DM> Well, that's better, but...

 DM> const size_t read_target = r[i]->target;
 DM> const short* p_typeid    = s[read_target].typeid;
 DM> fread(p_typeid, sizeof(*p_typeid), 1, f);

 DM> That I find to be much improved.  Just my opinion, of course.

 AR> Yes, but your C++ bias is showing ...

Which, the const's?  Actually, that's more than just C++ bias: it's me
giving the compiler a couple of hints about the code.  By telling the
compiler that this value isn't going to change, perhaps it may be able to
make a few optimisations.  For example, in this case, if neither
read_target nor p_typeid are used after the fread, perhaps the compiler (in
optimise mode) may decide not to allocate real stack space for it, but to
simply build the fread parameter list with the right values, discarding the
values on return.

Sure, C++ compilers are more apt to perform this optimisation (I think),
but with most current C compilers also being C++ compilers, I would imagine
that the optimisation engine would be common...  And those straight-C
compilers shouldn't really be that far off the mark.  :-)


---