UFFI was designed as a cross-implementation compatible Foreign Function Interface. Necessarily, only a common subset of functionality can be provided. Likewise, not every optimization for that a specific implementation provides can be supported. Wherever possible, though, implementation-specific optimizations are invoked.
The design of UFFI is dictated by the order of these priorities:
Code using UFFI must operate correctly on all supported implementations.
Take advantage of implementation-specific optimizations. Ideally, there will not a situation where an implementation-specific FFI will be chosen due to lack of optimizations in UFFI.
Provide a simple interface to developers using
UFFI. This priority is quite a bit lower than the above priorities.
This lower priority is manifest by programmers having to pass types in
pointer and array dereferencing, needing to use
cstring wrapper functions, and the use of
ensure-char-character and ensure-char-integer functions. My hope is
that the developer inconvenience will be outweighed by the generation
of optimized code that is cross-implementation compatible.