feat: Sync to Upstream 20240820

.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 +1,127 @@
+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=${{matrix.AVX512}} -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 4
+    - name: test
+      run: cd build && ctest --output-on-failure -j 4
+    - name: build-shared
+      run: |
+        mkdir build-shared && cd build-shared
+        cmake -DNCNN_SSE2=${{matrix.SSE2}} -DNCNN_AVX=${{matrix.AVX}} -DNCNN_AVX2=${{matrix.AVX2}} -DNCNN_AVX512=${{matrix.AVX512}} -DNCNN_SHARED_LIB=ON ..
+        cmake --build . -j 4
+    - name: build-noint8
+      run: |
+        mkdir build-noint8 && cd build-noint8
+        cmake -DNCNN_SSE2=${{matrix.SSE2}} -DNCNN_AVX=${{matrix.AVX}} -DNCNN_AVX2=${{matrix.AVX2}} -DNCNN_AVX512=${{matrix.AVX512}} -DNCNN_INT8=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 4
+    - name: test-noint8
+      run: cd build-noint8 && ctest --output-on-failure -j 4
+
+  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 4
+    - name: test-nostdio
+      run: cd build-nostdio && ctest --output-on-failure -j 4
+    - 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 4
+    - 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 4
+    - name: test-simplestl
+      run: cd build-simplestl && ctest --output-on-failure -j 4
+    - 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 4
+    - name: test-simplestl-simpleomp
+      run: cd build-simplestl-simpleomp && ctest --output-on-failure -j 4
+    - name: build-simplestl-simplemath
+      run: |
+        mkdir build-simplestl-simplemath && cd build-simplestl-simplemath
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/host-c.gcc.toolchain.cmake -DNCNN_STDIO=ON -DNCNN_STRING=ON -DNCNN_SIMPLESTL=ON -DNCNN_SIMPLEMATH=ON -DNCNN_BUILD_TESTS=ON -DNCNN_BUILD_BENCHMARK=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j 4
+    - name: test-simplestl-simplemath
+      run: cd build-simplestl-simplemath && ctest --output-on-failure -j 4
+

.ci/pnnx.yml

@@ -0,0 +1,207 @@
+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 }}
+
+variables:
+  protobuf_version: 21.12
+  libtorch_version: 2.4.0
+  libtorchvision_version: 0.19.0
+  onnxruntime_version: 1.18.1
+  cache_date: 20240804
+
+jobs:
+  ubuntu:
+    strategy:
+      matrix:
+        include:
+          - torch-version: 1.8.1
+            torchvision-version: 0.9.1
+
+          - torch-version: 1.9.1
+            torchvision-version: 0.10.1
+
+          - torch-version: 1.10.0
+            torchvision-version: 0.11.1
+
+          - torch-version: 1.11.0
+            torchvision-version: 0.12.0
+
+          - torch-version: 1.12.0
+            torchvision-version: 0.13.0
+
+          - torch-version: 1.13.0
+            torchvision-version: 0.14.0
+
+          - torch-version: 2.0.0
+            torchvision-version: 0.15.1
+
+          - torch-version: 2.1.0
+            torchvision-version: 0.16.0
+
+          - torch-version: 2.2.1
+            torchvision-version: 0.17.1
+
+          - torch-version: 2.3.0
+            torchvision-version: 0.18.0
+
+          - torch-version: 2.4.0
+            torchvision-version: 0.19.0
+
+    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
+        python3 -m pip install --upgrade pip
+        apt-get remove -y python3-setuptools
+        pip3 install -U setuptools==69.5.1
+        pip3 install -U pytest wheel twine requests einops numpy==1.26.4
+        cd ${{ci.workspace}}
+        wget -q https://github.com/Kitware/CMake/releases/download/v3.29.3/cmake-3.29.3-linux-x86_64.tar.gz
+        tar -xf cmake-3.29.3-linux-x86_64.tar.gz
+
+    - name: cache-pnnx-deps-torch
+      id: cache-pnnx-deps-torch
+      uses: cache@1.*
+      with:
+        cachePaths: pnnx-deps-torch-install
+        cacheKey: pnnx-deps-torch-linux-install-${{variables.cache_date}}
+
+    - name: cache-pnnx-deps-onnx
+      id: cache-pnnx-deps-onnx
+      uses: cache@1.*
+      with:
+        cachePaths: pnnx-deps-onnx-install
+        cacheKey: pnnx-deps-onnx-linux-install-${{variables.cache_date}}
+
+    - name: checkout-pnnx-patches
+      if: steps.cache-pnnx-deps-torch.outputs.cacheHit != 'true' || steps.cache-pnnx-deps-onnx.outputs.cacheHit != 'true'
+      checkout: https://github.com/pnnx/pnnx.git
+      with:
+        refName: main
+        localPath: pnnx-patches
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: pnnx-deps-torch-libtorch
+      if: steps.cache-pnnx-deps-torch.outputs.cacheHit != 'true'
+      run: |
+        export PATH=${{ci.workspace}}/cmake-3.29.3-linux-x86_64/bin:$PATH
+        wget -q https://github.com/pytorch/pytorch/releases/download/v${{variables.libtorch_version}}/pytorch-v${{variables.libtorch_version}}.tar.gz
+        tar -xf pytorch-v${{variables.libtorch_version}}.tar.gz
+        cd pytorch-v${{variables.libtorch_version}}
+        pip3 install -r requirements.txt
+        patch -p1 -i ${{ci.workspace}}/pnnx-patches/pytorch-v${{variables.libtorch_version}}-fix-mobile-build.patch
+        patch -p1 -i ${{ci.workspace}}/pnnx-patches/pytorch-v${{variables.libtorch_version}}-no-link-system-lib.patch
+        patch -p1 -i ${{ci.workspace}}/pnnx-patches/pytorch-v${{variables.libtorch_version}}-fix-pocketfft-build.patch
+        mkdir -p build && cd build
+        cmake -DCMAKE_INSTALL_PREFIX=${{ci.workspace}}/pnnx-deps-torch-install -DCMAKE_BUILD_TYPE=MinSizeRel -DBUILD_SHARED_LIBS=OFF -DBUILD_CAFFE2_OPS=OFF -DBUILD_CUSTOM_PROTOBUF=OFF -DBUILD_LITE_INTERPRETER=OFF -DBUILD_PYTHON=OFF -DINTERN_BUILD_MOBILE=ON -DINTERN_DISABLE_AUTOGRAD=ON -DINTERN_DISABLE_ONNX=ON -DUSE_CUDA=OFF -DUSE_DISTRIBUTED=OFF -DUSE_ITT=OFF -DUSE_KINETO=OFF -DUSE_LITE_INTERPRETER_PROFILER=OFF -DUSE_MKLDNN=OFF -DUSE_MPS=OFF -DUSE_NUMPY=OFF -DUSE_OPENMP=OFF -DUSE_SOURCE_DEBUG_ON_MOBILE=OFF -DUSE_XNNPACK=OFF ..
+        cmake --build . -j $(nproc)
+        cmake --build . -j $(nproc) --target install/strip
+    - name: pnnx-deps-torch-libtorchvision
+      if: steps.cache-pnnx-deps-torch.outputs.cacheHit != 'true'
+      run: |
+        export PATH=${{ci.workspace}}/cmake-3.29.3-linux-x86_64/bin:$PATH
+        wget -q https://github.com/pytorch/vision/archive/v${{variables.libtorchvision_version}}.zip -O vision-${{variables.libtorchvision_version}}.zip
+        unzip -q vision-${{variables.libtorchvision_version}}.zip
+        cd vision-${{variables.libtorchvision_version}}
+        patch -p1 -i ${{ci.workspace}}/pnnx-patches/vision-${{variables.libtorchvision_version}}-ops-only.patch
+        patch -p1 -i ${{ci.workspace}}/pnnx-patches/vision-${{variables.libtorchvision_version}}-no-cuda-version.patch
+        mkdir -p build && cd build
+        cmake -DCMAKE_INSTALL_PREFIX=${{ci.workspace}}/pnnx-deps-torch-install -DTorch_DIR=${{ci.workspace}}/pnnx-deps-torch-install/share/cmake/Torch -DCMAKE_BUILD_TYPE=MinSizeRel -DWITH_PNG=OFF -DWITH_JPEG=OFF ..
+        cmake --build . -j $(nproc)
+        cmake --build . -j $(nproc) --target install/strip
+
+    - name: pnnx-deps-onnx-protobuf
+      if: steps.cache-pnnx-deps-onnx.outputs.cacheHit != 'true'
+      run: |
+        export PATH=${{ci.workspace}}/cmake-3.29.3-linux-x86_64/bin:$PATH
+        wget -q https://github.com/protocolbuffers/protobuf/archive/v${{variables.protobuf_version}}.zip -O protobuf-${{variables.protobuf_version}}.zip
+        unzip -q protobuf-${{variables.protobuf_version}}.zip
+        cd protobuf-${{variables.protobuf_version}}
+        mkdir -p build2 && cd build2
+        cmake -DCMAKE_INSTALL_PREFIX=${{ci.workspace}}/pnnx-deps-onnx-install -Dprotobuf_BUILD_TESTS=OFF -DCMAKE_BUILD_TYPE=MinSizeRel -DCMAKE_POSITION_INDEPENDENT_CODE=ON ..
+        cmake --build . -j $(nproc)
+        cmake --build . --target install/strip
+    - name: pnnx-deps-onnx-onnxruntime
+      if: steps.cache-pnnx-deps-onnx.outputs.cacheHit != 'true'
+      run: |
+        export PATH=${{ci.workspace}}/cmake-3.29.3-linux-x86_64/bin:$PATH
+        wget -q https://github.com/microsoft/onnxruntime/archive/v${{variables.onnxruntime_version}}.zip -O onnxruntime-${{variables.onnxruntime_version}}.zip
+        unzip -q onnxruntime-${{variables.onnxruntime_version}}.zip
+        cd onnxruntime-${{variables.onnxruntime_version}}
+        patch -p1 -i ${{ci.workspace}}/pnnx-patches/onnxruntime-${{variables.onnxruntime_version}}-less-mlas-features.patch
+        patch -p1 -i ${{ci.workspace}}/pnnx-patches/onnxruntime-${{variables.onnxruntime_version}}-monolithic-static-library.patch
+        mkdir -p build && cd build
+        cmake -DCMAKE_INSTALL_PREFIX=${{ci.workspace}}/pnnx-deps-onnx-install -DCMAKE_BUILD_TYPE=MinSizeRel -Donnxruntime_USE_FULL_PROTOBUF=ON -Donnxruntime_BUILD_SHARED_LIB=ON -Donnxruntime_BUILD_UNIT_TESTS=OFF -Donnxruntime_ENABLE_CPUINFO=OFF -Donnxruntime_DISABLE_CONTRIB_OPS=ON -Donnxruntime_DISABLE_ML_OPS=ON -Donnxruntime_DISABLE_SPARSE_TENSORS=ON --compile-no-warning-as-error ../cmake
+        cmake --build . -j $(nproc)
+        cmake --build . -j $(nproc) --target install/strip
+
+    - name: setup-pytorch
+      run: |
+        export PYTHONUSERBASE=${{ci.workspace}}/torch-${{matrix.torch-version}}
+        pip3 install --user torch==${{matrix.torch-version}}+cpu torchvision==${{matrix.torchvision-version}}+cpu --index-url https://download.pytorch.org/whl/cpu
+        pip3 install --user onnx
+        pip3 install --user onnxscript
+
+    - name: build-ncnn
+      run: |
+        export PATH=${{ci.workspace}}/cmake-3.29.3-linux-x86_64/bin:$PATH
+        export PYTHONUSERBASE=${{ci.workspace}}/torch-${{matrix.torch-version}}
+        mkdir build && cd build
+        cmake -DCMAKE_BUILD_TYPE=Release -DNCNN_PYTHON=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF ..
+        cmake --build . -j 16
+        cd ..
+        export CMAKE_BUILD_PARALLEL_LEVEL=16
+        pip3 install --user .
+
+    - name: build-pnnx
+      run: |
+        export PATH=${{ci.workspace}}/cmake-3.29.3-linux-x86_64/bin:$PATH
+        export PYTHONUSERBASE=${{ci.workspace}}/torch-${{matrix.torch-version}}
+        cd tools/pnnx
+        mkdir build && cd build
+        cmake -DCMAKE_BUILD_TYPE=Release -DTorch_INSTALL_DIR=${{ci.workspace}}/pnnx-deps-torch-install -DTorchVision_INSTALL_DIR=${{ci.workspace}}/pnnx-deps-torch-install -Donnxruntime_INSTALL_DIR=${{ci.workspace}}/pnnx-deps-onnx-install -Dprotobuf_DIR=${{ci.workspace}}/pnnx-deps-onnx-install/lib/cmake/protobuf ..
+        cmake --build . -j 16 --verbose
+
+    - name: test
+      run: |
+        export PATH=${{ci.workspace}}/cmake-3.29.3-linux-x86_64/bin:$PATH
+        export PYTHONUSERBASE=${{ci.workspace}}/torch-${{matrix.torch-version}}
+        export OMP_NUM_THREADS=1
+        export MKL_NUM_THREADS=1
+        export MKL_ENABLE_INSTRUCTIONS=SSE4_2
+        cd tools/pnnx
+        cd build && ctest --output-on-failure -j 16
+
+    - name: python-pnnx
+      run: |
+        export PYTHONUSERBASE=${{ci.workspace}}/torch-${{matrix.torch-version}}
+        export PNNX_WHEEL_WITHOUT_BUILD=ON
+        cd tools/pnnx/python
+        cp ../build/src/pnnx pnnx/
+        python3 setup.py install --user
+        pytest tests