imx6q neon tensorlow lite

이번 빌드 옵션들을 뒤져보다 보니 빼먹고 안 넣은게 있었네

(라고 쓰고 실은 귀찮았던...)

 

-mvectorize-with-neon-quad -ffast-math

[링크 : http://stackoverflow.com/questions/14962447/gcc-options-for-a-freescale-imx6q-arm-processor]

 

 

In addition GCC offers the -ffast-math flag which is a shortcut for several options, presenting the least conforming but fastest math mode. It enables -fno-trapping-math, -funsafe-math-optimizations, -ffinite-math-only, -fno-errno-math, -fno-signaling-nans, -fno-rounding-math, -fcx-limited-range and -fno-signed-zeros. Each of these flags violates IEEE in a different way.

[링크 : http://gcc.gnu.org/wiki/FloatingPointMath]

 

 

Cortex-A9 -mcpu=cortex-a9 -mfpu=vfpv3-fp16 -mfpu=vfpv3-d16-fp16 -mfpu=neon-fp16

[링크 : http://www.programmersought.com/article/5320739655/]

 

-DENABLE_VFPV3=ON

ARM VFPv3 floating

[링크 : http://www.google.com/amp/s/learnopencv.com/build-and-install-opencv-4-for-raspberry-pi/amp/]