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]

-fpic and calls.c; sibcalls


This question concerns 3.4 BIB:

Maybe I'm not thinking clearly right now, but I can not understand why
we currently check for PIC-code in the machine descriptions (machine.c),
rather than in calls.c when considering sibcall candidates.

For example, in i386.c there are a lot of if-statements checking whether
the code is PIC, but wouldn't it be more sensible having this checked
merely once in calls.c, rather than consulting every platform supporting
sibcalls?

Just a thought...  I'd be interested to hear the rationale behind it,
even if it means I did not read the sources well enough.  :-)  Because to
me, it seems as if no architecture would allow sibcalls in PIC-code
anyway; correct?

Thank you in advance,
Andi.


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