What is the best way to maximize a discrete function?

Question

CUDA

There is a function f = (a, b, c, ... ) where
as arguments a, b, c - values ​​can be passed that change from k to n (different for each a, b, c)
1) you need to get the values a, b, c for which f is maximum
2) you need to get combinations of intervals a,b,c for which f is not lower than the fMin threshold Are there any libraries or ready-made high-performance algorithms that solve this problem on a GPU (CUDA)? Or maybe there are other tricky mathematical ways to at least approximately find the result?
arguments a,b,c - to some extent correlate with the value of the function
f