Clone this repo:
  1. a1652fd fix typo (#54) by Cuong V. Nguyen · 7 weeks ago master
  2. 3c9879b Fix vcvt_n functions to handle 32 fraction bits and clamping unsigned to zero (#50) by avostrik · 4 months ago
  3. 2cbbc2a vcvt(q)_n_s32_f32 and similar functions bug fix by Zhislina · 4 months ago
  4. 67ea942 fix: round position for vr{add,sub}hn_{s,u}{16,32,64} (#47) by Victoria · 7 months ago
  5. 42b2beb Update LICENSE by Victoria · 10 months ago

The NEON_2_SSE.h file is intended to simplify ARM->IA32 porting. It makes the correspondence (or a real porting) of ARM NEON intrinsics as defined in “arm_neon.h” header and x86 SSE (up to SSE4.2) intrinsic functions as defined in corresponding x86 compilers headers files.


To take advantage of this file just include it in your project that uses ARM NEON intinsics instead of “arm_neon.h”, compile it as usual and enjoy the result.

For significant performance improvement in some cases you might need to define USE_SSE4 in your project settings. Otherwise SIMD up to SSSE3 to be used.

If NEON2SSE_DISABLE_PERFORMANCE_WARNING macro is defined, then the performance warnings are disabled.

For more information and license please read the NEON_2_SSE.h content.

The unit tests set used for ARM NEON - x86 SSE conformance verification is https://github.com/christophe-lyon/arm-neon-tests