This is the mail archive of the gcc-patches@gcc.gnu.org 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]

Re: [patch] Fix PR tree-optimization/47056


> OTOH, is writing/reading to/from a function descriptor (or text) not rather
> similar to a call to it than to taking its address?  I think making it

Adding fake calls at places where call does not happen is also bad idea.  IPA
passes expect call edges to correspond to call statemnets and will try to do
stuff with them.

I didn't noticed it is with ia64 and function descriptors.  Here read from
function descriptor indeed makes more sense then read from random place in text
segment on other arches.  What exactly is the code sequence doing with the
descritor?  We might add read edges for functions for this reason, but it seems
to me that only reason to read function descriptor is to call it indirectly
later, so it is equivalent of taking address anyway?


Honza
> address-taken
> will make all indirect calls possibly calling this function whereas "calling it"
> would merely make sure it will stay around.
> 
> Richard.
> 
> > Honza
> >>
> >> Richard.
> >>
> >> >
> >> > Honza
> >> >
> >


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