diff --git a/.ci/test-coverage.yml b/.ci/test-coverage.yml
index ccd1f994ea0..6cd442850b3 100644
--- a/.ci/test-coverage.yml
+++ b/.ci/test-coverage.yml
@@ -22,329 +22,329 @@ 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-20240622
-  #
-  #   - name: checkout-swiftshader
-  #     if: steps.cache-swiftshader.outputs.cacheHit != 'true'
-  #     checkout: https://github.com/google/swiftshader.git
-  #     with:
-  #       pullType: COMMIT_ID
-  #       refName: de870ac7518fe2b6bb651ecc22fc36647cf7b986
-  #       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/* ${{ci.workspace}}/swiftshader-install
-  #
-  #   - name: build
-  #     run: |
-  #       mkdir build && cd build
-  #       cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_AVX2=ON -DNCNN_AVX512=OFF -DNCNN_XOP=OFF -DNCNN_OPENMP=OFF -DNCNN_VULKAN=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
-  #       cmake --build . -j 4
-  #   - name: test
-  #     run: |
-  #       printf "[Processor]\nThreadCount=4\n" > build/tests/SwiftShader.ini
-  #       export VK_ICD_FILENAMES="${{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 ${{ci.workspace}}/lavapipe-install/share/vulkan/icd.d/lvp_icd.x86_64.json
-  #
-  #   - name: build
-  #     run: |
-  #       mkdir build && cd build
-  #       cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_AVX2=ON -DNCNN_AVX512=OFF -DNCNN_XOP=OFF -DNCNN_OPENMP=OFF -DNCNN_VULKAN=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
-  #       cmake --build . -j 4
-  #   - name: test
-  #     run: |
-  #       export LP_NUM_THREADS=4
-  #       export VK_ICD_FILENAMES="${{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:
-  #     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: '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 4
-  #   - name: test
-  #     run: 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 --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 4
-  #   - 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 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 --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=${{matrix.GNU_INLINE_ASM}} -DNCNN_VFPV4=OFF -DNCNN_ARM82=OFF -DNCNN_OPENMP=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
-  #       cmake --build . -j 4
-  #   - name: test-armhf-vfpv3-d16
-  #     run: |
-  #       export PATH=${{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 4
-  #   - 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-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-20240622
+
+    - name: checkout-swiftshader
+      if: steps.cache-swiftshader.outputs.cacheHit != 'true'
+      checkout: https://github.com/google/swiftshader.git
+      with:
+        pullType: COMMIT_ID
+        refName: de870ac7518fe2b6bb651ecc22fc36647cf7b986
+        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/* ${{ci.workspace}}/swiftshader-install
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_AVX2=ON -DNCNN_AVX512=OFF -DNCNN_XOP=OFF -DNCNN_OPENMP=OFF -DNCNN_VULKAN=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 4
+    - name: test
+      run: |
+        printf "[Processor]\nThreadCount=4\n" > build/tests/SwiftShader.ini
+        export VK_ICD_FILENAMES="${{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 ${{ci.workspace}}/lavapipe-install/share/vulkan/icd.d/lvp_icd.x86_64.json
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_AVX2=ON -DNCNN_AVX512=OFF -DNCNN_XOP=OFF -DNCNN_OPENMP=OFF -DNCNN_VULKAN=ON -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 4
+    - name: test
+      run: |
+        export LP_NUM_THREADS=4
+        export VK_ICD_FILENAMES="${{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:
+      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: '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 4
+    - name: test
+      run: 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 --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 4
+    - 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 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 --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=${{matrix.GNU_INLINE_ASM}} -DNCNN_VFPV4=OFF -DNCNN_ARM82=OFF -DNCNN_OPENMP=OFF -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 4
+    - name: test-armhf-vfpv3-d16
+      run: |
+        export PATH=${{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 4
+    - 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
@@ -423,7 +423,7 @@ jobs:
             -DNCNN_ARM84BF16=${{matrix.ARM84BF16}} \
             -DNCNN_ARM84I8MM=${{matrix.ARM84I8MM}} \
             ..
-        cmake --build . -j 1 --verbose
+        cmake --build . -j 4
     - name: test
       run: |
         export PATH=${{ci.workspace}}/qemu-install/bin:$PATH
@@ -439,532 +439,532 @@ jobs:
     - 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=${{matrix.OPENMP}} -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
-  #       cmake --build . -j 4
-  #   - name: test
-  #     run: |
-  #       export PATH=${{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 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 --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 4
-  #   - 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 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 --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-powerpc:
-  #   name: linux-gcc-powerpc
-  #   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++-powerpc-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-ppc-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=ppc-linux-user --disable-system
-  #       make -j$(nproc)
-  #       make install
-  #
-  #   - name: build
-  #     run: |
-  #       mkdir build && cd build
-  #       cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/powerpc-linux-gnu.toolchain.cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_OPENMP=${{matrix.OPENMP}} -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
-  #       cmake --build . -j 4
-  #   - name: test
-  #     run: |
-  #       export PATH=${{ci.workspace}}/qemu-install/bin:$PATH
-  #       cd build
-  #       TESTS_EXECUTABLE_LOADER=qemu-ppc TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/powerpc-linux-gnu" 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 --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-powerpc64le:
-  #   name: linux-gcc-powerpc64le
-  #   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++-powerpc64le-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-ppc64le-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=ppc64le-linux-user --disable-system
-  #       make -j$(nproc)
-  #       make install
-  #
-  #   - name: build
-  #     run: |
-  #       mkdir build && cd build
-  #       cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/powerpc64le-linux-gnu.toolchain.cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_OPENMP=${{matrix.OPENMP}} -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
-  #       cmake --build . -j 4
-  #   - name: test
-  #     run: |
-  #       export PATH=${{ci.workspace}}/qemu-install/bin:$PATH
-  #       cd build
-  #       TESTS_EXECUTABLE_LOADER=qemu-ppc64le TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/powerpc64le-linux-gnu" ctest --output-on-failure -j 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 --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 4
-  #   - 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 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 --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-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=${{matrix.OPENMP}} -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
-  #       cmake --build . -j 4
-  #   - name: test
-  #     run: |
-  #       export PATH=${{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 4
-  #   - 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
-  #
-  # linux-gcc-x64-simplemath:
-  #   name: linux-gcc-x64-simplemath
-  #
-  #   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_TOOLCHAIN_FILE=../toolchains/host-c.gcc.toolchain.cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -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
-  #     run: 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 --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 4
+    - 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 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 --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 4
+    - 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 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 --list lcov.info
+    - name: codecov
+      run: ./codecov -t ${{settings.CODECOV_TOKEN.access_token}} -C ${{ ci.sha }} -B ${{ ci.head_ref }} -f build/lcov.info
+
+  linux-gcc-powerpc:
+    name: linux-gcc-powerpc
+    strategy:
+      matrix:
+        OPENMP: ['OFF', 'ON']
+
+    runs-on:
+      pool-name: docker
+      container:
+        image: bkci/ci:ubuntu
+    steps:
+    - name: checkout
+      checkout: self
+      with:
+        strategy: FRESH_CHECKOUT
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: install-deps
+      run: |
+        apt-get update
+        apt-get install -y lcov g++-powerpc-linux-gnu libcapstone4 libglib2.0-0
+        curl https://uploader.codecov.io/verification.gpg | gpg --no-default-keyring --keyring trustedkeys.gpg --import
+        curl -Os https://uploader.codecov.io/latest/linux/codecov
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM.sig
+        gpgv codecov.SHA256SUM.sig codecov.SHA256SUM
+        shasum -a 256 -c codecov.SHA256SUM
+        chmod +x codecov
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: cache@1.*
+      with:
+        cachePaths: qemu-install
+        cacheKey: qemu-ppc-install-20220831-1
+
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cacheHit != 'true'
+      checkout: https://github.com/qemu/qemu.git
+      with:
+        pullType: COMMIT_ID
+        refName: 621da7789083b80d6f1ff1c0fb499334007b4f51
+        localPath: qemu
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: qemu
+      if: steps.cache-qemu.outputs.cacheHit != 'true'
+      run: |
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye main' | tee -a /etc/apt/sources.list
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye-updates main' | tee -a /etc/apt/sources.list
+        apt-get update
+        apt-get build-dep -y qemu
+        cd qemu
+        ./configure --prefix=${{ci.workspace}}/qemu-install --target-list=ppc-linux-user --disable-system
+        make -j$(nproc)
+        make install
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/powerpc-linux-gnu.toolchain.cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_OPENMP=${{matrix.OPENMP}} -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 4
+    - name: test
+      run: |
+        export PATH=${{ci.workspace}}/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-ppc TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/powerpc-linux-gnu" 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 --list lcov.info
+    - name: codecov
+      run: ./codecov -t ${{settings.CODECOV_TOKEN.access_token}} -C ${{ ci.sha }} -B ${{ ci.head_ref }} -f build/lcov.info
+
+  linux-gcc-powerpc64le:
+    name: linux-gcc-powerpc64le
+    strategy:
+      matrix:
+        OPENMP: ['OFF', 'ON']
+
+    runs-on:
+      pool-name: docker
+      container:
+        image: bkci/ci:ubuntu
+    steps:
+    - name: checkout
+      checkout: self
+      with:
+        strategy: FRESH_CHECKOUT
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: install-deps
+      run: |
+        apt-get update
+        apt-get install -y lcov g++-powerpc64le-linux-gnu libcapstone4 libglib2.0-0
+        curl https://uploader.codecov.io/verification.gpg | gpg --no-default-keyring --keyring trustedkeys.gpg --import
+        curl -Os https://uploader.codecov.io/latest/linux/codecov
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM
+        curl -Os https://uploader.codecov.io/latest/linux/codecov.SHA256SUM.sig
+        gpgv codecov.SHA256SUM.sig codecov.SHA256SUM
+        shasum -a 256 -c codecov.SHA256SUM
+        chmod +x codecov
+
+    - name: cache-qemu
+      id: cache-qemu
+      uses: cache@1.*
+      with:
+        cachePaths: qemu-install
+        cacheKey: qemu-ppc64le-install-20220831-1
+
+    - name: checkout-qemu
+      if: steps.cache-qemu.outputs.cacheHit != 'true'
+      checkout: https://github.com/qemu/qemu.git
+      with:
+        pullType: COMMIT_ID
+        refName: 621da7789083b80d6f1ff1c0fb499334007b4f51
+        localPath: qemu
+        enableSubmodule: false
+        enableGitLfs: false
+
+    - name: qemu
+      if: steps.cache-qemu.outputs.cacheHit != 'true'
+      run: |
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye main' | tee -a /etc/apt/sources.list
+        echo 'deb-src http://mirrors.cloud.tencent.com/debian bullseye-updates main' | tee -a /etc/apt/sources.list
+        apt-get update
+        apt-get build-dep -y qemu
+        cd qemu
+        ./configure --prefix=${{ci.workspace}}/qemu-install --target-list=ppc64le-linux-user --disable-system
+        make -j$(nproc)
+        make install
+
+    - name: build
+      run: |
+        mkdir build && cd build
+        cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/powerpc64le-linux-gnu.toolchain.cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -DNCNN_RUNTIME_CPU=OFF -DNCNN_OPENMP=${{matrix.OPENMP}} -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 4
+    - name: test
+      run: |
+        export PATH=${{ci.workspace}}/qemu-install/bin:$PATH
+        cd build
+        TESTS_EXECUTABLE_LOADER=qemu-ppc64le TESTS_EXECUTABLE_LOADER_ARGUMENTS="-L;/usr/powerpc64le-linux-gnu" ctest --output-on-failure -j 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 --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 4
+    - 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 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 --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-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=${{matrix.OPENMP}} -DNCNN_BUILD_TOOLS=OFF -DNCNN_BUILD_EXAMPLES=OFF -DNCNN_BUILD_TESTS=ON ..
+        cmake --build . -j 4
+    - name: test
+      run: |
+        export PATH=${{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 4
+    - 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
+
+  linux-gcc-x64-simplemath:
+    name: linux-gcc-x64-simplemath
+
+    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_TOOLCHAIN_FILE=../toolchains/host-c.gcc.toolchain.cmake -DCMAKE_BUILD_TYPE=debug -DNCNN_COVERAGE=ON -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
+      run: 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 --list lcov.info
+    - name: codecov
+      run: ./codecov -t ${{settings.CODECOV_TOKEN.access_token}} -C ${{ ci.sha }} -B ${{ ci.head_ref }} -f build/lcov.info