Bugs or features?

[Carlos Carreras]

        In the C standard, the behaviour of right shift of negative
        numbers is
        undefined.  But in practice, almost every compiler defines it as
        rounding towards -oo.  The one exception I am aware of is Cray's
        compilers for their vector systems.  (They shift the numbers
        logically
        always, which is not very helpful.)
        
        I am inclined to changing gmpxx.h to make it follow the common
        behaviour, although that is not formally standardized.
        
        What do people think?  I am mainly interested in the opinion of
        people
        that actually use gmpxx.h.
        
        (The change is trivial, just change the t to an f in
        mpz_tdiv_q_2exp
        in gmpxx.h.)

The term arithmetic shift does not mean shifts are arithmetic
operations. I'd rather keep shifts (even if arithmetic) as bitwise
operations emulating 2's complement (i.e. >> implemented as
floor(a/2^c)).

I program in C++ and use shifts also on floating-point types, and
combine native and gmp types in the algorithms. If the question refers
to operator>>= (I don't remember the original post), I prefer functions
(rather than operators) to multiply and divide by powers of two, so I
can define them also for the native types and use the same function for
all types. This way, I can specify types as template arguments and use
the same code for all types. So far, the compiler doesn't let me define
operator>> for double...

Carlos