Stay organized with collections
Save and categorize content based on your preferences.
The NDK supports ARM Advanced SIMD, commonly known as Neon, an optional
instruction set extension for ARMv7 and ARMv8. Neon provides scalar/vector
instructions and registers (shared with the FPU) comparable to MMX/SSE/3DNow!
in the x86 world.
All ARMv8-based ("arm64") Android devices support Neon. Almost all ARMv7-based
("32-bit") Android devices support Neon, including all devices that shipped with
API level 21 or later. The NDK enables Neon by default for both Arm ABIs.
If you target very old devices, you can filter out incompatible devices on the
Google Play Console. You can also use the console for your app to see how many
devices this would affect.
Alternatively, for maximum compatibility, 32-bit code can perform runtime
detection to confirm that Neon code can be run on the target device. An app can
perform this check using any of the options mentioned in
CPU features.
You should not write explicit Neon intrinsics in your C/C++ code. Clang's
portable vector types will automatically use Neon instructions. Clang's Neon
intrinsics are actually just a non-portable wrapper around the portable types,
so writing Neon intrinsics will not make your code any faster than using the
portable types, just less portable.
Build
Disable Neon globally
ndk-build
ndk-build does not support disabling Neon globally. To disable Neon an entire
ndk-build application, apply the per-module steps to every module in your
application.
CMake
Pass -DANDROID_ARM_NEON=ON when invoking CMake. If building with Android
Studio/Gradle, set the following option in your build.gradle:
The vectorization sample demonstrates how to use a variety of vectorization
tools to implement a matrix multiply, and compares their performance.
Content and code samples on this page are subject to the licenses described in the Content License. Java and OpenJDK are trademarks or registered trademarks of Oracle and/or its affiliates.
Last updated 2025-08-25 UTC.
[[["Easy to understand","easyToUnderstand","thumb-up"],["Solved my problem","solvedMyProblem","thumb-up"],["Other","otherUp","thumb-up"]],[["Missing the information I need","missingTheInformationINeed","thumb-down"],["Too complicated / too many steps","tooComplicatedTooManySteps","thumb-down"],["Out of date","outOfDate","thumb-down"],["Samples / code issue","samplesCodeIssue","thumb-down"],["Other","otherDown","thumb-down"]],["Last updated 2025-08-25 UTC."],[],[],null,["# Neon support\n\nThe NDK supports ARM Advanced SIMD, commonly known as Neon, an optional\ninstruction set extension for ARMv7 and ARMv8. Neon provides scalar/vector\ninstructions and registers (shared with the FPU) comparable to MMX/SSE/3DNow!\nin the x86 world.\n\nAll ARMv8-based (\"arm64\") Android devices support Neon. Almost all ARMv7-based\n(\"32-bit\") Android devices support Neon, including all devices that shipped with\nAPI level 21 or later. The NDK enables Neon by default for both Arm ABIs.\n\nIf you target very old devices, you can filter out incompatible devices on the\nGoogle Play Console. You can also use the console for your app to see how many\ndevices this would affect.\n\nAlternatively, for maximum compatibility, 32-bit code can perform runtime\ndetection to confirm that Neon code can be run on the target device. An app can\nperform this check using any of the options mentioned in\n[CPU features](/ndk/guides/cpu-features).\n\nYou should not write explicit Neon intrinsics in your C/C++ code. Clang's\n[portable vector types](https://clang.llvm.org/docs/LanguageExtensions.html#vectors-and-extended-vectors) will automatically use Neon instructions. Clang's Neon\nintrinsics are actually just a non-portable wrapper around the portable types,\nso writing Neon intrinsics will not make your code any faster than using the\nportable types, just less portable.\n\nBuild\n-----\n\n| **Note:** For NDK r21 and newer Neon is enabled by default for all API levels. If you need to disable Neon to support non-Neon devices (which are rare), invert the settings described below. Alternatively, the Play Store console can be used to [exclude CPUs](https://support.google.com/googleplay/android-developer/answer/7353455) that do not support Neon to prevent your application from being installed on those devices.\n\nDisable Neon globally\n---------------------\n\n### ndk-build\n\nndk-build does not support disabling Neon globally. To disable Neon an entire\nndk-build application, apply the per-module steps to every module in your\napplication.\n\n### CMake\n\nPass `-DANDROID_ARM_NEON=ON` when invoking CMake. If building with Android\nStudio/Gradle, set the following option in your build.gradle: \n\n android {\n defaultConfig {\n externalNativeBuild {\n cmake {\n arguments \"-DANDROID_ARM_NEON=OFF\"\n }\n }\n }\n }\n\nDisable Neon per module\n-----------------------\n\n### ndk-build\n\nTo build all the source files in an ndk-build module without Neon, add the\nfollowing to the module definition in your Android.mk: \n\n LOCAL_ARM_NEON := false\n\n### CMake\n\nTo build all the source files in a CMake target without Neon, add the\nfollowing to your CMakeLists.txt: \n\n if(ANDROID_ABI STREQUAL armeabi-v7a)\n set_target_properties(${TARGET} PROPERTIES COMPILE_FLAGS -mfpu=vfpv3-d16)\n endif()\n\nWhere `${TARGET}` is replaced with the name of your library.\n\nCross-platform support for x86\n------------------------------\n\n\nNDK supports cross-platform compilation of your existing ARM SIMD (Neon)\ninstrinsic functions into x86 SSE code, through the use of the third-party\n[NEON_2_SSE.h](https://github.com/intel/ARM_NEON_2_x86_SSE).\nFor more information on this topic, see\n[From ARM NEON to Intel SSE-the automatic porting solution, tips and tricks](http://software.intel.com/en-us/blogs/2012/12/12/from-arm-neon-to-intel-mmxsse-automatic-porting-solution-tips-and-tricks).\n\nSample code\n-----------\n\nThe [vectorization sample](https://github.com/android/ndk-samples/tree/main/vectorization) demonstrates how to use a variety of vectorization\ntools to implement a matrix multiply, and compares their performance."]]