get install -y lcov g++-mipsisa64r6el-linux-gnuabi64 libcapstone4 libglib2.0-0
+        curl https://uploader.codecov.io/verification.gpg | gpg --no-default-keyring --keyring trustedkeys.gpg --import
+        curl -Os https://uploader.codecov.io/latest/linux/codecov
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM.sig
+        gpgv codecov.SHA256SUM.sig codecov.SHA256SUM
+        shasum -a 256 -c codecov.SHA256SUM
+        chmod +x codecov
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: cache@1.*
+      with:
+        cachePaths: qemu-install
+        cacheKey: qemu-mips64el-install-20220831-1
+
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cacheHit != 'true'
+      checkout: https://github.com/qemu/qemu.git
+      with:
+        pullType: COMMIT_ID
+        refName: 621da7789083b80d6f1ff1c0fb499334007b4f51
+        localPath: qemu
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: qemu
+      if: steps.cache-qemu.outputs.cacheHit != 'true'
+      run: |
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye main' | tee -a /etc/apt/sources.list
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye-updates main' | tee -a /etc/apt/sources.list
+        apt-get update
+        apt-get build-dep -y qemu
+        cd qemu
+        ./configure --prefix=${{ci.workspace}}/qemu-install --target-list=mips64el-linux-user --disable-system
+        make -j$(nproc)
+        make install
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/mipsisa64r6el-linux-gnuabi64.toolchain.cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_MSA=ON -DNCNN_MMI=OFF -DNCNN_OPENMP=${{matrix.OPENMP}} -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j $(nproc)
+    - name: test
+      run: |
+        export PATH=${{ci.workspace}}/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-mips64el TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/mipsisa64r6el-linux-gnuabi64" ctest --output-on-failure -j $(nproc)
+    - name: lcov-collect
+      run: |
+        cd build
+        lcov -d ./src -c -o lcov.info
+        lcov -r lcov.info '/usr/*' -o lcov.info
+        lcov -r lcov.info '*/build/*' -o lcov.info
+        lcov --list lcov.info
+    - name: codecov
+      run: ./codecov -t ${{settings.CODECOV_TOKEN.access_token}} -C ${{ ci.sha }} -B ${{ ci.head_ref }} -f build/lcov.info
+
+  linux-gcc-riscv64:
+    name: linux-gcc-riscv64
+    strategy:
+      matrix:
+        OPENMP: ['OFF', 'ON']
+
+    runs-on:
+      pool-name: docker
+      container:
+        image: bkci/ci:ubuntu
+    steps:
+    - name: checkout
+      checkout: self
+      with:
+        strategy: FRESH_CHECKOUT
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: install-deps
+      run: |
+        apt-get update
+        apt-get install -y lcov g++-riscv64-linux-gnu libcapstone4 libglib2.0-0
+        curl https://uploader.codecov.io/verification.gpg | gpg --no-default-keyring --keyring trustedkeys.gpg --import
+        curl -Os https://uploader.codecov.io/latest/linux/codecov
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM.sig
+        gpgv codecov.SHA256SUM.sig codecov.SHA256SUM
+        shasum -a 256 -c codecov.SHA256SUM
+        chmod +x codecov
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: cache@1.*
+      with:
+        cachePaths: qemu-install
+        cacheKey: qemu-riscv64-install-20230624-1
+
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cacheHit != 'true'
+      checkout: https://github.com/qemu/qemu.git
+      with:
+        pullType: COMMIT_ID
+        refName: b455ce4c2f300c8ba47cba7232dd03261368a4cb
+        localPath: qemu
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: qemu
+      if: steps.cache-qemu.outputs.cacheHit != 'true'
+      run: |
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye main' | tee -a /etc/apt/sources.list
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye-updates main' | tee -a /etc/apt/sources.list
+        apt-get update
+        apt-get build-dep -y qemu
+        apt-get install -y python3-pip
+        python3 -m pip install --upgrade pip
+        apt-get remove -y python3-setuptools
+        pip3 install -U setuptools
+        cd qemu
+        wget https://raw.githubusercontent.com/nihui/ncnn-assets/master/qemu-patches/0007-linux-user-Expose-risc-v-V-isa-bit-in-get_elf_hwcap.patch
+        patch -p1 -i 0007-linux-user-Expose-risc-v-V-isa-bit-in-get_elf_hwcap.patch
+        ./configure --prefix=${{ci.workspace}}/qemu-install --target-list=riscv64-linux-user --disable-system
+        make -j$(nproc)
+        make install
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/riscv64-linux-gnu.toolchain.cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_OPENMP=${{matrix.OPENMP}} -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j $(nproc)
+    - name: test
+      run: |
+        export PATH=${{ci.workspace}}/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-riscv64 TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/riscv64-linux-gnu" ctest --output-on-failure -j $(nproc)
+    - name: lcov-collect
+      run: |
+        cd build
+        lcov -d ./src -c -o lcov.info
+        lcov -r lcov.info '/usr/*' -o lcov.info
+        lcov -r lcov.info '*/build/*' -o lcov.info
+        lcov --list lcov.info
+    - name: codecov
+      run: ./codecov -t ${{settings.CODECOV_TOKEN.access_token}} -C ${{ ci.sha }} -B ${{ ci.head_ref }} -f build/lcov.info
+
+  linux-gcc-riscv64-rvv:
+    name: linux-gcc-riscv64-rvv
+    strategy:
+      matrix:
+        OPENMP: ['OFF', 'ON']
+
+    runs-on:
+      pool-name: docker
+      container:
+        image: bkci/ci:ubuntu
+    steps:
+    - name: checkout
+      checkout: self
+      with:
+        strategy: FRESH_CHECKOUT
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: install-deps
+      run: |
+        apt-get update
+        apt-get install -y lcov libcapstone4 libglib2.0-0
+        curl https://uploader.codecov.io/verification.gpg | gpg --no-default-keyring --keyring trustedkeys.gpg --import
+        curl -Os https://uploader.codecov.io/latest/linux/codecov
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM.sig
+        gpgv codecov.SHA256SUM.sig codecov.SHA256SUM
+        shasum -a 256 -c codecov.SHA256SUM
+        chmod +x codecov
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: cache@1.*
+      with:
+        cachePaths: qemu-install
+        cacheKey: qemu-riscv64-install-20230624-1
+
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cacheHit != 'true'
+      checkout: https://github.com/qemu/qemu.git
+      with:
+        pullType: COMMIT_ID
+        refName: b455ce4c2f300c8ba47cba7232dd03261368a4cb
+        localPath: qemu
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: qemu
+      if: steps.cache-qemu.outputs.cacheHit != 'true'
+      run: |
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye main' | tee -a /etc/apt/sources.list
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye-updates main' | tee -a /etc/apt/sources.list
+        apt-get update
+        apt-get build-dep -y qemu
+        apt-get install -y python3-pip
+        python3 -m pip install --upgrade pip
+        apt-get remove -y python3-setuptools
+        pip3 install -U setuptools
+        cd qemu
+        wget https://raw.githubusercontent.com/nihui/ncnn-assets/master/qemu-patches/0007-linux-user-Expose-risc-v-V-isa-bit-in-get_elf_hwcap.patch
+        patch -p1 -i 0007-linux-user-Expose-risc-v-V-isa-bit-in-get_elf_hwcap.patch
+        ./configure --prefix=${{ci.workspace}}/qemu-install --target-list=riscv64-linux-user --disable-system
+        make -j$(nproc)
+        make install
+
+    - name: cache-rv64gcv
+      id: cache-rv64gcv
+      uses: cache@1.*
+      with:
+        cachePaths: rv64gcv-install
+        cacheKey: rv64gcv-linux-install-20221029-1
+
+    - name: checkout-riscv-gnu-toolchain
+      if: steps.cache-rv64gcv.outputs.cacheHit != 'true'
+      checkout: https://github.com/riscv/riscv-gnu-toolchain.git
+      with:
+        pullType: COMMIT_ID
+        refName: da01ba455ce3802ffa84fdca3a089079996dbfc3
+        localPath: riscv-gnu-toolchain
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: riscv-gnu-toolchain
+      if: steps.cache-rv64gcv.outputs.cacheHit != 'true'
+      run: |
+        apt-get update
+        apt-get install -y autoconf automake autotools-dev curl python3 libmpc-dev libmpfr-dev libgmp-dev gawk build-essential bison flex texinfo gperf libtool patchutils bc zlib1g-dev libexpat-dev device-tree-compiler
+        cd riscv-gnu-toolchain
+        git submodule update --init --recursive --depth 1 glibc
+        git submodule update --init --recursive --depth 1 newlib
+        #git submodule update --init --recursive --depth 1 riscv-binutils
+        #git submodule update --init --recursive --depth 1 riscv-gcc
+        git submodule update --init --recursive --depth 1 riscv-dejagnu
+        git submodule update --init --recursive --depth 1 riscv-gdb
+        rm -rf riscv-binutils
+        git clone -b binutils-2_39-branch https://sourceware.org/git/binutils-gdb.git riscv-binutils
+        rm -rf riscv-gcc
+        git clone -b riscv-gcc-rvv-next https://github.com/riscv-collab/riscv-gcc.git riscv-gcc
+        cd riscv-gcc
+        git checkout 8a0c1b106f01c455a8fb478cfe52d859a69020fd
+        cd ..
+        sed -i '/__OBSOLETE_MATH/d' newlib/newlib/libm/common/math_errf.c
+        ./configure --prefix=${{ci.workspace}}/rv64gcv-install --with-arch=rv64gcv_zfh
+        make linux -j$(nproc)
+        find ${{ci.workspace}}/rv64gcv-install -type f | xargs -i strip -g {} || true
+
+    - name: build
+      run: |
+        export RISCV_ROOT_PATH=${{ci.workspace}}/rv64gcv-install
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/riscv64-unknown-linux-gnu.toolchain.cmake -DCMAKE_BUILD_TYPE=debug -DCMAKE_C_FLAGS="-O1" -DCMAKE_CXX_FLAGS="-O1" -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_RVV=ON -DNCNN_OPENMP=${{matrix.OPENMP}} -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j $(nproc)
+    - name: test-vlen128
+      run: |
+        export PATH=${{ci.workspace}}/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-riscv64 TESTS_EXECUTABLE_LOADER_ARGUMENTS="-cpu;rv64,v=true,Zfh=true,x-zvfh=true,vlen=128,elen=64,vext_spec=v1.0;-L;${{ci.workspace}}/rv64gcv-install/sysroot" ctest --output-on-failure -j $(nproc)
+    - name: lcov-collect-vlen128
+      run: |
+        cd build
+        lcov --gcov-tool ${{ci.workspace}}/rv64gcv-install/bin/riscv64-unknown-linux-gnu-gcov -d ./src -c -o lcov.info
+        lcov -r lcov.info '/usr/*' -o lcov.info
+        lcov -r lcov.info '*/build/*' -o lcov.info
+        lcov -r lcov.info '*/rv64gcv-install/*' -o lcov.info
+        lcov --list lcov.info
+    - name: codecov-vlen128
+      run: ./codecov -t ${{settings.CODECOV_TOKEN.access_token}} -C ${{ ci.sha }} -B ${{ ci.head_ref }} -f build/lcov.info
+    - name: test-vlen256
+      run: |
+        export PATH=${{ci.workspace}}/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-riscv64 TESTS_EXECUTABLE_LOADER_ARGUMENTS="-cpu;rv64,v=true,Zfh=true,x-zvfh=true,vlen=256,elen=64,vext_spec=v1.0;-L;${{ci.workspace}}/rv64gcv-install/sysroot" ctest --output-on-failure -j $(nproc)
+    - name: lcov-collect-vlen256
+      run: |
+        cd build
+        lcov --gcov-tool ${{ci.workspace}}/rv64gcv-install/bin/riscv64-unknown-linux-gnu-gcov -d ./src -c -o lcov.info
+        lcov -r lcov.info '/usr/*' -o lcov.info
+        lcov -r lcov.info '*/build/*' -o lcov.info
+        lcov -r lcov.info '*/rv64gcv-install/*' -o lcov.info
+        lcov --list lcov.info
+    - name: codecov-vlen256
+      run: ./codecov -t ${{settings.CODECOV_TOKEN.access_token}} -C ${{ ci.sha }} -B ${{ ci.head_ref }} -f build/lcov.info
+
+  linux-gcc-loongarch64:
+    name: linux-gcc-loongarch64
+    strategy:
+      matrix:
+        OPENMP: ['OFF', 'ON']
+
+    runs-on:
+      pool-name: docker
+      container:
+        image: bkci/ci:ubuntu
+    steps:
+    - name: checkout
+      checkout: self
+      with:
+        strategy: FRESH_CHECKOUT
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: install-deps
+      run: |
+        apt-get update
+        apt-get install -y lcov libcapstone4 libglib2.0-0 python3-venv
+        curl https://uploader.codecov.io/verification.gpg | gpg --no-default-keyring --keyring trustedkeys.gpg --import
+        curl -Os https://uploader.codecov.io/latest/linux/codecov
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM.sig
+        gpgv codecov.SHA256SUM.sig codecov.SHA256SUM
+        shasum -a 256 -c codecov.SHA256SUM
+        chmod +x codecov
+        wget https://github.com/sunhaiyong1978/CLFS-for-LoongArch/releases/download/8.0/loongarch64-clfs-8.0-cross-tools-gcc-full.tar.xz
+        tar -xf loongarch64-clfs-8.0-cross-tools-gcc-full.tar.xz
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: cache@1.*
+      with:
+        cachePaths: qemu-install
+        cacheKey: qemu-loongarch64-install-20230524-1
+
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cacheHit != 'true'
+      checkout: https://github.com/qemu/qemu.git
+      with:
+        pullType: COMMIT_ID
+        refName: 1c12355b31046a6b35a4f50c85c4f01afb1bd728
+        localPath: qemu
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: qemu
+      if: steps.cache-qemu.outputs.cacheHit != 'true'
+      run: |
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye main' | tee -a /etc/apt/sources.list
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye-updates main' | tee -a /etc/apt/sources.list
+        apt-get update
+        apt-get build-dep -y qemu
+        cd qemu
+        wget https://github.com/nihui/ncnn-assets/raw/master/qemu-patches/0001-linux-user-Expose-loongarch-lsx-isa-bit-in-get_elf_hwcap.patch
+        patch -p1 -i 0001-linux-user-Expose-loongarch-lsx-isa-bit-in-get_elf_hwcap.patch
+        ./configure --prefix=${{ci.workspace}}/qemu-install --target-list=loongarch64-linux-user --disable-system
+        make -j$(nproc)
+        make install
+
+    - name: build
+      run: |
+        export LOONGARCH64_ROOT_PATH=${{ci.workspace}}/cross-tools
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/loongarch64-unknown-linux-gnu.toolchain.cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_LSX=ON -DNCNN_LASX=OFF -DNCNN_OPENMP=${{matrix.OPENMP}} -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j $(nproc)
+    - name: test
+      run: |
+        export PATH=${{ci.workspace}}/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-loongarch64 TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;${{ci.workspace}}/cross-tools/target" ctest --output-on-failure -j $(nproc)
+    - name: lcov-collect
+      run: |
+        cd build
+        lcov --gcov-tool ${{ci.workspace}}/cross-tools/bin/loongarch64-unknown-linux-gnu-gcov -d ./src -c -o lcov.info
+        lcov -r lcov.info '/usr/*' -o lcov.info
+        lcov -r lcov.info '*/build/*' -o lcov.info
+        lcov -r lcov.info '*/cross-tools/*' -o lcov.info
+        lcov --list lcov.info
+    - name: codecov
+      run: ./codecov -t ${{settings.CODECOV_TOKEN.access_token}} -C ${{ ci.sha }} -B ${{ ci.head_ref }} -f build/lcov.info

.clang-format

@@ -0,0 +1,132 @@
+# find src/ tools/ tests/ examples/ benchmark/ -type f -name '*.c' -o -name '*.cpp' -o -name '*.h' | xargs -i clang-format -i {}
+
+# need clang-format >= 10.0
+
+AccessModifierOffset: -4
+AlignAfterOpenBracket: Align
+AlignConsecutiveAssignments: false
+# AlignConsecutiveBitFields: true
+AlignConsecutiveDeclarations: false
+AlignConsecutiveMacros: true
+AlignEscapedNewlines: Left
+# AlignOperands: AlignAfterOperator
+AlignTrailingComments: true
+AllowAllArgumentsOnNextLine: true
+AllowAllConstructorInitializersOnNextLine: true
+AllowAllParametersOfDeclarationOnNextLine: true
+AllowShortBlocksOnASingleLine: Always
+AllowShortCaseLabelsOnASingleLine: true
+# AllowShortEnumsOnASingleLine: true
+AllowShortFunctionsOnASingleLine: None
+AllowShortIfStatementsOnASingleLine: WithoutElse
+AllowShortLambdasOnASingleLine: All
+AllowShortLoopsOnASingleLine: true
+AlwaysBreakAfterReturnType: None
+AlwaysBreakBeforeMultilineStrings: false
+AlwaysBreakTemplateDeclarations: Yes
+BinPackArguments: true
+BinPackParameters: true
+BraceWrapping:
+  AfterCaseLabel: true
+  AfterClass: true
+  AfterControlStatement: Always
+  AfterEnum: true
+  AfterFunction: true
+  AfterNamespace: false
+  AfterObjCDeclaration: false
+  AfterStruct: true
+  AfterUnion: true
+  AfterExternBlock: false
+  BeforeCatch: true
+  BeforeElse: true
+#  BeforeLambdaBody: false
+#  BeforeWhile: false
+  IndentBraces: false
+  SplitEmptyFunction: true
+  SplitEmptyRecord: true
+  SplitEmptyNamespace: false
+BreakAfterJavaFieldAnnotations: true
+BreakBeforeBinaryOperators: All
+BreakBeforeBraces: Custom
+BreakBeforeTernaryOperators: true
+BreakConstructorInitializers: BeforeColon
+BreakInheritanceList: BeforeColon
+BreakStringLiterals: false
+ColumnLimit: 0
+# CommentPragmas:
+CompactNamespaces: false
+ConstructorInitializerAllOnOneLineOrOnePerLine: true
+ConstructorInitializerIndentWidth: 4
+ContinuationIndentWidth: 4
+Cpp11BracedListStyle: true
+DeriveLineEnding: false
+DerivePointerAlignment: false
+# DisableFormat:
+# ExperimentalAutoDetectBinPacking:
+FixNamespaceComments: true
+# ForEachMacros:
+IncludeBlocks: Regroup
+# IncludeCategories:
+# IncludeIsMainRegex:
+# IncludeIsMainSourceRegex:
+# IndentCaseBlocks: false
+IndentCaseLabels: false
+# IndentExternBlock: NoIndent
+IndentGotoLabels: false
+IndentPPDirectives: None
+IndentWidth: 4
+# IndentWrappedFunctionNames: 4
+# InsertTrailingCommas: None
+# JavaImportGroups:
+# JavaScriptQuotes
+# JavaScriptWrapImports:
+KeepEmptyLinesAtTheStartOfBlocks: false
+Language: Cpp
+# MacroBlockBegin:
+# MacroBlockEnd:
+MaxEmptyLinesToKeep: 1
+NamespaceIndentation: None
+# NamespaceMacros:
+# ObjCBinPackProtocolList:
+# ObjCBlockIndentWidth:
+# ObjCBreakBeforeNestedBlockParam:
+# ObjCSpaceAfterProperty:
+# ObjCSpaceBeforeProtocolList:
+# PenaltyBreakAssignment:
+# PenaltyBreakBeforeFirstCallParameter:
+# PenaltyBreakComment:
+# PenaltyBreakFirstLessLess:
+# PenaltyBreakString:
+# PenaltyBreakTemplateDeclaration:
+# PenaltyExcessCharacter:
+# PenaltyReturnTypeOnItsOwnLine:
+PointerAlignment: Left
+# RawStringFormats:
+ReflowComments: false
+SortIncludes: false
+SortUsingDeclarations: true
+SpaceAfterCStyleCast: false
+SpaceAfterLogicalNot: false
+SpaceAfterTemplateKeyword: false
+SpaceBeforeAssignmentOperators: true
+SpaceBeforeCpp11BracedList: false
+SpaceBeforeCtorInitializerColon: true
+SpaceBeforeInheritanceColon: true
+SpaceBeforeParens: ControlStatements
+SpaceBeforeRangeBasedForLoopColon: true
+SpaceBeforeSquareBrackets: false
+SpaceInEmptyBlock: false
+SpaceInEmptyParentheses: false
+SpacesBeforeTrailingComments: 1
+SpacesInAngles: false
+SpacesInCStyleCastParentheses: false
+SpacesInConditionalStatement: false
+SpacesInContainerLiterals: false
+SpacesInParentheses: false
+SpacesInSquareBrackets: false
+Standard: c++03
+#StatementMacros:
+TabWidth: 4
+# TypenameMacros:
+UseCRLF: false
+UseTab: Never

.gitattributes

@@ -1,35 +1,47 @@
-*.7z filter=lfs diff=lfs merge=lfs -text
-*.arrow filter=lfs diff=lfs merge=lfs -text
-*.bin filter=lfs diff=lfs merge=lfs -text
-*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
-*.ftz filter=lfs diff=lfs merge=lfs -text
-*.gz filter=lfs diff=lfs merge=lfs -text
-*.h5 filter=lfs diff=lfs merge=lfs -text
-*.joblib filter=lfs diff=lfs merge=lfs -text
-*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
-*.onnx filter=lfs diff=lfs merge=lfs -text
-*.ot filter=lfs diff=lfs merge=lfs -text
-*.parquet filter=lfs diff=lfs merge=lfs -text
-*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
-*.pt filter=lfs diff=lfs merge=lfs -text
-*.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* filter=lfs diff=lfs merge=lfs -text
-*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text
-*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
-*.xz filter=lfs diff=lfs merge=lfs -text
-*.zip filter=lfs diff=lfs merge=lfs -text
-*.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text
+*.comp linguist-language=GLSL
+build/CMakeFiles/3.27.4/CMakeDetermineCompilerABI_C.bin filter=lfs diff=lfs merge=lfs -text
+build/CMakeFiles/3.27.4/CMakeDetermineCompilerABI_CXX.bin filter=lfs diff=lfs merge=lfs -text
+build/benchmark/benchncnn filter=lfs diff=lfs merge=lfs -text
+build/examples/fasterrcnn filter=lfs diff=lfs merge=lfs -text
+build/examples/mobilenetssd filter=lfs diff=lfs merge=lfs -text
+build/examples/mobilenetv2ssdlite filter=lfs diff=lfs merge=lfs -text
+build/examples/nanodet filter=lfs diff=lfs merge=lfs -text
+build/examples/nanodetplus_pnnx filter=lfs diff=lfs merge=lfs -text
+build/examples/p2pnet filter=lfs diff=lfs merge=lfs -text
+build/examples/peleenetssd_seg filter=lfs diff=lfs merge=lfs -text
+build/examples/retinaface filter=lfs diff=lfs merge=lfs -text
+build/examples/rfcn filter=lfs diff=lfs merge=lfs -text
+build/examples/rvm filter=lfs diff=lfs merge=lfs -text
+build/examples/scrfd filter=lfs diff=lfs merge=lfs -text
+build/examples/scrfd_crowdhuman filter=lfs diff=lfs merge=lfs -text
+build/examples/shufflenetv2 filter=lfs diff=lfs merge=lfs -text
+build/examples/simplepose filter=lfs diff=lfs merge=lfs -text
+build/examples/squeezenet filter=lfs diff=lfs merge=lfs -text
+build/examples/squeezenet_c_api filter=lfs diff=lfs merge=lfs -text
+build/examples/squeezenetssd filter=lfs diff=lfs merge=lfs -text
+build/examples/yolact filter=lfs diff=lfs merge=lfs -text
+build/examples/yolov2 filter=lfs diff=lfs merge=lfs -text
+build/examples/yolov3 filter=lfs diff=lfs merge=lfs -text
+build/examples/yolov4 filter=lfs diff=lfs merge=lfs -text
+build/examples/yolov5 filter=lfs diff=lfs merge=lfs -text
+build/examples/yolov5_pnnx filter=lfs diff=lfs merge=lfs -text
+build/examples/yolov7 filter=lfs diff=lfs merge=lfs -text
+build/examples/yolov7_pnnx filter=lfs diff=lfs merge=lfs -text
+build/examples/yolox filter=lfs diff=lfs merge=lfs -text
+build/install/bin/caffe2ncnn filter=lfs diff=lfs merge=lfs -text
+build/install/bin/ncnn2int8 filter=lfs diff=lfs merge=lfs -text
+build/install/bin/ncnn2mem filter=lfs diff=lfs merge=lfs -text
+build/install/bin/ncnn2table filter=lfs diff=lfs merge=lfs -text
+build/install/bin/ncnnoptimize filter=lfs diff=lfs merge=lfs -text
+build/install/lib/libncnn.a filter=lfs diff=lfs merge=lfs -text
+build/src/CMakeFiles/ncnn.dir/layer.cpp.o filter=lfs diff=lfs merge=lfs -text
+build/src/libncnn.a filter=lfs diff=lfs merge=lfs -text
+build/tools/caffe/CMakeFiles/caffe2ncnn.dir/caffe.pb.cc.o filter=lfs diff=lfs merge=lfs -text
+build/tools/caffe/caffe2ncnn filter=lfs diff=lfs merge=lfs -text
+build/tools/ncnn2mem filter=lfs diff=lfs merge=lfs -text
+build/tools/ncnnoptimize filter=lfs diff=lfs merge=lfs -text
+build/tools/quantize/ncnn2int8 filter=lfs diff=lfs merge=lfs -text
+build/tools/quantize/ncnn2table filter=lfs diff=lfs merge=lfs -text
+examples/squeezenet_v1.1.bin filter=lfs diff=lfs merge=lfs -text
+examples/squeezenet_v1.1.caffemodel filter=lfs diff=lfs merge=lfs -text
+examples/squeezenet_v1.1.param.bin filter=lfs diff=lfs merge=lfs -text

.github/ISSUE_TEMPLATE/bug.md

@@ -0,0 +1,15 @@
+---
+name: "\U0001F41B bug issue"
+about: submit a bug report +_+
+---
+
+## error log | 日志或报错信息 | ログ
+
+## context | 编译/运行环境 | バックグラウンド
+
+## how to reproduce | 复现步骤 | 再現方法
+1.
+2.
+3.
+
+## more | 其他 | その他

.github/ISSUE_TEMPLATE/model-convert.md

@@ -0,0 +1,14 @@
+---
+name: "\U0001F6B8 model convert issue"
+about: "Life is Short, Use pnnx and convertmodel.com"
+---
+
+## error log | 日志或报错信息 | ログ
+
+## model | 模型 | モデル
+1. original model
+
+## how to reproduce | 复现步骤 | 再現方法
+1.
+2.
+3.

.github/ISSUE_TEMPLATE/others.md

@@ -0,0 +1,6 @@
+---
+name: "\U0001F4DD others"
+about: discussion, suggestion and question
+---
+
+## detail | 详细描述 | 詳細な説明

.github/ISSUE_TEMPLATE/quantization.md

@@ -0,0 +1,13 @@
+---
+name: "\U0001F4C8 quantization"
+about: best wishes for your low bit quantization has a low accuracy loss...\(^▽^)/...2333... 
+---
+
+## expectation | 诉求 | 期待する
+1. speed 
+2. precision
+
+## model | 模型 | モデル
+1. model.param and model.bin
+
+## detail | 详细描述 | 詳細な説明

.github/dependabot.yml

@@ -0,0 +1,6 @@
+version: 2
+updates:
+  - package-ecosystem: "github-actions"
+    directory: "/"
+    schedule:
+      interval: "daily"

.github/workflows/android-armv7-cpu.yml

@@ -0,0 +1,56 @@
+name: android-armv7-cpu
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/android-armv7-cpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/android-armv7-cpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+concurrency:
+  group: android-armv7-cpu-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  android-armv7:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="armeabi-v7a" -DANDROID_ARM_NEON=ON -DANDROID_PLATFORM=android-14 ..
+        cmake --build . -j 2
+    - name: build-shared
+      run: |
+        mkdir build-shared && cd build-shared
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="armeabi-v7a" -DANDROID_ARM_NEON=ON -DANDROID_PLATFORM=android-14 -DNCNN_SHARED_LIB=ON -DNCNN_ENABLE_LTO=OFF ..
+        cmake --build . -j 2
+
+    - name: ndk-r16b
+      run: |
+        wget https://dl.google.com/android/repository/android-ndk-r16b-linux-x86_64.zip -O $GITHUB_WORKSPACE/android-ndk-r16b-linux-x86_64.zip
+        cd $GITHUB_WORKSPACE && unzip -q android-ndk-r16b-linux-x86_64.zip
+    - name: build-noneon
+      run: |
+        mkdir build-noneon && cd build-noneon
+        cmake -DCMAKE_TOOLCHAIN_FILE=$GITHUB_WORKSPACE/android-ndk-r16b/build/cmake/android.toolchain.cmake -DANDROID_ABI="armeabi-v7a" -DANDROID_ARM_NEON=OFF -DANDROID_PLATFORM=android-14 ..
+        cmake --build . -j 2
+    - name: build-noneon-shared
+      run: |
+        mkdir build-noneon-shared && cd build-noneon-shared
+        cmake -DCMAKE_TOOLCHAIN_FILE=$GITHUB_WORKSPACE/android-ndk-r16b/build/cmake/android.toolchain.cmake -DANDROID_ABI="armeabi-v7a" -DANDROID_ARM_NEON=OFF -DANDROID_PLATFORM=android-14 -DNCNN_SHARED_LIB=ON -DNCNN_ENABLE_LTO=OFF ..
+        cmake --build . -j 2

.github/workflows/android-armv7-gpu.yml

@@ -0,0 +1,45 @@
+name: android-armv7-gpu
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/android-armv7-gpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+    - 'src/layer/vulkan/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/android-armv7-gpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+    - 'src/layer/vulkan/**'
+concurrency:
+  group: android-armv7-gpu-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  android-armv7-gpu:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v4
+      with:
+        submodules: true
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="armeabi-v7a" -DANDROID_ARM_NEON=ON -DANDROID_PLATFORM=android-24 -DNCNN_VULKAN=ON ..
+        cmake --build . -j 2
+    - name: build-shared
+      run: |
+        mkdir build-shared && cd build-shared
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="armeabi-v7a" -DANDROID_ARM_NEON=ON -DANDROID_PLATFORM=android-24 -DNCNN_VULKAN=ON -DNCNN_SHARED_LIB=ON -DNCNN_ENABLE_LTO=OFF ..
+        cmake --build . -j 2

.github/workflows/android-armv8-cpu.yml

@@ -0,0 +1,41 @@
+name: android-armv8-cpu
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/android-armv8-cpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/android-armv8-cpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+concurrency:
+  group: android-armv8-cpu-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  android-aarch64:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v4
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="arm64-v8a" -DANDROID_PLATFORM=android-21 ..
+        cmake --build . -j 2
+    - name: build-shared
+      run: |
+        mkdir build-shared && cd build-shared
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="arm64-v8a" -DANDROID_PLATFORM=android-21 -DNCNN_SHARED_LIB=ON ..
+        cmake --build . -j 2

.github/workflows/android-armv8-gpu.yml

@@ -0,0 +1,75 @@
+name: android-armv8-gpu
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/android-armv8-gpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+    - 'src/layer/vulkan/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/android-armv8-gpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+    - 'src/layer/vulkan/**'
+concurrency:
+  group: android-armv8-gpu-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  android-aarch64-gpu:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v4
+      with:
+        submodules: true
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="arm64-v8a" -DANDROID_PLATFORM=android-24 -DNCNN_VULKAN=ON ..
+        cmake --build . -j 2
+    - name: build-shared
+      run: |
+        mkdir build-shared && cd build-shared
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="arm64-v8a" -DANDROID_PLATFORM=android-24 -DNCNN_VULKAN=ON -DNCNN_SHARED_LIB=ON ..
+        cmake --build . -j 2
+    - name: build-termux
+      run: |
+        mkdir build-termux && cd build-termux
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="arm64-v8a" -DANDROID_PLATFORM=android-24 -DNCNN_VULKAN=ON -DNCNN_PLATFORM_API=OFF ..
+        cmake --build . -j 2
+    - name: build-android-29
+      run: |
+        mkdir build-android-29 && cd build-android-29
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="arm64-v8a" -DANDROID_PLATFORM=android-29 -DNCNN_VULKAN=ON ..
+        cmake --build . -j 2
+    - name: build-android-29-shared
+      run: |
+        mkdir build-android-29-shared && cd build-android-29-shared
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="arm64-v8a" -DANDROID_PLATFORM=android-29 -DNCNN_VULKAN=ON -DNCNN_SHARED_LIB=ON ..
+        cmake --build . -j 2
+
+  android-aarch64-gpu-ndk-r16b:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+      with:
+        submodules: true
+    - name: ndk-r16b
+      run: |
+        wget https://dl.google.com/android/repository/android-ndk-r16b-linux-x86_64.zip -O $GITHUB_WORKSPACE/android-ndk-r16b-linux-x86_64.zip
+        cd $GITHUB_WORKSPACE && unzip -q android-ndk-r16b-linux-x86_64.zip
+    - name: configure
+      run: mkdir build && cd build && cmake -DCMAKE_TOOLCHAIN_FILE=$GITHUB_WORKSPACE/android-ndk-r16b/build/cmake/android.toolchain.cmake -DANDROID_ABI="arm64-v8a" -DANDROID_PLATFORM=android-24 -DNCNN_VULKAN=ON ..
+    - name: build
+      run: cmake --build build -j 2

.github/workflows/android-x64-cpu.yml

@@ -0,0 +1,41 @@
+name: android-x64-cpu
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/android-x64-cpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/android-x64-cpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+concurrency:
+  group: android-x64-cpu-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  android-x86_64:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v4
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="x86_64" -DANDROID_PLATFORM=android-21 ..
+        cmake --build . -j 2
+    - name: build-shared
+      run: |
+        mkdir build-shared && cd build-shared
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="x86_64" -DANDROID_PLATFORM=android-21 -DNCNN_SHARED_LIB=ON ..
+        cmake --build . -j 2

.github/workflows/android-x64-gpu.yml

@@ -0,0 +1,45 @@
+name: android-x64-gpu
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/android-x64-gpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'src/layer/vulkan/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/android-x64-gpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'src/layer/vulkan/**'
+concurrency:
+  group: android-x64-gpu-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  android-x86_64-gpu:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v4
+      with:
+        submodules: true
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="x86_64" -DANDROID_PLATFORM=android-24 -DNCNN_VULKAN=ON ..
+        cmake --build . -j 2
+    - name: build-shared
+      run: |
+        mkdir build-shared && cd build-shared
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="x86_64" -DANDROID_PLATFORM=android-24 -DNCNN_VULKAN=ON -DNCNN_SHARED_LIB=ON ..
+        cmake --build . -j 2

.github/workflows/android-x86-cpu.yml

@@ -0,0 +1,41 @@
+name: android-x86-cpu
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/android-x86-cpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/android-x86-cpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+concurrency:
+  group: android-x86-cpu-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  android-x86:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v4
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="x86"  -DANDROID_PLATFORM=android-14 ..
+        cmake --build . -j 2
+    - name: build-shared
+      run: |
+        mkdir build-shared && cd build-shared
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="x86"  -DANDROID_PLATFORM=android-14 -DNCNN_SHARED_LIB=ON ..
+        cmake --build . -j 2

.github/workflows/android-x86-gpu.yml

@@ -0,0 +1,45 @@
+name: android-x86-gpu
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/android-x86-gpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'src/layer/vulkan/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/android-x86-gpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'src/layer/vulkan/**'
+concurrency:
+  group: android-x86-gpu-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  android-x86-gpu:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v4
+      with:
+        submodules: true
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="x86" -DANDROID_PLATFORM=android-24 -DNCNN_VULKAN=ON ..
+        cmake --build . -j 2
+    - name: build-shared
+      run: |
+        mkdir build-shared && cd build-shared
+        cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake -DANDROID_ABI="x86" -DANDROID_PLATFORM=android-24 -DNCNN_VULKAN=ON -DNCNN_SHARED_LIB=ON ..
+        cmake --build . -j 2

.github/workflows/code-format.yml

@@ -0,0 +1,61 @@
+name: code-format
+
+on: [push, pull_request, pull_request_target]
+
+concurrency:
+  group: code-format-${{ github.ref }}
+  cancel-in-progress: true
+
+permissions:
+  contents: read
+
+jobs:
+  code-format:
+    permissions:
+      contents: write  # for stefanzweifel/git-auto-commit-action to push code in repo
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-clang-format
+      id: cache-clang-format
+      uses: actions/cache@v3
+      with:
+        path: clang-format-install
+        key: clang-format-install-4
+    - name: clang-format
+      if: steps.cache-clang-format.outputs.cache-hit != 'true'
+      run: |
+        wget https://github.com/llvm/llvm-project/releases/download/llvmorg-10.0.1/llvm-project-10.0.1.tar.xz
+        tar -xf llvm-project-10.0.1.tar.xz
+        cd llvm-project-10.0.1
+        mkdir build
+        cd build
+        cmake -DCMAKE_INSTALL_PREFIX=install -DCMAKE_BUILD_TYPE=Release -DBUILD_SHARED_LIBS=OFF -DLLVM_ENABLE_PROJECTS="clang" -DLLVM_TARGETS_TO_BUILD="" -DLLVM_INCLUDE_EXAMPLES=OFF -DLLVM_INCLUDE_TESTS=OFF -DLLVM_INCLUDE_DOCS=OFF ../llvm/
+        make -j2 clang-format
+        mkdir $GITHUB_WORKSPACE/clang-format-install
+        cp -r bin/clang-format $GITHUB_WORKSPACE/clang-format-install
+        cd ../../
+        rm -rf llvm-project-10.0.1
+        rm llvm-project-10.0.1.tar.xz
+
+    - name: astyle
+      run: |
+        sudo apt-get update
+        sudo apt-get install -y astyle
+
+    - name: code-format
+      run: |
+        mkdir -p ~/bin
+        mv $GITHUB_WORKSPACE/clang-format-install/clang-format ~/bin/clang-format
+        rm -rf $GITHUB_WORKSPACE/clang-format-install
+        export PATH=~/bin:$PATH
+        sh codeformat.sh
+    - uses: stefanzweifel/git-auto-commit-action@v4
+      with:
+        commit_message: apply code-format changes
+
+    - name: restore-clang-format-cache
+      run: |
+        mkdir $GITHUB_WORKSPACE/clang-format-install
+        cp -r ~/bin/clang-format $GITHUB_WORKSPACE/clang-format-install

.github/workflows/codeql-analysis.yml

@@ -0,0 +1,84 @@
+# For most projects, this workflow file will not need changing; you simply need
+# to commit it to your repository.
+#
+# You may wish to alter this file to override the set of languages analyzed,
+# or to provide custom queries or build logic.
+name: "CodeQL"
+
+on:
+  push:
+    branches: [master]
+    paths-ignore: ['**.md']
+  pull_request:
+    # The branches below must be a subset of the branches above
+    branches: [master]
+    paths-ignore: ['**.md']
+  schedule:
+    - cron: '0 20 * * 4'
+
+concurrency:
+  group: CodeQL-${{ github.ref }}
+  cancel-in-progress: true
+
+permissions:
+  contents: read
+
+jobs:
+  analyze:
+    permissions:
+      actions: read  # for github/codeql-action/init to get workflow details
+      contents: read  # for actions/checkout to fetch code
+      security-events: write  # for github/codeql-action/autobuild to send a status report
+    name: Analyze
+    runs-on: ubuntu-latest
+
+    strategy:
+      fail-fast: false
+      matrix:
+        # Override automatic language detection by changing the below list
+        # Supported options are ['csharp', 'cpp', 'go', 'java', 'javascript', 'python']
+        language: ['cpp']
+        # Learn more...
+        # https://docs.github.com/en/github/finding-security-vulnerabilities-and-errors-in-your-code/configuring-code-scanning#overriding-automatic-language-detection
+
+    steps:
+    - name: Checkout repository
+      uses: actions/checkout@v4
+      with:
+        # We must fetch at least the immediate parents so that if this is
+        # a pull request then we can checkout the head.
+        fetch-depth: 2
+
+    # If this run was triggered by a pull request event, then checkout
+    # the head of the pull request instead of the merge commit.
+    - run: git checkout HEAD^2
+      if: ${{ github.event_name == 'pull_request' }}
+
+    # Initializes the CodeQL tools for scanning.
+    - name: Initialize CodeQL
+      uses: github/codeql-action/init@v2
+      with:
+        languages: ${{ matrix.language }}
+        # If you wish to specify custom queries, you can do so here or in a config file.
+        # By default, queries listed here will override any specified in a config file. 
+        # Prefix the list here with "+" to use these queries and those in the config file.
+        # queries: ./path/to/local/query, your-org/your-repo/queries@main
+
+    # Autobuild attempts to build any compiled languages  (C/C++, C#, or Java).
+    # If this step fails, then you should remove it and run the build manually (see below)
+    - name: Autobuild
+      uses: github/codeql-action/autobuild@v2
+
+    # ℹ️ Command-line programs to run using the OS shell.
+    # 📚 https://git.io/JvXDl
+
+    # ✏️ If the Autobuild fails above, remove it and uncomment the following three lines
+    #    and modify them (or add more) to build your code if your project
+    #    uses a compiled language
+
+    #- run: |
+    #   make bootstrap
+    #   make release
+
+    - name: Perform CodeQL Analysis
+      uses: github/codeql-action/analyze@v2

.github/workflows/elf-riscv32-cpu-gcc.yml

@@ -0,0 +1,126 @@
+name: elf-riscv32-cpu-gcc
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/elf-riscv32-cpu-gcc.yml'
+    - 'toolchains/riscv32-unknown-elf.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/riscv/**'
+    - 'tests/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/elf-riscv32-cpu-gcc.yml'
+    - 'toolchains/riscv32-unknown-elf.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/riscv/**'
+    - 'tests/**'
+concurrency:
+  group: elf-riscv32-cpu-gcc-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  newlib-rv32imc-gcc:
+    runs-on: [self-hosted, linux, centos]
+    steps:
+    - uses: actions/checkout@v4
+
+    #- name: cache-riscv
+      #id: cache-riscv
+      #uses: actions/cache@v3
+      #with:
+        #path: rv32imc-install
+        #key: rv32imc-newlib-install-20210425
+
+    #- name: install-riscv-build-deps
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: |
+        #sudo apt-get update
+        #sudo apt-get install autoconf automake autotools-dev curl python3 libmpc-dev libmpfr-dev libgmp-dev gawk build-essential bison flex texinfo gperf libtool patchutils bc zlib1g-dev libexpat-dev device-tree-compiler
+
+    #- name: checkout-riscv-gnu-toolchain
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #uses: actions/checkout@v4
+      #with:
+        #repository: riscv/riscv-gnu-toolchain
+        #path: riscv-gnu-toolchain
+        #ref: b715e4f01b43efef487166f75d5d85d3c33fa7ef
+    #- name: checkout-riscv-gnu-toolchain-submodules
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: |
+        #cd riscv-gnu-toolchain
+        #git submodule update --init --recursive --depth 1 riscv-binutils
+        #git submodule update --init --recursive --depth 1 riscv-gcc
+        #git submodule update --init --recursive --depth 1 riscv-glibc
+        #git submodule update --init --recursive --depth 1 riscv-dejagnu
+        #git submodule update --init --recursive --depth 1 riscv-newlib
+        #git submodule update --init --recursive --depth 1 riscv-gdb
+    #- name: riscv-gnu-toolchain
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: |
+        #cd riscv-gnu-toolchain
+        #./configure --prefix=$GITHUB_WORKSPACE/rv32imc-install --with-arch=rv32imc
+        #make -j2
+
+    #- name: checkout-riscv-pk
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #uses: actions/checkout@v4
+      #with:
+        #repository: riscv/riscv-pk
+        #path: riscv-pk
+        #ref: ef7bebaf9bf24d3e90bcaae96387ce418e136b6d
+    #- name: riscv-pk
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: |
+        #cd riscv-pk
+        #mkdir build
+        #cd build
+        #export PATH=$GITHUB_WORKSPACE/rv32imc-install/bin:$PATH
+        #../configure --prefix=$GITHUB_WORKSPACE/rv32imc-install --with-arch=rv32imc --host=riscv32-unknown-elf
+        #make -j2
+        #make install
+
+    #- name: checkout-riscv-isa-sim
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #uses: actions/checkout@v4
+      #with:
+        #repository: riscv/riscv-isa-sim
+        #path: riscv-isa-sim
+        #ref: 9d4f45c2ebf105503974fc80a42590ca1584c354
+    #- name: riscv-isa-sim
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: |
+        #cd riscv-isa-sim
+        #mkdir build
+        #cd build
+        #export PATH=$GITHUB_WORKSPACE/rv32imc-install/bin:$PATH
+        #../configure --prefix=$GITHUB_WORKSPACE/rv32imc-install --with-isa=rv32imc
+        #make -j2
+        #make install
+
+    #- name: riscv-strip-install
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: find $GITHUB_WORKSPACE/rv32imc-install -type f | xargs -i strip -g {} || true
+
+    - name: configure
+      run: export RISCV_ROOT_PATH=/data/action/osd/rv32imc-install && mkdir build && cd build && cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/riscv32-unknown-elf.toolchain.cmake -DNCNN_THREADS=OFF -DNCNN_OPENMP=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+    - name: build
+      run: cmake --build build -j 4
+
+    # too slow for softfloat arch :(
+    #- name: test
+      #run: |
+        #sudo apt-get update
+        #sudo apt-get install device-tree-compiler
+        #export PATH=/data/action/osd/rv32imc-install/bin:$PATH
+        #cd build
+        #TESTS_EXECUTABLE_LOADER=spike TESTS_EXECUTABLE_LOADER_ARGUMENTS=/data/action/osd/rv32imc-install/riscv32-unknown-elf/bin/pk ctest --output-on-failure -j 2

.github/workflows/elf-riscv64-cpu-gcc.yml

@@ -0,0 +1,123 @@
+name: elf-riscv64-cpu-gcc
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/elf-riscv64-cpu-gcc.yml'
+    - 'toolchains/riscv64-unknown-elf.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/riscv/**'
+    - 'tests/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/elf-riscv64-cpu-gcc.yml'
+    - 'toolchains/riscv64-unknown-elf.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/riscv/**'
+    - 'tests/**'
+concurrency:
+  group: elf-riscv64-cpu-gcc-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  newlib-rv64gc-gcc:
+    runs-on: [self-hosted, linux, centos]
+    steps:
+    - uses: actions/checkout@v4
+
+    #- name: cache-riscv
+      #id: cache-riscv
+      #uses: actions/cache@v3
+      #with:
+        #path: rv64gc-install
+        #key: rv64gc-newlib-install-20210425
+
+    #- name: install-riscv-build-deps
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: |
+        #sudo apt-get update
+        #sudo apt-get install autoconf automake autotools-dev curl python3 libmpc-dev libmpfr-dev libgmp-dev gawk build-essential bison flex texinfo gperf libtool patchutils bc zlib1g-dev libexpat-dev device-tree-compiler
+
+    #- name: checkout-riscv-gnu-toolchain
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #uses: actions/checkout@v4
+      #with:
+        #repository: riscv/riscv-gnu-toolchain
+        #path: riscv-gnu-toolchain
+        #ref: b715e4f01b43efef487166f75d5d85d3c33fa7ef
+    #- name: checkout-riscv-gnu-toolchain-submodules
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: |
+        #cd riscv-gnu-toolchain
+        #git submodule update --init --recursive --depth 1 riscv-binutils
+        #git submodule update --init --recursive --depth 1 riscv-gcc
+        #git submodule update --init --recursive --depth 1 riscv-glibc
+        #git submodule update --init --recursive --depth 1 riscv-dejagnu
+        #git submodule update --init --recursive --depth 1 riscv-newlib
+        #git submodule update --init --recursive --depth 1 riscv-gdb
+    #- name: riscv-gnu-toolchain
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: |
+        #cd riscv-gnu-toolchain
+        #sed -i '/__OBSOLETE_MATH/d' riscv-newlib/newlib/libm/common/math_errf.c
+        #./configure --prefix=$GITHUB_WORKSPACE/rv64gc-install --with-arch=rv64gc
+        #make -j2
+
+    #- name: checkout-riscv-pk
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #uses: actions/checkout@v4
+      #with:
+        #repository: riscv/riscv-pk
+        #path: riscv-pk
+        #ref: ef7bebaf9bf24d3e90bcaae96387ce418e136b6d
+    #- name: riscv-pk
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: |
+        #cd riscv-pk
+        #mkdir build
+        #cd build
+        #export PATH=$GITHUB_WORKSPACE/rv64gc-install/bin:$PATH
+        #../configure --prefix=$GITHUB_WORKSPACE/rv64gc-install --with-arch=rv64gc --host=riscv64-unknown-elf
+        #make -j2
+        #make install
+
+    #- name: checkout-riscv-isa-sim
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #uses: actions/checkout@v4
+      #with:
+        #repository: riscv/riscv-isa-sim
+        #path: riscv-isa-sim
+        #ref: 9d4f45c2ebf105503974fc80a42590ca1584c354
+    #- name: riscv-isa-sim
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: |
+        #cd riscv-isa-sim
+        #mkdir build
+        #cd build
+        #export PATH=$GITHUB_WORKSPACE/rv64gc-install/bin:$PATH
+        #../configure --prefix=$GITHUB_WORKSPACE/rv64gc-install --with-isa=rv64gc
+        #make -j2
+        #make install
+
+    #- name: riscv-strip-install
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: find $GITHUB_WORKSPACE/rv64gc-install -type f | xargs -i strip -g {} || true
+
+    - name: configure
+      run: export RISCV_ROOT_PATH=/data/action/osd/rv64gc-install && mkdir build && cd build && cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/riscv64-unknown-elf.toolchain.cmake -DNCNN_THREADS=OFF -DNCNN_OPENMP=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+    - name: build
+      run: cmake --build build -j 4
+    - name: test
+      run: |
+        export PATH=/data/action/osd/rv64gc-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=spike TESTS_EXECUTABLE_LOADER_ARGUMENTS=/data/action/osd/rv64gc-install/riscv64-unknown-elf/bin/pk ctest --output-on-failure -j 4

.github/workflows/ios-arm64-gpu.yml

@@ -0,0 +1,146 @@
+name: ios-arm64-gpu
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/ios-arm64-gpu.yml'
+    - 'toolchains/ios.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+    - 'src/layer/vulkan/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/ios-arm64-gpu.yml'
+    - 'toolchains/ios.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+    - 'src/layer/vulkan/**'
+concurrency:
+  group: ios-arm64-gpu-${{ github.ref }}
+  cancel-in-progress: true
+env:
+  DEVELOPER_DIR: /Applications/Xcode_13.4.1.app/Contents/Developer
+  IOS_DEPLOYMENT_TARGET: '9.0'
+  ENABLE_BITCODE: OFF
+  ENABLE_ARC: OFF
+  ENABLE_VISIBILITY: OFF
+permissions:
+  contents: read
+
+jobs:
+  ios-iphone-os-gpu:
+    runs-on: macos-12
+    steps:
+    - uses: actions/checkout@v4
+      with:
+        submodules: true
+
+    - name: cache-openmp
+      id: cache-openmp
+      uses: actions/cache@v3
+      with:
+        path: openmp-install
+        key: openmp-ios-install-20230504
+    - name: openmp
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        wget https://github.com/llvm/llvm-project/releases/download/llvmorg-11.0.0/openmp-11.0.0.src.tar.xz
+        tar -xf openmp-11.0.0.src.tar.xz
+        cd openmp-11.0.0.src
+        sed -i'' -e '/.size __kmp_unnamed_critical_addr/d' runtime/src/z_Linux_asm.S
+        sed -i'' -e 's/__kmp_unnamed_critical_addr/___kmp_unnamed_critical_addr/g' runtime/src/z_Linux_asm.S
+    - name: openmp-build-armv7
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-armv7 && cd build-armv7
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=OS -DARCHS="armv7" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-build-arm64
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-arm64 && cd build-arm64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=OS64 -DARCHS="arm64" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-build-arm64e
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-arm64e && cd build-arm64e
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=OS64 -DARCHS="arm64e" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-merge-fat-library
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        mkdir -p $GITHUB_WORKSPACE/openmp-install
+        cp -a openmp-11.0.0.src/build-arm64/install/include $GITHUB_WORKSPACE/openmp-install
+        mkdir -p $GITHUB_WORKSPACE/openmp-install/lib
+        lipo -create \
+            openmp-11.0.0.src/build-armv7/install/lib/libomp.a \
+            openmp-11.0.0.src/build-arm64/install/lib/libomp.a \
+            openmp-11.0.0.src/build-arm64e/install/lib/libomp.a \
+            -o $GITHUB_WORKSPACE/openmp-install/lib/libomp.a
+
+    - name: install-openmp
+      run: |
+        sudo cp $GITHUB_WORKSPACE/openmp-install/include/* $DEVELOPER_DIR/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS.sdk/usr/include
+        sudo cp $GITHUB_WORKSPACE/openmp-install/lib/libomp.a $DEVELOPER_DIR/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS.sdk/usr/lib
+
+    - name: vulkansdk
+      run: |
+        wget https://sdk.lunarg.com/sdk/download/1.3.236.0/mac/vulkansdk-macos-1.3.236.0.dmg?Human=true -O vulkansdk-macos-1.3.236.0.dmg
+        hdiutil attach vulkansdk-macos-1.3.236.0.dmg
+        sudo /Volumes/vulkansdk-macos-1.3.236.0/InstallVulkan.app/Contents/MacOS/InstallVulkan --root $GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0 --accept-licenses --default-answer --confirm-command install
+        hdiutil detach /Volumes/vulkansdk-macos-1.3.236.0
+
+    - name: build-arm64
+      run: |
+        export VULKAN_SDK=$GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0/macOS
+        mkdir build-arm64 && cd build-arm64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/ios.toolchain.cmake -DPLATFORM=OS64 -DARCHS="arm64" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DOpenMP_C_FLAGS="-Xclang -fopenmp" -DOpenMP_CXX_FLAGS="-Xclang -fopenmp" \
+            -DOpenMP_C_LIB_NAMES="libomp" -DOpenMP_CXX_LIB_NAMES="libomp" \
+            -DOpenMP_libomp_LIBRARY="$DEVELOPER_DIR/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS.sdk/usr/lib/libomp.a" \
+            -DVulkan_INCLUDE_DIR=$GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0/MoltenVK/include \
+            -DVulkan_LIBRARY=$GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0/MoltenVK/dylib/iOS/libMoltenVK.dylib \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DNCNN_VULKAN=ON -DNCNN_BUILD_BENCHMARK=OFF ..
+        cmake --build . -j 3
+    - name: build-arm64e
+      run: |
+        export VULKAN_SDK=$GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0/macOS
+        mkdir build-arm64e && cd build-arm64e
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/ios.toolchain.cmake -DPLATFORM=OS64 -DARCHS="arm64e" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DOpenMP_C_FLAGS="-Xclang -fopenmp" -DOpenMP_CXX_FLAGS="-Xclang -fopenmp" \
+            -DOpenMP_C_LIB_NAMES="libomp" -DOpenMP_CXX_LIB_NAMES="libomp" \
+            -DOpenMP_libomp_LIBRARY="$DEVELOPER_DIR/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS.sdk/usr/lib/libomp.a" \
+            -DVulkan_INCLUDE_DIR=$GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0/MoltenVK/include \
+            -DVulkan_LIBRARY=$GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0/MoltenVK/dylib/iOS/libMoltenVK.dylib \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DNCNN_VULKAN=ON -DNCNN_BUILD_BENCHMARK=OFF ..
+        cmake --build . -j 3

.github/workflows/ios-cpu.yml

@@ -0,0 +1,140 @@
+name: ios-cpu
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/ios-cpu.yml'
+    - 'toolchains/ios.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/ios-cpu.yml'
+    - 'toolchains/ios.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+concurrency:
+  group: ios-cpu-${{ github.ref }}
+  cancel-in-progress: true
+env:
+  DEVELOPER_DIR: /Applications/Xcode_13.4.1.app/Contents/Developer
+  IOS_DEPLOYMENT_TARGET: '9.0'
+  ENABLE_BITCODE: OFF
+  ENABLE_ARC: OFF
+  ENABLE_VISIBILITY: OFF
+permissions:
+  contents: read
+
+jobs:
+  ios-iphone-os:
+    runs-on: macos-12
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-openmp
+      id: cache-openmp
+      uses: actions/cache@v3
+      with:
+        path: openmp-install
+        key: openmp-ios-install-20230504
+    - name: openmp
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        wget https://github.com/llvm/llvm-project/releases/download/llvmorg-11.0.0/openmp-11.0.0.src.tar.xz
+        tar -xf openmp-11.0.0.src.tar.xz
+        cd openmp-11.0.0.src
+        sed -i'' -e '/.size __kmp_unnamed_critical_addr/d' runtime/src/z_Linux_asm.S
+        sed -i'' -e 's/__kmp_unnamed_critical_addr/___kmp_unnamed_critical_addr/g' runtime/src/z_Linux_asm.S
+    - name: openmp-build-armv7
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-armv7 && cd build-armv7
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=OS -DARCHS="armv7" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-build-arm64
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-arm64 && cd build-arm64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=OS64 -DARCHS="arm64" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-build-arm64e
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-arm64e && cd build-arm64e
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=OS64 -DARCHS="arm64e" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-merge-fat-library
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        mkdir -p $GITHUB_WORKSPACE/openmp-install
+        cp -a openmp-11.0.0.src/build-arm64/install/include $GITHUB_WORKSPACE/openmp-install
+        mkdir -p $GITHUB_WORKSPACE/openmp-install/lib
+        lipo -create \
+            openmp-11.0.0.src/build-armv7/install/lib/libomp.a \
+            openmp-11.0.0.src/build-arm64/install/lib/libomp.a \
+            openmp-11.0.0.src/build-arm64e/install/lib/libomp.a \
+            -o $GITHUB_WORKSPACE/openmp-install/lib/libomp.a
+
+    - name: install-openmp
+      run: |
+        sudo cp $GITHUB_WORKSPACE/openmp-install/include/* $DEVELOPER_DIR/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS.sdk/usr/include
+        sudo cp $GITHUB_WORKSPACE/openmp-install/lib/libomp.a $DEVELOPER_DIR/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS.sdk/usr/lib
+
+    - name: build-armv7
+      run: |
+        mkdir build-armv7 && cd build-armv7
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/ios.toolchain.cmake -DPLATFORM=OS -DARCHS="armv7" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DOpenMP_C_FLAGS="-Xclang -fopenmp" -DOpenMP_CXX_FLAGS="-Xclang -fopenmp" \
+            -DOpenMP_C_LIB_NAMES="libomp" -DOpenMP_CXX_LIB_NAMES="libomp" \
+            -DOpenMP_libomp_LIBRARY="$DEVELOPER_DIR/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS.sdk/usr/lib/libomp.a" \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            ..
+        cmake --build . -j 3
+    - name: build-arm64
+      run: |
+        mkdir build-arm64 && cd build-arm64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/ios.toolchain.cmake -DPLATFORM=OS64 -DARCHS="arm64" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DOpenMP_C_FLAGS="-Xclang -fopenmp" -DOpenMP_CXX_FLAGS="-Xclang -fopenmp" \
+            -DOpenMP_C_LIB_NAMES="libomp" -DOpenMP_CXX_LIB_NAMES="libomp" \
+            -DOpenMP_libomp_LIBRARY="$DEVELOPER_DIR/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS.sdk/usr/lib/libomp.a" \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            ..
+        cmake --build . -j 3
+    - name: build-arm64e
+      run: |
+        mkdir build-arm64e && cd build-arm64e
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/ios.toolchain.cmake -DPLATFORM=OS64 -DARCHS="arm64e" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DOpenMP_C_FLAGS="-Xclang -fopenmp" -DOpenMP_CXX_FLAGS="-Xclang -fopenmp" \
+            -DOpenMP_C_LIB_NAMES="libomp" -DOpenMP_CXX_LIB_NAMES="libomp" \
+            -DOpenMP_libomp_LIBRARY="$DEVELOPER_DIR/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS.sdk/usr/lib/libomp.a" \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            ..
+        cmake --build . -j 3

.github/workflows/ios-simulator-gpu.yml

@@ -0,0 +1,146 @@
+name: ios-simulator-gpu
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/ios-simulator-gpu.yml'
+    - 'toolchains/ios.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'src/layer/arm/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/ios-simulator-gpu.yml'
+    - 'toolchains/ios.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'src/layer/arm/**'
+concurrency:
+  group: ios-simulator-gpu-${{ github.ref }}
+  cancel-in-progress: true
+env:
+  DEVELOPER_DIR: /Applications/Xcode_13.4.1.app/Contents/Developer
+  IOS_DEPLOYMENT_TARGET: '9.0'
+  ENABLE_BITCODE: OFF
+  ENABLE_ARC: OFF
+  ENABLE_VISIBILITY: OFF
+permissions:
+  contents: read
+
+jobs:
+  ios-iphone-simulator-gpu:
+    runs-on: macos-12
+    steps:
+    - uses: actions/checkout@v4
+      with:
+        submodules: true
+
+    - name: cache-openmp
+      id: cache-openmp
+      uses: actions/cache@v3
+      with:
+        path: openmp-install
+        key: openmp-ios-simulator-install-20230504
+    - name: openmp
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        wget https://github.com/llvm/llvm-project/releases/download/llvmorg-11.0.0/openmp-11.0.0.src.tar.xz
+        tar -xf openmp-11.0.0.src.tar.xz
+        cd openmp-11.0.0.src
+        sed -i'' -e '/.size __kmp_unnamed_critical_addr/d' runtime/src/z_Linux_asm.S
+        sed -i'' -e 's/__kmp_unnamed_critical_addr/___kmp_unnamed_critical_addr/g' runtime/src/z_Linux_asm.S
+    - name: openmp-build-i386
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-i386 && cd build-i386
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=SIMULATOR -DARCHS="i386" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-build-x86_64
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-x86_64 && cd build-x86_64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=SIMULATOR64 -DARCHS="x86_64" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-build-arm64
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-arm64 && cd build-arm64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=SIMULATORARM64 -DARCHS="arm64" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-merge-fat-library
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        mkdir -p $GITHUB_WORKSPACE/openmp-install
+        cp -a openmp-11.0.0.src/build-x86_64/install/include $GITHUB_WORKSPACE/openmp-install
+        mkdir -p $GITHUB_WORKSPACE/openmp-install/lib
+        lipo -create \
+            openmp-11.0.0.src/build-i386/install/lib/libomp.a \
+            openmp-11.0.0.src/build-x86_64/install/lib/libomp.a \
+            openmp-11.0.0.src/build-arm64/install/lib/libomp.a \
+            -o $GITHUB_WORKSPACE/openmp-install/lib/libomp.a
+
+    - name: openmp-install
+      run: |
+        sudo cp $GITHUB_WORKSPACE/openmp-install/include/* $DEVELOPER_DIR/Platforms/iPhoneSimulator.platform/Developer/SDKs/iPhoneSimulator.sdk/usr/include
+        sudo cp $GITHUB_WORKSPACE/openmp-install/lib/libomp.a $DEVELOPER_DIR/Platforms/iPhoneSimulator.platform/Developer/SDKs/iPhoneSimulator.sdk/usr/lib
+
+    - name: vulkansdk
+      run: |
+        wget https://sdk.lunarg.com/sdk/download/1.3.236.0/mac/vulkansdk-macos-1.3.236.0.dmg?Human=true -O vulkansdk-macos-1.3.236.0.dmg
+        hdiutil attach vulkansdk-macos-1.3.236.0.dmg
+        sudo /Volumes/vulkansdk-macos-1.3.236.0/InstallVulkan.app/Contents/MacOS/InstallVulkan --root $GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0 --accept-licenses --default-answer --confirm-command install
+        hdiutil detach /Volumes/vulkansdk-macos-1.3.236.0
+
+    - name: build-x86_64
+      run: |
+        export VULKAN_SDK=$GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0/macOS
+        mkdir build-x86_64 && cd build-x86_64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/ios.toolchain.cmake -DPLATFORM=SIMULATOR64 -DARCHS="x86_64" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DOpenMP_C_FLAGS="-Xclang -fopenmp" -DOpenMP_CXX_FLAGS="-Xclang -fopenmp" \
+            -DOpenMP_C_LIB_NAMES="libomp" -DOpenMP_CXX_LIB_NAMES="libomp" \
+            -DOpenMP_libomp_LIBRARY="$DEVELOPER_DIR/Platforms/iPhoneSimulator.platform/Developer/SDKs/iPhoneSimulator.sdk/usr/lib/libomp.a" \
+            -DVulkan_INCLUDE_DIR=$GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0/MoltenVK/include \
+            -DVulkan_LIBRARY=$GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0/MoltenVK/dylib/macOS/libMoltenVK.dylib \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DNCNN_VULKAN=ON -DNCNN_BUILD_BENCHMARK=OFF ..
+        cmake --build . -j 3
+    - name: build-arm64
+      run: |
+        export VULKAN_SDK=$GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0/macOS
+        mkdir build-arm64 && cd build-arm64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/ios.toolchain.cmake -DPLATFORM=SIMULATORARM64 -DARCHS="arm64" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DOpenMP_C_FLAGS="-Xclang -fopenmp" -DOpenMP_CXX_FLAGS="-Xclang -fopenmp" \
+            -DOpenMP_C_LIB_NAMES="libomp" -DOpenMP_CXX_LIB_NAMES="libomp" \
+            -DOpenMP_libomp_LIBRARY="$DEVELOPER_DIR/Platforms/iPhoneSimulator.platform/Developer/SDKs/iPhoneSimulator.sdk/usr/lib/libomp.a" \
+            -DVulkan_INCLUDE_DIR=$GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0/MoltenVK/include \
+            -DVulkan_LIBRARY=$GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0/MoltenVK/dylib/macOS/libMoltenVK.dylib \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DNCNN_VULKAN=ON -DNCNN_BUILD_BENCHMARK=OFF ..
+        cmake --build . -j 3

.github/workflows/ios-simulator.yml

@@ -0,0 +1,142 @@
+name: ios-simulator
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/ios-simulator.yml'
+    - 'toolchains/ios.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'src/layer/arm/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/ios-simulator.yml'
+    - 'toolchains/ios.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'src/layer/arm/**'
+concurrency:
+  group: ios-simulator-${{ github.ref }}
+  cancel-in-progress: true
+env:
+  DEVELOPER_DIR: /Applications/Xcode_13.4.1.app/Contents/Developer
+  IOS_DEPLOYMENT_TARGET: '9.0'
+  ENABLE_BITCODE: OFF
+  ENABLE_ARC: OFF
+  ENABLE_VISIBILITY: OFF
+permissions:
+  contents: read
+
+jobs:
+  ios-iphone-simulator:
+    runs-on: macos-12
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-openmp
+      id: cache-openmp
+      uses: actions/cache@v3
+      with:
+        path: openmp-install
+        key: openmp-ios-simulator-install-20230504
+    - name: openmp
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        wget https://github.com/llvm/llvm-project/releases/download/llvmorg-11.0.0/openmp-11.0.0.src.tar.xz
+        tar -xf openmp-11.0.0.src.tar.xz
+        cd openmp-11.0.0.src
+        sed -i'' -e '/.size __kmp_unnamed_critical_addr/d' runtime/src/z_Linux_asm.S
+        sed -i'' -e 's/__kmp_unnamed_critical_addr/___kmp_unnamed_critical_addr/g' runtime/src/z_Linux_asm.S
+    - name: openmp-build-i386
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-i386 && cd build-i386
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=SIMULATOR -DARCHS="i386" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-build-x86_64
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-x86_64 && cd build-x86_64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=SIMULATOR64 -DARCHS="x86_64" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-build-arm64
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-arm64 && cd build-arm64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=SIMULATORARM64 -DARCHS="arm64" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-merge-fat-library
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        mkdir -p $GITHUB_WORKSPACE/openmp-install
+        cp -a openmp-11.0.0.src/build-x86_64/install/include $GITHUB_WORKSPACE/openmp-install
+        mkdir -p $GITHUB_WORKSPACE/openmp-install/lib
+        lipo -create \
+            openmp-11.0.0.src/build-i386/install/lib/libomp.a \
+            openmp-11.0.0.src/build-x86_64/install/lib/libomp.a \
+            openmp-11.0.0.src/build-arm64/install/lib/libomp.a \
+            -o $GITHUB_WORKSPACE/openmp-install/lib/libomp.a
+
+    - name: openmp-install
+      run: |
+        sudo cp $GITHUB_WORKSPACE/openmp-install/include/* $DEVELOPER_DIR/Platforms/iPhoneSimulator.platform/Developer/SDKs/iPhoneSimulator.sdk/usr/include
+        sudo cp $GITHUB_WORKSPACE/openmp-install/lib/libomp.a $DEVELOPER_DIR/Platforms/iPhoneSimulator.platform/Developer/SDKs/iPhoneSimulator.sdk/usr/lib
+
+    - name: build-i386
+      run: |
+        mkdir build-i386 && cd build-i386
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/ios.toolchain.cmake -DPLATFORM=SIMULATOR -DARCHS="i386" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DOpenMP_C_FLAGS="-Xclang -fopenmp" -DOpenMP_CXX_FLAGS="-Xclang -fopenmp" \
+            -DOpenMP_C_LIB_NAMES="libomp" -DOpenMP_CXX_LIB_NAMES="libomp" \
+            -DOpenMP_libomp_LIBRARY="$DEVELOPER_DIR/Platforms/iPhoneSimulator.platform/Developer/SDKs/iPhoneSimulator.sdk/usr/lib/libomp.a" \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            ..
+        cmake --build . -j 3
+    - name: build-x86_64
+      run: |
+        mkdir build-x86_64 && cd build-x86_64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/ios.toolchain.cmake -DPLATFORM=SIMULATOR64 -DARCHS="x86_64" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DOpenMP_C_FLAGS="-Xclang -fopenmp" -DOpenMP_CXX_FLAGS="-Xclang -fopenmp" \
+            -DOpenMP_C_LIB_NAMES="libomp" -DOpenMP_CXX_LIB_NAMES="libomp" \
+            -DOpenMP_libomp_LIBRARY="$DEVELOPER_DIR/Platforms/iPhoneSimulator.platform/Developer/SDKs/iPhoneSimulator.sdk/usr/lib/libomp.a" \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            ..
+        cmake --build . -j 3
+    - name: build-arm64
+      run: |
+        mkdir build-arm64 && cd build-arm64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/ios.toolchain.cmake -DPLATFORM=SIMULATORARM64 -DARCHS="arm64" \
+            -DDEPLOYMENT_TARGET=$IOS_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DOpenMP_C_FLAGS="-Xclang -fopenmp" -DOpenMP_CXX_FLAGS="-Xclang -fopenmp" \
+            -DOpenMP_C_LIB_NAMES="libomp" -DOpenMP_CXX_LIB_NAMES="libomp" \
+            -DOpenMP_libomp_LIBRARY="$DEVELOPER_DIR/Platforms/iPhoneSimulator.platform/Developer/SDKs/iPhoneSimulator.sdk/usr/lib/libomp.a" \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            ..
+        cmake --build . -j 3

.github/workflows/linux-aarch64-cpu-gcc.yml

@@ -0,0 +1,193 @@
+name: linux-aarch64-cpu-gcc
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-aarch64-cpu-gcc.yml'
+    - 'toolchains/aarch64-linux-gnu.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+    - 'tests/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-aarch64-cpu-gcc.yml'
+    - 'toolchains/aarch64-linux-gnu.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+    - 'tests/**'
+concurrency:
+  group: linux-aarch64-cpu-gcc-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-gcc:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: actions/cache@v3
+      with:
+        path: qemu-install
+        key: qemu-aarch64-install-20220502-ubuntu-2004-2
+    - name: install-qemu-build-deps
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        sudo apt-get update
+        sudo apt-get install autoconf automake autotools-dev ninja-build
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      uses: actions/checkout@v4
+      with:
+        repository: qemu/qemu
+        path: qemu
+        ref: f5643914a9e8f79c606a76e6a9d7ea82a3fc3e65
+    - name: qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        cd qemu
+        ./configure --prefix=$GITHUB_WORKSPACE/qemu-install --target-list=aarch64-linux-user --disable-system
+        make -j2
+        make install
+
+    - name: aarch64-gnu-toolchain
+      run: |
+        sudo apt-get update
+        sudo apt-get install g++-aarch64-linux-gnu
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/aarch64-linux-gnu.toolchain.cmake -DNCNN_ARM82=OFF -DNCNN_ARM82DOT=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-aarch64 TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/aarch64-linux-gnu" ctest --output-on-failure -j 2
+
+    - name: build-noint8
+      run: |
+        mkdir build-noint8 && cd build-noint8
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/aarch64-linux-gnu.toolchain.cmake -DNCNN_ARM82=OFF -DNCNN_ARM82DOT=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test-noint8
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build-noint8
+        TESTS_EXECUTABLE_LOADER=qemu-aarch64 TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/aarch64-linux-gnu" ctest --output-on-failure -j 2
+
+  linux-gcc-arm82:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: actions/cache@v3
+      with:
+        path: qemu-install
+        key: qemu-aarch64-install-20220502-ubuntu-2004-2
+    - name: install-qemu-build-deps
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        sudo apt-get update
+        sudo apt-get install autoconf automake autotools-dev ninja-build
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      uses: actions/checkout@v4
+      with:
+        repository: qemu/qemu
+        path: qemu
+        ref: f5643914a9e8f79c606a76e6a9d7ea82a3fc3e65
+    - name: qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        cd qemu
+        ./configure --prefix=$GITHUB_WORKSPACE/qemu-install --target-list=aarch64-linux-user --disable-system
+        make -j2
+        make install
+
+    - name: aarch64-gnu-toolchain
+      run: |
+        sudo apt-get update
+        sudo apt-get install g++-aarch64-linux-gnu
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/aarch64-linux-gnu.toolchain.cmake -DNCNN_ARM82=ON -DNCNN_ARM82DOT=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-aarch64 TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/aarch64-linux-gnu" ctest --output-on-failure -j 2
+
+    - name: build-noint8
+      run: |
+        mkdir build-noint8 && cd build-noint8
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/aarch64-linux-gnu.toolchain.cmake -DNCNN_ARM82=ON -DNCNN_ARM82DOT=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON -DNCNN_INT8=OFF ..
+        cmake --build . -j 2
+    - name: test-noint8
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build-noint8
+        TESTS_EXECUTABLE_LOADER=qemu-aarch64 TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/aarch64-linux-gnu" ctest --output-on-failure -j 2
+
+  linux-gcc-arm86:
+    runs-on: ubuntu-22.04
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: actions/cache@v3
+      with:
+        path: qemu-install
+        key: qemu-aarch64-install-20230717
+    - name: install-qemu-build-deps
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        sudo apt-get update
+        sudo apt-get install autoconf automake autotools-dev ninja-build
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      uses: actions/checkout@v4
+      with:
+        repository: qemu/qemu
+        path: qemu
+        ref: ed8ad9728a9c0eec34db9dff61dfa2f1dd625637
+    - name: qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        cd qemu
+        ./configure --prefix=$GITHUB_WORKSPACE/qemu-install --target-list=aarch64-linux-user --disable-system
+        make -j2
+        make install
+
+    - name: aarch64-gnu-toolchain
+      run: |
+        sudo apt-get update
+        sudo apt-get install g++-aarch64-linux-gnu
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/aarch64-linux-gnu.toolchain.cmake -DNCNN_ARM82=ON -DNCNN_ARM82DOT=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-aarch64 TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/aarch64-linux-gnu" ctest --output-on-failure -j 2

.github/workflows/linux-arm-cpu-gcc.yml

@@ -0,0 +1,206 @@
+name: linux-arm-cpu-gcc
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-arm-cpu-gcc.yml'
+    - 'toolchains/arm-linux-gnueabi.toolchain.cmake'
+    - 'toolchains/arm-linux-gnueabihf.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+    - 'tests/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-arm-cpu-gcc.yml'
+    - 'toolchains/arm-linux-gnueabi.toolchain.cmake'
+    - 'toolchains/arm-linux-gnueabihf.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+    - 'tests/**'
+concurrency:
+  group: linux-arm-cpu-gcc-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-gcc-arm:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: actions/cache@v3
+      with:
+        path: qemu-install
+        key: qemu-arm-install-20220502-2
+    - name: install-qemu-build-deps
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        sudo apt-get update
+        sudo apt-get install autoconf automake autotools-dev ninja-build
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      uses: actions/checkout@v4
+      with:
+        repository: qemu/qemu
+        path: qemu
+        ref: f5643914a9e8f79c606a76e6a9d7ea82a3fc3e65
+    - name: qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        cd qemu
+        ./configure --prefix=$GITHUB_WORKSPACE/qemu-install --target-list=arm-linux-user --disable-system
+        make -j2
+        make install
+
+    - name: arm-gnu-toolchain
+      run: |
+        sudo apt-get update
+        sudo apt-get install g++-arm-linux-gnueabi
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/arm-linux-gnueabi.toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-arm TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/arm-linux-gnueabi" ctest --output-on-failure -j 2
+
+    - name: build-noint8
+      run: |
+        mkdir build-noint8 && cd build-noint8
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/arm-linux-gnueabi.toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test-noint8
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build-noint8
+        TESTS_EXECUTABLE_LOADER=qemu-arm TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/arm-linux-gnueabi" ctest --output-on-failure -j 2
+
+  linux-gcc-armhf:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: actions/cache@v3
+      with:
+        path: qemu-install
+        key: qemu-arm-install-20220502-2
+    - name: install-qemu-build-deps
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        sudo apt-get update
+        sudo apt-get install autoconf automake autotools-dev ninja-build
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      uses: actions/checkout@v4
+      with:
+        repository: qemu/qemu
+        path: qemu
+        ref: f5643914a9e8f79c606a76e6a9d7ea82a3fc3e65
+    - name: qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        cd qemu
+        ./configure --prefix=$GITHUB_WORKSPACE/qemu-install --target-list=arm-linux-user --disable-system
+        make -j2
+        make install
+
+    - name: arm-gnu-toolchain
+      run: |
+        sudo apt-get update
+        sudo apt-get install g++-arm-linux-gnueabihf
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/arm-linux-gnueabihf.toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-arm TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/arm-linux-gnueabihf" ctest --output-on-failure -j 2
+
+    - name: build-noint8
+      run: |
+        mkdir build-noint8 && cd build-noint8
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/arm-linux-gnueabihf.toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON -DNCNN_INT8=OFF ..
+        cmake --build . -j 2
+    - name: test-noint8
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build-noint8
+        TESTS_EXECUTABLE_LOADER=qemu-arm TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/arm-linux-gnueabihf" ctest --output-on-failure -j 2
+
+  linux-gcc-armhf-vfpv3-d16:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: actions/cache@v3
+      with:
+        path: qemu-install
+        key: qemu-arm-install-20220502-2
+    - name: install-qemu-build-deps
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        sudo apt-get update
+        sudo apt-get install autoconf automake autotools-dev ninja-build
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      uses: actions/checkout@v4
+      with:
+        repository: qemu/qemu
+        path: qemu
+        ref: f5643914a9e8f79c606a76e6a9d7ea82a3fc3e65
+    - name: qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        cd qemu
+        ./configure --prefix=$GITHUB_WORKSPACE/qemu-install --target-list=arm-linux-user --disable-system
+        make -j2
+        make install
+
+    - name: arm-gnu-toolchain
+      run: |
+        sudo apt-get update
+        sudo apt-get install g++-arm-linux-gnueabihf
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/arm-linux-gnueabihf-vfpv3-d16.toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-arm TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/arm-linux-gnueabihf" ctest --output-on-failure -j 2
+
+    - name: build-noint8
+      run: |
+        mkdir build-noint8 && cd build-noint8
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/arm-linux-gnueabihf-vfpv3-d16.toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON -DNCNN_INT8=OFF ..
+        cmake --build . -j 2
+    - name: test-noint8
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build-noint8
+        TESTS_EXECUTABLE_LOADER=qemu-arm TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/arm-linux-gnueabihf" ctest --output-on-failure -j 2

.github/workflows/linux-loongarch64-cpu-gcc.yml

@@ -0,0 +1,77 @@
+name: linux-loongarch64-cpu-gcc
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-loongarch64-cpu-gcc.yml'
+    - 'toolchains/loongarch64-linux-gnu.toolchain.cmake'
+    - 'toolchains/loongarch64-unknown-linux-gnu.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/loongarch/**'
+    - 'tests/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-loongarch64-cpu-gcc.yml'
+    - 'toolchains/loongarch64-linux-gnu.toolchain.cmake'
+    - 'toolchains/loongarch64-unknown-linux-gnu.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/loongarch/**'
+    - 'tests/**'
+concurrency:
+  group: linux-loongarch64-cpu-gcc-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-gcc-loongarch64:
+    runs-on: [self-hosted, linux, centos]
+
+    steps:
+    - uses: actions/checkout@v4
+
+    # - name: cache-qemu
+    #   id: cache-qemu
+    #   uses: actions/cache@v3
+    #   with:
+    #     path: qemu-install
+    #     key: qemu-loongarch64-install-20230524
+    # - name: checkout-qemu
+    #   if: steps.cache-qemu.outputs.cache-hit != 'true'
+    #   uses: actions/checkout@v4
+    #   with:
+    #     repository: qemu/qemu
+    #     path: qemu
+    #     ref: 1c12355b31046a6b35a4f50c85c4f01afb1bd728
+    # - name: qemu
+    #   if: steps.cache-qemu.outputs.cache-hit != 'true'
+    #   run: |
+    #     cd qemu
+    #     ./configure --prefix=$GITHUB_WORKSPACE/qemu-install --target-list=loongarch64-linux-user --disable-system
+    #     make -j2
+    #     make install
+
+    # - name: loongarch64-toolchain
+    #   run: |
+    #     wget https://github.com/sunhaiyong1978/CLFS-for-LoongArch/releases/download/8.0/loongarch64-clfs-8.0-cross-tools-gcc-full.tar.xz
+    #     tar -xf loongarch64-clfs-8.0-cross-tools-gcc-full.tar.xz
+
+    - name: build
+      run: |
+        export LOONGARCH64_ROOT_PATH=/data/action/osd/cross-tools
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/loongarch64-unknown-linux-gnu.toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 4
+
+    - name: test
+      run: |
+        export PATH=/data/action/osd/qemu-loongson/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-loongarch64 TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/data/action/osd/cross-tools/target" ctest --output-on-failure -j 4

.github/workflows/linux-mips-cpu-gcc.yml

@@ -0,0 +1,126 @@
+name: linux-mips-cpu-gcc
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-mips-cpu-gcc.yml'
+    - 'toolchains/mipsel-linux-gnu.toolchain.cmake'
+    - 'toolchains/mipsisa32r6el-linux-gnu.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/mips/**'
+    - 'tests/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-mips-cpu-gcc.yml'
+    - 'toolchains/mipsel-linux-gnu.toolchain.cmake'
+    - 'toolchains/mipsisa32r6el-linux-gnu.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/mips/**'
+    - 'tests/**'
+concurrency:
+  group: linux-mips-cpu-gcc-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-gcc-mipsel:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: actions/cache@v3
+      with:
+        path: qemu-install
+        key: qemu-mipsel-install-20220502-2
+    - name: install-qemu-build-deps
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        sudo apt-get update
+        sudo apt-get install autoconf automake autotools-dev ninja-build
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      uses: actions/checkout@v4
+      with:
+        repository: qemu/qemu
+        path: qemu
+        ref: f5643914a9e8f79c606a76e6a9d7ea82a3fc3e65
+    - name: qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        cd qemu
+        ./configure --prefix=$GITHUB_WORKSPACE/qemu-install --target-list=mipsel-linux-user --disable-system
+        make -j2
+        make install
+
+    - name: mipsel-gnu-toolchain
+      run: |
+        sudo apt-get update
+        sudo apt-get install g++-mipsel-linux-gnu
+
+    - name: configure
+      run: mkdir build && cd build && cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/mipsel-linux-gnu.toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+    - name: build
+      run: cmake --build build -j 2
+
+    - name: test
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-mipsel TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/mipsel-linux-gnu" ctest --output-on-failure -j 2
+
+  linux-gcc-mipsisa32r6el:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: actions/cache@v3
+      with:
+        path: qemu-install
+        key: qemu-mipsel-install-20220502-2
+    - name: install-qemu-build-deps
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        sudo apt-get update
+        sudo apt-get install autoconf automake autotools-dev ninja-build
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      uses: actions/checkout@v4
+      with:
+        repository: qemu/qemu
+        path: qemu
+        ref: f5643914a9e8f79c606a76e6a9d7ea82a3fc3e65
+    - name: qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        cd qemu
+        ./configure --prefix=$GITHUB_WORKSPACE/qemu-install --target-list=mipsel-linux-user --disable-system
+        make -j2
+        make install
+
+    - name: mipsisa32r6el-gnu-toolchain
+      run: |
+        sudo apt-get update
+        sudo apt-get install g++-mipsisa32r6el-linux-gnu
+
+    - name: configure
+      run: mkdir build && cd build && cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/mipsisa32r6el-linux-gnu.toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+    - name: build
+      run: cmake --build build -j 2
+
+    - name: test
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-mipsel TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/mipsisa32r6el-linux-gnu" ctest --output-on-failure -j 2

.github/workflows/linux-mips64-cpu-gcc.yml

@@ -0,0 +1,138 @@
+name: linux-mips64-cpu-gcc
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-mips64-cpu-gcc.yml'
+    - 'toolchains/mips64el-linux-gnuabi64.toolchain.cmake'
+    - 'toolchains/mipsisa64r6el-linux-gnuabi64.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/mips/**'
+    - 'tests/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-mips64-cpu-gcc.yml'
+    - 'toolchains/mips64el-linux-gnuabi64.toolchain.cmake'
+    - 'toolchains/mipsisa64r6el-linux-gnuabi64.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/mips/**'
+    - 'tests/**'
+concurrency:
+  group: linux-mips64-cpu-gcc-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-gcc-mips64el:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: actions/cache@v3
+      with:
+        path: qemu-install
+        key: qemu-mips64el-install-20220502-2
+    - name: install-qemu-build-deps
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        sudo apt-get update
+        sudo apt-get install autoconf automake autotools-dev ninja-build
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      uses: actions/checkout@v4
+      with:
+        repository: qemu/qemu
+        path: qemu
+        ref: f5643914a9e8f79c606a76e6a9d7ea82a3fc3e65
+    - name: qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        cd qemu
+        ./configure --prefix=$GITHUB_WORKSPACE/qemu-install --target-list=mips64el-linux-user --disable-system
+        make -j2
+        make install
+
+    - name: mips64el-gnuabi64-toolchain
+      run: |
+        sudo apt-get update
+        sudo apt-get install g++-mips64el-linux-gnuabi64
+
+    - name: configure
+      run: mkdir build && cd build && cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/mips64el-linux-gnuabi64.toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+    - name: build
+      run: cmake --build build -j 2
+
+    - name: test
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-mips64el TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/mips64el-linux-gnuabi64" ctest --output-on-failure -j 2
+
+  linux-gcc-mipsisa64r6el:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: actions/cache@v3
+      with:
+        path: qemu-install
+        key: qemu-mips64el-install-20220502-4
+    - name: install-qemu-build-deps
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        sudo apt-get update
+        sudo apt-get install autoconf automake autotools-dev ninja-build
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      uses: actions/checkout@v4
+      with:
+        repository: qemu/qemu
+        path: qemu
+        ref: f5643914a9e8f79c606a76e6a9d7ea82a3fc3e65
+    - name: qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        cd qemu
+        wget https://raw.githubusercontent.com/nihui/ncnn-assets/master/qemu-patches/0001-target-mips-Fix-SAT_S-trans-helper.patch
+        wget https://raw.githubusercontent.com/nihui/ncnn-assets/master/qemu-patches/0002-target-mips-Fix-df_extract_val-and-df_extract_df-dfe.patch
+        wget https://raw.githubusercontent.com/nihui/ncnn-assets/master/qemu-patches/0003-target-mips-Fix-msa-checking-condition-in-trans_msa_.patch
+        wget https://raw.githubusercontent.com/nihui/ncnn-assets/master/qemu-patches/0004-target-mips-Do-not-treat-msa-INSERT-as-NOP-when-wd-i.patch
+        wget https://raw.githubusercontent.com/nihui/ncnn-assets/master/qemu-patches/0005-target-mips-Fix-FTRUNC_S-and-FTRUNC_U-trans-helper.patch
+        wget https://raw.githubusercontent.com/nihui/ncnn-assets/master/qemu-patches/0006-target-mips-Fix-store-adress-of-high-64bit-in-helper.patch
+        patch -p1 -i 0001-target-mips-Fix-SAT_S-trans-helper.patch
+        patch -p1 -i 0002-target-mips-Fix-df_extract_val-and-df_extract_df-dfe.patch
+        patch -p1 -i 0003-target-mips-Fix-msa-checking-condition-in-trans_msa_.patch
+        patch -p1 -i 0004-target-mips-Do-not-treat-msa-INSERT-as-NOP-when-wd-i.patch
+        patch -p1 -i 0005-target-mips-Fix-FTRUNC_S-and-FTRUNC_U-trans-helper.patch
+        patch -p1 -i 0006-target-mips-Fix-store-adress-of-high-64bit-in-helper.patch
+        ./configure --prefix=$GITHUB_WORKSPACE/qemu-install --target-list=mips64el-linux-user --disable-system
+        make -j2
+        make install
+
+    - name: mipsisa64r6el-gnuabi64-toolchain
+      run: |
+        sudo apt-get update
+        sudo apt-get install g++-mipsisa64r6el-linux-gnuabi64
+
+    - name: configure
+      run: mkdir build && cd build && cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/mipsisa64r6el-linux-gnuabi64.toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+    - name: build
+      run: cmake --build build -j 2
+
+    - name: test
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-mips64el TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/mipsisa64r6el-linux-gnuabi64" ctest --output-on-failure -j 2

.github/workflows/linux-ppc64-cpu-gcc.yml

@@ -0,0 +1,167 @@
+name: linux-ppc64-cpu-gcc
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-ppc64-cpu-gcc.yml'
+    - 'toolchains/powerpc64le-linux-gnu.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'tests/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-ppc64-cpu-gcc.yml'
+    - 'toolchains/powerpc64le-linux-gnu.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'tests/**'
+concurrency:
+  group: linux-ppc64-cpu-gcc-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-gcc-ppc64le:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: actions/cache@v3
+      with:
+        path: qemu-install
+        key: qemu-ppc64le-install-20220502-2
+    - name: install-qemu-build-deps
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        sudo apt-get update
+        sudo apt-get install autoconf automake autotools-dev ninja-build
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      uses: actions/checkout@v4
+      with:
+        repository: qemu/qemu
+        path: qemu
+        ref: f5643914a9e8f79c606a76e6a9d7ea82a3fc3e65
+    - name: qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        cd qemu
+        ./configure --prefix=$GITHUB_WORKSPACE/qemu-install --target-list=ppc64le-linux-user --disable-system
+        make -j2
+        make install
+
+    - name: powerpc64le-gnu-toolchain
+      run: |
+        sudo apt-get update
+        sudo apt-get install g++-powerpc64le-linux-gnu
+
+    - name: configure
+      run: mkdir build && cd build && cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/powerpc64le-linux-gnu.toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+    - name: build
+      run: cmake --build build -j 2
+
+    - name: test
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-ppc64le TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/powerpc64le-linux-gnu" ctest --output-on-failure -j 2
+  linux-gcc-power8le-vsx:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: actions/cache@v3
+      with:
+        path: qemu-install
+        key: qemu-ppc64le-install-20220502-2
+    - name: install-qemu-build-deps
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        sudo apt-get update
+        sudo apt-get install autoconf automake autotools-dev ninja-build
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      uses: actions/checkout@v4
+      with:
+        repository: qemu/qemu
+        path: qemu
+        ref: f5643914a9e8f79c606a76e6a9d7ea82a3fc3e65
+    - name: qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        cd qemu
+        ./configure --prefix=$GITHUB_WORKSPACE/qemu-install --target-list=ppc64le-linux-user --disable-system
+        make -j2
+        make install
+
+    - name: powerpc64le-gnu-toolchain
+      run: |
+        sudo apt-get update
+        sudo apt-get install g++-powerpc64le-linux-gnu
+
+    - name: configure
+      run: mkdir build && cd build && cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/power8le-linux-gnu-vsx.toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+    - name: build
+      run: cmake --build build -j 2
+
+    - name: test
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-ppc64le TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/powerpc64le-linux-gnu;-cpu;power8_v2.0" ctest --output-on-failure -j 2
+  linux-gcc-power9le-vsx:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: actions/cache@v3
+      with:
+        path: qemu-install
+        key: qemu-ppc64le-install-20220502-2
+    - name: install-qemu-build-deps
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        sudo apt-get update
+        sudo apt-get install autoconf automake autotools-dev ninja-build
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      uses: actions/checkout@v4
+      with:
+        repository: qemu/qemu
+        path: qemu
+        ref: f5643914a9e8f79c606a76e6a9d7ea82a3fc3e65
+    - name: qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        cd qemu
+        ./configure --prefix=$GITHUB_WORKSPACE/qemu-install --target-list=ppc64le-linux-user --disable-system
+        make -j2
+        make install
+
+    - name: powerpc64le-gnu-toolchain
+      run: |
+        sudo apt-get update
+        sudo apt-get install g++-powerpc64le-linux-gnu
+
+    - name: configure
+      run: mkdir build && cd build && cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/power9le-linux-gnu-vsx.toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+    - name: build
+      run: cmake --build build -j 2
+
+    - name: test
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-ppc64le TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/powerpc64le-linux-gnu;-cpu;power9_v2.0" ctest --output-on-failure -j 2

.github/workflows/linux-riscv64-cpu-gcc.yml

@@ -0,0 +1,192 @@
+name: linux-riscv64-cpu-gcc
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-riscv64-cpu-gcc.yml'
+    - 'toolchains/riscv64-linux-gnu.toolchain.cmake'
+    - 'toolchains/riscv64-unknown-linux-gnu.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/riscv/**'
+    - 'tests/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-riscv64-cpu-gcc.yml'
+    - 'toolchains/riscv64-linux-gnu.toolchain.cmake'
+    - 'toolchains/riscv64-unknown-linux-gnu.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/riscv/**'
+    - 'tests/**'
+concurrency:
+  group: linux-riscv64-cpu-gcc-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-gcc-riscv64:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: actions/cache@v3
+      with:
+        path: qemu-install
+        key: qemu-riscv64-install-20220502-4
+    - name: install-qemu-build-deps
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        sudo apt-get update
+        sudo apt-get install autoconf automake autotools-dev ninja-build
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      uses: actions/checkout@v4
+      with:
+        repository: qemu/qemu
+        path: qemu
+        ref: f5643914a9e8f79c606a76e6a9d7ea82a3fc3e65
+    - name: qemu
+      if: steps.cache-qemu.outputs.cache-hit != 'true'
+      run: |
+        cd qemu
+        wget https://raw.githubusercontent.com/nihui/ncnn-assets/master/qemu-patches/0007-linux-user-Expose-risc-v-V-isa-bit-in-get_elf_hwcap.patch
+        patch -p1 -i 0007-linux-user-Expose-risc-v-V-isa-bit-in-get_elf_hwcap.patch
+        ./configure --prefix=$GITHUB_WORKSPACE/qemu-install --target-list=riscv64-linux-user --disable-system
+        make -j2
+        make install
+
+    - name: riscv64-gnu-toolchain
+      run: |
+        sudo apt-get update
+        sudo apt-get install g++-riscv64-linux-gnu
+
+    - name: configure
+      run: mkdir build && cd build && cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/riscv64-linux-gnu.toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+    - name: build
+      run: cmake --build build -j 2
+
+    - name: test
+      run: |
+        export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-riscv64 TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/riscv64-linux-gnu" ctest --output-on-failure -j 2
+
+  linux-gcc-riscv64-c906:
+    runs-on: [self-hosted, linux, centos]
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: configure
+      run: |
+        export RISCV_ROOT_PATH=/data/action/osd/Xuantie-900-gcc-linux-5.10.4-glibc-x86_64-V2.6.1
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/c906-v226.toolchain.cmake -DCMAKE_BUILD_TYPE=release -DNCNN_OPENMP=OFF -DNCNN_THREADS=OFF -DNCNN_RUNTIME_CPU=OFF -DNCNN_RVV=ON -DNCNN_SIMPLEOCV=ON -DNCNN_BUILD_EXAMPLES=ON -DNCNN_BUILD_TESTS=ON ..
+    - name: build
+      run: cmake --build build -j 4
+
+    - name: test
+      run: |
+        export PATH=/data/action/osd/xuantie-qemu-x86_64-Ubuntu-18.04-20230413-0706/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-riscv64 TESTS_EXECUTABLE_LOADER_ARGUMENTS="-cpu;c906fdv" ctest --output-on-failure -j 4
+
+  linux-gcc-riscv64-rvv:
+    runs-on: [self-hosted, linux, centos]
+    steps:
+    - uses: actions/checkout@v4
+
+    #- name: cache-qemu
+      #id: cache-qemu
+      #uses: actions/cache@v3
+      #with:
+        #path: qemu-install
+        #key: qemu-riscv64-install-20220502-3
+    #- name: install-qemu-build-deps
+      #if: steps.cache-qemu.outputs.cache-hit != 'true'
+      #run: |
+        #sudo apt-get update
+        #sudo apt-get install autoconf automake autotools-dev ninja-build
+    #- name: checkout-qemu
+      #if: steps.cache-qemu.outputs.cache-hit != 'true'
+      #uses: actions/checkout@v4
+      #with:
+        #repository: qemu/qemu
+        #path: qemu
+        #ref: f5643914a9e8f79c606a76e6a9d7ea82a3fc3e65
+    #- name: qemu
+      #if: steps.cache-qemu.outputs.cache-hit != 'true'
+      #run: |
+        #cd qemu
+        #wget https://raw.githubusercontent.com/nihui/ncnn-assets/master/qemu-patches/0007-linux-user-Expose-risc-v-V-isa-bit-in-get_elf_hwcap.patch
+        #patch -p1 -i 0007-linux-user-Expose-risc-v-V-isa-bit-in-get_elf_hwcap.patch
+        #./configure --prefix=$GITHUB_WORKSPACE/qemu-install --target-list=riscv64-linux-user --disable-system
+        #make -j2
+        #make install
+
+    #- name: cache-riscv
+      #id: cache-riscv
+      #uses: actions/cache@v3
+      #with:
+        #path: rv64gcv-install-next
+        #key: rv64gcv-linux-install-20210504
+
+    #- name: install-riscv-build-deps
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: |
+        #sudo apt-get update
+        #sudo apt-get install autoconf automake autotools-dev curl python3 libmpc-dev libmpfr-dev libgmp-dev gawk build-essential bison flex texinfo gperf libtool patchutils bc zlib1g-dev libexpat-dev device-tree-compiler
+
+    #- name: checkout-riscv-gnu-toolchain
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #uses: actions/checkout@v4
+      #with:
+        #repository: riscv/riscv-gnu-toolchain
+        #path: riscv-gnu-toolchain
+        #ref: da01ba455ce3802ffa84fdca3a089079996dbfc3
+    #- name: checkout-riscv-gnu-toolchain-submodules
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: |
+        #cd riscv-gnu-toolchain
+        #git submodule update --init --recursive --depth 1 glibc
+        #git submodule update --init --recursive --depth 1 newlib
+        #git submodule update --init --recursive --depth 1 riscv-binutils
+        #git submodule update --init --recursive --depth 1 riscv-gcc
+        #git submodule update --init --recursive --depth 1 riscv-dejagnu
+        #git submodule update --init --recursive --depth 1 riscv-gdb
+    #- name: riscv-gnu-toolchain
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: |
+        #cd riscv-gnu-toolchain
+        #sed -i '/__OBSOLETE_MATH/d' newlib/newlib/libm/common/math_errf.c
+        #./configure --prefix=$GITHUB_WORKSPACE/rv64gcv-install-next --with-arch=rv64gcv_zfh
+        #make linux
+
+    #- name: riscv-strip-install
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: find $GITHUB_WORKSPACE/rv64gcv-install-next -type f | xargs -i strip -g {} || true
+
+    - name: configure
+      run: export RISCV_ROOT_PATH=/data/action/osd/rv64gcv-install-next && mkdir build && cd build && cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/riscv64-unknown-linux-gnu.toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+    - name: build
+      run: cmake --build build -j 4
+
+    - name: test-vlen256
+      run: |
+        export PATH=/data/action/osd/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-riscv64 TESTS_EXECUTABLE_LOADER_ARGUMENTS="-cpu;rv64,v=true,Zfh=true,vlen=256,elen=64,vext_spec=v1.0;-L;/data/action/osd/rv64gcv-install-next/sysroot" ctest --output-on-failure -j 4
+
+    - name: test-vlen128
+      run: |
+        export PATH=/data/action/osd/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-riscv64 TESTS_EXECUTABLE_LOADER_ARGUMENTS="-cpu;rv64,v=true,Zfh=true,vlen=128,elen=64,vext_spec=v1.0;-L;/data/action/osd/rv64gcv-install-next/sysroot" ctest --output-on-failure -j 4

.github/workflows/linux-riscv64-cpu-gnu-clang.yml

@@ -0,0 +1,142 @@
+name: linux-riscv64-cpu-gnu-clang
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-riscv64-cpu-gnu-clang.yml'
+    - 'toolchains/riscv64-unknown-linux-gnu.llvm-toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/riscv/**'
+    - 'tests/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-riscv64-cpu-gnu-clang.yml'
+    - 'toolchains/riscv64-unknown-linux-gnu.llvm-toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/riscv/**'
+    - 'tests/**'
+concurrency:
+  group: linux-riscv64-cpu-gnu-clang-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-gcc-riscv64-rvv:
+    runs-on: [self-hosted, linux, centos]
+    steps:
+    - uses: actions/checkout@v4
+
+    #- name: cache-qemu
+      #id: cache-qemu
+      #uses: actions/cache@v3
+      #with:
+        #path: qemu-install
+        #key: qemu-riscv64-install-20220502-3
+    #- name: install-qemu-build-deps
+      #if: steps.cache-qemu.outputs.cache-hit != 'true'
+      #run: |
+        #sudo apt-get update
+        #sudo apt-get install autoconf automake autotools-dev ninja-build
+    #- name: checkout-qemu
+      #if: steps.cache-qemu.outputs.cache-hit != 'true'
+      #uses: actions/checkout@v4
+      #with:
+        #repository: qemu/qemu
+        #path: qemu
+        #ref: f5643914a9e8f79c606a76e6a9d7ea82a3fc3e65
+    #- name: qemu
+      #if: steps.cache-qemu.outputs.cache-hit != 'true'
+      #run: |
+        #cd qemu
+        #wget https://raw.githubusercontent.com/nihui/ncnn-assets/master/qemu-patches/0007-linux-user-Expose-risc-v-V-isa-bit-in-get_elf_hwcap.patch
+        #patch -p1 -i 0007-linux-user-Expose-risc-v-V-isa-bit-in-get_elf_hwcap.patch
+        #./configure --prefix=$GITHUB_WORKSPACE/qemu-install --target-list=riscv64-linux-user --disable-system
+        #make -j2
+        #make install
+
+    #- name: cache-riscv
+      #id: cache-riscv
+      #uses: actions/cache@v3
+      #with:
+        #path: rv64gcv-install-next
+        #key: rv64gcv-linux-install-20210504
+
+    #- name: install-riscv-build-deps
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: |
+        #sudo apt-get update
+        #sudo apt-get install autoconf automake autotools-dev curl python3 libmpc-dev libmpfr-dev libgmp-dev gawk build-essential bison flex texinfo gperf libtool patchutils bc zlib1g-dev libexpat-dev device-tree-compiler
+
+    #- name: checkout-riscv-gnu-toolchain
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #uses: actions/checkout@v4
+      #with:
+        #repository: riscv/riscv-gnu-toolchain
+        #path: riscv-gnu-toolchain
+        #ref: da01ba455ce3802ffa84fdca3a089079996dbfc3
+    #- name: checkout-riscv-gnu-toolchain-submodules
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: |
+        #cd riscv-gnu-toolchain
+        #git submodule update --init --recursive --depth 1 glibc
+        #git submodule update --init --recursive --depth 1 newlib
+        #git submodule update --init --recursive --depth 1 riscv-binutils
+        #git submodule update --init --recursive --depth 1 riscv-gcc
+        #git submodule update --init --recursive --depth 1 riscv-dejagnu
+        #git submodule update --init --recursive --depth 1 riscv-gdb
+    #- name: riscv-gnu-toolchain
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: |
+        #cd riscv-gnu-toolchain
+        #sed -i '/__OBSOLETE_MATH/d' newlib/newlib/libm/common/math_errf.c
+        #./configure --prefix=$GITHUB_WORKSPACE/rv64gcv-install-next --with-arch=rv64gcv_zfh
+        #make linux
+
+    #- name: riscv-strip-install
+      #if: steps.cache-riscv.outputs.cache-hit != 'true'
+      #run: find $GITHUB_WORKSPACE/rv64gcv-install-next -type f | xargs -i strip -g {} || true
+
+    # - name: install-clang
+    #   run: |
+    #     wget https://github.com/llvm/llvm-project/releases/download/llvmorg-15.0.1/llvm-project-15.0.1.src.tar.xz
+    #     tar -xf llvm-project-15.0.1.src.tar.xz
+    #     cd llvm-project-15.0.1.src
+    #     mkdir build
+    #     cd build
+    #     cmake -DCMAKE_INSTALL_PREFIX=install -DCMAKE_BUILD_TYPE=Release -DBUILD_SHARED_LIBS=ON -DLLVM_ENABLE_PROJECTS="clang" -DLLVM_TARGETS_TO_BUILD="RISCV" -DLLVM_INCLUDE_EXAMPLES=OFF -DLLVM_INCLUDE_TESTS=OFF ../llvm/
+    #     make -j16
+    #     make install
+
+    - name: build
+      env:
+        LD_LIBRARY_PATH: /data/action/install/lib64
+      run: |
+        export RISCV_ROOT_PATH=/data/action/osd/rv64gcv-install-next
+        export PATH=/data/action/osd/llvm-project-15.0.1.src/build/install/bin:$PATH
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/riscv64-unknown-linux-gnu.llvm-toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 4
+
+    - name: test-vlen256
+      env:
+        LD_LIBRARY_PATH: /data/action/install/lib64
+      run: |
+        export PATH=/data/action/osd/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-riscv64 TESTS_EXECUTABLE_LOADER_ARGUMENTS="-cpu;rv64,v=true,Zfh=true,vlen=256,elen=64,vext_spec=v1.0;-L;/data/action/osd/rv64gcv-install-next/sysroot" ctest --output-on-failure -j 4
+
+    - name: test-vlen128
+      env:
+        LD_LIBRARY_PATH: /data/action/install/lib64
+      run: |
+        export PATH=/data/action/osd/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-riscv64 TESTS_EXECUTABLE_LOADER_ARGUMENTS="-cpu;rv64,v=true,Zfh=true,vlen=128,elen=64,vext_spec=v1.0;-L;/data/action/osd/rv64gcv-install-next/sysroot" ctest --output-on-failure -j 4

.github/workflows/linux-x64-cpu-clang-python.yml

@@ -0,0 +1,68 @@
+name: linux-x64-cpu-clang-python
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x64-cpu-clang-python.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'python/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x64-cpu-clang-python.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'python/**'
+concurrency:
+  group: linux-x64-cpu-clang-python-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-clang-python:
+    runs-on: ubuntu-20.04
+    strategy:
+      matrix:
+        python-version: [3.7, 3.9, 3.11]
+
+    steps:
+    - uses: actions/checkout@v4
+      with:
+        submodules: true
+    - name: set up python ${{ matrix.python-version }}
+      uses: actions/setup-python@v4
+      with:
+        python-version: ${{ matrix.python-version }}
+    - name: Install dependencies
+      run: |
+        python -m pip install --upgrade pip
+        pip install pytest setuptools wheel twine
+    - name: configure
+      env:
+        CC: clang
+        CXX: clang++
+      run: mkdir build && cd build && cmake -DNCNN_PYTHON=ON -DNCNN_DISABLE_RTTI=OFF -DNCNN_DISABLE_EXCEPTION=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+    - name: build
+      run: cmake --build build -j 2
+    - name: install python
+      run: cd python && pip install .
+    - name: test
+      run: cd python && pytest tests
+    - name: build and publish
+      if: startsWith(github.ref, 'refs/tags')
+      env:
+        TWINE_USERNAME: __token__
+        TWINE_PASSWORD: ${{ secrets.TEST_PYPI_API_TOKEN }}
+        TWINE_REPOSITORY_URL: "https://test.pypi.org/legacy/"
+      run: |
+        cd python
+        python setup.py bdist_wheel
+        twine upload dist/*

.github/workflows/linux-x64-cpu-clang.yml

@@ -0,0 +1,128 @@
+name: linux-x64-cpu-clang
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x64-cpu-clang.yml'
+    - 'toolchains/host-c.clang.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'tests/**'
+    - 'tools/**'
+    - '!tools/pnnx/**'
+    - 'examples/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x64-cpu-clang.yml'
+    - 'toolchains/host-c.clang.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'tests/**'
+    - 'tools/**'
+    - '!tools/pnnx/**'
+    - 'examples/**'
+concurrency:
+  group: linux-x64-cpu-clang-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-clang:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v4
+    - name: update
+      run: sudo apt-get update
+    - name: protobuf
+      run: sudo apt-get install libprotobuf-dev protobuf-compiler libopencv-dev
+    - name: build-sse2
+      env:
+        CC: clang
+        CXX: clang++
+      run: |
+        mkdir build-sse2 && cd build-sse2
+        cmake -DNCNN_AVX=OFF -DNCNN_AVX2=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test-sse2
+      run: cd build-sse2 && ctest --output-on-failure -j 2
+    - name: build-shared
+      env:
+        CC: clang
+        CXX: clang++
+      run: |
+        mkdir build-shared && cd build-shared
+        cmake -DNCNN_AVX2=ON -DNCNN_SHARED_LIB=ON ..
+        cmake --build . -j 2
+    - name: build-avx2
+      env:
+        CC: clang
+        CXX: clang++
+      run: |
+        mkdir build-avx2 && cd build-avx2
+        cmake -DNCNN_AVX2=ON -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test-avx2
+      run: cd build-avx2 && ctest --output-on-failure -j 2
+    - name: build-avx
+      env:
+        CC: clang
+        CXX: clang++
+      run: |
+        mkdir build-avx && cd build-avx
+        cmake -DNCNN_AVX2=OFF -DNCNN_AVX=ON -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test-avx
+      run: cd build-avx && ctest --output-on-failure -j 2
+    - name: build-avx1-2
+      env:
+        CC: clang
+        CXX: clang++
+      run: |
+        mkdir build-avx1-2 && cd build-avx1-2
+        cmake -DNCNN_AVX2=ON -DNCNN_AVX=ON -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test-avx1-2
+      run: cd build-avx1-2 && ctest --output-on-failure -j 2
+    - name: build-noint8
+      env:
+        CC: clang
+        CXX: clang++
+      run: |
+        mkdir build-noint8 && cd build-noint8
+        cmake -DNCNN_INT8=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test-noint8
+      run: cd build-noint8 && ctest --output-on-failure -j 2
+
+  linux-clang-simplestl:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v4
+    - name: build-simplestl
+      env:
+        CC: clang
+        CXX: clang++
+      run: |
+        mkdir build-simplestl && cd build-simplestl
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/host-c.clang.toolchain.cmake -DNCNN_STDIO=ON -DNCNN_STRING=ON -DNCNN_SIMPLESTL=ON -DNCNN_BUILD_TESTS=ON -DNCNN_BUILD_BENCHMARK=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j 2
+    - name: test-simplestl
+      run: cd build-simplestl && ctest --output-on-failure -j 2
+    - name: build-simplestl-simpleomp
+      env:
+        CC: clang
+        CXX: clang++
+      run: |
+        mkdir build-simplestl-simpleomp && cd build-simplestl-simpleomp
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/host-c.clang.toolchain.cmake -DNCNN_STDIO=ON -DNCNN_STRING=ON -DNCNN_SIMPLESTL=ON -DNCNN_SIMPLEOMP=ON -DNCNN_BUILD_TESTS=ON -DNCNN_BUILD_BENCHMARK=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j 2
+    - name: test-simplestl-simpleomp
+      run: cd build-simplestl-simpleomp && ctest --output-on-failure -j 2

.github/workflows/linux-x64-cpu-gcc-musl.yml

@@ -0,0 +1,67 @@
+name: linux-x64-cpu-gcc-musl
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x64-cpu-gcc-musl.yml'
+    - 'toolchains/host-c.gcc.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'tests/**'
+    - 'tools/**'
+    - '!tools/pnnx/**'
+    - 'examples/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x64-cpu-gcc-musl.yml'
+    - 'toolchains/host-c.gcc.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'tests/**'
+    - 'tools/**'
+    - '!tools/pnnx/**'
+    - 'examples/**'
+concurrency:
+  group: linux-x64-cpu-gcc-musl-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-gcc-musl:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: jirutka/setup-alpine@v1
+      with:
+        packages: >
+          cmake
+          clang
+          clang-dev
+          make
+          gcc
+          g++
+          libc-dev
+          linux-headers
+
+    - uses: actions/checkout@v4
+    - name: build
+      shell: alpine.sh {0}
+      run: |
+        mkdir build && cd build
+        cmake -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test
+      shell: alpine.sh {0}
+      run: cd build && ctest --output-on-failure -j 2
+    - name: build-shared
+      run: |
+        mkdir build-shared && cd build-shared
+        cmake -DNCNN_SHARED_LIB=ON ..
+        cmake --build . -j 2

.github/workflows/linux-x64-cpu-gcc-san.yml

@@ -0,0 +1,42 @@
+name: linux-x64-cpu-gcc-san
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x64-cpu-gcc-san.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'tests/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x64-cpu-gcc-san.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'tests/**'
+concurrency:
+  group: linux-x64-cpu-gcc-san-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-gcc-san:
+    runs-on: ubuntu-22.04
+    steps:
+    - uses: actions/checkout@v4
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_ASAN=ON -DNCNN_BUILD_TESTS=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j 2
+    - name: test
+      run: |
+        cd build
+        ctest --output-on-failure -j 2

.github/workflows/linux-x64-cpu-gcc-sde.yml

@@ -0,0 +1,57 @@
+name: linux-x64-cpu-gcc-sde
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x64-cpu-gcc-sde.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'tests/**'
+    - 'tools/**'
+    - '!tools/pnnx/**'
+    - 'examples/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x64-cpu-gcc-sde.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'tests/**'
+    - 'tools/**'
+    - '!tools/pnnx/**'
+    - 'examples/**'
+concurrency:
+  group: linux-x64-cpu-gcc-sde-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-gcc-sde:
+    runs-on: ubuntu-22.04
+    steps:
+    - uses: actions/checkout@v4
+    - name: update
+      run: sudo apt-get update
+    - name: gcc12
+      run: sudo apt-get install gcc-12 g++-12
+    - name: Setup SDE binaries
+      uses: petarpetrovt/setup-sde@v2
+    - name: build-avx512-spr
+      env:
+        CC: gcc-12
+        CXX: g++-12
+      run: |
+        mkdir build-avx512-spr && cd build-avx512-spr
+        cmake -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test-avx512-spr
+      run: |
+        cd build-avx512-spr
+        TESTS_EXECUTABLE_LOADER=$SDE_PATH/sde64 TESTS_EXECUTABLE_LOADER_ARGUMENTS="-spr;--" ctest --output-on-failure -j 2

.github/workflows/linux-x64-cpu-gcc.yml

@@ -0,0 +1,134 @@
+name: linux-x64-cpu-gcc
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x64-cpu-gcc.yml'
+    - 'toolchains/host-c.gcc.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'tests/**'
+    - 'tools/**'
+    - '!tools/pnnx/**'
+    - 'examples/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x64-cpu-gcc.yml'
+    - 'toolchains/host-c.gcc.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'tests/**'
+    - 'tools/**'
+    - '!tools/pnnx/**'
+    - 'examples/**'
+concurrency:
+  group: linux-x64-cpu-gcc-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-gcc:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+    - name: update
+      run: sudo apt-get update
+    - name: protobuf
+      run: sudo apt-get install libprotobuf-dev protobuf-compiler libopencv-dev
+    - name: build-sse2
+      run: |
+        mkdir build-sse2 && cd build-sse2
+        cmake -DNCNN_AVX=OFF -DNCNN_AVX2=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test-sse2
+      run: cd build-sse2 && ctest --output-on-failure -j 2
+    - name: build-shared
+      run: |
+        mkdir build-shared && cd build-shared
+        cmake -DNCNN_AVX2=ON -DNCNN_SHARED_LIB=ON ..
+        cmake --build . -j 2
+    - name: build-avx2
+      run: |
+        mkdir build-avx2 && cd build-avx2
+        cmake -DNCNN_AVX2=ON -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test-avx2
+      run: cd build-avx2 && ctest --output-on-failure -j 2
+    - name: build-avx
+      run: |
+        mkdir build-avx && cd build-avx
+        cmake -DNCNN_AVX2=OFF -DNCNN_AVX=ON -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test-avx
+      run: cd build-avx && ctest --output-on-failure -j 2
+    - name: build-avx1-2
+      run: |
+        mkdir build-avx1-2 && cd build-avx1-2
+        cmake -DNCNN_AVX2=ON -DNCNN_AVX=ON -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test-avx1-2
+      run: cd build-avx1-2 && ctest --output-on-failure -j 2
+    - name: build-noint8
+      run: |
+        mkdir build-noint8 && cd build-noint8
+        cmake -DNCNN_INT8=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 2
+    - name: test-noint8
+      run: cd build-noint8 && ctest --output-on-failure -j 2
+
+  linux-gcc-cpp03-nostdio-nostring-simplestl:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+    - name: build-nostdio
+      run: |
+        mkdir build-nostdio && cd build-nostdio
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/host.gcc-c++03.toolchain.cmake -DNCNN_BUILD_TESTS=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j 2
+    - name: test-nostdio
+      run: cd build-nostdio && ctest --output-on-failure -j 2
+    - name: build-nostdio-nostring
+      run: |
+        mkdir build-nostdio-nostring && cd build-nostdio-nostring
+        cmake -DNCNN_STDIO=OFF -DNCNN_STRING=OFF -DNCNN_BUILD_TESTS=OFF -DNCNN_BUILD_BENCHMARK=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j 2
+    - name: build-simplestl
+      run: |
+        mkdir build-simplestl && cd build-simplestl
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/host-c.gcc.toolchain.cmake -DNCNN_STDIO=ON -DNCNN_STRING=ON -DNCNN_SIMPLESTL=ON -DNCNN_BUILD_TESTS=ON -DNCNN_BUILD_BENCHMARK=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j 2
+    - name: test-simplestl
+      run: cd build-simplestl && ctest --output-on-failure -j 2
+    - name: build-simplestl-simpleomp
+      run: |
+        mkdir build-simplestl-simpleomp && cd build-simplestl-simpleomp
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/host-c.gcc.toolchain.cmake -DNCNN_STDIO=ON -DNCNN_STRING=ON -DNCNN_SIMPLESTL=ON -DNCNN_SIMPLEOMP=ON -DNCNN_BUILD_TESTS=ON -DNCNN_BUILD_BENCHMARK=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j 2
+    - name: test-simplestl-simpleomp
+      run: cd build-simplestl-simpleomp && ctest --output-on-failure -j 2
+
+  linux-gcc-avx512:
+    runs-on: [self-hosted, linux, t4]
+    steps:
+    - uses: actions/checkout@v4
+    - name: build
+      env:
+        CC: gcc
+        CXX: g++
+        LD_LIBRARY_PATH: /data/action/install/lib64
+      run: |
+        mkdir build && cd build
+        cmake -DNCNN_AVX2=ON -DNCNN_AVX512=ON -DNCNN_AVX512VNNI=ON -DNCNN_BUILD_TESTS=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j 4
+    - name: test
+      env:
+        LD_LIBRARY_PATH: /data/action/install/lib64
+      run: cd build && ctest --output-on-failure -j 4

.github/workflows/linux-x64-gpu-clang-python.yml

@@ -0,0 +1,113 @@
+name: linux-x64-gpu-clang-python
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x64-gpu-clang-python.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'src/layer/vulkan/**'
+    - 'python/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x64-gpu-clang-python.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'src/layer/vulkan/**'
+    - 'python/**'
+concurrency:
+  group: linux-x64-gpu-clang-python-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-clang-gpu:
+    runs-on: ubuntu-20.04
+    strategy:
+      matrix:
+        python-version: [3.7, 3.9, 3.11]
+    steps:
+    - uses: actions/checkout@v4
+      with:
+        submodules: true
+    - name: cache-vulkansdk
+      id: cache-vulkansdk
+      uses: actions/cache@v3
+      with:
+        path: "1.3.236.0"
+        key: vulkansdk-linux-x86_64-1.3.236.0
+    - name: vulkansdk
+      if: steps.cache-vulkansdk.outputs.cache-hit != 'true'
+      run: |
+        wget https://sdk.lunarg.com/sdk/download/1.3.236.0/linux/vulkansdk-linux-x86_64-1.3.236.0.tar.gz?Human=true -O vulkansdk-linux-x86_64-1.3.236.0.tar.gz
+        tar -xf vulkansdk-linux-x86_64-1.3.236.0.tar.gz
+        rm -rf 1.3.236.0/source 1.3.236.0/samples
+        find 1.3.236.0 -type f | grep -v -E 'vulkan|glslang' | xargs rm
+    - name: cache-swiftshader
+      id: cache-swiftshader
+      uses: actions/cache@v3
+      with:
+        path: swiftshader-install
+        key: swiftshader-linux-install-20230420
+    - name: checkout-swiftshader
+      if: steps.cache-swiftshader.outputs.cache-hit != 'true'
+      uses: actions/checkout@v4
+      with:
+        repository: google/swiftshader
+        path: swiftshader
+        ref: dd55e592406dc0bae219df11adec6363840aff4a
+    - name: checkout-swiftshader-submodules
+      if: steps.cache-swiftshader.outputs.cache-hit != 'true'
+      run: |
+        cd swiftshader
+        git -c submodule."third_party/git-hooks".update=none submodule update --init --recursive
+    - name: swiftshader
+      if: steps.cache-swiftshader.outputs.cache-hit != 'true'
+      run: |
+        cd swiftshader
+        mkdir -p build; cd build
+        cmake -DCMAKE_INSTALL_PREFIX=install -DSWIFTSHADER_BUILD_EGL=FALSE -DSWIFTSHADER_BUILD_GLESv2=FALSE -DSWIFTSHADER_BUILD_GLES_CM=FALSE -DSWIFTSHADER_BUILD_VULKAN=TRUE -DSWIFTSHADER_BUILD_PVR=FALSE -DSWIFTSHADER_BUILD_TESTS=FALSE -DSWIFTSHADER_ENABLE_ASTC=FALSE -DSWIFTSHADER_WARNINGS_AS_ERRORS=FALSE -DREACTOR_BACKEND=Subzero -DREACTOR_DEFAULT_OPT_LEVEL=Default -DCMAKE_BUILD_TYPE=Release ..
+        cmake --build . -j 2
+        mkdir $GITHUB_WORKSPACE/swiftshader-install
+        cp Linux/* $GITHUB_WORKSPACE/swiftshader-install
+    - name: set up python ${{ matrix.python-version }}
+      uses: actions/setup-python@v4
+      with:
+        python-version: ${{ matrix.python-version }}
+    - name: Install dependencies
+      run: |
+        python -m pip install --upgrade pip
+        pip install pytest setuptools wheel twine
+    - name: build
+      env:
+        CC: clang
+        CXX: clang++
+      run: |
+        export VULKAN_SDK=`pwd`/1.3.236.0/x86_64
+        mkdir build && cd build
+        cmake -DNCNN_VULKAN=ON -DNCNN_PYTHON=ON -DNCNN_DISABLE_RTTI=OFF -DNCNN_DISABLE_EXCEPTION=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j 2
+    - name: install python
+      run: cd python && pip install .
+    - name: test
+      run: |
+        export VK_ICD_FILENAMES="$GITHUB_WORKSPACE/swiftshader-install/vk_swiftshader_icd.json"
+        cd python && pytest tests
+    - name: build and publish
+      if: startsWith(github.ref, 'refs/tags')
+      env:
+        TWINE_USERNAME: __token__
+        TWINE_PASSWORD: ${{ secrets.TEST_PYPI_API_TOKEN }}
+        TWINE_REPOSITORY_URL: "https://test.pypi.org/legacy/"
+      run: |
+        cd python
+        python setup.py bdist_wheel
+        twine upload dist/*

.github/workflows/linux-x64-gpu-clang.yml

@@ -0,0 +1,91 @@
+name: linux-x64-gpu-clang
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x64-gpu-clang.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'src/layer/vulkan/**'
+    - 'tests/**'
+    - 'tools/**'
+    - '!tools/pnnx/**'
+    - 'examples/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x64-gpu-clang.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'src/layer/vulkan/**'
+    - 'tests/**'
+    - 'tools/**'
+    - '!tools/pnnx/**'
+    - 'examples/**'
+concurrency:
+  group: linux-x64-gpu-clang-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-clang-gpu:
+    runs-on: [self-hosted, linux, cvm]
+    steps:
+    - uses: actions/checkout@v4
+      with:
+        submodules: true
+    - name: cache-swiftshader
+      id: cache-swiftshader
+      uses: actions/cache@v3
+      with:
+        path: swiftshader-install
+        key: swiftshader-linux-install-20230420
+    - name: checkout-swiftshader
+      if: steps.cache-swiftshader.outputs.cache-hit != 'true'
+      uses: actions/checkout@v4
+      with:
+        repository: google/swiftshader
+        path: swiftshader
+        ref: dd55e592406dc0bae219df11adec6363840aff4a
+    - name: checkout-swiftshader-submodules
+      if: steps.cache-swiftshader.outputs.cache-hit != 'true'
+      run: |
+        cd swiftshader
+        git -c submodule."third_party/git-hooks".update=none submodule update --init --recursive
+    - name: swiftshader
+      if: steps.cache-swiftshader.outputs.cache-hit != 'true'
+      run: |
+        cd swiftshader
+        mkdir -p build; cd build
+        cmake -DCMAKE_INSTALL_PREFIX=install -DSWIFTSHADER_BUILD_EGL=FALSE -DSWIFTSHADER_BUILD_GLESv2=FALSE -DSWIFTSHADER_BUILD_GLES_CM=FALSE -DSWIFTSHADER_BUILD_VULKAN=TRUE -DSWIFTSHADER_BUILD_PVR=FALSE -DSWIFTSHADER_BUILD_TESTS=FALSE -DSWIFTSHADER_ENABLE_ASTC=FALSE -DSWIFTSHADER_WARNINGS_AS_ERRORS=FALSE -DREACTOR_BACKEND=Subzero -DREACTOR_DEFAULT_OPT_LEVEL=Default -DCMAKE_BUILD_TYPE=Release ..
+        cmake --build . -j 4
+        mkdir $GITHUB_WORKSPACE/swiftshader-install
+        cp Linux/* $GITHUB_WORKSPACE/swiftshader-install
+    - name: build
+      env:
+        CC: clang
+        CXX: clang++
+      run: |
+        mkdir build && cd build
+        cmake -DNCNN_VULKAN=ON -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 4
+    - name: test
+      run: |
+        printf "[Processor]\nThreadCount=1\n" > build/tests/SwiftShader.ini
+        export VK_ICD_FILENAMES="$GITHUB_WORKSPACE/swiftshader-install/vk_swiftshader_icd.json"
+        cd build && ctest --output-on-failure -j 4
+    - name: build-shared
+      env:
+        CC: clang
+        CXX: clang++
+      run: |
+        mkdir build-shared && cd build-shared
+        cmake -DNCNN_VULKAN=ON -DNCNN_SHARED_LIB=ON ..
+        cmake --build . -j 4

.github/workflows/linux-x64-gpu-gcc.yml

@@ -0,0 +1,128 @@
+name: linux-x64-gpu-gcc
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x64-gpu-gcc.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'src/layer/vulkan/**'
+    - 'tests/**'
+    - 'tools/**'
+    - '!tools/pnnx/**'
+    - 'examples/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x64-gpu-gcc.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'src/layer/vulkan/**'
+    - 'tests/**'
+    - 'tools/**'
+    - '!tools/pnnx/**'
+    - 'examples/**'
+concurrency:
+  group: linux-x64-gpu-gcc-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-gcc-gpu:
+    runs-on: [self-hosted, linux, cvm]
+    steps:
+    - uses: actions/checkout@v4
+      with:
+        submodules: true
+    - name: cache-swiftshader
+      id: cache-swiftshader
+      uses: actions/cache@v3
+      with:
+        path: swiftshader-install
+        key: swiftshader-linux-install-20230420
+    - name: checkout-swiftshader
+      if: steps.cache-swiftshader.outputs.cache-hit != 'true'
+      uses: actions/checkout@v4
+      with:
+        repository: google/swiftshader
+        path: swiftshader
+        ref: dd55e592406dc0bae219df11adec6363840aff4a
+    - name: checkout-swiftshader-submodules
+      if: steps.cache-swiftshader.outputs.cache-hit != 'true'
+      run: |
+        cd swiftshader
+        git -c submodule."third_party/git-hooks".update=none submodule update --init --recursive
+    - name: swiftshader
+      if: steps.cache-swiftshader.outputs.cache-hit != 'true'
+      run: |
+        cd swiftshader
+        mkdir -p build; cd build
+        cmake -DCMAKE_INSTALL_PREFIX=install -DSWIFTSHADER_BUILD_EGL=FALSE -DSWIFTSHADER_BUILD_GLESv2=FALSE -DSWIFTSHADER_BUILD_GLES_CM=FALSE -DSWIFTSHADER_BUILD_VULKAN=TRUE -DSWIFTSHADER_BUILD_PVR=FALSE -DSWIFTSHADER_BUILD_TESTS=FALSE -DSWIFTSHADER_ENABLE_ASTC=FALSE -DSWIFTSHADER_WARNINGS_AS_ERRORS=FALSE -DREACTOR_BACKEND=Subzero -DREACTOR_DEFAULT_OPT_LEVEL=Default -DCMAKE_BUILD_TYPE=Release ..
+        cmake --build . -j 4
+        mkdir $GITHUB_WORKSPACE/swiftshader-install
+        cp Linux/* $GITHUB_WORKSPACE/swiftshader-install
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DNCNN_VULKAN=ON -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 4
+    - name: test
+      run: |
+        printf "[Processor]\nThreadCount=1\n" > build/tests/SwiftShader.ini
+        export VK_ICD_FILENAMES="$GITHUB_WORKSPACE/swiftshader-install/vk_swiftshader_icd.json"
+        cd build && ctest --output-on-failure -j 4
+    - name: build-shared
+      run: |
+        mkdir build-shared && cd build-shared
+        cmake -DNCNN_VULKAN=ON -DNCNN_SHARED_LIB=ON ..
+        cmake --build . -j 4
+
+  linux-gcc-gpu-system-glslang:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: install-deps
+      run: |
+        sudo apt-get update
+        sudo apt-get install libprotobuf-dev protobuf-compiler libopencv-dev libvulkan-dev glslang-dev spirv-tools
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DNCNN_VULKAN=ON -DNCNN_SYSTEM_GLSLANG=ON -DGLSLANG_TARGET_DIR=/usr/lib/x86_64-linux-gnu/cmake ..
+        cmake --build . -j 4
+    - name: build-shared
+      run: |
+        mkdir build-shared && cd build-shared
+        cmake -DNCNN_VULKAN=ON -DNCNN_SYSTEM_GLSLANG=ON -DGLSLANG_TARGET_DIR=/usr/lib/x86_64-linux-gnu/cmake -DNCNN_SHARED_LIB=ON ..
+        cmake --build . -j 4
+
+  linux-gcc-gpu-t4:
+    runs-on: [self-hosted, linux, t4]
+    steps:
+    - uses: actions/checkout@v4
+      with:
+        submodules: true
+    - name: build
+      env:
+        CC: gcc
+        CXX: g++
+        LD_LIBRARY_PATH: /data/action/install/lib64
+      run: |
+        export VULKAN_SDK=/data/action/osd/1.2.189.0/x86_64
+        mkdir build && cd build
+        cmake -DNCNN_VULKAN=ON -DNCNN_BUILD_TESTS=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j 4
+    - name: test
+      env:
+        LD_LIBRARY_PATH: /data/action/install/lib64
+      run: |
+        cd build && ctest --output-on-failure -j 4

.github/workflows/linux-x86-cpu-clang.yml

@@ -0,0 +1,67 @@
+name: linux-x86-cpu-clang
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x86-cpu-clang.yml'
+    - 'toolchains/host.clang-m32.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'tests/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x86-cpu-clang.yml'
+    - 'toolchains/host.clang-m32.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'tests/**'
+concurrency:
+  group: linux-x86-cpu-clang-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-clang:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v4
+    - name: update
+      run: sudo apt-get update
+    - name: gcc-multilib
+      run: sudo apt-get install gcc-multilib g++-multilib
+    - name: build
+      env:
+        CC: clang
+        CXX: clang++
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/host.clang-m32.toolchain.cmake -DNCNN_BUILD_TESTS=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j 2
+    - name: test
+      run: cd build && ctest --output-on-failure -j 2
+    - name: build-shared
+      env:
+        CC: clang
+        CXX: clang++
+      run: |
+        mkdir build-shared && cd build-shared
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/host.clang-m32.toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_SHARED_LIB=ON ..
+        cmake --build . -j 2
+    - name: build-noint8
+      env:
+        CC: clang
+        CXX: clang++
+      run: |
+        mkdir build-noint8 && cd build-noint8
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/host.clang-m32.toolchain.cmake -DNCNN_BUILD_TESTS=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_INT8=OFF ..
+        cmake --build . -j 2
+    - name: test-noint8
+      run: cd build-noint8 && ctest --output-on-failure -j 2

.github/workflows/linux-x86-cpu-gcc.yml

@@ -0,0 +1,65 @@
+name: linux-x86-cpu-gcc
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x86-cpu-gcc.yml'
+    - 'toolchains/host.gcc-m32.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'tests/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/linux-x86-cpu-gcc.yml'
+    - 'toolchains/host.gcc-m32.toolchain.cmake'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'tests/**'
+concurrency:
+  group: linux-x86-cpu-gcc-${{ github.ref }}
+  cancel-in-progress: true
+permissions:
+  contents: read
+
+jobs:
+  linux-gcc:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v4
+    - name: update
+      run: sudo apt-get update
+    - name: gcc-multilib
+      run: sudo apt-get install gcc-multilib g++-multilib
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/host.gcc-m32.toolchain.cmake -DNCNN_BUILD_TESTS=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j 2
+    - name: test
+      run: cd build && ctest --output-on-failure -j 2
+    - name: build-nosse
+      run: |
+        mkdir build-nosse && cd build-nosse
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/host.gcc-m32.toolchain.cmake -DNCNN_RUNTIME_CPU=OFF -DNCNN_SSE2=OFF -DNCNN_AVX=OFF -DNCNN_BUILD_TESTS=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j 2
+    - name: test-nosse
+      run: cd build-nosse && ctest --output-on-failure -j 2
+    - name: build-shared
+      run: |
+        mkdir build-shared && cd build-shared
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/host.gcc-m32.toolchain.cmake -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_SHARED_LIB=ON ..
+        cmake --build . -j 2
+    - name: build-noint8
+      run: |
+        mkdir build-noint8 && cd build-noint8
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/host.gcc-m32.toolchain.cmake -DNCNN_BUILD_TESTS=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_INT8=OFF ..
+        cmake --build . -j 2
+    - name: test-noint8
+      run: cd build-noint8 && ctest --output-on-failure -j 2

.github/workflows/mac-catalyst-arm64-cpu.yml

@@ -0,0 +1,103 @@
+name: mac-catalyst-arm64-cpu
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/mac-catalyst-arm64-cpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/mac-catalyst-arm64-cpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+concurrency:
+  group: mac-catalyst-arm64-cpu-${{ github.ref }}
+  cancel-in-progress: true
+env:
+  DEVELOPER_DIR: /Applications/Xcode_13.4.1.app/Contents/Developer
+  MAC_CATALYST_DEPLOYMENT_TARGET: '13.1'
+  ENABLE_BITCODE: OFF
+  ENABLE_ARC: OFF
+  ENABLE_VISIBILITY: OFF
+permissions:
+  contents: read
+
+jobs:
+  mac-catalyst-clang:
+    runs-on: macos-12
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-openmp
+      id: cache-openmp
+      uses: actions/cache@v3
+      with:
+        path: openmp-install
+        key: openmp-mac-catalyst-install-20230504
+    - name: openmp
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        wget https://github.com/llvm/llvm-project/releases/download/llvmorg-11.0.0/openmp-11.0.0.src.tar.xz
+        tar -xf openmp-11.0.0.src.tar.xz
+        cd openmp-11.0.0.src
+        sed -i'' -e '/.size __kmp_unnamed_critical_addr/d' runtime/src/z_Linux_asm.S
+        sed -i'' -e 's/__kmp_unnamed_critical_addr/___kmp_unnamed_critical_addr/g' runtime/src/z_Linux_asm.S
+    - name: openmp-build-x86_64
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-x86_64 && cd build-x86_64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=MAC_CATALYST -DARCHS="x86_64" \
+            -DDEPLOYMENT_TARGET=$MAC_CATALYST_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-build-arm64
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-arm64 && cd build-arm64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=MAC_CATALYST_ARM64 -DARCHS="arm64" \
+            -DDEPLOYMENT_TARGET=$MAC_CATALYST_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-merge-fat-library
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        mkdir -p $GITHUB_WORKSPACE/openmp-install
+        cp -a openmp-11.0.0.src/build-x86_64/install/include $GITHUB_WORKSPACE/openmp-install
+        mkdir -p $GITHUB_WORKSPACE/openmp-install/lib
+        lipo -create \
+            openmp-11.0.0.src/build-x86_64/install/lib/libomp.a \
+            openmp-11.0.0.src/build-arm64/install/lib/libomp.a \
+            -o $GITHUB_WORKSPACE/openmp-install/lib/libomp.a
+
+    - name: install-openmp
+      run: |
+        sudo cp $GITHUB_WORKSPACE/openmp-install/include/* $DEVELOPER_DIR/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/usr/include
+        sudo cp $GITHUB_WORKSPACE/openmp-install/lib/libomp.a $DEVELOPER_DIR/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/usr/lib
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/ios.toolchain.cmake -DPLATFORM=MAC_CATALYST_ARM64 -DARCHS="arm64" \
+            -DDEPLOYMENT_TARGET=$MAC_CATALYST_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DOpenMP_C_FLAGS="-Xclang -fopenmp" -DOpenMP_CXX_FLAGS="-Xclang -fopenmp" \
+            -DOpenMP_C_LIB_NAMES="libomp" -DOpenMP_CXX_LIB_NAMES="libomp" \
+            -DOpenMP_libomp_LIBRARY="$DEVELOPER_DIR/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/usr/lib/libomp.a" \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j 3

.github/workflows/mac-catalyst-arm64-gpu.yml

@@ -0,0 +1,117 @@
+name: mac-catalyst-arm64-gpu
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/mac-catalyst-arm64-gpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+    - 'src/layer/vulkan/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/mac-catalyst-arm64-gpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/arm/**'
+    - 'src/layer/vulkan/**'
+concurrency:
+  group: mac-catalyst-arm64-gpu-${{ github.ref }}
+  cancel-in-progress: true
+env:
+  DEVELOPER_DIR: /Applications/Xcode_13.4.1.app/Contents/Developer
+  MAC_CATALYST_DEPLOYMENT_TARGET: '13.1'
+  ENABLE_BITCODE: OFF
+  ENABLE_ARC: OFF
+  ENABLE_VISIBILITY: OFF
+permissions:
+  contents: read
+
+jobs:
+  mac-catalyst-clang-gpu:
+    runs-on: macos-12
+    steps:
+    - uses: actions/checkout@v4
+      with:
+        submodules: true
+
+    - name: cache-openmp
+      id: cache-openmp
+      uses: actions/cache@v3
+      with:
+        path: openmp-install
+        key: openmp-mac-catalyst-install-20230504
+    - name: openmp
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        wget https://github.com/llvm/llvm-project/releases/download/llvmorg-11.0.0/openmp-11.0.0.src.tar.xz
+        tar -xf openmp-11.0.0.src.tar.xz
+        cd openmp-11.0.0.src
+        sed -i'' -e '/.size __kmp_unnamed_critical_addr/d' runtime/src/z_Linux_asm.S
+        sed -i'' -e 's/__kmp_unnamed_critical_addr/___kmp_unnamed_critical_addr/g' runtime/src/z_Linux_asm.S
+    - name: openmp-build-x86_64
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-x86_64 && cd build-x86_64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=MAC_CATALYST -DARCHS="x86_64" \
+            -DDEPLOYMENT_TARGET=$MAC_CATALYST_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-build-arm64
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-arm64 && cd build-arm64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=MAC_CATALYST_ARM64 -DARCHS="arm64" \
+            -DDEPLOYMENT_TARGET=$MAC_CATALYST_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-merge-fat-library
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        mkdir -p $GITHUB_WORKSPACE/openmp-install
+        cp -a openmp-11.0.0.src/build-x86_64/install/include $GITHUB_WORKSPACE/openmp-install
+        mkdir -p $GITHUB_WORKSPACE/openmp-install/lib
+        lipo -create \
+            openmp-11.0.0.src/build-x86_64/install/lib/libomp.a \
+            openmp-11.0.0.src/build-arm64/install/lib/libomp.a \
+            -o $GITHUB_WORKSPACE/openmp-install/lib/libomp.a
+
+    - name: install-openmp
+      run: |
+        sudo cp $GITHUB_WORKSPACE/openmp-install/include/* $DEVELOPER_DIR/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/usr/include
+        sudo cp $GITHUB_WORKSPACE/openmp-install/lib/libomp.a $DEVELOPER_DIR/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/usr/lib
+
+    - name: vulkansdk
+      run: |
+        wget https://sdk.lunarg.com/sdk/download/1.3.236.0/mac/vulkansdk-macos-1.3.236.0.dmg?Human=true -O vulkansdk-macos-1.3.236.0.dmg
+        hdiutil attach vulkansdk-macos-1.3.236.0.dmg
+        sudo /Volumes/vulkansdk-macos-1.3.236.0/InstallVulkan.app/Contents/MacOS/InstallVulkan --root $GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0 --accept-licenses --default-answer --confirm-command install
+        hdiutil detach /Volumes/vulkansdk-macos-1.3.236.0
+
+    - name: build
+      run: |
+        export VULKAN_SDK=$GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0/macOS
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/ios.toolchain.cmake -DPLATFORM=MAC_CATALYST_ARM64 -DARCHS="arm64" \
+            -DDEPLOYMENT_TARGET=$MAC_CATALYST_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DOpenMP_C_FLAGS="-Xclang -fopenmp" -DOpenMP_CXX_FLAGS="-Xclang -fopenmp" \
+            -DOpenMP_C_LIB_NAMES="libomp" -DOpenMP_CXX_LIB_NAMES="libomp" \
+            -DOpenMP_libomp_LIBRARY="$DEVELOPER_DIR/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/usr/lib/libomp.a" \
+            -DVulkan_INCLUDE_DIR=$GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0/MoltenVK/include \
+            -DVulkan_LIBRARY=$GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0/MoltenVK/dylib/macOS/libMoltenVK.dylib \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DNCNN_VULKAN=ON -DNCNN_BUILD_BENCHMARK=OFF ..
+        cmake --build . -j 3

.github/workflows/mac-catalyst-x64-cpu.yml

@@ -0,0 +1,111 @@
+name: mac-catalyst-x64-cpu
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/mac-catalyst-x64-cpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'tests/**'
+    - 'tools/**'
+    - '!tools/pnnx/**'
+    - 'examples/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/mac-catalyst-x64-cpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'tests/**'
+    - 'tools/**'
+    - '!tools/pnnx/**'
+    - 'examples/**'
+concurrency:
+  group: mac-catalyst-x64-cpu-${{ github.ref }}
+  cancel-in-progress: true
+env:
+  DEVELOPER_DIR: /Applications/Xcode_13.4.1.app/Contents/Developer
+  MAC_CATALYST_DEPLOYMENT_TARGET: '13.1'
+  ENABLE_BITCODE: OFF
+  ENABLE_ARC: OFF
+  ENABLE_VISIBILITY: OFF
+permissions:
+  contents: read
+
+jobs:
+  mac-catalyst-clang:
+    runs-on: macos-12
+    steps:
+    - uses: actions/checkout@v4
+
+    - name: cache-openmp
+      id: cache-openmp
+      uses: actions/cache@v3
+      with:
+        path: openmp-install
+        key: openmp-mac-catalyst-install-20230504
+    - name: openmp
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        wget https://github.com/llvm/llvm-project/releases/download/llvmorg-11.0.0/openmp-11.0.0.src.tar.xz
+        tar -xf openmp-11.0.0.src.tar.xz
+        cd openmp-11.0.0.src
+        sed -i'' -e '/.size __kmp_unnamed_critical_addr/d' runtime/src/z_Linux_asm.S
+        sed -i'' -e 's/__kmp_unnamed_critical_addr/___kmp_unnamed_critical_addr/g' runtime/src/z_Linux_asm.S
+    - name: openmp-build-x86_64
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-x86_64 && cd build-x86_64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=MAC_CATALYST -DARCHS="x86_64" \
+            -DDEPLOYMENT_TARGET=$MAC_CATALYST_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-build-arm64
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-arm64 && cd build-arm64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=MAC_CATALYST_ARM64 -DARCHS="arm64" \
+            -DDEPLOYMENT_TARGET=$MAC_CATALYST_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-merge-fat-library
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        mkdir -p $GITHUB_WORKSPACE/openmp-install
+        cp -a openmp-11.0.0.src/build-x86_64/install/include $GITHUB_WORKSPACE/openmp-install
+        mkdir -p $GITHUB_WORKSPACE/openmp-install/lib
+        lipo -create \
+            openmp-11.0.0.src/build-x86_64/install/lib/libomp.a \
+            openmp-11.0.0.src/build-arm64/install/lib/libomp.a \
+            -o $GITHUB_WORKSPACE/openmp-install/lib/libomp.a
+
+    - name: install-openmp
+      run: |
+        sudo cp $GITHUB_WORKSPACE/openmp-install/include/* $DEVELOPER_DIR/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/usr/include
+        sudo cp $GITHUB_WORKSPACE/openmp-install/lib/libomp.a $DEVELOPER_DIR/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/usr/lib
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/ios.toolchain.cmake -DPLATFORM=MAC_CATALYST -DARCHS="x86_64" \
+            -DDEPLOYMENT_TARGET=$MAC_CATALYST_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DOpenMP_C_FLAGS="-Xclang -fopenmp" -DOpenMP_CXX_FLAGS="-Xclang -fopenmp" \
+            -DOpenMP_C_LIB_NAMES="libomp" -DOpenMP_CXX_LIB_NAMES="libomp" \
+            -DOpenMP_libomp_LIBRARY="$DEVELOPER_DIR/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/usr/lib/libomp.a" \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            ..
+        cmake --build . -j 3

.github/workflows/mac-catalyst-x64-gpu.yml

@@ -0,0 +1,125 @@
+name: mac-catalyst-x64-gpu
+on:
+  push:
+    branches: [master]
+    paths:
+    - '.github/workflows/mac-catalyst-x64-gpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'src/layer/vulkan/**'
+    - 'tests/**'
+    - 'tools/**'
+    - '!tools/pnnx/**'
+    - 'examples/**'
+  pull_request:
+    branches: [master]
+    paths:
+    - '.github/workflows/mac-catalyst-x64-gpu.yml'
+    - 'CMakeLists.txt'
+    - 'cmake/**'
+    - 'src/*'
+    - 'src/layer/*'
+    - 'src/layer/x86/**'
+    - 'src/layer/vulkan/**'
+    - 'tests/**'
+    - 'tools/**'
+    - '!tools/pnnx/**'
+    - 'examples/**'
+concurrency:
+  group: mac-catalyst-x64-gpu-${{ github.ref }}
+  cancel-in-progress: true
+env:
+  DEVELOPER_DIR: /Applications/Xcode_13.4.1.app/Contents/Developer
+  MAC_CATALYST_DEPLOYMENT_TARGET: '13.1'
+  ENABLE_BITCODE: OFF
+  ENABLE_ARC: OFF
+  ENABLE_VISIBILITY: OFF
+permissions:
+  contents: read
+
+jobs:
+  mac-catalyst-clang-gpu:
+    runs-on: macos-12
+    steps:
+    - uses: actions/checkout@v4
+      with:
+        submodules: true
+
+    - name: cache-openmp
+      id: cache-openmp
+      uses: actions/cache@v3
+      with:
+        path: openmp-install
+        key: openmp-mac-catalyst-install-20230504
+    - name: openmp
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        wget https://github.com/llvm/llvm-project/releases/download/llvmorg-11.0.0/openmp-11.0.0.src.tar.xz
+        tar -xf openmp-11.0.0.src.tar.xz
+        cd openmp-11.0.0.src
+        sed -i'' -e '/.size __kmp_unnamed_critical_addr/d' runtime/src/z_Linux_asm.S
+        sed -i'' -e 's/__kmp_unnamed_critical_addr/___kmp_unnamed_critical_addr/g' runtime/src/z_Linux_asm.S
+    - name: openmp-build-x86_64
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-x86_64 && cd build-x86_64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=MAC_CATALYST -DARCHS="x86_64" \
+            -DDEPLOYMENT_TARGET=$MAC_CATALYST_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-build-arm64
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        cd openmp-11.0.0.src
+        mkdir -p build-arm64 && cd build-arm64
+        cmake -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake -DPLATFORM=MAC_CATALYST_ARM64 -DARCHS="arm64" \
+            -DDEPLOYMENT_TARGET=$MAC_CATALYST_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DPERL_EXECUTABLE=/usr/local/bin/perl \
+            -DLIBOMP_ENABLE_SHARED=OFF -DLIBOMP_OMPT_SUPPORT=OFF -DLIBOMP_USE_HWLOC=OFF ..
+        cmake --build . -j 3
+        cmake --build . --target install
+    - name: openmp-merge-fat-library
+      if: steps.cache-openmp.outputs.cache-hit != 'true'
+      run: |
+        mkdir -p $GITHUB_WORKSPACE/openmp-install
+        cp -a openmp-11.0.0.src/build-x86_64/install/include $GITHUB_WORKSPACE/openmp-install
+        mkdir -p $GITHUB_WORKSPACE/openmp-install/lib
+        lipo -create \
+            openmp-11.0.0.src/build-x86_64/install/lib/libomp.a \
+            openmp-11.0.0.src/build-arm64/install/lib/libomp.a \
+            -o $GITHUB_WORKSPACE/openmp-install/lib/libomp.a
+
+    - name: install-openmp
+      run: |
+        sudo cp $GITHUB_WORKSPACE/openmp-install/include/* $DEVELOPER_DIR/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/usr/include
+        sudo cp $GITHUB_WORKSPACE/openmp-install/lib/libomp.a $DEVELOPER_DIR/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/usr/lib
+
+    - name: vulkansdk
+      run: |
+        wget https://sdk.lunarg.com/sdk/download/1.3.236.0/mac/vulkansdk-macos-1.3.236.0.dmg?Human=true -O vulkansdk-macos-1.3.236.0.dmg
+        hdiutil attach vulkansdk-macos-1.3.236.0.dmg
+        sudo /Volumes/vulkansdk-macos-1.3.236.0/InstallVulkan.app/Contents/MacOS/InstallVulkan --root $GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0 --accept-licenses --default-answer --confirm-command install
+        hdiutil detach /Volumes/vulkansdk-macos-1.3.236.0
+
+    - name: build
+      run: |
+        export VULKAN_SDK=$GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0/macOS
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/ios.toolchain.cmake -DPLATFORM=MAC_CATALYST -DARCHS="x86_64" \
+            -DDEPLOYMENT_TARGET=$MAC_CATALYST_DEPLOYMENT_TARGET -DENABLE_BITCODE=$ENABLE_BITCODE -DENABLE_ARC=$ENABLE_ARC -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
+            -DOpenMP_C_FLAGS="-Xclang -fopenmp" -DOpenMP_CXX_FLAGS="-Xclang -fopenmp" \
+            -DOpenMP_C_LIB_NAMES="libomp" -DOpenMP_CXX_LIB_NAMES="libomp" \
+            -DOpenMP_libomp_LIBRARY="$DEVELOPER_DIR/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/usr/lib/libomp.a" \
+            -DVulkan_INCLUDE_DIR=$GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0/MoltenVK/include \
+            -DVulkan_LIBRARY=$GITHUB_WORKSPACE/vulkansdk-macos-1.3.236.0/MoltenVK/dylib/macOS/libMoltenVK.dylib \
+            -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=install \
+            -DNCNN_VULKAN=ON -DNCNN_BUILD_BENCHMARK=OFF ..
+        cmake --build . -j 3