Merge branch 'master' into fix-bof

Author: junwha

.ci/pnnx.yml

@@ -19,10 +19,10 @@ concurrency:
 
 variables:
   protobuf_version: 21.12
-  libtorch_version: 2.4.0
-  libtorchvision_version: 0.19.0
-  onnxruntime_version: 1.18.1
-  cache_date: 20240804
+  libtorch_version: 2.5.0
+  libtorchvision_version: 0.20.0
+  onnxruntime_version: 1.19.2
+  cache_date: 20241018
 
 jobs:
   ubuntu:
@@ -31,36 +31,51 @@ jobs:
         include:
           - torch-version: 1.8.1
             torchvision-version: 0.9.1
+            torchaudio-version: 0.8.1
 
           - torch-version: 1.9.1
             torchvision-version: 0.10.1
+            torchaudio-version: 0.9.1
 
           - torch-version: 1.10.0
             torchvision-version: 0.11.1
+            torchaudio-version: '0.10.0+cpu'
 
           - torch-version: 1.11.0
             torchvision-version: 0.12.0
+            torchaudio-version: '0.11.0+cpu'
 
           - torch-version: 1.12.0
             torchvision-version: 0.13.0
+            torchaudio-version: '0.12.0+cpu'
 
           - torch-version: 1.13.0
             torchvision-version: 0.14.0
+            torchaudio-version: '0.13.0+cpu'
 
           - torch-version: 2.0.0
             torchvision-version: 0.15.1
+            torchaudio-version: '2.0.0+cpu'
 
           - torch-version: 2.1.0
             torchvision-version: 0.16.0
+            torchaudio-version: '2.1.0+cpu'
 
           - torch-version: 2.2.1
             torchvision-version: 0.17.1
+            torchaudio-version: '2.2.1+cpu'
 
           - torch-version: 2.3.0
             torchvision-version: 0.18.0
+            torchaudio-version: '2.3.0+cpu'
 
           - torch-version: 2.4.0
             torchvision-version: 0.19.0
+            torchaudio-version: '2.4.0+cpu'
+
+          - torch-version: 2.5.0
+            torchvision-version: 0.20.0
+            torchaudio-version: '2.5.0+cpu'
 
     runs-on:
       pool-name: docker
@@ -157,6 +172,7 @@ jobs:
         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
+        patch -p1 -i ${{ci.workspace}}/pnnx-patches/onnxruntime-${{variables.onnxruntime_version}}-fix-gcc-avxvnni-check.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 $(nproc)
@@ -165,7 +181,7 @@ jobs:
     - name: setup-pytorch
       run: |
         export PYTHONUSERBASE=${{ci.workspace}}/torch-${{matrix.torch-version}}
-        pip3 install --user torch==${{matrix.torch-version}}+cpu torchvision==${{matrix.torchvision-version}}+cpu --index-url https://download.pytorch.org/whl/cpu
+        pip3 install --user torch==${{matrix.torch-version}}+cpu torchvision==${{matrix.torchvision-version}}+cpu torchaudio==${{matrix.torchaudio-version}} --index-url https://download.pytorch.org/whl/cpu
         pip3 install --user onnx
         pip3 install --user onnxscript
 
@@ -196,8 +212,8 @@ jobs:
         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
+        cd tools/pnnx/build
+        ctest --output-on-failure -j 16
 
     - name: python-pnnx
       run: |

.ci/test-coverage.yml

@@ -96,87 +96,6 @@ jobs:
     - 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 ${{ci.workspace}}/lavapipe-install/share/vulkan/icd.d/lvp_icd.x86_64.json
-
-    - name: build
-      run: |
-        mkdir build && cd build
-        cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_AVX2=ON -DNCNN_AVX512=OFF -DNCNN_XOP=OFF -DNCNN_OPENMP=OFF -DNCNN_VULKAN=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
-        cmake --build . -j 4
-    - name: test
-      run: |
-        export LP_NUM_THREADS=4
-        export VK_ICD_FILENAMES="${{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:
@@ -840,149 +759,6 @@ jobs:
     - 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=${{matrix.OPENMP}} -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
-        cmake --build . -j 4
-    - name: test-vlen128
-      run: |
-        export PATH=${{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 4
-    - 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 4
-    - 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:

.github/workflows/android.yml

@@ -9,6 +9,7 @@ on:
     - 'src/*'
     - 'src/layer/*'
     - 'src/layer/arm/**'
+    - 'src/layer/riscv/**'
     - 'src/layer/x86/**'
     - 'src/layer/vulkan/**'
   pull_request:
@@ -20,6 +21,7 @@ on:
     - 'src/*'
     - 'src/layer/*'
     - 'src/layer/arm/**'
+    - 'src/layer/riscv/**'
     - 'src/layer/x86/**'
     - 'src/layer/vulkan/**'
 concurrency:
@@ -35,6 +37,7 @@ jobs:
       NCNN_CMAKE_OPTIONS: |
         -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK_LATEST_HOME/build/cmake/android.toolchain.cmake \
         -DANDROID_PLATFORM=android-21 \
+        -DANDROID_SUPPORT_FLEXIBLE_PAGE_SIZES=ON \
         -DCMAKE_INSTALL_PREFIX=install \
         -DCMAKE_BUILD_TYPE=Release \
         -DNCNN_VULKAN=ON \
@@ -64,6 +67,11 @@ jobs:
         mkdir build-x86_64 && cd build-x86_64
         cmake ${{ env.NCNN_CMAKE_OPTIONS }} -DANDROID_ABI="x86_64" ..
         cmake --build . -j $(nproc)
+    - name: riscv64
+      run: |
+        mkdir build-riscv64 && cd build-riscv64
+        cmake ${{ env.NCNN_CMAKE_OPTIONS }} -DANDROID_ABI="riscv64" ..
+        cmake --build . -j $(nproc)
 
     - name: armeabi-v7a-shared
       run: |
@@ -85,6 +93,11 @@ jobs:
         mkdir build-x86_64-shared && cd build-x86_64-shared
         cmake ${{ env.NCNN_CMAKE_OPTIONS }} -DANDROID_ABI="x86_64" -DNCNN_SHARED_LIB=ON ..
         cmake --build . -j $(nproc)
+    - name: riscv64-shared
+      run: |
+        mkdir build-riscv64-shared && cd build-riscv64-shared
+        cmake ${{ env.NCNN_CMAKE_OPTIONS }} -DANDROID_ABI="riscv64" -DNCNN_SHARED_LIB=ON ..
+        cmake --build . -j $(nproc)
 
   ndk-r16b:
     runs-on: ubuntu-latest
@@ -105,8 +118,10 @@ jobs:
       env:
         DEBIAN_FRONTEND: noninteractive
       run: |
-        sudo apt-get update
-        sudo apt-get install -y libncurses5
+        pushd /usr/lib/x86_64-linux-gnu/
+        sudo ln -s libncurses.so.6 libncurses.so.5
+        sudo ln -s libtinfo.so.6 libtinfo.so.5
+        popd
         wget -q 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