This is the mail archive of the gcc@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: g++ pointer tracking question


> Does this fix the problem?

Yes, my code correctly compiles with this change.  On IA64 HP-UX, in
ILP32 mode, it does not do an optimization that I want because the
pointer that the offset is being added to is not marked as a pointer but
that may be a bug in my own IA64 specific code or it could be a bug in
the more generic POINTERS_EXTEND_UNSIGNED code, I am not sure.  When I
generate LP64 code everything seems marked correctly as to whether it is
a pointer or not.  At any rate, while I miss an opportunity for a
platform specific optimization, I do not generate any bad code, so I
like the change and think it is better then my local change which was to
just turn off the optimization entirely when compiling C++.  This will
only turn off the optimization in certain cases where it is known that
it cannot do it safely.

Steve Ellcey
sje@cup.hp.com


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