Zlib-ng is zlib replacement with optimizations for "next generation" systems. Its optimization may benifits image library decode and encode speed such as libpng. In our tests, if using zlib-ng and libpng combination on a x86_64 machine with AVX2, the time of `imdecode` amd `imencode` will drop 20% approximately. This patch enables zlib-ng's optimization if `CV_DISABLE_OPTIMIZATION` is OFF. Since Zlib-ng can dispatch intrinsics on the fly, port work is much easier. Related discussion: https://github.com/opencv/opencv/issues/22573
1.7 KiB
Building RISC-V Target with Cmake
Warning
Runtime rvv detection (using
hwcap
) requires linux kernel 6.5 or newer.When running on older kernels, we fall back to compile-time detection, potentially this can cause crashes if rvv is enabled at compile but not supported by the target cpu. Therefore if older kernel support is needed, rvv should be disabled if the target cpu does not support it.
Prerequisite: Build RISC-V Clang Toolchain and QEMU
If you don't have prebuilt clang and riscv64 qemu, you can refer to the script to get the source. Copy the script to the zlib-ng root directory, and run it to download the source and build them. Modify the content according to your conditions (e.g., toolchain version).
./prepare_riscv_toolchain_qemu.sh
After running script, clang & qemu are built in build-toolchain-qemu/riscv-clang/
& build-toolchain-qemu/riscv-qemu/
.
build-toolchain-qemu/riscv-clang/
is your TOOLCHAIN_PATH
.
build-toolchain-qemu/riscv-qemu/bin/qemu-riscv64
is your QEMU_PATH
.
You can also download the prebuilt toolchain & qemu from the release page, and enjoy using them.
Cross-Compile for RISC-V Target
cmake -G Ninja -B ./build-riscv \
-D CMAKE_TOOLCHAIN_FILE=./cmake/toolchain-riscv.cmake \
-D CMAKE_INSTALL_PREFIX=./build-riscv/install \
-D TOOLCHAIN_PATH={TOOLCHAIN_PATH} \
-D QEMU_PATH={QEMU_PATH} \
.
cmake --build ./build-riscv
Disable the option if there is no RVV support:
-D WITH_RVV=OFF
Run Unittests on User Mode QEMU
cd ./build-riscv && ctest --verbose