As I understand, there are two ways to do that. Either Haskell code is called from C, or C code is called for Haskell. So my questions are: 1. Are they both possible?
Yep.
2. If yes, which is better performance-wise? (C function is performance-critical). If generated function is called many times, how big an overhead is going to be?
I'm also interested in an answer to this, since I have a program that does both, quite a bit. Of course the real answer is to try to profile it, but I'm also interested in the theory of what's going on in both directions (I'm using safe imports of course, since there are callbacks). I have a bit of a performance problem with the C->haskell part, but that may also be all the marshalling the haskell side has to do (both via the with / alloca approach, and malloc in haskell and free() in C), and I'd like to know a bit more before I go wild trying to cache data in C that comes from haskell.