- verify thread support (clone() and various locks)
- signals
- optimize translated cache chaining (DLL PLT-like system)
- vm86 syscall support
- overrides/16bit for string ops
- more syscalls (in particular all 64 bit ones)
- make it self runnable (use same trick as ld.so : include its own relocator and libc)
- improved 16 bit support 
- fix FPU exceptions (in particular: gen_op_fpush not before mem load)