Hello, On Jan 3, 2008 4:45 PM, Manish Katiyar <[EMAIL PROTECTED]> wrote: > As everyone mentioned, it is to upgrade the data structure, without > affecting the existing code. It also prevents coders who liberally use > int with other datatypes like long etc. from doing so and thus also > prevents some unnoticed errors. > > Hope that helps......... > > One of the reasons > > > On Jan 3, 2008 7:59 PM, Thanos McAtos <[EMAIL PROTECTED]> wrote: > > Because that way the data type is more generic. If, in the future, it is > > decided to be changed from int to something else, you won't have to change > > anything in your code. This is a foundamental concept in ADTs and > > generally a good way of programming. > > > > > > > I wonder why atomic_t is defined as a structure containing an integer > > > rather than just a simple integer? > > > What are the advantages? > > > > > > Hope somebody can explain this, please cc me. > > >
I would like to add to the other replies the following: even if the atomic_t type would be int on every architecture, without special requirements (24 bit ...), and this would be guaranteed not to change in the future, even then there is a point in using atomic_t instead of int. The point is that atomic_t is a so-called opaque type, with the purpose of hiding the actual type (int) to the user. The only operations you are allowed to do on atomic_t variables are through the atomic interface, for example atomic_inc(). You should never do integer operations on variables of type atomic_t, like atomic_t my_atom; my_atom = 3; my_atom++; because you would loose the atomicity.If you would know that atomic_t is actually an int, you could be tempted to do it this way. By making atomic_t a single-field structure, this type of programming is even made impossible, and trying to to my_atom++ will be caught by the compiler. Best regards, Thomas -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to [EMAIL PROTECTED] Please read the FAQ at http://kernelnewbies.org/FAQ
