Unsolved intel intrinsic 8bit int to float
-
How to convert 8bit uint to float using most optimal AVX512 or older features?
Could be vectorized operation but i am new to intel intrinsics and cannot figure it out. -
Isn't compiling with '-O2 -funroll-loops' enough? According to godbolt.org I would say yes. And when you're sure the target supports AVX instructions then add ' -mavx'
-
@Christian-Ehrlicher Thanks for the flags , will check the results.