This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]


Hi Jim, and thanks for the informative reply.

> > Is it possible to get FUNCTION_OUTGOING_VALUE to return something else
> > than a fixed value based on the type of the return operand? I'd like to
> > have all functions returning their value into a pseudo-register instead
> > of a fixed hard one.
> Using a pseudo for the return value makes no sense.  The only way to
> find the value after the function exits is to make sure it is in a known
> fixed location.  A pseudo is not a known fixed location.

The trick is my function calls on my target is done this way:

result <- function

i.e. the result of a function is directly put into the destination operand. So 
it is not a fixed place, and being obliged to use a fixed place with GCC I 
get suboptimal code.

> There is also an optimization problem here.  A store into a pseudo will
> be deleted as dead if it isn't used, so if you put the return value in a
> pseudo, the optimizer will delete it as dead.

The idea behind is of course to use this value after ;)


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]