search

intel / ARM_NEON_2_x86_SSE

The platform independent header allowing to compile any C/C++ code containing ARM NEON intrinsic functions for x86 target systems using SIMD up to AVX2 intrinsic functions
Other
431 stars 150 forks source link

Could there be a precision error between sse and neon? #38

Closed forMwish closed 4 years ago

forMwish commented 4 years ago

For example, floating point Numbers? Thanks. My project has some errors between arm and intel, and I'm not sure where that happened

Zvictoria commented 4 years ago

Hi. Sure some error on floating point numbers treatment exists, especially on reciprocal, sqrt, floqt->int conversion etc. Moreover unfortunately original ARM and my implementation treat NAN and over-underflow differently. It might be the point. If you make the whole list of functions used in your app I may point you the suspicious ones. For the purposes just compile your app without this header and the list of errors (functions undefined) will be the thing to share

forMwish commented 4 years ago

I am glad to receive your reply. Luckily I find the error (vrecpe / vrsqrte) by android web .Finally, I used C to implement the corresponding pseudo-code on the arm architecture manual but not implement the exception handler. Now, it work. Thank you very much.