Patch to allow targets to prevent inlining

Jeffrey A Law law@cygnus.com
Tue Feb 29 09:08:00 GMT 2000


  In message < 200002281824.KAA01951@elmo.cygnus.com >you write:
  > : Sounds basically sensible, but I suspect we would still need an override.
  >   
  > : What if a target tags every function with an attribute by default (I'm 
  > : thinking of something like the -mlong-calls patch that is being worked on
  >  
  > : for the arm?  Then we'd never get inlining.
  > 
  > I agree with Richard.  I think that even if we do default to not
  > inlining functions with target specific attributes we are still going
  > to need the equivalent of a TARGET_CAN_INLINE_P macro to override this
  > default.
  > 
  > I am happy to rewrite my patch to go this way if this is what the
  > concensus suggests.
Let's go this way.  Feel free to create a macro which indicates that a
function with a machine specific attribute can be inlined.

[ TARGET_CAN_INLINE_P is a marginal name, since what we're really checking
  is whether or not a target specific *attribute* should not inhibit inlining,
  not whether the target supports inlining. ]

jeff



More information about the Gcc-patches mailing list