2 - optimize translated cache chaining (DLL PLT like system)
3 - segment ops (minimal LDT/GDT support for wine)
4 - improved 16 bit support
5 - optimize inverse flags propagation (easy by generating intermediate
6 micro operation array).
9 - make it self runnable (use same trick as ld.so : include its own relocator and libc)
10 - fix FPU exceptions (in particular: gen_op_fpush not before mem load)