I am trying to build OpenCV's master for my i.MX6q with the following flags:
-mthumb -O3 -march=armv7-a -mcpu=cortex-a9 -mtune=cortex-a9 -mfpu=neon mvectorize-with-neon-quad -mfloat-abi=hard
- is it going to make a difference, or is this kind of stuff supposed to be detected automatically?
The build does not work, I get this message:
intrin_neon.hpp:1205:42: error: 'vcvt_f16_f32' was not declared in this scope
I tried to remove -mfloat-abi=hard altogether, it didn't help. cmake refers to libs there /usr/lib/arm-linux-gnueabihf/, so I supposed that hard was the correct option? Anyway, both with and without give the same error message.
What can I do to build OpenCV optimized for i.MX6?