What do ENABLE_FAST_MATH and CUDA_FAST_MATH do?

C++
1

I’ve seen some tutorials to compile OpenCV with CUDA support and many of them enable the ENABLE_FAST_MATH and CUDA_FAST_MATH flags. But what exactly those flags do? Is it actually recommended to use them?

2

CUDA_FAST_MATH sets the --use_fast_math flag which enables certain optimizations

Essentially it can speed up some compute workloads at the expense of accuracy.

3

Thanks for the reply!
Do you know what is the difference with ENABLE_FAST_MATH?

4

That’s a CPU flag, see

5

Thank you so much! I see this is mostly for MSVC.