public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug other/49666] New: passing nested function to inline function causes a trampoline call
@ 2011-07-07  3:46 kevin.waugh at gmail dot com
  2011-07-07  9:22 ` [Bug tree-optimization/49666] " rguenth at gcc dot gnu.org
  0 siblings, 1 reply; 2+ messages in thread
From: kevin.waugh at gmail dot com @ 2011-07-07  3:46 UTC (permalink / raw)
  To: gcc-bugs

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49666

           Summary: passing nested function to inline function causes a
                    trampoline call
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: enhancement
          Priority: P3
         Component: other
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: kevin.waugh@gmail.com


#include <stdio.h>
static inline int identity(int (*f)()) {
  return f();
}

int n = 0;
int f() {
  return n;
}

int main() {
  int m = 0;
  int g() {
    return m;
  }
  printf("%d\n", identity(f)); /* inlines all the way */
  printf("%d\n", identity(g)); /* calls g through trampoline */
  printf("%d\n", g()); /* inline's g */
}


^ permalink raw reply	[flat|nested] 2+ messages in thread

* [Bug tree-optimization/49666] passing nested function to inline function causes a trampoline call
  2011-07-07  3:46 [Bug other/49666] New: passing nested function to inline function causes a trampoline call kevin.waugh at gmail dot com
@ 2011-07-07  9:22 ` rguenth at gcc dot gnu.org
  0 siblings, 0 replies; 2+ messages in thread
From: rguenth at gcc dot gnu.org @ 2011-07-07  9:22 UTC (permalink / raw)
  To: gcc-bugs

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49666

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |missed-optimization
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2011.07.07 09:22:08
          Component|other                       |tree-optimization
     Ever Confirmed|0                           |1

--- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-07-07 09:22:08 UTC ---
We don't know how to "devirtualize" trampolines.  In fact trampoline setup
isn't very SSA analyzer friendly (look at the hoops IPA points-to analysis
jumps through).  We end up with

<bb 2>:
  __builtin_init_trampoline (&FRAME.0.g, g, &FRAME.0);
  FRAME.0.m = 0;
  D.3251_20 = n;
  printf (&"%d\n"[0], D.3251_20);
  D.3230_3 = __builtin_adjust_trampoline (&FRAME.0.g);
  D.3231_4 = (int (*<T50a>) (void)) D.3230_3;
  D.3245_18 = D.3231_4 ();
  printf (&"%d\n"[0], D.3245_18);

Now we could "devirtualize" the above based on the fact the trampoline
doesn't escape.


^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2011-07-07  9:22 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2011-07-07  3:46 [Bug other/49666] New: passing nested function to inline function causes a trampoline call kevin.waugh at gmail dot com
2011-07-07  9:22 ` [Bug tree-optimization/49666] " rguenth at gcc dot gnu.org

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).