Home | History | Annotate | Download | only in text 
      1 Android NDK & ARM NEON instruction set extension support
      2 
      3 Introduction:
      4 ====
      5 
      6 Android NDK r3 added support for the new 'armeabi-v7a' ARM-based ABI
      7 that allows native code to use two useful instruction set extensions:
      8 
      9 - Thumb-2, which provides performance comparable to 32-bit ARM
     10   instructions with similar compactness to Thumb-1
     11 
     12 - VFPv3, which provides hardware FPU registers and computations,
     13   to boost floating point performance significantly.
     14 
     15   More specifically, by default 'armeabi-v7a' only supports
     16   VFPv3-D16 which only uses/requires 16 hardware FPU 64-bit registers.
     17 
     18 More information about this can be read in docs/CPU-ARCH-ABIS.html
     19 
     20 The ARMv7 Architecture Reference Manual also defines another optional
     21 instruction set extension known as "ARM Advanced SIMD", nick-named
     22 "NEON". It provides:
     23 
     24 - A set of interesting scalar/vector instructions and registers
     25   (the latter are mapped to the same chip area as the FPU ones),
     26   comparable to MMX/SSE/3DNow! in the x86 world.
     27 
     28 - VFPv3-D32 as a requirement (i.e. 32 hardware FPU 64-bit registers,
     29   instead of the minimum of 16).
     30 
     31 Not all ARMv7-based Android devices will support NEON, but those that
     32 do may benefit in significant ways from the scalar/vector instructions.
     33 
     34 The NDK supports the compilation of modules or even specific source
     35 files with support for NEON. What this means is that a specific compiler
     36 flag will be used to enable the use of GCC ARM Neon intrinsics and
     37 VFPv3-D32 at the same time. The intrinsics are described here:
     38 
     39 > http://gcc.gnu.org/onlinedocs/gcc/ARM-NEON-Intrinsics.html
     40 
     41 
     42 Using LOCAL_ARM_NEON:
     43 ---------------------
     44 
     45 Define LOCAL_ARM_NEON to 'true' in your module definition, and the NDK
     46 will build all its source files with NEON support. This can be useful if
     47 you want to build a static or shared library that specifically contains
     48 NEON code paths.
     49 
     50 
     51 Using the .neon suffix:
     52 -----------------------
     53 
     54 When listing sources files in your LOCAL_SRC_FILES variable, you now have
     55 the option of using the .neon suffix to indicate that you want to
     56 corresponding source(s) to be built with Neon support. For example:
     57 
     58         LOCAL_SRC_FILES := foo.c.neon bar.c
     59 
     60 Will only build 'foo.c' with NEON support.
     61 
     62 Note that the .neon suffix can be used with the .arm suffix too (used to
     63 specify the 32-bit ARM instruction set for non-NEON instructions), but must
     64 appear after it.
     65 
     66 In other words, 'foo.c.arm.neon' works, but 'foo.c.neon.arm' does NOT.
     67 
     68 
     69 Build Requirements:
     70 ------------------
     71 
     72 Neon support only works when targeting the 'armeabi-v7a' ABI, otherwise the
     73 NDK build scripts will complain and abort. It is important to use checks like
     74 the following in your Android.mk:
     75 
     76         # define a static library containing our NEON code
     77         ifeq ($(TARGET_ARCH_ABI),armeabi-v7a)
     78             include $(CLEAR_VARS)
     79             LOCAL_MODULE    := mylib-neon
     80             LOCAL_SRC_FILES := mylib-neon.c
     81             LOCAL_ARM_NEON  := true
     82             include $(BUILD_STATIC_LIBRARY)
     83         endif # TARGET_ARCH_ABI == armeabi-v7a
     84 
     85 
     86 Runtime Detection:
     87 ------------------
     88 
     89 As said previously, NOT ALL ARMv7-BASED ANDROID DEVICES WILL SUPPORT NEON !
     90 It is thus crucial to perform runtime detection to know if the NEON-capable
     91 machine code can be run on the target device.
     92 
     93 To do that, use the 'cpufeatures' library that comes with this NDK. To learn
     94 more about it, see docs/CPU-FEATURES.html.
     95 
     96 You should explicitly check that android_getCpuFamily() returns
     97 ANDROID_CPU_FAMILY_ARM, and that android_getCpuFeatures() returns a value
     98 that has the ANDROID_CPU_ARM_FEATURE_NEON flag set, as  in:
     99 
    100           #include <cpu-features.h>
    101 
    102           ...
    103           ...
    104 
    105           if (android_getCpuFamily() == ANDROID_CPU_FAMILY_ARM &&
    106               (android_getCpuFeatures() & ANDROID_CPU_ARM_FEATURE_NEON) != 0)
    107           {
    108               // use NEON-optimized routines
    109               ...
    110           }
    111           else
    112           {
    113               // use non-NEON fallback routines instead
    114               ...
    115           }
    116 
    117           ...
    118 
    119 Sample code:
    120 ------------
    121 
    122 Look at the source code for the "hello-neon" sample in this NDK for an example
    123 on how to use the 'cpufeatures' library and Neon intrinsics at the same time.
    124 
    125 This implements a tiny benchmark for a FIR filter loop using a C version, and
    126 a NEON-optimized one for devices that support it.
    127